<%BANNER%>

Strain Relaxation and Solid Phase Epitaxial Regrowth in Ion Implanted Strained Silicon and Strained Silicon Germanium

Permanent Link: http://ufdc.ufl.edu/UFE0021930/00001

Material Information

Title: Strain Relaxation and Solid Phase Epitaxial Regrowth in Ion Implanted Strained Silicon and Strained Silicon Germanium
Physical Description: 1 online resource (181 p.)
Language: english
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: amorphization, implantation, si, sige, sper, strain
Materials Science and Engineering -- Dissertations, Academic -- UF
Genre: Materials Science and Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The relaxation process of ion-implanted strained silicon films and strained Si1-xGex alloys was studied to determine the magnitude of critical strain necessary for the breakdown of solid phase epitaxial recrystallization in both biaxial tension and compression. Tensile strained silicon layers 50 nm thick were grown via Molecular Beam Epitaxy on relaxed Si1-xGex virtual substrates. Substrate alloy compositions ranged from 10 to 30% Ge. Compressively strained 50 nm Si1-xGex layers were grown on Si substrate via Chemical Vapor Deposition with Ge compositions ranging from 16 to 26%. All samples underwent a 5, 12, or 18 keV Si+ implant at a fluence of 1x10^15 atoms/cm^2 to generate amorphous layers ~15, 30, or 40 nm thick, confining them within the strained layers. The regrowth process, defect morphology, and the effect of implant damage proximity to the Si/SiGe interface was then studied between 500 and 800 degrees C. Strain relaxation of the layers post processing was quantified by High-Resolution X-Ray Diffraction rocking curves and reciprocal space maps. Upon annealing, the solid phase epitaxial regrowth (SPER) process broke down for the highest level of tensile strain and for all levels of compressive strain. Additionally, regrowth related defects were observed in the relaxed samples using cross-section and plan-view Transmission Electron Microscopy (TEM). In tension, regrowth related defects were nucleated as the amorphous-crystalline front advanced to the surface. Once regrowth was complete, the regrowth related defects propagated down to the strained interface and formed stacking faults which promoted further relaxation. In compression, the advancing amorphous-crystalline front roughened and nucleated an extended dislocation network. The density of these dislocations were stable and did not depend on temperature or duration of anneals. The results from this study conclude that the SPER process can be achieved without strain loss or defect nucleation for moderate strain values in tension. However, in compression all strain levels in this study nucleated defects and exhibited strain relaxation.
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, 2008.
Local: Adviser: Jones, Kevin S.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-05-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0021930:00001

Permanent Link: http://ufdc.ufl.edu/UFE0021930/00001

Material Information

Title: Strain Relaxation and Solid Phase Epitaxial Regrowth in Ion Implanted Strained Silicon and Strained Silicon Germanium
Physical Description: 1 online resource (181 p.)
Language: english
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: amorphization, implantation, si, sige, sper, strain
Materials Science and Engineering -- Dissertations, Academic -- UF
Genre: Materials Science and Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The relaxation process of ion-implanted strained silicon films and strained Si1-xGex alloys was studied to determine the magnitude of critical strain necessary for the breakdown of solid phase epitaxial recrystallization in both biaxial tension and compression. Tensile strained silicon layers 50 nm thick were grown via Molecular Beam Epitaxy on relaxed Si1-xGex virtual substrates. Substrate alloy compositions ranged from 10 to 30% Ge. Compressively strained 50 nm Si1-xGex layers were grown on Si substrate via Chemical Vapor Deposition with Ge compositions ranging from 16 to 26%. All samples underwent a 5, 12, or 18 keV Si+ implant at a fluence of 1x10^15 atoms/cm^2 to generate amorphous layers ~15, 30, or 40 nm thick, confining them within the strained layers. The regrowth process, defect morphology, and the effect of implant damage proximity to the Si/SiGe interface was then studied between 500 and 800 degrees C. Strain relaxation of the layers post processing was quantified by High-Resolution X-Ray Diffraction rocking curves and reciprocal space maps. Upon annealing, the solid phase epitaxial regrowth (SPER) process broke down for the highest level of tensile strain and for all levels of compressive strain. Additionally, regrowth related defects were observed in the relaxed samples using cross-section and plan-view Transmission Electron Microscopy (TEM). In tension, regrowth related defects were nucleated as the amorphous-crystalline front advanced to the surface. Once regrowth was complete, the regrowth related defects propagated down to the strained interface and formed stacking faults which promoted further relaxation. In compression, the advancing amorphous-crystalline front roughened and nucleated an extended dislocation network. The density of these dislocations were stable and did not depend on temperature or duration of anneals. The results from this study conclude that the SPER process can be achieved without strain loss or defect nucleation for moderate strain values in tension. However, in compression all strain levels in this study nucleated defects and exhibited strain relaxation.
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, 2008.
Local: Adviser: Jones, Kevin S.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-05-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0021930:00001


This item has the following downloads:


Full Text

PAGE 1

1 STRAIN RELAXATION AND SOLID P HASE EPITAXIAL REGROWTH IN ION IMPLANTED STRAINED SILICON AN D STRAINED SILICON GERMANIUM By MICHELLE S. PHEN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008

PAGE 2

2 2008 Michelle Phen

PAGE 3

3 In loving memory of my grandfather and fellow engineer, Jack A. DSousa

PAGE 4

4 ACKNOWLEDGMENTS I acknowledge the Sem iconductor Research Corporation (SRC) and Intel/SRC scholarship for financial support for this work and my graduate studies. To all of my industrial liaisons who have played a vital role in my professional de velopment and have extended assistance with acquisition of materials and implants, I exte nd the utmost gratitude to Lenny Rubin, Hans Weijtmans, and Karuppanan Sekar. I am also inde bted to A.N. Larsen and J.L. Hansen at the University of Aarhus, Denmark for growing the stra ined Si on relaxed SiGe films. I also thank my committee member, Kevin Jones, Mark Law, Stephen Pearton, Valentin Craciun, and Ant Ural for their support. I extend my appreciation to all the memb er of the SWAMP group whom were always willing to assist and nurture me as a young resear cher. I thank Scott Thompson for many helpful discussion regarding industry and future prospe cts. I especially thank Robert Crosby for mentoring me, Renata Camillo-Castillo for her friendship and assistance with perfecting my WBDF imaging skills, Erik Kuryliw and Mark Clark with assistance in developing my first research project. I also tha nk Ljubo Radic for his friendship a nd assistance with Floops, Chad Lindfors and Jian Chen for teach ing me how to use the Hall sy stem, Heather Randall, Nicole Staszkiewicz, Jeannette Jacques, and Diane Hi ckey for their friendship and encouragement during the most challenging year s of my studies. Furthermore, I thank Tony Saavedra and Danny Zeenberg for many helpful discussions and their support along the way. I also thank Sam Moore for editing this document and for his su pport, encouragement, and friendship throughout the years. I thank the current members of th e group who have fostered a pleasant working environment both in the lab and the office. These members include David Horton, Saurabh Morarka, Nick Rudawski, Leah Edelman, Nicole Rowsey, Sidan Jin, Ray Holzworth, and Lucia Romano.

PAGE 5

5 During the course of my studies I held two internships with Texas Instruments, Inc. I meet a number of people there who have influen ced me and have contributed to my professional growth. I thank the following who work for the High Performance Analog (HPA) packaging group: Tony Coyle and Jeffrey Holloway for ment oring me and teaching me so much about both package development and time management, and KK for his encouragement and inspiration. I also thank my former co-worker at Silicon Technology Development (SiTD). They include Hans Weijtmans and Mark Visokay for being great mentors, and Elizabeth Marley-Koontz and Majid Mansoori for many helpful discussions regarding future experiments and techniques. In the course of my experiments I used a variety of characterizat ion methods at Major Analytical Instrumentation Center (MAIC). The MAIC staff has always been very helpful and I extend my deepest appreciation. I especially th ank, Amelia Dempere, the director of MAIC, Kerry Siebein for assistance and training on the TE Ms, Jerry Bourne for assistance with the FIB, Valentin Craciun for many helpful discussion an d training on new analysis methods using the XPert system, and finally, Rosabel Ruiz for assi stant with administrative matters at MAIC. Finally, I thank my family for their love a nd encouragement, especially my parents for always believing that I could accomplish anything.

PAGE 6

6 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........9 LIST OF FIGURES.......................................................................................................................10 ABSTRACT...................................................................................................................................15 CHAP TER 1 INTRODUCTION..................................................................................................................17 Motivation...............................................................................................................................17 Objective.................................................................................................................................19 Dissertation Organization...................................................................................................... .19 Background and Literature Review........................................................................................ 20 Ion Implantation.............................................................................................................. 20 Point defects.............................................................................................................21 Amorphization.......................................................................................................... 21 Solid phase epitaxial regrowth.................................................................................22 Extended defects......................................................................................................23 Si/SiGe Heterostructures.................................................................................................24 Strained Structures................................................................................................... 25 Misfit Dislocations...................................................................................................26 Critical thickness...................................................................................................... 30 Morphology during growth......................................................................................33 Thermal stability...................................................................................................... 33 SPER and Strain..............................................................................................................34 Strained silicon SPER..............................................................................................34 Strained silicon germanium SPER........................................................................... 35 Relaxed silicon germanium SPER........................................................................... 36 Outstanding Questions............................................................................................................37 2 EXPERIMENTAL METHODS.............................................................................................. 47 Material Processing................................................................................................................47 Growth Conditions: Strained Si on Relaxed SiGe..........................................................47 Growth Conditions: Strained SiGe on Si.........................................................................48 Ion Implantation and Annealing...................................................................................... 48 Material Characterization.......................................................................................................49 Transmission Electron Microscopy................................................................................. 49 X-Ray Diffraction............................................................................................................54

PAGE 7

7 3 TENSILE STRAIN: DEFECT NUCLEATION AND TEMPERATURE DEPENDENCE ..................................................................................................................... .71 Experimental Design............................................................................................................ ..72 Results and Discussion......................................................................................................... ..75 As-Grown and As-Implanted Sample Analysis.............................................................. 75 As-Implanted Sample Analysis....................................................................................... 75 Critical Strain: SPER Breakdown Study.........................................................................76 Defect Nucleation Study.................................................................................................. 81 Relaxation: Thermally Activ ated Glide Process Study ................................................... 83 Conclusion..............................................................................................................................85 4 TENSILE STRAIN: PROXIMITY EFFECT.......................................................................104 Experimental Design............................................................................................................ 104 Results...................................................................................................................................107 As-Implanted Sample Analysis..................................................................................... 107 Proximity Effect Experiment.........................................................................................107 Discussion.............................................................................................................................109 Conclusions...........................................................................................................................111 5 COMPRESSIVE STRAIN: DEFECT NU CLEATION AND STRAIN RELAXATION .... 116 Experimental Design............................................................................................................ 116 Results and Discussion......................................................................................................... 118 As-Grown Sample Analysis.......................................................................................... 118 As-Implanted Sample Analysis..................................................................................... 119 Defect Nucleation/Interface Roughness Study.............................................................. 119 Temperature Dependence Study.................................................................................... 121 Conclusion............................................................................................................................124 6 COMPRESSIVE STRAIN: PROXIMITY EFFECT............................................................137 Experimental Design............................................................................................................ 137 Results...................................................................................................................................138 As-Implanted Sample Analysis..................................................................................... 138 Proximity Effect Experiment.........................................................................................139 Discussion.............................................................................................................................142 Conclusion............................................................................................................................145 7 SUMMARY AND COMPARISION.................................................................................... 156 Overview....................................................................................................................... ........156 Summary...............................................................................................................................156 Tensile Strain Case........................................................................................................156 Compressive Strain Case...............................................................................................159 Comparison..................................................................................................................... ......163 Future Work..........................................................................................................................163

PAGE 8

8 APPENDIX A STRESS-STRAIN CONVERSION..................................................................................... 165 B AMORPHOUS-CRYSTALLINE INTERFACE RELAXATION...................................... 166 C BORON CLUSTER EXPERIMENT: EF FECT OF EOR POPULATION ......................... 168 LIST OF REFERENCES.............................................................................................................174 BIOGRAPHICAL SKETCH.......................................................................................................181

PAGE 9

9 LIST OF TABLES Table page 5-1. Summary of XRR and RC spectra simu lations com pared to XTEM measurements......... 126 A-1. Stress-strain conversi ons for strained Si on Si1-xGex.........................................................165 A-2. Stress-strain conversions for strained Si1-xGex on Si.........................................................165

PAGE 10

10 LIST OF FIGURES Figure page 1-1. XTEM image of P-type a nd n-type MOSFET channel [Tho04]. .........................................39 1-2. Arrheiuns plot of regrowth rate for Si sam ples oriented along <111>, <110>, <100> directions [Cse78]............................................................................................................. .40 1-3. Valence band of (a) unstrained and (b) strained Si showing decreased in light hole effective m ass for strained case [Tho04]........................................................................... 41 1-4. Binary phase diagram of Si-Ge system showing com plete solid solubility [Mas90]........... 42 1-5. Schematic representation of strained f ilm interface versus a relaxed film interface............ 43 1-6. Microstructure of mi s fit dislocation with threadi ng arms in cross-section ......................... 44 1-7. Si cap critical thickness as a functi on of Ge% in the unifor m SiGe layer [Sam99].............45 1-8. Critical thickness as a func tion of Ge concentration [Peo85]. ............................................. 46 2-1. Schematic representation of biaxial tensile strain stru cture used in this work. .................... 61 2-2. Schematic representation biaxial com pressive strain stru cture used in this work................ 62 2-3. Principle behind weak beam dark fi eld im aging technique in TEM for an edge dislocation. High intensity occurs close to dislocation core because planes are bent back to Bragg condition [Wil96].......................................................................................63 2-4. Strained (left) and relaxed (right) laye rs for film layer in compression (top) and in tension (bottom) [Bir03].................................................................................................... 64 2-5. Schematic represen tation of (a) symmetric /2 scan, (b) rocking cu rve scan and (c) radian scan in plane of momentum transfer [Bir03].......................................................... 65 2-6. Combined plot of Qx,Qz plane and positio n of Bragg peaks for an epitaxial film on Si (001) substrate [Bir03].......................................................................................................66 2-7. Two ways by which reciprocal space maps m ay be recorded are by either (a) subsequent rocking curves or (b) subsequent radial scans [Bir03].................................... 67 2-8. Representation of a reciprocal space map s hows the relative positions of the layer with respect to the substrate for fully strained and completely relaxed system in reciprocal space [Pie04]......................................................................................................................68 2-9. (113) Reciprocal space map of as-grown Si on Si0.7Ge0.3 indicating the location of each peak....................................................................................................................................69

PAGE 11

11 2-10. Relative d spacings and precision acco rding to Bartel [Bar83, Bir03].............................. 70 3-1. HRXRD (004) rocking curve of a ll as-grown strained Si on relaxed Si1-xGex samples.......87 3-2. XTEM of as-implanted strained Si/ Si0.7Ge0.3 sample implanted with 12 keV Si+..............88 3-3. HRXRD (004) rocking curves of strained Si/Si0.7Ge0.3 structures annealed at 500 C for 30 minutes................................................................................................................. ...89 3-4a. HRXRD (113) reciprocal spa ce m ap of as-grown strained Si/Si0.7Ge0.3 structure.............90 3-4b. HRXRD (113) reciprocal space m ap of strained Si/Si0.7Ge0.3 structure implanted with 12 keV Si+ and annealed at 500 C for 30 minutes............................................................ 91 3-4c. HRXRD (113) reciprocal space m ap of strained Si/Si0.7Ge0.3 structure implanted with 12 keV Si+ and annealed at 800 C for 30 minutes............................................................ 92 3-5. Strain Relaxation ca lculated via HRXRD RSM for strained Si on relaxed Si1-xGex samples implanted with 12 keV Si+ and annealed for 30 minute at temperatures indicated.............................................................................................................................93 3-6. XTEM (top row) and PTEM (bottom row) for all strained Si on relaxed Si1-xGex samples implanted with 12 keV Si+ and annealed for 30 minutes at 650 C..................... 94 3-7. PTEM images for AG, AG plus anneal, and implanted with 12 keV Si+ plus anneal for all strained Si on relaxed Si1-xGex samples. All anneals carried out for 30 minutes at 800 C................................................................................................................................95 3-8. Linear defect density quantified using PTEM as a function of Ge concentration after 30 m inute anneal at 800 C................................................................................................ 96 3-9. Percent strain relaxation quantified us ing H RXRD SM measurements as a function of Ge concentration after a 30 minute anneal at 800 C........................................................97 3-10. PTEM images for Si cont rol and strained Si on relaxed Si1-xGex samples implanted with 12 keV Si+ and annealed for 5, 30, and 300 minutes at 800 C................................. 98 3-11. Trapped interstitial concentration as a function of anneal tim e for Si, Si on Si0.8Ge0.2, and Si on Si0.7Ge0.3 implanted with 12 keV Si+ and annealed at 800 C........................... 99 3-12. XTEM of strained silicon layer of the Si0.7Ge0.3 implanted with 12keV Si+ and annealed at (a) and (b) 500 C for 30 minut es and for (c) and (d) 45 minutes. XTEM are imaged using (a) and (c) bright-field and in (b) and (d) dark-field mode..................100 3-13a. XTEM of stacking faults observed in the strained silicon layer of the Si0.7Ge0.3 sample implanted with 12 keV Si+ and annealed at 800 C for 30 minutes.................... 101

PAGE 12

12 3-13b. High-Resolution XTEM of stacking faults observed in the strained silicon layer of the Si0.7Ge0.3 sample implanted with 12 keV Si+ and annealed at 800 C for 30 minutes.............................................................................................................................102 3-14. Plot of ln relaxation rate vs. 1/kT for 12keV Si+ implanted strained silicon on relaxed Si0.7Ge0.3 sample. Linear regression of data is shown in red.......................................... 103 4-1. PTEM images for (a) Si control sample and Si on Si0.7Ge0.3 samples implanted with (b) 5 keV Si+ (c) 12 keV Si+ and (d) 18 keV Si+ and annealed for 30 minutes at 800 C......................................................................................................................................112 4-2. Strain relaxation (circles ) and linear defect density (squ ares) as a function of implant energy after 800 C 30 minute anneal. Zero im plant energy indicates data for asgrown plus anneal case.................................................................................................... 113 4-3. PTEM images for strained Si on Si0.7Ge0.3 samples implanted with 5, 12, and 18 keV Si+ and annealed for 5, 30, and 300 minutes at 800 C.................................................... 114 4-4. Linear dislocation de nsity quantified usi ng PTEM for strained S i on Si0.7Ge0.3 samples implanted with 5, 12, and 18 ke V Si+ and annealed for 5, 30, and 300 minutes at 800 C............................................................................................................. 115 5-1. XRD (004) rocking curves for all as-grown strained SiGe on Si samples. ........................127 5-2. Cross-section of as-implant ed strained SiGe on Si sam ples............................................... 128 5-3. XTEM of strained Si1-xGex on Si samples implanted with 12keV Si+ annealed at 500 C for 45 minutes (a) x= 16 (b) x= 26. Arrows indicate a-c interface and hatched white lines indicate epitaxial interface............................................................................. 129 5-4. PTEM image of strained Si0.74Ge0.26 on Si samples implanted with 12 keV Si+ and annealed for 45 minutes at 500 C................................................................................... 130 5-5. PTEM image of strained Si0.74Ge0.26 on Si samples implanted with 12 keV Si+ and annealed showing region etched down from surface to the SiGe/Si interface, reveling the misfit dislocations that cont ribute to strain relaxation...............................................131 5-6. PTEM images of strained Si0.74Ge0.26 on Si samples implanted with 12 keV Si+ and annealed at (a) 500C (b) 575C (c) 650C (d) 725C (e) 800C all for 30 minutes........ 132 5-7. Linear dislocation density qu antified using PTEM for strained S i0.74Ge0.26 on Si samples implanted with 12 keV Si+ and annealed for 30 mi nutes at temperatures ranging from 500 to 800 C..............................................................................................133 5-8. XRD RC spectra for strained Si0.74Ge0.26 on Si samples implanted with 12 keV Si+ and annealed for 30 minutes at temperat ures ranging from 500 to 800 C............................ 134

PAGE 13

13 5-9. HRXRD (113) reciprocal space m ap for strained Si0.74Ge0.26 on Si samples implanted with 12 keV Si+ and annealed at 800 C for 30 minutes.................................................. 135 5-10. Average strain relaxati on from RC data for strained S i0.74Ge0.26 on Si samples implanted with 12 keV Si+ and annealed for 30 minutes at temperatures ranging from 500 to 800 C...................................................................................................................136 6-1. Amorphous depth measurements using XTEM for all strained Si1-xGex on Si samples implanted with 5, 12, and 18 keV Si+ shown as a function of Ge concentrations........... 148 6-2. All 12keV Si+ implanted with varying Ge con centration (a) 16% (b) 22% (c) 26% XRD RC on the left with corre sponding XTEM on the right..........................................149 6-3. (a) XRD (004) RC of 5, 12, and 18keV 16% Ge sam ples. PTEM of 16% Ge sample implanted with (b) 5keV (c) 12keV and (d) 18keV Si+ implants. All samples annealed for 30 minutes at 800 C................................................................................... 150 6-4. PTEM of 5, 12, and 18keV Si+ implanted samples annealed at 800 C for 30 minutes as a function of germanium concentration.......................................................................151 6-5. Linear dislocation density (filled mark ers) a nd %Strain Relaxation (unfilled markers) for 5, 12, and 18keV Si+ implanted samples annealed at 800 C for 30 minutes as a function of germanium concentration.............................................................................. 152 6-6. Linear disloca tion density of 18keV Si+ implanted samples anne aled at 800 C for all germanium concentration.................................................................................................153 6-7. Summary of %Strain relaxation obtained fro m XRD RC data for all samples annealed at 800 C 30 minutes........................................................................................................ 154 6-8. Critical thickness, measured using X TEM, for all sam ples conditions annealed at 800 C for 30 minutes............................................................................................................. 155 B-1. XTEM of samples implanted with B18H22 cluster (a) 26% Ge annealed for 400 C 60 minutes (b) 26% Ge annealed for 400 C 60 minutes followed by 500 C 30 minutes (c) 30% Ge annealed for 400 C 60 minutes (d) 30% Ge annealed for 400 C 60 minutes followed by 500 C 30 minutes. Arrows indicate a-c interface and hatched white lines indicate epitaxial interface............................................................................. 167 C-1. XTEM of as-implanted with B18H22 cluster at equivalent ener gy of 4 keV into Si (100) substrate...........................................................................................................................170 C-2. High-resolution XTEM of as-implanted B18H22 at equivalent energy of 4 keV into Si (100) substrate showi ng a rough a-c interface................................................................. 171 C-3. PTEM images of (a) strained Si0.74Ge0.26 on Si implanted with 5 keV Si+ and (b) implanted with B18H22 and of (c) strained Si on Si0.7Ge0.3 implanted with 5 keV Si+ and (d) implanted with B18H22. All samples annealed at 800 C for 30 minutes............ 172

PAGE 14

14 C-4. Strain relaxation quantified using XRD da t a for highest tensile and compressive strain samples implanted with 5 keV Si+ and 4keV equivalent B18H22 annealed for 30 minutes at 800 C.............................................................................................................173

PAGE 15

15 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy STRAIN RELAXATION AND SOLID P HASE EPITAXIAL REGROWTH IN ION IMPLANTED STRAINED SILICON AN D STRAINED SILICON GERMANIUM By Michelle Phen May 2008 Chair: Kevin Jones Major: Materials Science and Engineering The relaxation process of ion-implanted strained silicon films and strained Si1-xGex alloys was studied to determine the magnitude of criti cal strain necessary fo r the breakdown of solid phase epitaxial regrowth in both biaxial tension and compression. Tensile st rained silicon layers 50 nm thick were grown via Mol ecular Beam Epitaxy on relaxed Si1-xGex virtual substrates. Substrate alloy compositions ranged from 10 to 30% Ge. Compressively strained 50 nm Si1-xGex layers were grown on Si substrate via Chemical Vapor Deposition with Ge compositions ranging from 16 to 26%. All samples underwent a 5, 12, or 18 keV Si+ implant at a fluence of 1x1015 atoms/cm2 to generate amorphous layers ~15, 30, or 40 nm thick, confining them within the strained layers. The regrowth process, def ect morphology, and the effect of implant damage proximity to the Si/SiGe interface was th en studied between 500 and 800 C. Strain relaxation of the layers post pro cessing was quantified by High-Resolution X-Ray Diffraction rocking curves and reci procal space maps. Upon ann ealing, the solid phase epitaxial regrowth (SPER) process broke down for the highest level of tensile strain and for all levels of compressive strain. Additionally, regrowth relate d defects were observed in the relaxed samples using cross-section and plan-v iew Transmission Electron Microscopy (TEM). In tension,

PAGE 16

16 regrowth related defects were nucleated as the amorphous-crystalline fr ont advanced to the surface. Once regrowth was complete, the regr owth related defects propagated down to the strained interface and formed stacking faults which promoted further re laxation. In compression, the advancing amorphous-crystal line front roughened and nuclea ted an extended dislocation network. The density of these dislocations were stable and did not depend on temperature or duration of anneals. The results from this study conclude that the SPER process ca n be achieved without strain loss or defect nucleation for moderate strain va lues in tension. However, in compression all strain levels in this study nucleated defects and exhibited strain relaxation.

PAGE 17

17 CHAPTER 1 INTRODUCTION The sem iconductor industry, encompassing both silicon and compound materials, has sales in excess of $200 billion dollars per year. While compound semiconductors, specifically III-V technologies, show superior performance to silico n, silicon still dominates the market due to its abundance and low manufacturing costs [Mil05]. Decreasing the manufacturing cost to improve both capability and profit margins has driven the miniaturization of transistor size, which, in turn, decreases the cost per chip while improving device performance. This trend has doubled the number of transistors on a chip every 18 to 24 months, following the prediction of Moores Law [Moo65]. Until recently, this trend has been satisfied by decreasing (scaling down) the transistor in size. The limit of scaling, however, is being approached and other options to improve device performance must be investigat ed. One such option is the use of strain engineering in the channel region of the devi ces. The limitations of strain technology, though, and the impact of strain on the individual fabrication steps are not fully understood. These limitations will be explored in this work. Motivation The processing lim its of strain ed silicon technology, which was introduced in the 90 nm technology node, are of great intere st to researchers and indust ry alike. Strained silicon technology improves performance by increasing carrier mobility in the channel of the device through decreases in both the average effective mass and inter-band scattering [Nay96]. Strain can be introduced using two main approaches, substrate-induced a nd process-induced. The focus of this work will be mainly on process-induced strain. The strain in the P-type Metal Oxide Semi conductor (PMOS) channel is created by using Si1-xGex Source/Drain (S/D) wells. Alternativel y, in N-type Metal Oxide Semiconductor

PAGE 18

18 (NMOS) devices, the strain is induced by silicon nitride overlays forming tensile in the channel region. Examples of both devices are shown in Fi gure 1-1. Strain can also be process-induced via Stress Memorization Techniques (SMT). This utilizes conventional fabrication processing of amorphizing the S/D region. In the stre ss memorization process, the activation and recrystallization anneal is carri ed out post deposition of a nitride tensile stressor capping layer over the gate region. During the regrowth proce ss, the stress induced by the nitride capping layer is memorized [Cha05]. The n itride layer is then removed and silicidation is carried out. The degree of stress is dependent on the thickne ss of the capping layer; the drive current improvement can be increased up to 15% by in creasing the nitride th ickness [Che04]. Also, because of the simplicity of the SMT and no adde d etch steps or mask levels it can be easily incorporated into current Si and Silicon-On-Insulator (SOI) tr ansistor fabrication at low fabrication cost [Hor05, Sin05, Sin06]. Additionall y, in conjunction with other process-induced strain such as dual-stress liners and SiGe or SiC source-drain regions the device performance benefits are additive [Hor05]. Once strained, these device struct ures must undergo additional pr ocessing, in particular the final activation anneal. Subsequent thermal processing can cause the strain energy in the heterostructures to decrease a nd cause strain relaxation by propaga tion of threading dislocations and formation of misfits [Koe01, Sam99, Sug01]. Additionally, relaxation can also be caused by Ge diffusion from the Si1-xGex into the strained region. Ge interdiffusion and ensuing strain relaxation has shown to be dramatic when th e collision cascade of an amorphizing implant overlaps a Si/ Si1-xGex interface [Van05]. In any case, as th e level of strain in the structure decreases due to relaxation and dislocation nucleation, the mob ility enhancements provided by strained silicon technolo gy also decreases. The defects creat ed in the relaxation process also

PAGE 19

19 decreases device performance by ac ting as scattering sites. Thus it is important to understand the exact response of strained regions due to subsequent processing. Objective The effects of perfor ming an amorphizing impl ant contained within a strained layer have not yet been studied, nor has the degree of relaxation, regrowth qua lity, or thermal stability of such an amorphized region been investigated. These effects may be important for future device structures, as arsenic and phosphorus, both self-amorphizing implants are often used to create channel extensions in NMOS devices. Additionally, any effects may also be important for stress memorization techniques as they carry out recr ystallization of amorphous regions under strain. The purpose of this work, therefore, is to study the effect of an amorphizing implant contained within a strained layer as a function of strai n, especially concerning the degree of relaxation, stability after amorphization and re crystallization, crystalline quali ty of the regrown layer, and proximity of implant to the heterostructure interface. Dissertation Organization The contents of this work are organized into seven chapters and three appendixes. T his chapter, Chapter One, outlines the motivation and objectives of this work as well as provides a literature review of pertinent to pics. Chapter Two provides an overview of material deposition and characterization techni ques utilized to carry out the experime nts in this work. The next four chapters discuss experimental results. Chapters Three and Four discuss the work carried out under biaxial tensile strain using st ructures with strained Si on rela xed SiGe virtual substrates. The third chapter discusses the critical strain necessary for SPER breakdown and the mechanism of defect nucleation once the critic al strain has been met. The experiments in this chapter discuss results for samples solely implanted with 12 keV Si+ implants. The fourth chapter discusses the effect of the proximity of the implant to the Si/SiGe interface and how the proximity affects

PAGE 20

20 strain relaxation and defect nucl eation. Samples in this chapte r were implanted with varying energies of 5, 12, and 18 keV to alter the implan t damage proximity to the Si/SiGe interface. Chapters Five and Six mirror the previous two chapters with the exception that the implanted material is under biaxial comp ressive strain using SiGe on Si structures. Chapter Seven summarizes the results from all experiments and compares the tensile and compressive cases. There are three appendices in this work. The first provides a conversion between stress and strain for the heterostructures used in this wor k. The second presents th e results of using a low temperature pre-anneal to planarize and relax th e a-c interface of implanted samples prior to regrowth. The third appendix discusses an experi ment carried out on the highest strained tensile and compressive films. In this experiment, the end-of-range damage from conventional beamline implants was eliminated with the use of octadecaborane cluster implants to generate amorphous layers with the strained epilayer. Background and Literature Review Ion Implantation For the past 50 years, ion im plantation has b een used by the semiconductor industry as the preferred method for incorporating dopants into s ilicon. The process offers advantages including reliability, reproducibility and c ontrol of dopant dose and distri bution. In this non-equilibrium process, ions bombard the host lattice and, thro ugh nuclear and electroni c interactions, lose energy until they come to rest. This process introduces primary defects consisting of a vacancy rich region near the surface and an interstitial rich region deeper within the substrate where the implanted ions come to rest. For the implan ted dopants to be electr ically active, a high temperature anneal is required; this process also allows Frenkel pairs, which are interstitial and vacancy pairs, to recombine. At lower implant doses the host lattice can maintain its integrity with isolated regions of defect s around Rp, the projected range of the incoming ion. However, at

PAGE 21

21 higher doses when more than 10% of the hos t atoms are displaced, the implanted region undergoes a first order crystalline-to-amorphous transition, termed amorphization [Chr81]. The amorphized region is completely damaged and no periodic lattice exists. The defective region in this case lies just beyond the amorphous-crystalline (a-c) interface. Point defects Point defects are categorized into native de f ects and impurity relate d defects. Native defects exist in the crystalline lattice above th e absolute zero temperature. Impurity related defects arise from the incorpora tion of foreign atoms as can be the case in ion implantation processing. There are two types of native point def ects: vacancy and intersti tial. The simplest of the point defects is a vacant lattice site norma lly occupied by an atom, termed a vacancy. An interstitial is an atom that occ upies a site in the lattice that un der ordinary circumstance is not occupied, i.e. a void between host atoms. Diff usion in semiconductors is mediated via point defects; primarily by either an interstitial mechanism or vacancy mechanism. Amorphization Am orphization takes place when a sufficient level of fluence and ion mass has been reached. Si and Ge atoms are often used to creat e amorphized regions in Si substrates [Dea73]. The incoming implanted ions and recoil ions from the surface act as point defects within the host lattice. Upon annealing, thes e point defect agglomerate into {311} defects and dislocation loops which are located just beyond the amorphous/crystalline (a /c) interface. These are termed EndOf-Range (EOR) defects or type II defects [Jon88]. There are some advantages of amorphizing the substrate prior to dopant incorporation. In particular, boron implants for p-type devices ar e usually performed after a pre-amorphizing step to avoid channeling effects. The amorphized region, however, must be then be regrown to restore the host lattice as well as activate the dopants by placing them in substitutional sites.

PAGE 22

22 Other dopants, such as arsenic and phosphorus fo r n-type devices, are self-amorphizing due to their large size. Regions implanted with these io ns will also need to be regrown to restore the lattice and activate the dopants. Solid phase epitaxial regrowth The recrystallization process of an amorphous layer in contact with a crystalline substrate is te rmed Solid Phase Epitaxial Regrowth (SPER). This process requires thermal energy for the rearrangement of the atoms in the amorphous region onto the template provided by the crystalline substrate. Rearrangement of the atoms begins at the a/c interface and progresses towards the surface of the material. The regrow th velocity is influenced by and dependent on temperature [Cse78, Ols88], substrate orientatio n [Cse78], and type of dopant incorporated [Ols88]. Typically, SPER in Si commences around 450 C and proceeds up to temperatures just below the melting point [Poa84]. Use of a SPER process is advantageous in that it yields more ab rupt junctions, less transient enhanced diffusion eff ects, and higher activation at a relatively lower temperature. Disadvantages include poor elec trical characteristics of the ju nction, specifically higher leakage current [Bul79, Lin03]. Substrate orientation. Several experiments were carri ed out by Csepregi et al. [Cse76, Cse78] to study the temperature and substrate orientation dependence of regrowth rate in silicon substrates. Amorphous layers we re created by implantation of Si+ ions at energies ranging from 50 to 250 keV. The samples were annealed and growth velocities were measured via backscattering yield. The growth velocity wa s found to follow Arrhen ius relations and was dependent on the substrate orientation, as show in Figure 1-2 [Cse78]. The growth rate at all orientations was found to have identical activation energies of 2.35 eV over the temperature range of 450-575 C. Csepregi found that an <100> orientation resulted in the fastest growth

PAGE 23

23 velocity, <110> growth was approximately th ree times slower and <111> was the slowest [Cse78]. He proposed a model to explain the orientation dependence. The model proposed a bond-breaking mechanism which transferred atoms at the a/c interface to regul ar lattice sites. The model also suggested that the transference required at least two nearest-neighboring atoms to be on regular lattice sites. Later similar experiments were carried out by Olson et al. using more accurate measurement techniques [Ols85, Ols88]. These e xperiments were carried out by Time Resolved Reflectivity (TRR) measurements which allowe d simultaneous measurement of the amorphous layer depth. This experiment f ound the activation energy to be 2.68 eV and is the most accepted value for SPER in Si. Extended defects Extended defects such as dislocations loops, stacking faults and m icrotwins have been noted after recrystallization [Cse 76]. Dislocation loops were ob served beyond the a/c interface for <100> and <110> oriented substrates whereas microtwins and stackin g faults were observed throughout the regrown layer in the <111> orie nted substrates afte r recrystallization. End of range defects. End-of Range defects (EOR) def ects result after an amorphizing implant and form just below the a-c interface. These defects consist of {311}s at low thermal budgets and dislocations loops at higher thermal budgets. EOR defects are extrinsic in character and have been studied extensively [Cof00, Ea g94, Mau94, Pan96]. The main sources of these defects are transmitted ions that stop below the a-c interface and the recoil of excess interstitials as a result of ion bombardment. Excess interstitials exist after SPER because they were unable to undergo Frenkel pair recombination, since th e shallower depth of the implant where excess vacancies reside is amorphized. The evolution of these defects at various annealing temperatures and under various ambient atmospheres has also been widely studied [Gil99, Liu95]. At

PAGE 24

24 annealing temperatures below 800 C, EOR defect s coarsen and decrease in density at the expense of smaller loops through a process ca lled Ostwald ripening [Bon98]. Above 800 C, these loops become unstable and di ssolve releasing trapped intersti tials [Liu95]. The release of these interstitials may lead to enhanced dopant diffusion, known as Transient Enhanced Diffusion (TED) which can drive an implanted junction deeper than desired [Cow94, Fah89, Hof74, Sto95]. Regrowth-related defects. Imperfect regrowth of the amorphous layer leads to the formation of hairpin dislocatio ns or microtwins, also termed type III defects. Hairpin dislocations, found in {100} oriented substrates, nucleate when the a/c growth front encounters misoriented microcrystalline regions and forms a pe rfect dislocation segment. This segment then wraps around the misoriented material forming a half loop which consists of the base of the hairpin. As annealing continues, the hairpin arms diverge as they advance past the microcrystal regions forming a V shape dislocation [San84]. These defects have been shown to be easily avoidable [Jon88, San84]. On the other hand, microt win formation is observed in {111} oriented substrates during amorphous regrowth. Various models have been proposed to explain the formation of these defects [Cse78, Dro82, Nar82]. Most of these models are based on the bond arrangements of the different orientations. The formation of two distorted bonds defines the difference between an atom in the amorphous and crystalline phase. On the {100} surface, an atom can add anywhere and form two undistor ted bonds. However, the {111} surface requires simultaneous addition of three adjacent atoms. Th ese three atoms can either add in the correct positions or with a twin orientation, forming a microtwin [Jon88]. Si/SiGe Heterostructures Heterostruc tures offers the ability to construct a variety of device configurations and has become the basis behind bandgap engineering [Cap83] At first, this field was dominated by III-

PAGE 25

25 V materials. However, compared to silicon th ese materials are ten times more costly. The obvious solution was to apply the benefits of heterostructures junctions to silicon technology with the use of germanium. The silicon-germanium system offers many advantages such as the ability to alter strain and bandga p by addition of germanium to silicon. The use of this technique has opened the SiGe/Si system to be applied to a variety of applications such as photodetectors, modulation-doped transistors, and heterojunction bipolar transistors. Complimentary metal oxide semiconductor (CMOS) devices dominate the i ndustry and the use of S i/SiGe to this field will be the focus of this work. The use of strained silicon in PMOS devices offers enhanced hole mobility due to a decrease in the average effec tive mass and decreased inter-val ley scattering [Nay96]. Both biaxial tensile and longitudinal compressive strain can be used to lift the degeneracy in the valence band causing it to shift and become light hole like, as illustrated in Figure 1-3 [Tho04]. This decreases the hole effectiv e mass thereby, increasing mobilit y. This can be implemented in the CMOS flow by the use of epitaxially deposite d B-SiGe S/D wells in p-type transistors. Additionally, in situ boron deposition with SiGe growth allows higher dopant activa tion without the need of high temperature activation anneal because B occupies substitutional sites upon deposition. A higher activation concentration of B also lead s to lower contact resistance [Raa99]. Strained Structures The Si/SiGe system possesses several attractive pr operties. First, Ge and Si have similar properties including crystal struct ure, atomic size factor, valen ce, and electronegativity which allow a complete solid solution to be formed wh en they are mixed. The phase diagram of Si-Ge is presented in Figure 1-4. Second, the lattice pa rameter of SiGe is a function of Ge composition that shows a slight deviation from the linear func tion stated in Vegards law [Dis64]. Therefore,

PAGE 26

26 since epitaxial growth of pseudomorphic films causes the lattice consta nt of the film and substrate to be in perfect atomic registry, Figure 1-5, a variable amount of strain can be created in the system by varying the SiGe composition. Biaxial tensile strained silicon is obtained by using Si1-xGex to serve as a virtual substr ate. Additionally, by growing Si1-xGex on Si substrates one can grow a biaxial compressive st rain material. The maximum mismatch between the layers is based on the difference between the equilibrium lattice spacing of Si (5.43 ) and Ge (5.66 ) and has a maximum value of approxima tely 4.2%. The amount of Si that can be grown on top of a SiGe virtual s ubstrate or SiGe on Si, however, mu st be kept below a critical thickness to prevent relaxation [Mat 74, Peo85]. Below this critical th ickness, the strain is stored in the film elastically. Above this thickness, the strain is accomm odated by strain-relieving defects, termed misfit disl ocations, causing relaxation. Misfit Dislocations The elastic strain which is stored in the pseudomorphic film can cau se the formation of interfacial misfit dislocations, whic h act to relieve the elastic stra in after sufficient annealing or critical magnitude of strain. The generati on, propagation, and velocity of these dislocations have been studied extensively [Bea87, Dod88, Dod89, Fri87, Hou91,Hul89, Hul92, Kas75, Peo85, Tsa87,Van63]. The following sections will summarize the prior work involving misfit dislocation microstructure, mechanism of strain re lief, and calculation of the critical thickness. Microstructure. Si, Ge, and Si1-xGex alloys all have diamond cubic lattice structures in which dislocations are known to glide primarily on {111} planes. Geometrically, a dislocation cannot terminate within the bulk of a crystal. In stead, it must terminate upon itself, with another defect, or at the nearest free su rface. Most commonly, misfit disl ocations terminate by forming threading arms that extend to the surface, as show n in Figure 1-6. Misfit di slocations are perfect dislocations with a Burgers vect or of a/2<110>. The dislocations are energetically unstable and

PAGE 27

27 are hence known to dissoci ate into Shockley partials [Hir68, Rea53], according to the reaction in Equation 1-1. a/2<110> = a/6<1-12> + a/6<211> (1-1) The Shockley partials (right si de of equation) are mutually re pulsive and glide away from each other on the {111} glide plane. This causes form ation of a stacking fault. The equilibrium partial spacing, so, is determined by the balance of the repulsive energy betw een the interacting partials and the energy of the st acking fault. The glide motion on the {111} plane also results in an angle, between the Burgers vector and the disloca tion line direction. For the a/2<110> type misfit dislocation the angle is 60 and for the two Shockley part ials, 90 and 30. The latticemismatch stress is resolved differently onto th ese two partials [Mar87]. Thus, the critical resolved shear stress is different for each case; th e stress is higher on the 90 partial than the 30 partial. For a case where the (100) interface is under compressive st rain, the 30 partial leads and the 90 partial trails. The traili ng partial experiences a greater reso lved shear stress than the 30, thus, reducing the partial separation, so, and may result in zero separation. For the tensile case, the 90 partial leads and th e 30 trails, increasing so to levels which can approach infinity. In this case, the separation causes the misfit dislocations to consist of 90 a/2<211> type partials, which leave stacking faults behind as th ey propagate through the crystal; this has been observed in the SiGe/Ge (100) system [Weg90]. Nucleation. Misfit dislocation nucleation mechanism in the SiGe/Si system is still debated among researchers. There are three generic mech anisms for nucleation of misfit dislocations discussed in literature: homoge neous, heterogeneous, and dislocation multiplication events. Homogeneous nucleation occurs when intrinsic st rain is high enough to allow finite rate of dislocation loop nucleation within the epilayer, or half-loop nuclea tion at the free surface. This

PAGE 28

28 requires higher strains to produ ce a significant nucleation rate. The earliest calculation of activation energy for homogeneous dislocation loop nucleation goes back several decades [Hir68]. For the case of the stra ined layer, activation energy is th e sum of the self energy of the dislocation loop and the strain en ergy relaxed by it. For the fr ee surface half-loop, the energy is a function of loop radius. The activation barrie r for half-loop calculated by researcher [Hul89, Weg90] is about 5 eV at 600 C for strain values in excess of 0.02 or a 50% Ge. Heterogeneous nucleation must be considered for lower levels of strain where homogeneous nucleation is not possible. Heterogeneous nucleation occurs at a local site within the crystal where strain is higher thus; the probability of disloca tion nucleation is higher. Such sites correspond to defects such as precipita tes, grain boundaries, contamination, etc. The predominant nucleation source can differ within a given material system, growth condition, and/or growth techniques. Lastly, dislocation multiplic ation nucleation is recognized as the dominant nucleation mechanism at relatively low and high epilayer th icknesses. This type of dislocation nucleation concept goes back to Frank-Read sources [Hir68 ] and the first documented source in strained (Ge/GaAs) layers [Hag78]. In recent years, se veral works have reported multiplication event in the SiGe/Si system [Alb95, Cap92, Tup90]. Typically, the sources for this mechanism require thick epilayers which translate to relatively low strain. Tuppen et al. [Tup90] used layers with minimum thickness of 700 nm and strain of 0.005 (Si0.87Ge0.13). LeGoues and Mooneys group [LeG93, Moo94] determined a Frank-Read-like mu ltiplication source activation energy of 4-5 eV for in SiGe/Si layers with Ge concentrations from 5 to 20% and a thickness of 380 nm. In summary, homogeneous mechanism is thought to dominate at higher strains and dislocation multiplication at lower strains with gr eater epilayer thicknesses. For relatively low

PAGE 29

29 strain and lower thicknesses, only heterogeneous sources are av ailable and misfit dislocations become nucleation limited. Experimentally, there is still a small collection of misfit nucleation mechanism data in the SiGe/Si system. A wider ra nge of strain and epilay er thicknesses data is necessary to fully understand the regime in which each of the nuc leation mechanisms occur. Propagation. The mechanism of misfit propagation in the SiGe/Si heterostructure systems have been well researched and is accep ted to be mediated via dislocation glide. Dislocation glide in bulk Si and Ge crystals from plastic defo rmation is also well documented [Ale68, Far86, Ima83, Pat66]. Pure intrinsic Si a nd Ge crystals have activation energies of 2.2 and 1.6 eV, respectively, at low stress (up to 100 MPa). The pre-expon ential factor for both materials is the same and thus dislocation glide is much faster in Ge than Si. Additionally, glide activation energy in SiGe alloys is expected to decrease with increasing Ge while glide velocity is expected to increase with increasing Ge content. In addition, dislocation glide in bulk SiGe alloys of low Ge or Si concentrations have been reported [Yon96, Yon99] and agree with extrapol ation of experiment from the elemental compounds. Comparatively, an activation energy fo r glide has been documented for a Si0.7Ge0.3 on Si thin film relaxation with a value around 1.1 eV [Hul89, Dod88] and 1.38 eV [Lei01], all values are lower than the 1.8 eV determined for SiGe bulk material. The lower activation energy observed in thin film versus bulk materials coul d be caused by lower formation energy due to the close proximity of the interface and free surface [Hul89]. Since activation energy is the sum of the formation and migration energies, it is reduce d. An alternative explanation by Hull et al. [Hul89] to be more likely in thei r work, is that as the dislocati on density and anneals temperature increases, the propagating dislocations have to cross more and more orthogonal dislocations in a given length. These intersecting events are observed to impede propagation. Thus, the low

PAGE 30

30 observed activation energy would be a function of re duced velocity at high temperatures rather than an enhancement at low temperatures [Hul89]. Dislocation kink model. The Si lattice is periodic and dislocations tend to follow the low index, low energy Peierls valleys of the crysta l. Ideally, dislocations are linear; however, dislocation interactions cause cu rvature. Two straight dislocat ion segments in the same glide plane lying in neighboring Peierls valleys are connected by a kink where the dislocation jumps the Peierls barrier. Thus, disloc ation glide is controlled by the Perierls-Nabaro mechanism, such that a dislocation can move from one Peierl s valley to a neighbori ng one by nucleation and migration of kink pairs along the dislocation. Esse ntially, this results in the kink formation and migration determining the dislocation velocity [Hir68]. Hirth and Lothe describe a double kink model based on Perierls-Nabaro mechanism which is the generally accepted model for dislocation motion in Si/SiGe hete rostructures [Hir68]. Furthermore, this model also predicts a lowering of the activation energy with increased driving stress wh ich has been observed experimentally. Critical thickness Pseudom orphic epilayers are obtained when the thickness of the strained overlay is below a critical thickness and the atoms are arranged in perfect registry on either side of the heterostructure interface [Mat74, Peo85]. This prevents pe rformance-degrading and strain relaxing misfit dislocations with a Burgers vector of a/2 <110> fr om forming at the interface. Many researchers have undertaken work to understand the limitations of strained layer epigrowth. Previous work has shown that there exists a critical thickness at which no misfit dislocations are generated and the strain is accommodated completely by the elastic strain energy. The first to propose a theory for crit ical thickness for mismatch layers was Van der

PAGE 31

31 Merwe [Van63]. His theory was based on an energy balance criteria in which the stability is described by relative energies of two competing interfacial structures First, the lattice mismatch is accommodated by elastic strain only and the second, accommodated by bot h elastic strain and misfit dislocations. The balance of these two energies can yield a critical thickness. The energy to generate a dislocation is described in Equation 1-2. b h Gb Endislocatioln 2 )1(42 (1-2) G is the shear modulus, assumed to be the same in the film and substrate, b is the Burgers vector, s is Poissons ratio and 2/l is the disl ocation length per unit area of the epilayer. The energy stored in strained layer is presented in Equation 1-3. 2MhEstrain (1-3) Where M is the biaxial elastic modulus of the epilayer, e is the stra in and h is the thickness of the epilayer. Matthews and Blakeslee based their calculation on the energy balance criteria of Van der Merwe with one exception. They stated that mi sfit dislocations are ge nerated through the glide of threading dislocations from the underlying subs trate [Mat74]. Additionally, they made some assumptions about the two materials properties: crystal structures are cubi c, elastically isotropic and they both have similar elastic constants. The critical thickness data ca n, therefore, be applied to either Si on SiGe or vice versa. Figure 1-7 shows the critical thickne ss of strained silicon on SiGe calculated using Matthews-Blakeslee theory which is similar to the SiGe/Si system based on the material property assumption. These cal culations can be applied to silicon on Si1-xGex according to these assumptions. For metal films, the theoretical critical th ickness agrees well with the predictions of Matthews-Blakeslee [Kuk83]. This is not the case in semiconductors. Jain et al. [Jai90] offered

PAGE 32

32 two possible explanations for this discrepancy. The first, due to the lack of resolution of the measuring technique and the second suggest ed that equilibrium is not reached under experimental conditions. These two explanations will be discussed below. Fritz compared the experimental values of the critical thickness in InGaAs/GaAs and GeSi/Si systems by photoluminescence and XRD ro cking curve measurements [Fri87]. He found that the critical thickness determined by photoluminescence was in good agreement with theoretical values, whereas, the XRD rocking curves (sensitivity of 10-3 strain) yielded higher critical values [Bir03]. They also found that relaxation is slow in its initial stage and a high resolution measurement technique would be needed to measure the onset of relaxation which can occur as low as 10-7 strain. People and Bean [Peo85] made the first attempt to calculate critical thickness taking into account the extra strain energy need ed to overcome the barrier to relaxation in order to explain the large observed values of critical thickness expe rimentally. By equatin g the strain energy to the dislocation energy at critical thickness and assuming the dislocation width to be 5b, they obtained an equation that describe s their experimental results well. Their results are plotted in Figure 1-8 along with the predictions of other theori es. Furthermore, it is theorized by many that the larger observed values ar e due to non-equilibrium growth carried out at relatively low temperatures. Extensive research and various models have been proposed to provide explanation of the discrepancy in the calculation of the critic al thickness [Dod87, Jai90, Kas75, Mat74]. The discrepancy lies mostly in growth temperature and measurement technique yielding different critical thicknesses. However, the Matthews-Bl akeslee mechanical equilibrium theory is the most accepted among researchers today.

PAGE 33

33 Morphology during growth Different growth m odes have been observed dur ing epitaxially film growth of mismatched materials. Frank Van der Merwe growth consis ting of layer by layer growth, Volmer-Weber consisting of island growth, and Stranski-Krast anov growth consisting of layer then island growth. Prior work has shown that the type of growth observed is dependent on the relative interfacial energy and strain energy contributions. In the case, of an epitaxially growth film on a substrate with mismatched lattice constants such as SiGe on Si or vice versa, the strain in the film is the given as (as-af)/af. An energetic analysis shows the e xpected instability of the film due to strain [Sro89]. Consider a step wave surf ace morphology where the sample is stressed in the in-plane direction. The change in energy going from the flat surface to the rough morphology is given by: cc E F 2 222 (1.4) Where E is the elastic modulus, is the wavelength of wave morphology, c is the amplitude of the wave, is in-plane stress, and is the surface energy. Equation 1.3 shows that forming a rough surface will lower the ove rall energy of the system pr ovided that the wavelength, > 8 E/ 2 [Sro89]. This crude estimate by Srolovitz demonstrates why surface roughness is observed in stressed films. Thermal stability The stability of strained silicon epilayer s after thermal proce ssing alone has been investigated by many [Koe01, Sam99, Sug01]. Stra in loss has been observed after high thermal processing, above 950 C, through th reading dislocation propagation and misfit formation. This relaxation behavior shows both a te mperature and time dependence. It also strongly depends on the initial level of strain set by the thickness of the silicon cap. Additionally, relaxation can be

PAGE 34

34 caused by Ge diffusion into the strained overlay. Ge diffusion into the strained layer, termed interdiffusion, becomes significant after 950 C for 1 hour [Sug01]. SPER and Strain Since the introduction of strain into the CMOS process, the effect of SPER under strain has been a great topic of interest. SPER has been used to produce strain by strain m emorization techniques by regrowing the amorphized layer po st deposition of a strained capping layer to memorize the strain produced by the cap. Als o, strain has been induced by implantation of high dose Ge or C ions to induce strain by changing the lattice paramete r within the projected range of the implant. This section will disc uss previous work conducted using SPER to produce residual strain within the implanted substrate. Strained silicon SPER The effects of an a morphizing implant on strained silicon structures yield more detrimental results to the strained layers than therma l processing alone. Chilton et al. amorphized a heterostructure with 33 nm Si on 30-35 nm Si1-xGex deposited by MBE on a (001) Si substrate. The structure was amorphized to a depth of 130 nm with a 120 keV As+ implant. Coherent epitaxial growth was observed if the Ge frac tion was below x = 0.16, however, at x = 0.29 after a 600 C anneal the regrown layer exhibited a 75% reduction in stra in. The silicon on this now relaxed Si0.69Ge0.29 exhibited 15% relaxation. The relaxation was measured using Rutherford backscattering spectroscopy and no defect density analysis was performed [Chi89]. It is important to note that no graded buffer la yer was used to relax the SiGe layer. Vandervorst et al. [Van05] implanted arsenic with varying energies of 2 to 15 keV through a 10 nm Si cap grown on relaxed Si0.78Ge0.22 and monitored Ge diffusion using secondary ion mass spectrometry. He showed that Ge interdi ffusion and ensuing strain relaxation was more dramatic when the collision cascade of an amorphizing implant overlaps the Si/ Si1-xGex

PAGE 35

35 interface. Therefore, the implant should be cont ained in the capping layer in order to avoid Ge redistribution, which can ultimately cause relaxati on. Again, the defect microstructure was not characterized. Strained silicon germanium SPER A great deal of work has been performed using structures containing strained silicon germanium grown on silicon. This section, how ever, will focus on the stability of these structures after an amorphizi ng implant. Pseudomorphic Si1-xGex structures with epilayer thicknesses that exceeded the cr itical thickness but maintained atomic registry due to nonequilibrium low-T growth are among the structures that have been studied. Lee and Hong both studied SPER in metastable Si1-xGex layers, x < 12 at% [Hon92, Lee 93]. The amorphous layer, in both cases, was generated through the Si1-xGex /Si interface using Si+ implants. Upon annealing, as the a/c front translated thr ough the interface and the amorphous layer regrew defect-free until the critical thickness [Mat74] was reached. Beyond this thickness the film reached an energy state where defe cts were energetically favorable. The energetically favorable state caused defects to nucleate and grow within the remaining regrown layer. Paine et al. conducted similar experiments with germanium compositions up to 17 % [Pai91]. Paine analyzes the preferred defect configuration using Matthew-Blakesl ee [Mat74] and Freund [Fre87] criteria. He concludes that beyond a critical th ickness, the strained crystalline portion of the alloy can fully relax via stacki ng fault bounded by a 90 partials. Some SPER experiments were conducted in nearly stable SiGe layers and conducted amorphizing implants through the Si/SiGe interface. Rodriguez et al. conducted an experiment using 30nm CVD grown SiGe with concentrations of 21, 26, and 34% Ge. An amorphous layer was generated using a 200keV Ge+ implant at a fluence of 1x1015 atoms/cm2 which extended through the Si/SiGe layer. Half of the samples were also implanted with B+ 5-20 keV at a

PAGE 36

36 fluence of 7x1015 atoms/cm2 and annealed at 600 C. The samp les regrew defect free for the undoped case as follows: 18-20 nm for 21%, 12-15 nm for 26%, and 0 nm for 34%. The doped samples regrew with a thicker defect free layer than the undoped case, 30 nm, 18-20 nm, and 0 nm for the 21, 26, and 34 % Ge respectively [Rod97]. The defects were in good agreement with the calculation done by Paine et al. consisting of the 90 partial accompanied with a stacking fault as described by Paine et al. [Pai91]. During annealing B+ is competing for a substitutional site and thus reduces the overall strain in the material. By decreasing the strain the critical thickness of the system is increased as observed. Chilton et al. conducted similar experiment using 30-35 nm SiGe with concentrations ra nging from 16-29% [Chi89]. Amorphization was conducted using 120 keV As+ implant. Strain recovered for the lowest case, 16%, however, coherency was destroyed in the 29% Ge case. Characterization was carri ed out using RBS and no finding of the defect micr ostructure was discussed. Prior work has also shown instability in th e advancing amorphous-crystalline (a-c) front, mostly under compressive strain. This roughening effect has been observed in compressively strained Si films [Bar04] and in metastable strained SiGe films [Ang07, Cor96] during SPER and in high dose Ge+ implants into Si to form SiGe [Cri96, El l96]. A defect-free planar a-c interface, however, has been shown in rela xed SiGe films with up to a 38% Ge concentration [Kri95]. Therefore, the a-c interface mor phology is connected to strain and not due to alloying with germanium. Additionally, Antonell et al. [Ant96] showed that incorporating C into SiGe prior to amorphization delayed the onset of dislocation formation and promoted planar a-c interface growth due to strain compensation effects. Relaxed silicon germanium SPER Defects. Defects were o bserved after amorphizati on and regrowth in some cases with strained SiGe. However, the relaxed SiGe case is quite different. Relaxed psuedomorphic SiGe

PAGE 37

37 grown using graded layers on Si substrate wa s used to study the SPER rate and crystalline quality post regrowth. In each cas e, regrowth was observed to be defect-free [Hay95, Kri95a]. A defect-free regrown layer and a planar advanci ng a-c interface has been shown in relaxed SiGe films with up to a 38% Ge concentration [Ell95, Kri95]. Therefore, we concluded that SPER breakdown observed in the metastable st rained SiGe is a strain effect and not a chemical effect. Velocity. Several researchers have reported SPER ra tes of strained and unstrained SiGe with respect to pure Si. Haynes et al. [Hay95] studied SPER rate in unstrained alloys 8 m thick with germanium compositions varying from 2 to 87%, including pure Si and Ge. Amorphous layers 300-400 nm thick were produced using Si+ implants. For each composition, the measured SPE rates span two orders of magnitude and are related to the pure elements. The Si-rich alloys displayed activation energies for SPER higher than pur e Si and same to be true for Ge-rich alloys and pure Ge. The a-c interface of high germaniu m alloys was confirmed to be planar using XTEM. Lee at al. [Lee93] measured SPER rates of strained SiGe with germanium composition of 12%. Rates were measured using Time Resolved Reflectivity (TRR) and real-time measurements demonstrate SPER rate is not consta nt over a fixed temperatur e but varies with the position of the a-c interface. The activation barrie r is higher than pure Si and ranges from 2.94 to 3.11 eV for temperatures between 503 and 603 C. Paine et al. [Pai91] reported an activation energy of 3.2 eV independent of germanium using in situ TEM. Structures used in this experiment were 200 nm SiGe grown via CVD with germanium compositions of 5.4, 11.6, and 17%. Outstanding Questions It is im portant to note that prior experiments observed re growth related defects for structures that were thermodynamically metast able with film thicknesses exceeding 200 nm. Also, most implants carried out were conducted through the Si1-xGex /Si interface. Crosby et al.

PAGE 38

38 [Cro04], however, has shown that implants con ducted through-the-layer and within-the-layer impact the overall strain state of the material post anneal. Thus there is a gap in understanding the effect of SPER within a strained layer vs. through a strained layer an d how the proximity of the a-c to interface to the heterostructure interf ace affects relaxation. Al so, the microstructure and stability of regrowth rela ted defects have not been well studied. This work aims to understand the impact of both proximity and st rain level on defect nucleation and strain relaxation. This work also aims to explore the relationship of the critical thickness with defectfree regrowth under biaxial compressive and tensile strain.

PAGE 39

39 Figure 1-1. XTEM image of P-type and n-type MOSFET channel [Tho04].

PAGE 40

40 Figure 1-2. Arrheiuns plot of regrowth rate fo r Si samples oriented along <111>, <110>, <100> directions [Cse78].

PAGE 41

41 Figure 1-3. Valence band of (a) unstrained and (b) strained Si showing decreased in light hole effective mass for strained case [Tho04].

PAGE 42

42 Figure 1-4. Binary phase diagram of Si-Ge syst em showing complete solid solubility [Mas90].

PAGE 43

43 Figure 1-5. Schematic representation of strained film interface versus a relaxed film interface. substrate film Strained Relaxed

PAGE 44

44 Figure 1-6. Microstructure of misfit dislocation with thread ing arms in cross-section. Misfit dislocation Threading arms

PAGE 45

45 Figure 1-7. Si cap critical th ickness as a function of Ge% in the uniform SiGe layer [Sam99].

PAGE 46

46 Figure 1-8. Critical thickness as a fu nction of Ge concentration [Peo85].

PAGE 47

47 CHAPTER 2 EXPERIMENTAL METHODS The experim ents carried out in this work re quired application of di fferent processing and growth techniques. A brief description of these techniques will be discussed in this chapter. The strained films and alloys were grown by Mol ecular Beam Epitaxy (MBE) and Chemical Vapor Deposition (CVD) techniques. Post deposition, the strained laye rs were then implanted and annealed. The film morphology and strain rela xation post processing and anneal were studied and characterized using Transmission Electr on Microscopy (TEM) and High-Resolution X-Ray Diffraction (HRXRD) techniques. Material Processing Growth Conditions: Strained Si on Relaxed SiGe The struc ture with Si in biaxial tensile stra in were grown using a Molecular Beam Epitaxy (MBE) chamber at the University of Aarhus in Denmark. MBE is a deposition technique that uses a ultra-high vacuum chamber. Material is deposited by heating the solid source until it evaporates using an electron-beam. The evaporated material then condenses on the substrate, which spins during deposition to promote uniformit y. The chamber walls are cooled with liquid N2 to reduce outgassing. The experimental structures were grown with compositionall y graded buffer layers at 750 C incorporating Ge at a rate of 10 at.% per micr ometer on silicon substrates to compositions of 0, 10, 20 and 30 at.% Ge with low threading dislocation densities of 1x105 cm2. A 630 nm thick fully relaxed SiGe layer of corresponding compos ition was then grown on top of the buffer layer at 550 C, followed by a pseudomorphically strained silicon capping layer 50 nm thick. A schematic of the multi-layered structure is shown in Figure 2-1. The details of the growth conditions and threading dislocation density are presented by Ga iduk et al. [Gai00].

PAGE 48

48 Growth Conditions: Strained SiGe on Si The struc tures with SiGe in biaxial compre ssive strain structures were grown using a reduced pressure Chemical Vapor Deposition (C VD) chamber at Texas Instruments, Inc. Specifically, a RP-CVD tool from ASM, the Eps ilon 3200, was used to grow all samples. The gaseous precursors used for silicon and germaniu m were dichlorosilane and germane. Hydrogen was used as the carrier gas at a flow of 40 standard liters per mi nute. The growth was carried out at 700 C and a fixed pressure of 10 Torr. Prior to deposition, an HF clean and a pre-bake of 1050 C 3 minutes was carried out to remove oxi de and all contaminants on the wafer surface. Strained 50 nm Si1-xGex was deposited on Si (001) substrate with alloy compositions of 0, 16, 22, and 26 at% germanium. A schematic of th e structure is shown in Figure 2-2. Ion Implantation and Annealing Ion im plantation has been used by the semic onductor industry as the preferred method for incorporating dopants into silicon. The pro cess offers advantages including reliability, reproducibility and control of dopant dose and distri bution. In this non-equilibrium process, ions bombard the host lattice and, through nuclear and electronic interactions, lose energy until they come to rest. This process introduces primary defects consisting of a vacancy rich region near the surface and an interst itial rich region deeper within the substrate where the implanted ions come to rest. For the implanted dopants to be electrically active, a high temperature anneal is required; this process also allo ws Frenkel pairs, which are interstitial and vacancy pairs, to recombine. At lower implant doses the host latt ice can maintain its integrity with isolated regions of defects around Rp, the projected range of the incoming ion. However, at higher doses when more than 10% of the host atoms are di splaced, the implanted region undergoes a first order crystalline-to-amorphous transition, term ed amorphization [Chr81]. The amorphized region is completely damaged and no periodic lattice exists. The defective region in this case

PAGE 49

49 lies just beyond the amorphous-crystalline (a -c) interface. Annealing is required post implantation to repair the damaged lattice. All Si+ implants were carried out at Axcelis Technologies in Beve rly, Massachusetts. First, the ions are accelerated to a potential of 100 keV before be ing mass analyzed by a magnet. The analyzed ions are ejected in to the accelerator stage where they are accelerated further. For energies below 100 keV, the implant is operated in acceleration /deceleration mode. During this mode, a N2 stripper gas canal is evacuated such th at the ions do not exchange charges while passing through the accelerator. Th e ions are accelerated into th e terminal potential and then decelerated. The ions then exit the acceleration stage at the energy desired. The implant energy used in this work were 5, 12, and 18 keV at a fluence of 1 x 1015/cm2 which generated continuous amorphous layers 15, 30, and 40 nm, respectively. Material Characterization Transmission Electron Microscopy Sample preparation. The m orphology of the samples was monitored using TEM. Post implantation and regrowth damage can be determin ed in three dimensions with the use of planview and cross-sectional sample preparation. In order for the samples to be imaged using the TEM, it is imperative to polish th em to a thickness such that el ectron transmission occurs (~200 nm). Plan-view (PTEM) samples were prepared cut ting a 3 mm disc from the sample using a Gatan ultrasonic disc cutter with the aid of SiC cutting slurry. The disc was then thinned down to approximately 100 micrometers using 15 m particle size Al2O3 slurry on a glass plate. The implanted side was next coated with paraffin wax and the samp le is then mounted on a Teflon mount termed Johnny. The sample was subs equently jet etched with an acid solution consisting of 25%HF:75% HNO3 [Ste07] until a small hole appeared. The sample is removed

PAGE 50

50 from the Johnny and placed in a beaker filled with heptane left to soak overnight. Heptane is used to dissolve and remove the wax off the sample Regions of the sample in proximity of the hole are electron transparent and can be imaged using TEM. These plan-view samples allow planar observation of the dislocations and the ability to quantify them. Cross-sectional (XTEM) samples were prepar ed using a FEI Strata DB 235 Dual-beam Focused Ion Beam (FIB). The sample was m ounted on a FIB stub using conductive carbon paint and then coated with 50 nm of puttered carbon fi lm prior to FIB milling. The sample was then put into the FIB chamber and th e chamber pumped down to vac uum. A strip of GIS Pt was deposited on the area of interest to protect it from beam damage. Two wedges were then milled on either side of the Pt strip using 5000 pA beam current. The Pt protected area was next thinned with consecutively smaller beam currents until a final thickness of about 150 nm was achieved. The sample was extracted using an ex-situ mi cromanipulator system attached to a light microscope. Finally, the sample was placed on to a carbon film on a Cu grid array, ready for imaging. Cross-section TEM analysis primarily allows observation of the depth of the epitaxial layers and defects. Imaging conditions. A JEOL 2010F high-resolution tr ansmission electron microscope (HRTEM) and a JEOL 200CX TEM operating at 200 kV were used to analyze the defects in the regrown layer. XTEM measurements were used to confirm the amorphous layer depth and layer thicknesses. PTEM samples were prepared and im aged to quantify dislocation density. TEM is the only technique that allows analys is to determine if a defect is extrinsic/intrinsic in nature and determine Burgers vector. Imaging of the extended defects post impl antation using diffraction contrast [Edi75, Wil96] was employed to determine the position and quantity of the defects. For the implant dose

PAGE 51

51 used in this study, end-of-range (EOR ), also known as type II defects, will form. Also, type III or regrowth related defects are probable after annealing. Diffracti on contrast theory states that defects are visible if g b x u 0 where g is the reciprocal lattice vector corresponding to diffracting plane, b is the Burgers vector for a dislocation, and u is the line direction of the dislocation. The Burgers vector for dislocation loops in Si ar e a/2<110> and a/3<111> type, and for {311} defects it is a/24<116> type. A two-beam bright field (BF) condition with s > 0, s is deviation from Bragg conditi on, is used to image the PTEM and XTEM. The g220 reflection is used to image dislocation loops and regrowth related defects in plan -view and cross-section because the defect contrast is greatest for this reflection. Th e resolution of the defects was increased using weak-beam dark field (WBDF) im aging, where the sample is tilted such that s (deviation from Bragg condition) is large, the plan es in most of the specimen is titled away from Bragg condition. However, as seen in Figure 2-3, th e planes near the core of the dislocation are bent back into Bragg condition to yield contrast [Wil96]. This technique is termed Weak-Beam Dark-Field (WBDF) and was taken at g.3g conditi on. This imaging condition was used to image all PTEM samples in this work. Regrowth related defect density analysis. PTEM quantification was carried out to obtain dislocation density in both Strain ed Si and Strained SiGe samples. Strained Si samples exhibited misfit dislocations that span in the <110> di rection and Strained SiGe samples exhibited dislocation networks in the form of a large connected array rather than individual dislocations. In order to quantify both types of dislocations in a similar manner, a linear (rather than area) quantity was measured. PTEM images were taken under g.3g WBDF mode Each sample was imaged ten times in different areas of the specimen, in order to get a good statistical account of the defect density.

PAGE 52

52 Each sample condition was taken at the same magnification of 20,000X, 50,000X, 100,000X, or 150,000X, depending on the density of the dislocations. The imag es were printed out on 8 x 10 inch sheets. For the strained Si samples, a ruler was placed in th e <110> directions, both parallel and perpendicular to the g vector the number of defects intersect ing the ruler was counted over a length and recorded. This was done for all ten im ages. Each of the ten micrographs was counted three times to include varia tion within a single microgra ph. The average of these 30 observations was used to calculate the value of linear defect density for each sample condition. Error bars equal to one standard deviation for a ll observations was applied to each data point. For the strained SiGe samples, the measurements were taken perp endicular to the g vector. The dislocation network was imaged using g220 and the images show a stronger linear alignment along one direction. A ruler was placed perpendicu lar to this direction an d the number of defects intersected with the ru ler was then counted per unit length. The results and error bars were determined in similar manner as for the strained Si samples. For clarification, it should be noted that the defect density quantified for the strained Si was misfit dislocations while the strained SiGe the defects were regrowth related defects. The misfit dislocations in the strained SiGe were not visible due to the high density of regrowth related defects. This is discussed further in Chapter Five. Trapped interstitial concentration analysis. PTEM were used to image end of ranges defects consisting of dislocati on loop and {311}s. The trapped in terstitials cont ent in both of these defects can be quantified by the following method. First, the quantification of the dislocation loops will be discussed followed by the {311} defects. End-of-range loops are most co mmonly images using WBDF g220 imaging mode. The images were taken at 150,000X and printed on 8 x 10 inch sheets. A grid consisting of 0.5 cm

PAGE 53

53 squares were printed out on a transparency and used to determine the area fraction of loops. This transparency was superimposed onto the printed out image and used to count the defect. The number of nodes that intersect a loop were c ounted and divided by the total number of nodes to yield an area fraction occupied by individual l oops. The area fraction is then multiplied by the planar atomic density of the {111} plane (~1.5x1015 atoms/cm2) will give the concentration of trapped interstitials in (atoms/cm2) bound by dislocation loops. Th e error of this measurement can be reduced by using finer grid size a nd choosing multiple areas and averaging the concentration of loops. Three areas per sample were counted and determined to have a 10% error within a sample set. Next, the {311} defects ar e counted in a different manner. Again, the g220 WBDF condition was used to image the PTEMs. However, some families of these defects will be invisible for g parallel to th eir length. Only using the g220 reflection will lead to lower defect count and higher error in calcu lation roughly 30%. The images were taken at 150,000X and printed out on 8 x 10 inch sheets. A transpar ency was placed over the image and the length of the {311} was traced with a fine-tipped marker. Note that the {311}s parallel to the sample surface are actual length and the {311}s that are not parallel are projecte d lengths (45 to the surface). The length of the defects that are not parallel to the surf ace will be multiplied by 1.4 and then added to the rest to yield total length of interstitials in the sampling area. Assuming 26 atoms per nm of the {311}s, the total number of trapped interstitials/cm2 can be determined. The total trapped interstitial concentration in this work is th e sum of both dislocation loops and {311}s. Note that there were no end-of-range def ects observed in the strained SiGe samples. Therefore, this calculation was only done for the strained Si samples.

PAGE 54

54 X-Ray Diffraction X-Ray diffr action (XRD) is a versatile, non-dest ructive analytical te chnique used to study crystallographic properties, chemical composition, and physical properties of thin film and bulk materials. XRD was used in this work to monitor the change in strain magnitude post processing. The strained layers in this work are pseudom orphically grown with the substrate lattice continuing into the thin film. This continuati on of substrate lattice parameter into the film (different equilibrium lattice parameter) is associat ed with incorporation of strain into the film layer. A measure for strain in the material is the difference in lattice parameter of the layer and substrate. The adaptation of the layers lattice parameter to match the parameter of the substrate will cause a tetragonal distortion in the film unit ce ll. The layer unit cell extends or shrinks from its original value of aL to a in the out-of-plane direction and aL to a in the in-plane direction, depending on whether aL is larger or smaller than aS, as shown in Figure 2-4. The degree of relaxation of the film is described by Equation 2-1. Strain Relaxation = (aaS) / (aL aS) (2-1) Braggs Law. In XRD, diffraction of the beam occu rs when Braggs Law is satisfied. Braggs Law states that parallel planes with a distance of d will constructively interfere when the distance traveled is equal to n times the x-rays wavelength, this relati onship is presented in Equation 2-2. n = 2d sin (2-2) A substrate containing epitaxial layers will result in additional p eaks corresponding to the difference in lattice parameter to the substrate. The positions of these peaks are used to determine the d spacing for the laye r and substrate. The lattice pa rameter can then be calculated

PAGE 55

55 using the relation of d spacing to lattice parame ter, a, shown by Equation 2-3. The miller index of the reflection is h, k, and l. dhkl = a/ (h2 + k2 + l2) (2-3) The degree of relaxation can be calculated using the relations hip shown in Equation 2-1. The quantified strain relaxation in the proceeding chapters are always made with reference to the asgrown state of the sample and not the theoretical strain state. Most of the materials were partially relaxed with misfit disl ocation linear density of 5 x 104/cm equal to less than 1% strain relaxation as received. Instrument and setup. A Pananalytical MRD XPert was used to obtain rocking curves and reciprocal space maps. Previous studies ha ve also employed rocking curves to study the strain relaxation observed in the strained si licon capping layer using pseudomorphic Si/ Si1-xGex structures [Cro04, Phe05] and strained SiGe on Si [Few82, Mat91]. XRD rocking curves and reciprocal space maps scan across a range of /2 angles and report the angle where diffraction occurs, called the Bragg angles. The x-rays have a wavelength of 1.54 (CuK 1 transition) and are directed through a series of (220) Ge crystals to produce a monochromatic beam. The incident angle of the x-ray to the sample is and the diffracted beam is 2 relative to the incident b eam. For both rocking curves and space maps, the Hybrid Mirror is used as the primary optic. It gives the best resolution and monochromates the beam such that only Cu K 1 radiation is used. Only single crystals will give any detectable intensity using this setup. For th e strained Si sample measurement, the secondary optics used were the rocking curve attachment with the triple axis. The triple axis mode places a channel cut germanium analyzer crystal before the detector. Under th is configuration, the diffracted beam undergoes three (022) reflections before entering the detector. The acceptance

PAGE 56

56 angle of the germanium analyzer crystal is 12 arcseconds. For the strained SiGe samples measurement, the secondary optics consisted of the open detector of the rocking curve attachment and the brass slit. Rocking curves. Rocking curves (RC) are acquire d by rocking the sample through the incident angle, while moving the detector, 2 Doing so allows the diffracted peaks of the substrate and film to be distinguished. The peak positions and intensity are dependent on the materials composition, crystalline quality, and thickness. The peak width is broadened by imperfections in the crystal such as faults and defects. Thus, the peak width can account for the film thickness in the case of a ps eudomorphic layer. The peak width is observed to decrease with increasing thickness. Additionally, for very th in layers fringing is observed to the left and right of the main peak. The fringing is due to x-ray interference reflec ted from surface of the film and the substrate-film interface. The distance of the fringes can also be used to determine the layer thickness and were observed in the stra ined SiGe samples. These fringes can be washed out if crystalline disorder associated with faults, dislocations, or surface roughness exceeds a critical level. The thickness, Ge concentration, and roughness of the layers can be simulated using the Epitaxy software designed for the XPert system. To simulate these parameters accurately, the degree of relaxation of the film must be known and measured using a different technique. Q scattering vector. The XRD scans are presented in terms of scattering vectors, Q. A symmetric /2 scan, rocking curve scan, and radial s can is represented in Figure 2-5. The symmetric /2 scan only has a nonzero component fo r the scattering vector normal to the substrate surface. A rocking curve, howev er, also has nonzero in-plane component, Qx, for all angular positions except when = 2 /2. The last scan type, radi al scan, the Q vector has the

PAGE 57

57 same tilt with respect to the sample surface durin g the entire measurement. It is sometimes convenient to report analysis in Q vector notation because it is more directly translatable for comparison with other measurement methods. Th e relation between instrumental coordinates, and 2 and the scattering vectors, Qx a nd Qz, is presented in Equation 2-4. Qx = K [cos( ) cos( + )] Qz = K [sin( ) + sin( + )] (2-4) The Q notation and the reciprocal space map comb ined in one plot is shown in Figure 2-6. The region that is accessible for data collection can be represented by a hemisphere in Q space having -2K Qx 2K and 0 Qz 2K and obeying Qx 2 + Qz 2 4K2. The region of reflection is divided from region of transmission by two sma ller hemispheres of radius K (shaded in gray) that are centered at Qx = K and K. These two gray areas cannot be accessed because they lie in the transmission region. In this figure there ar e point labeled by the Miller indices of the Bragg reflections that are observed for the alignment of Si (001). The 00l refl ections are allowed and occur along the Qz axis and can be measured only with a symmetric /2 scan. This combination plot can be generated fo r any substrate with orientation ( hkl ) using the instrument software. The set of Miller indices that are observed will depend on both the substrate orientation ( hkl ) and the direction of alignment of the sample with the x axis of the diffractometer. Reciprocal space map. Reciprocal Space Maps (RSM) was used for samples with lower film peak intensities such that the peak position from the substrate can be distinguished. RSM are subsequent /2 rocking curves over a range of angles or subseque nt radial scans, presented in Figure 2-7. The RSM obtained in this study were from subsequent rocking curves. RSM is performed such that the investigated Br agg reflection is fully mapped in Q space. In other words, it is not onl y monitored by one rocking curve crossing it rather the entire vicinity of

PAGE 58

58 the Bragg reflection is measured. RSM can provide further structural information of epilayers than a rocking curve alone. RSM measuremen ts were preferred over RC for the asymmetric reflections for the strained Si samples for two main reasons. First, the X-ray intensities for the strained peak post processing were too low to de termine the exact positio n of the peak. Second, if relaxation process also induced some tilt or twist within the la yer, the position of the strained layer peak will no longer be accessible using RC. The RC measurements are aligned to the substrate peak due to its high intensity and know n Bragg reflection position. If the layer does undergo some change in twist or tilt angle, the direction of elongation of Bragg reflection will also change in reciprocal space with respect to the substrate and ma y not lie in the area of the RC measurement. The RSM are generally plotted in terms of the scattering vector, Q. An example of a RSM plot, presented in Figure 2-8, shows the position of the layer with respect to the substrate in terms of the layer being in a fully strained or fully relaxed state. One can draw a line between the fully strained (R=0) and fully relaxed (R=1) peak positions, this line is called the Relaxation line. The layer will appear within the area of this line depending on the degree of relaxation observed. The percent strain relaxation observe d by the layer can also be determined by the position of the peak within the re laxation line by taking the ratio of the change in position from fully strained over the full length of the relaxa tion line. The as-grown (113) RSM for Si on Si0.7Ge0.3 is presented in Figure 2-9. The figure indicates the position of each layers peak. The out-of-plane lattice misfit appears as a peak se paration between the laye r and substrate along Qz. and the in-plane lattice mi sfit is measured along Qx [Pie04]. Thus, the strained film is fully strained when the strained Si and relaxed SiGe peak are aligned vertical ly, as indicated by the

PAGE 59

59 vertical line. Strain relaxation post processing is observed by movement of the strained Si peak towards the Si substrate peak. Errors. The errors associated with X-ray measurement are dependent on the user, measurement step size, peak broadening and vari ation within the wafer. Before acquiring any data, all axes of the instrument must be aligned, us ually to the substrate peak or sample surface. One can introduce a large error if the instrument is misaligned. This is especially true for measurements during an absolute scan. Howe ver, the RC and RSM measurements are relative scans. The RSM and RC measur ements carried out used the Si substrate Bragg position as reference and all axes of the instrument were a ligned to it. Thus, the epilayers were always measured relative to the Si substrate peak. The error of measurement due to user variability and alignment variations were tested. The varia tion between users was less than % Ge which roughly translates to a 0.03% strain relaxation error. The error in troduced by the step size for the RSM measurement was less than 0.5% strain rela xation. Another error to consider is the determination of the peak position when the peaks are relative broad. Thus, the degree of relaxation was also measured by average misfit spacing using PTEM micrographs, discussed in Chapter Three, and compared to the XRD measur ements. There exists a ~1% strain relaxation difference between the two measurement tec hniques. Additionally, there was a 0.5% Ge distribution across the wafer whic h was also considered. Finall y, taking all the errors into considering an error of % strain relaxa tion was estimated for the XRD measurements. Sensitivity of measurement. Epitaxial thin films exhibit a high degree of crystalline perfection. However, deviations from a perfect crystalline lattice do occur and the sensitivity of the measurement of this deviation is the focus of this section. In order for accurate measurement of closely spaced peaks that occur in the diffr action of epilayer-substrate materials, special

PAGE 60

60 experimental equipment must be introduced. Cr ystal monochromators and analyzers have to be introduced in the beam path which allow selection of x-rays that are only Bragg reflected from a single crystal or a set of th em. Figure 2-10 clearly shows the need for achieving higher resolution to investigate thin films used in elect ronics [Bar83]. The highe st resolution possible for the measurement setup used in this work is 10-4 for measured d-spacings. However, for onset of relaxation to be measured a sensitivity of 10-7 is needed. Therefore, it should be noted that Xray measurement cannot determine very low changes in strain relaxation nor can it be used to determine the onset of relaxation.

PAGE 61

61 Figure 2-1. Schematic representation of biaxial tensile strain structure used in this work. Note: not to scale 50 nm Strained Si Graded SiGe buffer layer 630 nm SiGe Si substrate

PAGE 62

62 Figure 2-2. Schematic representati on biaxial compressive strain structure used in this work.Note: not to scale 50 nm SiGe Si substrate

PAGE 63

63 Figure 2-3. Principle behind weak beam dark field imaging technique in TEM for an edge dislocation. High intensity occurs close to dislocation co re because planes are bent back to Bragg condition [Wil96].

PAGE 64

64 Figure 2-4. Strained (left) and relaxed (right) layers for film layer in compression (top) and in tension (bottom) [Bir03].

PAGE 65

65 Figure 2-5. Schematic repres entation of (a) symmetric /2 scan, (b) rocking curve scan and (c) radian scan in plane of momentum transfer [Bir03].

PAGE 66

66 Figure 2-6. Combined plot of Q x,Qz plane and position of Bragg peaks for an epitaxial film on Si (001) substrate [Bir03].

PAGE 67

67 Figure 2-7. Two ways by which reciprocal space maps may be recorded are by either (a) subsequent rocking curves or (b) subsequent radial scans [Bir03].

PAGE 68

68 Figure 2-8. Representation of a reciprocal space map shows the relative positions of the layer with respect to the substrate for fully st rained and completely relaxed system in reciprocal space [Pie04].

PAGE 69

69 Figure 2-9. (113) Reciprocal sp ace map of as-grown Si on Si0.7Ge0.3 indicating the location of each peak. Strained Si Si substrate Relaxed SiGe Compositionally graded layer

PAGE 70

70 Figure 2-10. Relative d spaci ngs and precision according to Bartel [Bar83, Bir03].

PAGE 71

71 CHAPTER 3 TENSILE STRAIN: DEFECT NUCLEATION AND TEMPERATURE DEPENDENCE Strain ed silicon technology offers enhanced hol e mobility due to a decrease in the average effective mass and decreased inter-valley scatteri ng [Nay96]. Strain can be induced in a Si capping layer through lattice mismatch between a Si1-xGex virtual substrate and the silicon film, provided the strained overlay is grown less than the critical thickness [Peo85]. Obeying this criterion, the interface will be in perfect atomic registry and maximum strain will be present. After thermal processing, however the strain energy in some he terostructures can decrease through strain relaxation via the pr opagation of threading disloca tions and formation of misfits [Koe01, Sam99, and Sug01]. Additionally, relaxati on can also be caused by Ge diffusion into the strained overlay du ring thermal processing. While numerous studies have reported results on silicon recrystallization and regrowth in Si1-xGex and Si bulk systems [Atz94, Pai91, Sug04], the stability of strained silicon af ter amorphization and r ecrystallization is not well understood. Ge interdiffusion and ensuing strain relaxation has been shown to be especially dramatic when the collision cascade of an amorphi zing implant overlaps the Si/Si1-xGex interface. Previous studies suggest that the implantation through the hetero structure interface process creates point defects which act as nucleation sites for relaxation-induced dislocations and/or may assist Ge interdiffusion by an interstitial mediated mechanism [Van05]. The behavior of strained structures where th e implant is constrained within the strained layer, however, is not well known. This experi ment focuses on determining the stability and degree of relaxation after amorphization and recrystallization with in a strained silicon layer. This knowledge will assist in defining the process window for Stress Memorization Technique (SMT) and other process-induced strain applications.

PAGE 72

72 Several variables are likely to affect the degree of strain relaxation. Some of these variables include point defect population, initial strai n, critical strain fo r enhanced strain relaxation, and proximity of point defects to heterostru cture interface. This chapter will discuss the effect of initial strain a nd point defect population on strain relaxation. The next chapter will discuss the effect of implant proximity to the heterostructure interface on relaxation. The experiments in this chapter will determine if there is a critical strain for Solid Phase Epitaxial Regrowth (SPER) breakdown and, if so, by what mechanism. The primary concern centers on the effect of the EOR on the relaxation process. Thus, an isochronal study was carried out at a temperature at which th e evolution of excess interstitials into dislocation loops was sampled. By monitoring the quantity of EOR dama ge in the strained vs. unstrained layers and corroborating the measurement with the relaxation defects known as misf its the role of the extended defects on strain relaxation could be understood. Second, how any defects present were nucleated. Finally, the thermal behavi or of the dislocations was explored. Experimental Design Strained Si structures were grown on relaxed Si1-xGex (Ge fractions of 0, 10, 20, and 30) virtual substrates via Molecular Beam Epitaxy (M BE) at the University of Aarhus, Denmark. 100mm 0.005-0.020 ohm-cm (100) n-t ype Czochralski-grown silic on wafers were used for substrate wafers. Virtual substrates with low threading disloca tion densities were grown with compositionally graded buffer layers which incorpor ated Ge at a rate of 10 at.% per micrometer at temperatures of 750 C to 800 C [F it91]. A 630 nm thick fully relaxed Si1-xGex layer of corresponding composition was then grown on top of the buffer layer, followed by a 50 nm strained silicon capping layer. Th e strain of the structures was experimentally determined using the in-plane lattice parameters, aSi-cap and aSiGe, obtained from HRXRD. Th e change of strain is reported in terms of % strain relaxation, obtained by the relationship in Equation 3-1.

PAGE 73

73 % Strain Relaxation = (aSi-cap aSiGe)/ aSi aSiGe x 100% (3-1) The silicon results from this analysis show that the cappi ng layers grown on Si1-xGex with Ge fractions of 0, 10, 20 and 30 correspond to a Si layer strain of 0, 0.37, 0.74, and 1.1% strain, respectively. XTEM analysis was also used in orde r to verify proper growth of strained films. The structures were then implanted with Si+ at an energy of 12 keV and a fluence of 1x1015 atoms/cm2 to generate a 30 nm continuous amorphous laye r confined within the strained layer. Next, isothermal and isochronal anneals were perf ormed in a quartz-tube furnace under an inert N2 ambient environment. Three experiments were designed to study spec ific factors that may affect the relaxation process. The first experiment consisted of 30 minute isochronal anneals performed at 500, 650, and 800 C and isothermal anneals at 800 C for 5, 30 and 300 minutes. In implanted Si, the supersaturation of interstitials leads to the evol ution and agglomeration of extended defects when annealed above 650 C. This experiment studied the effect of the evol ution of these extended defects within a strained structure. This experiment was also used to determine the critical strain necessary for breakdown of SPER. The second experiment consisted of an isothe rmal study at 500 C which was performed to investigate defect nucleation and propagation in the strained film during the regrowth process. Anneal times for this study were 15, 30, and 45 mi nutes. For these times and temperature, the regrowth velocity was slow enough that the amorphous-crystalline (a-c) interface could be captured before complete regrowth had taken place. This allowed observation of defects as they nucleated and grew with the us e of XTEM and PTEM. The third experiment consisted of a 30 minute isochronal study to investigate the thermal behavior of the relaxation process and how it comp ared to bulk processes. Temperatures of 500,

PAGE 74

74 575, 650, 725 and 800 C were used for this experiment. The highest strained sample, 30% Ge, was used in this isochronal study as it would show the maximum effect, if any. For all experiments, a Pananalytical MRD XPert was used to obtain HRXRD rocking curves and reciprocal space maps to study the st rain relaxation. Previous studies have also employed rocking curves to study the strain rela xation observed in the strained silicon capping layer using pseudomorphic Si/ Si1-xGex structures [Cro04, Phe05]. However, further investigation determined space maps to be a preferable technique for the measurement of annealed samples. Reciprocal Space Map (RSM)s are compilations of rock ing curves taken at a range of positions to create a 2D map of intensity in the vicinity of the Bragg reflection. The RSMs allows the observation of very low intensit y peaks from the strained silicon layer that rocking curves do not clearly obta in. Using these maps, the relaxati on of the layer can be directly observed by monitoring the shift of the strained silicon peak to ward the silicon substrate. Additionally, the crystalline integrity of the layer can also be monitored by observing any broadening of the peak that would indicate a disordered or defect ed layer [Pie04]. Therefore, space maps are preferred for multilayered structur es because the lattice parameters can be obtained from peak position and the degree of re laxation obtained direct ly from the relaxation line (refer to Chapter Two). A JEOL 2010F high-resolution transmission electron microscope (HRTEM) and a JEOL 200CX TEM were used to study cross-sections of the regrown layers (XTEM). Samples for XTEM were prepared using a Dual-Beam FE I Strata DB 235 Focused Ion Beam. XTEM measurements were used to confirm the amorpho us layer depth and layer thicknesses as well as identify defects present. Plan-view (PTEM) samples were also prepared and imaged using a

PAGE 75

75 JEOL 200CX operating at 200kV using g220 weak-beam dark-field imaging to quantify dislocation density and type. Results and Discussion As-Grown and As-Implanted Sample Analysis The HRXRD (004) rocking curves for the as-grown structures are show n in Figure 3-1. The position of the diffraction peak depends on Ge composition and strain relaxation in the layer. The strained silicon peak appear s on the right of the high intensity silicon peak and the relaxed Si1-xGex is to the left. The plateau between the Si and the relaxed Si1-xGex peaks is caused by the superposition of peaks from the compositionall y graded buffer layer. The HRXRD results obtained indicate proper growth a nd strain levels for the hetero structures. However, PTEM investigation observed the presen ce of misfit dislocations corres ponding to a relaxation of less than 0.01% strain, for all as-grown sample conditi ons. Therefore, the samples are only partially pseudomorphic. The strain relaxation observed in these as-grown samples, however, is below the detection limit of the X-Ray measurement tec hnique and was not accurately measured using X-Ray methods [Bir03]. The sensi tivity of the XRD measurements is discussed in Chapter Two. Note that the initial strain magnitude of the as-g rown samples was used as reference to calculate the relative degree of relaxation post annealing and implantation. As-Implanted Sample Analysis All sam ples discussed in this chapte r were implanted with a 12 keV Si+ implant at a fluence of 1x1015 atoms/cm2. The implant was tailored to generate a continuous amorphous region confined within the strained layer as sh own in Figure 3-2. XTEM of as-implanted Si on SiGe and Si control samples confirmed that am orphous depth did not vary with %Ge for this implant condition. The amorphous depths were within 1 nm for all samples within the error of XTEM measurement.

PAGE 76

76 Critical Strain: SPER Breakdown Study The relaxation behavior of all stra ined sam ples implanted with 12keV Si+ was studied using a series of 30 minute is ochronal anneals at 500, 650, and 800 C and isothermal anneals at 800 C. This experiment was used to determine the critical strain necessary for breakdown and the thermal behavior of relaxation (if observed). The thermal stability of the highest st rained as-grown (not implanted) Si0.7Ge0.3 structure was studied by subjecting samples to a 500 or 800 C anneal for 30 minutes. In both cases, no strain relaxation was observed. This result is in agreement with previously reported results [Koe01, Sam99, Sug01]. However, post-implantati on and anneal, relaxation is observed for the highest strain structure as observed in the rocking curve spectra. Figure 3-3 shows a sequence of rocking curves for the as-grown, as-grown ann ealed and implanted annealed sequence for the Si0.7Ge0.3 structure at 500 C. Note that the rocking curves for the strained silicon peak postimplant and anneal have a weak signal, theref ore, samples were further investigated with reciprocal space maps. The reciprocal space ma ps confirm that the rocking curves were only collecting the shoulder of the peak, or in the asymmetrical geometry cases, not collecting it at all. This is shown in the reciprocal space map in Fi gure 3-4, where the strain silicon peak is not in the area of the rocking curve a nd therefore gives a poor signal. Accordingly, lattice parameters were extracted from the peaks on the reciprocal space maps using Braggs Law. From these lattice parameters, the percent relaxation is calculated and is shown in Figure 3-5. The Si0.9Ge0.1 and Si0.8Ge0.2 samples showed 100% retention of the original strain for all thermal conditions. However, after the 500 C anneal the Si0.7Ge0.3 sample showed partial relaxation in the (224) map and no relaxation in the (113) map. Thus, this observation indicates that the relaxation process is not homogeneous at 500 C. The av erage between these results was taken to calculate the percent re laxation. The PTEM analysis of the highest strained sample

PAGE 77

77 (Si0.7Ge0.3) showed regrowth related de fects as well as confirming the results from HRXRD. After the 650 and 800 C anneals, the Si0.7Ge0.3 samples showed a clear trend of higher anneal temperatures yielding a higher de gree of relaxation. The strain ed silicon peak also broadened significantly with higher temperature anneals indica ting higher defect density. The crystalline quality of the regrown layer was further studied us ing XTEM. After the 500 C 30 minute anneal it was found that 10 nm of amorphous material remained in the strained layer. This may have attributed to the varying results in the (113) a nd (224) reciprocal space maps for the Si0.7Ge0.3 sample. An assessment of a fully re grown layer at 500 C would give a better indication of its true degree of relaxation. After the 650 and 800 C anneals, the TEM micrographs indicate a fully regrown layer for all samples. XTEM analysis for all conditions of the Si0.9Ge0.1 and Si0.8Ge0.2 samples show that the regrown la yer was of good crystalline quality. This confirms the HRXRD results. However, the Si0.7Ge0.3 samples show defects as a result of relaxation and SPER breakdown. The XTEM images, taken under g220 WBDF condition, and PTEM images taken at g220 BF for the Si0.9Ge0.1 and Si0.8Ge0.2 sample and WBDF for the 30% Ge sample are shown in Figure 3-6. All samples were annealed at 650 C for 30 minutes. Note that the PTEM magnification for the Si0.7Ge0.3 sample is different than the other samples to better show defect morphology (refer to scale bars). The Si0.9Ge0.1 and Si0.8Ge0.2 samples exhibit misfit dislocations (as seen in the as-grown and anneal alone, not shown) in pl an-view and a defect-free regrown layer in cross-section. The Si0.7Ge0.3 sample, however, shows the regrown layer full of defects which, using PTEM analys is, appear to be stacking fau lts. The XTEM micrograph show that these stacking faults extend between the surface and the heterostructure interface. PTEM was used to study the strain relieving de fects, specifically misfit dislocations, for both un-implanted and implanted samples post ann ealing to differentiate between the thermal

PAGE 78

78 and implant contributions to relaxation. PTEM images, taken under g220 bright field, of the asgrown, as-grown anneal, and implant plus anne al samples are presente d in Figure 3-7. The corresponding linear misfit dislocation densities are presented in Figure 3-8. The 10 and 20% Ge samples exhibit similar misfit dislocation densitie s for both the anneal only case and the implant plus anneal case. However, after implant plus anneal, the 30% Ge sample displayed misfit dislocation spacing much smaller than the equivalent anneal only case, this concludes that further relaxation took place after ion implantation. Upon re laxation, stacking faults are also observed in conjunction with an increase in misfit dislocat ions indicating further strain relaxation of the strained film than annealing alone. The nucleation of these defects will be discussed in the next section. XRD/TEM discrepancy. Comparing the HRXRD results to the PTEM quantitative results for the anneal case only, a relative differen ce of 1-2 % relaxation is observed. Further analysis of these results sheds light on the cause of this discrepancy. The quantitative relaxation observed via HRXRD is shown in Figure 3-9. Thes e results can be compared to the relaxation quantified using misfit spacing from PTEM analysis in Figure 3-8. The relaxation observed via PTEM is on the order of 1-2% of the initial strain for all sample s after anneal only. The relative discrepancy between the HRXRD and the PTEM misfit spacing calculations are explained by both the manner in which these defects affect the X-ray spectra and relative sampling area. The quantitative HRXRD results in this work are calculated using the relative peak positions of the SiGe and strained Si peaks. Additionally, the presence of misfit dislocations may also be observed in the spectra via broadening of the peak which was observed for all annealed samples. Due to excessive peak broadening from the presence of misfit dislocations an additional margin of error is introduced into the HRXRD measurement technique. Also, the sampling area of the

PAGE 79

79 PTEM and HRXRD differ greatly. The HRXRD re sults are an average of the lattice spacing over a cm2 range whereas the PTEM result s are site specific (a few m2). Thus, PTEM results are more susceptible to local vari ations in the wafer. Therefore, it should be noted that even though the HRXRD measurements indicate no rela xation, all samples undergo ~ 1-2% strain relaxation of the initial strain after annealing on ly. Additionally, for the 10 and 20% Ge samples post implant and anneal, the relaxation observed is al so on the order of 1-2% decrease in strain. Thus, implantation related relaxati on cannot be concluded to have occurred. The results from both TEM and HRXRD, however, conclude that the strain re laxation observed for the 30% sample is much greater than 12% of the initial value and is hence due to the break down of SPER via nucleation of stacking faults and a re lated increase in misfit dislocation density. Furthermore, by comparing the quantitative analys is of both techniques in Figures 3-8 and 3-9 similar trends between defect density a nd degree of strain relaxation is observed. EOR evolution in the presence of tension. Further analysis of the PTEM samples at higher magnification indicates the presence of EOR damage in the strained samples similar to that seen in pure Si samples post implantation and anneal. The EOR evolution and population, however, does seem to be affected by strain. Figure 3-10 shows an anneal series at 800 C for Si control, 20% and 30% Ge samples. After si milar anneal conditions, the EOR damage is observed to be much smaller in bo th size and number in the strained sample than the Si control. The total interstitial concen tration trapped by both {311} defect s and dislocation loops are quantified and presented in Figure 3-11. The Si control and unrelaxe d Si on 20% Ge have similar concentration of trapped interstitials init ially then deviate as s een in the figure. The {311} defects appear to be stabili zed under tension but do not lead to dislocation loop formation, as seen by comparing the EOR evolution in Figu re 3-10 for the pure Si and the Si on 20% Ge.

PAGE 80

80 Tension seems to suppress the {311} to dislocation loop process which causes the deviation of trapped interstitial density from the initial concen tration at 5 minutes. However, for the relaxed Si on 30% Ge, a lower number of interstitials are initially trappe d indicating that some of the excess interstitials may be contributing to the re laxation process. These excess interstitials, a result of the implant, may be providing a source for the generation of extrinsic stacking faults which then nucleate more misfit dislocation. If this hypothesis is true, the proximity of the implant damage to the interface should affect relaxation by acting as a sink for the excess interstitials. As the implant damage is placed closer to the Si/SiGe interface, relaxation will increase and a higher density of misfit disloc ation should be observed. This hypothesis was further analyzed by varying implant energy to cha nge the proximity of the implant damage to the Si/SiGe interface and will be further discussed in the next chapter. In summary, this study suggests that strain ed silicon can be amorphized and regrown without strain relaxation for all Si1-xGex compositions up to Si0.8Ge0.2 for anneal temperature up to 800 C for 30 minutes. For the Si0.7Ge0.3 strained sample, the solid phase regrowth of the amorphous layer breaks down and results in the form ation of regrowth rela ted defects within the amorphous region and extending back down to the strained Si/ Si1-xGex interface. This is the first observation of regrowth re lated defects extending below the original amorphous/crystalline interface in amorphized silicon. Furthermore, these defects are found to be primarily stacking faults rather than the traditionally observed hairpin dislocations. Since relaxation was only observed when the regrowth related defects were present, the result s also suggest that the defects are contributing to the relaxation of the strained layer. Addition ally, the EOR evolution seems to be altered in the presence of te nsile strain when compared to Si control; the evolved dislocation loops in the strained structures are smaller in size and numbe r. This indicates that, in

PAGE 81

81 conjunction with a certain level of strain, excess interstitials, a result of the implant, are providing a source for the generation of extrinsi c stacking faults which then nucleate misfit dislocations. The additional mi sfit dislocations then promote fu rther relaxation than annealing alone. The possibility of excess interstitials acting as a source for misfit dislocation nucleation will be further explored in the next chapter. In conclusion, these results i ndicate that a critical misfit strain between 0.74% a nd 1.1% results in a breakdown of the SPER process and the formation of extended regrowth related defects that promote further relaxation. Defect Nucleation Study Strain relaxation and def ect nucleation post SPER were obser ved solely for the highest strained samples. The mech anism of defect formation a nd the morphology of the amorphouscrystalline interface was studied using a series of isothermal anneals performed at 500 C. At this temperature, the regrowth velocity is slow enough such that images of the a-c interface at various stages prior to complete regrowth can be captured. Si0.7Ge0.3 samples were annealed for 15, 30, and 45 minutes and XTEM was used to observe the progression of SPER. XTEM images of the samples are presented in Figure 3-12 in both bright and dark-field imaging mode. SPER at higher strain results in relaxation by nucleating re growth related defects which propagate towards the surf ace with the advancing a-c interf ace. These defects are then observed to extend towards the Si/SiGe interf ace once SPER is complete. At 30 minutes, presented in Figure 3-12 (a) and (b), the a-c inte rface is observed to be planar and the regrowth related defect is located above th e original a-c interface. At 45 minutes, as seen in (c) and (d), these defects begin to extend down to the epilayer interface promoting further strain relaxation. At higher anneal temperatures, these regrowth de fects are determined to be stacking faults via XTEM and PTEM imaging. Figure 3-13 (a) and (b) show high-resolution XTEM images of

PAGE 82

82 these faults extending from the surface to the epil ayer interface. However, after a 500 C 45 minute anneal when regrowth is complete, no stac king faults were seen in PTEM micrographs. The absence of stacking faults immediately af ter regrowth and presence of them after further annealing suggests that 30 and 90 Shockley pa rtial misfit dislocation pairs are nucleated upon SPER to promote relaxation. Th ese partials can then extend down to the epilayer interface via point defect diffusion. Upon further annealin g, these partials glide apart and leave a fault between them. These faults are observed in PTEM to be relatively wide in length after anneals around 650 C. After further annealing at 800 C, the width of these faults is observed to decrease in conjunction with an increase in the 60 misfit dislocation. This observation suggests that the presence of stacking fa ults is promoting further misf it dislocation nucleation. The decrease in the width of the st acking fault in conjunction with increased perfect misfit density indicates that this process in conservative. The generation of partial misf it dislocation to provide strain relaxation in other stra ined bicrystal systems have b een previously reported [Gut98, Gut01a, Hwa91] and were first theorized and observed by Cherns [Che74, Mat75]. Relaxation via partial dislocation generation usually occurs at higher misfit st rain than the strain magnitude sampled in this work. Furthermore, the transformation of partial misfit dislocation to perfect misfit dislocations has been reported experime ntally by Tamura in the GaAs on Si system [Tam96]. Plastic deformation in bulk Si via partial dislocations is well known. The deformation behavior has been studied using two different techniques [Rab00]. The first is microindentation, the drawbacks of this technique are that the st ress tensor is unknown and the plastic region is localized in a small area making TEM analysis difficult. The second is deformation under a confining pressure which allows control of deviatoric and hydrostatic pre ssure. Both techniques

PAGE 83

83 induce a compressive stress. At the lowest temp erature of 500 C used in this work, the yield stress is ~400 MPa. For the highest temperature of 800 C this value dr ops significantly to ~20 MPa. According to this data, one might expect the thin films in this work to deform via dislocation at lower strain levels However, from prior work in thin films it is known that higher stress in the films can be attain ed than predicted by th e stress-strain studies in bulk silicon. The heteroepitaxy system is more complex and have other factors that contribute to the strain relaxation i.e. deformation. In bulk Si under ~500 MPa uniaxial stress induced by 4-point bend, plastic deformation was observed at low temperat ure (~550 C) [Phe04]. Th is results correlates well with the prior deformation studies carried out. In summary, this experiment showed that re growth related defects are observed for the highest strain case only. These regrowth relate d defects are confirmed to be stacking faults above 500 C. These defects are nucleated due to excess stress in the film that cannot support the SPER process. At lower anneal temperature and times; the regrowth related defects are observed to extend between the original a-c interface and the surface. After further annealing (at higher temperatures and/or longer times), the defects appear to extend from the surface down to the heterostructure inte rface and promote strain relaxation. Pr opagation of defect s below the initial a-c interface has not been previous ly reported. Furthermore, it shoul d be noted that the result in this work confirm prior work indicating that regr owth velocity is not affected by biaxial tensile strain [Phe05]. Relaxation: Thermally Acti vated Glide Pro cess Study The relaxation behavior of 30% Ge, 12keV Si+ implanted samples was studied using a series of 30 minute isochronal anneals perf ormed at 500, 575, 650, 725 and 800 C. Additional anneal times were carried out at the low and hi gh temperatures to confirm that the relaxation process is linear at constant temperature. Th is experiment was designed to study the thermally

PAGE 84

84 activated glide process of the SPER related defect s and compare it to dislocation glide observed in bulk Si. Since this process is being meas ured via XRD it should be noted that the misfit dislocation propagation via glide is being indirectly measured through monitoring of degree of strain relaxation. Prior work has used in situ TEM measur ements to determine the activation energy for glide [Hul89]. Prior studies pertaining to the plastic deformation of Si show that dislocation propagation follows Arrhenius behavior and occurs via dislo cation glide. The activation energy for plastic deformation via dislocation glide process is 2.2 eV for pure Si [Ale68]. The activation energy for relaxation in strained Si measured via XRD is presented in Figure 3-14 as rate of relaxation versus 1/kT. The activation energy is found to be 0.7 0.2 eV. This is much lower than the activation energy for plastic defo rmation in bulk Si. Compara tively, an activation energy for glide has been documented for a Si0.7Ge0.3 on Si thin film relaxation after an anneal only with a value around 1.1 eV [Hul89, Dod88] and 1.38 eV [Lei 01], lower than the 1.8 eV determined for SiGe bulk material. The explana tion proposed by Hull et al. [Hul89] is that as the dislocation density and anneal temperature increase, the prop agating dislocations have to cross more and more orthogonal dislocations in a given lengt h. These intersecting events are observed to impede propagation. Activation energies for glide are the sum of kink formation and migration energies. The lower activation energy in thin films versus bulk ma terial has been attributed to lower formation energies of kinks, due to the close proximity of the free surface and the re laxing interface. For the work presented here, since th e nucleation of the defect is a result of the SPER breakdown, the activation energy may be dominated by the migration term. The formation energy term is thought to be negligible due to the defects fo rming as a result of SPER breakdown. This is

PAGE 85

85 especially true since the activation energy is being measured by the re laxation rate which, in turn, depends on the rate at which misfit dislocations propagate. In conclusion, the relaxation observed is determined to be a result of disl ocation nucleation and pr opagation initiated by the SPER breakdown with activation energy of 0.7 0.2 eV. Conclusion The results in this chapter have show n three ma in points. First, unde r biaxial tension there is a critical strain for SPER breakdown that lies between 0.74 and 1.1% st rain. Thus, strained silicon can be amorphized and regrown without strain relaxation for all Si1-xGex compositions up to Si0.8Ge0.2 (0.74% strain). For the 1.1% strained (Si0.7Ge0.3) sample, the solid phase regrowth of the amorphous layer breaks dow n and results in the formation of regrowth related defects within the amorphous region. These defects th en extend back down to the strained Si/ Si1-xGex interface. This is the first observation of regrowth related defects extending below the amorphous/crystalline interface. Additionally, th e EOR evolution and population seems to be altered in the presence of tensile strain; the evolved dislocation loops are fewer in number especially in the relaxed Si on Si0.7Ge0.3 sample. This indicates that, in conjunction with a certain level of initial strain, excess interstitials, a result of the implant, are providing a source for the generation of extrinsic stacking faults wh ich then nucleate misfit dislocation. Thus, implantation increases misfit dislocation density and promotes further relaxation than annealing alone. The possibility of excess interstitials acting as a source for misfit dislocation nucleation will be further explored in the next chapter. In summary, these results indicate that a critical misfit strain between 0.74% a nd 1.1% results in a breakdown of the SPER process and the formation of extended regrowth related defects. Second, defects are nucleated as the a-c inte rface progresses towards the interface during SPER, forming regrowth related defects. Once fully regrown, these defects propagate down to

PAGE 86

86 the interface, promoting enhanced strain relaxa tion of the layer than annealing alone. The regrowth related defects are pr imarily stacking faults rather than the traditionally observed hairpin dislocations. Since st rain relaxation was only observed when these regrowth related defects were present, the results suggest that these defects are cont ributing significantly to enhanced relaxation of th e strained layer. Finally, the relaxation process is a thermally activated glide process with activation energy of 0.7 eV. This value is signifi cantly less than th e 2.2 eV glide energy observed in bulk Si. The difference is theorized to be due to the formati on term in this work being negligible since the defects are formed as a result of SPER breakdown. Thus, the activation energy in this work is dominated by the migration term. Overall, the re laxation process is concluded to be a result of dislocation nucleation and propagation via dislocation glide initiated by the SPER breakdown.

PAGE 87

87 10010110210310410510610734 34.5 35 50 nm Si on Si0.9Ge0.1 50 nm Si on Si0.8Ge0.2 50 nm Si on Si0.7Ge0.3 SiIntensity (counts/sec)Omega (deg) Figure 3-1. HRXRD (004) rocking curve of all as-grown strained Si on relaxed Si1-xGex samples.

PAGE 88

88 Figure 3-2. XTEM of as-implanted strained Si/ Si0.7Ge0.3 sample implanted with 12 keV Si+. Surface End of Range Damage Amorphous Region 20 nm SiGe Si

PAGE 89

89 0.1 10 103105107109101110133434.234.434.634.835 Implanted + 500oC 30 min As-grown + 500oC 30 min As-grownIntensity (counts/sec)Omega (o) Figure 3-3. HRXRD (004) rocki ng curves of strained Si/Si0.7Ge0.3 structures annealed at 500 C for 30 minutes.

PAGE 90

90 Figure 3-4a. HRXRD (113) reciprocal space map of as-grown strained Si/Si0.7Ge0.3 structure.

PAGE 91

91 Figure 3-4b. HRXRD (113) reciproc al space map of strained Si/Si0.7Ge0.3 structure implanted with 12 keV Si+ and annealed at 500 C for 30 minutes.

PAGE 92

92 Figure 3-4c. HRXRD (113) reciproc al space map of strained Si/Si0.7Ge0.3 structure implanted with 12 keV Si+ and annealed at 800 C for 30 minutes.

PAGE 93

93 0 20 40 60 80 100 10 20 30 500oC 650oC 800oC% Strain Relaxation% Ge Figure 3-5. Strain Relaxation calculated via HRXRD RSM fo r strained Si on relaxed Si1-xGex samples implanted with 12 keV Si+ and annealed for 30 minute at temperatures indicated.

PAGE 94

94 Figure 3-6. XTEM (top row) and PTEM (botto m row) for all strained Si on relaxed Si1-xGex samples implanted with 12 keV Si+ and annealed for 30 minutes at 650 C. Si0.8Ge0.2 Si0.7Ge0.3 Si0.9Ge0.1 Surface Si/SiGe interface Regrowth Defects 50 nm

PAGE 95

95 Figure 3-7. PTEM images for AG, AG plus anneal, and implanted with 12 keV Si+ plus anneal for all strained Si on relaxed Si1-xGex samples. All anneals carried out for 30 minutes at 800 C. 12 keV + anneal AG 10% Ge 20% Ge 30% Ge AG + anneal 500 nm

PAGE 96

96 0 5 1041 1051.5 1052 10510 20 30 AG AG + anneal 12keV + annealLinear Defect Density (#/cm)% Ge Figure 3-8. Linear defect densit y quantified using PTEM as a func tion of Ge concentration after 30 minute anneal at 800 C.

PAGE 97

97 0 20 40 60 80 100 10 20 30 AG AG + annealed 12keV + anneal% Strain Relaxation% Ge Figure 3-9. Percent strain relaxation quantified using HRXRD SM measurements as a function of Ge concentration after a 30 minute anneal at 800 C.

PAGE 98

98 Figure 3-10. PTEM images for Si co ntrol and strained Si on relaxed Si1-xGex samples implanted with 12 keV Si+ and annealed for 5, 30, and 300 minutes at 800 C. 5 min 30 min 300 min 30% Ge 20% Ge Si

PAGE 99

99 1012101310141015050100150200250300350 Si Si on 20%Ge Si on 30% GeTotal Trapped Interstitals (#/cm2)Anneal time (min) Figure 3-11. Trapped interstitia l concentration as a function of anneal time for Si, Si on Si0.8Ge0.2, and Si on Si0.7Ge0.3 implanted with 12 keV Si+ and annealed at 800 C.

PAGE 100

100 Figure 3-12. XTEM of stra ined silicon layer of the Si0.7Ge0.3 implanted with 12keV Si+ and annealed at (a) and (b) 500 C for 30 minut es and for (c) and (d) 45 minutes. XTEM are imaged using (a) and (c) bright-field and in (b) and (d) dark-field mode. (a) (d) (c) (b) 50 nm

PAGE 101

101 Figure 3-13a. XTEM of stacking faults obser ved in the strained si licon layer of the Si0.7Ge0.3 sample implanted with 12 keV Si+ and annealed at 800 C for 30 minutes. Surface Strained Si/SiGe interface

PAGE 102

102 Figure 3-13b. High-Resolution XTEM of stacking fa ults observed in the strained silicon layer of the Si0.7Ge0.3 sample implanted with 12 keV Si+ and annealed at 800 C for 30 minutes. 5 nm

PAGE 103

103 0 1 2 3 4 5 6 7 10111213141516ln Relaxation Rate1/kT Figure 3-14. Plot of ln rela xation rate vs. 1/kT for 12keV Si+ implanted strained silicon on relaxed Si0.7Ge0.3 sample. Linear regression of data is shown in red. Ea = 0.7 0.2 eV

PAGE 104

104 CHAPTER 4 TENSILE STRAIN: PROXIMITY EFFECT W ith the introduction of strain technology in CMOS devices, the eff ects of strain on the individual fabrication processes ha ve been increasingly important to determine. The work in this chapter investigates the use of amorphizing implants in tensile st rained Si regions. Specifically, to investigate whether the proximity of the ac interface and Si/SiGe interfaces affect strain relaxation. Epitaxially strained silicon laye rs were grown on virtual SiGe substrates and were used as test vehicles to study the effect of strain on SPER. Since epitaxial layers were used to study this effect, it is also important to design a study explore the effect of the Si/SiGe interface on the relaxation process as the strain is elastically accommodated by the latt ice mismatch at the heterostructure interface. Results in the previous chapter showed that the EOR evolution seemed to be altered in the presence of tensile strain when compared to Si control; the evolved dislocation loops in the strained structures were smaller in size and number. This indicated that, in conjunction with a certain leve l of strain, excess interstitials were providing a source for the generation of extrinsic stacking faults which then nucleated misfit dislocations. The additional misfit dislocations then promoted further relaxa tion than annealing alone. The possibility of excess interstitials acting as a source for misfit di slocation nucleation will be further explored in this chapter by varying the impl ant energy thereby, varying the im plant damage proximity to the Si/SiGe interface. Experimental Design Strained Si structures were grown on relaxed Si1-xGex (Ge fractions of 0, 10, 20, and 30) virtual substrates via Molecular Beam Epitaxy (M BE) at the University of Aarhus, Denmark. 100mm 0.005-0.020 ohm-cm (100) n-t ype Czochralski-grown silic on wafers were used for

PAGE 105

105 substrate wafers. Virtual substrates with low threading disloca tion densities were grown with compositionally graded buffer layers which incorpor ated Ge at a rate of 10 at.% per micrometer at temperatures of 750 C to 800 C [F it91]. A 630 nm thick fully relaxed Si1-xGex layer of corresponding composition was then grown on top of the buffer layer, followed by a 50 nm strained silicon capping layer. Th e strain of the structures was experimentally determined using the in-plane lattice parameters, aSi-cap and aSiGe, obtained from HRXRD a nd the relationship in Equation 4-1. % Strain Relaxation = (aSi-cap aSiGe)/ aSi aSiGe x 100% (4-1) The silicon results from this analysis show that the cappi ng layers grown on Si1-xGex with Ge fractions of 0, 10, 20 and 30 correspond to a Si layer strain of 0, 0.37, 0.74, and 1.1% strain, respectively. XTEM analysis was also used in or der to verify proper growth of strained films. The structures were then ion-im planted with 5, 12, and 18 keV Si+ ions at a fluence of 1x1015 atoms/cm2, generating an amorphous layer ~15, 30, 40 nm thick, with varying proximity to the epitaxial interface while confining them within the stra ined layer. An isothermal study at 800C for 5, 30, and 300 minutes was then carried out. This experiment studied the effect of impl antation damage proximity to the epitaxial interface on strain relaxation for the highest strain cas e only. Ion implantation is a nonconservative process which introduces excess point defects into the material. At temperatures between 600 and 800 C, excess point defects reco mbine and the remainin g excess interstitials agglomerate to form 311 defects. On further a nnealing, the 311 defects di ssolve and contribute to the formation and coarsening of dislocation loops. These defect s at all stages are termed EndOf-Range (EOR) damage ]. Since it is likely that strain relieving misfit dislocations may be nucleated at or near the EOR damage created by ion

PAGE 106

106 implantation, the proximity of the EOR damage to the Si/SiGe interf ace may influence the number and type of defects formed and hence aff ect the degree of strain relaxation. Thus, this experiment varied both strain and implant ener gy in order to investigate the effect of the proximity of the EOR to the epitaxial interface on strain relaxation as a func tion of initial strain. Once annealed, the strain state and crystalline quality was characterized. A Pananalytical MRD XPert was used to obt ain HRXRD rocking curves and reciprocal space maps to determine the degree of strain relaxation for each sample. Previous studies have also employed rocking curves to study the strain relaxation observed in the strained silicon capping layer using pseudomorphic Si/ Si1-xGex structures [Cro04, Phe05]. However, further investigation has determined spac e maps to be a preferable technique for the measurement of annealed samples. Reciprocal Space Maps (RSM)s are compilations of rock ing curves taken at a range of positions to create a 2D map of intensity in the vicinity of the Bragg reflection. The RSMs allows the observation of very low intensit y peaks from the strained silicon layer that rocking curves do not clearly obta in. Using these maps, the relaxati on of the layer can be directly observed by monitoring the shift of the strained silicon peak to ward the silicon substrate. Additionally, the crystalline integrity of the layer can also be monitored by observing any broadening of the peak that would indicate a disordered or def ected layer [Pie04]. Therefore, space maps are preferred for multilayered structur es because the lattice parameters can be obtained from peak position and the de gree of relaxation ob tained directly. A JEOL 2010F high-resolution transmission electron microscope (HRTEM) and a JEOL 200CX TEM were used to study cross-sections of the regrown layer (XTEM). Samples were prepared using a Dual-Beam FEI Strata DB 235 Focused Ion Beam. XTEM measurements were used to confirm the amorphous layer depth and layer thicknesses. Plan-view (PTEM) samples

PAGE 107

107 were also prepared and imaged using a JEOL 200CX operating at 200kV using g220 weak-beam dark-field (WBDF) imaging to quantify dislocation density. Results As-Implanted Sample Analysis Analysis of the as-grown sam ples can be found in the previous chapter. The experiment presented in this chapter varied the implant ener gy in order to change the proximity of the a-c interface to the epitax ial interface. A ll implants were tailored to be strictly confined within the strained layer. Previous work suggests th at the amorphization thre shold decreases with increasing Ge concentration due to a decrea se in binding energy [H ay92, Lie93, ORa96]. However, since all implants were conducted with in the Si capping layer and not through to the SiGe layer, the amorphous depths were constant as a function of Ge fraction. XTEM of asimplanted Si on SiGe and Si control samples c onfirmed that amorphous de pth did not vary with %Ge for this implant condition. The amorphous de pths were within 1 nm for all samples within the error of XTEM measurement. Proximity Effect Experiment The ef fect of implant proximity to the defect density and degree of strain relaxation will be discussed for the highest strain case only. The PTEM micrographs, taken at g220 BF and WBDF, of the samples post implant and anneal at 800 C for 30 minutes is presented in Figure 4-1. The micrograph in (a) is a BF image after anneal alon e, (b) is after 5keV implant and anneal, (c) is after 12keV implant and anneal, and (d) is after 18 keV Si+ implant and anneal. In comparing these results, notice the misfit dislocation spaci ng after each anneal and implant condition. The anneal only and the 5 keV implant and anneal samples both have similar low densities and the largest average misfit spacing, i.e. lowest degree of relaxation. Once the samples are implanted with higher energies (closer proximity) and annealed, the average misfit spacing decreases

PAGE 108

108 indicating a higher degree of stra in relaxation is observed. The misfit dislocation density was quantified in terms of linear density as outlined in the procedure in Chapter Two. The quantitative HRXRD and PTEM for sample s annealed at 800 C for 30 minutes is presented in Figure 4-2 as a func tion of implant energy. The zero implant energy indicates data for the anneal only case. The HRXRD results show a similar trend to that seen in the quantitative PTEM results, indicating that the defect density is rela ted to the degree of relaxation. The defect density of the anneal only and the 5 keV samples are similar in quantity and misfit spacing. Thus, if relaxation is contributed by form ation of misfit dislocat ions, these two samples should exhibit the same degree of relaxation via XRD measurements. However, this is not the case. The as-grown plus anneal sample exhi bited a 1% relaxation wh ereas the 5 keV samples exhibited a ~17% relaxation. In addition to misf it dislocations, the 5 keV samples also exhibited presence of stacking faults in the PTEM micrographs. Since both of these samples exhibited similar misfit dislocation densities, yet the 5 ke V sample observed a higher degree of relaxation, suggests that the presence of stacking faults al so contribute to a la rge percentage of the relaxation. The two higher implant energies, 12 and 18 keV, exhibit a greater density of misfit dislocations in conjunction with hi gher degree of relaxation. From these results, it is evident that with increasing proximity to the interface, a highe r degree of strain relaxation is observed which is confirmed with the higher density of misfit dislocations. Further investigation of the P TEM for the three different im plant conditions in terms of anneal time allows better understa nding of the dependence of st rain relaxation on proximity. Figure 4-3 presents PTEM micrographs, taken under g220 WBDF conditions, for the highest strain case implanted with 5, 12, and 18 keV Si+ and annealed for various times at 800 C. All micrographs were taken at the same magnificati on. Comparing samples annealed at the same

PAGE 109

109 time, the misfit densities increase with increasing pr oximity (implant energy). This effect is also evident at longer anneal times. This trend of increasing misfit density with increasing proximity is more clearly depicted in the quantitative analysis of PTEM presented in Figure 4-4. Furthermore, the misfit density is observed to in crease up to a saturation point at which it levels off. At this point, the strained layer is comp letely relaxed and there is no driving force for dislocations generation. The higher the implant energy the closer the exce ss interstitials are placed with respect to the Si/SiGe interface. Assuming that the net excess interstitials are constant at these energies [Gut01b], the only variable to be considered is proximity of the excess interstitials to the interface. These results indicate that the placement of the a-c interface closer to the epitaxial Si/SiGe interface is highly in fluential on the degree of rela xation and density of misfit dislocations generated. These observations indi cate that the Si/SiGe in terface may acting as a sink for the excess interstitials and promote further relaxation as a result. Discussion The results discussed in the previous chap te r suggested that strained silicon can be amorphized and regrown without strain relaxation for all Si1-xGex compositions up to Si0.8Ge0.2 for anneals up to 800 C for 30 minutes. For the strained Si sample on Si0.7Ge0.3, the solid phase regrowth of the amorphous layer breaks down and results in the formati on of regrowth related defects within the amorphous region which exte nded back down to the strained Si/ Si1-xGex interface. Furthermore, the EO R evolution and population seemed to be altered in the presence of tensile strain when compared to the Si control; the evolved di slocation loops in the strained structures are smaller both in si ze and quantity. This indicates th at, in conjunction with a certain level of strain, excess interstitials, a resu lt of the implant, are pr oviding a source for the generation of extrinsic stacking faults which then nucleate misfit dislocations The possibility of

PAGE 110

110 ion implantation produced excess interstitials acti ng as a source for misfit dislocation nucleation was further explored by varying proximity of the implant damage to the epitaxial interface while still confining it within the strained layer. In developing this experiment, one must cons ider all variable para meters. A change in implant energy with constant fluence and sp ecies will change the a-c interface depth. Additionally, the net interstitial profile will also change. However, with the energy range used in this experiment previous work by Gutierrez et al. [Gut01b] suggests that the net interstitial population is constant. Therefore, the net excess interstitial popul ation of the implants in this experiment is assumed to be constant. Theref ore, placing the implant cl oser to the interface places the net excess interstitials closer to th e interface while the net interstitial population is held constant. This experiment was specifically designed to determine if the interface acts as a sink for point defects. Results in this chapter have shown that proximity of the implant damage to the heterostructure interface influences both misf it dislocation density and degree of strain relaxation. The closer the damage proximity to th e heterostructure interface the higher degree of relaxation combined with a highe r misfit dislocation density is observed. Thus, it is plausible that the interface may be acting as a sink for the excess interstitials which then drives further relaxation when the damage is placed closer to the heterostructure inte rface. Additionally, the results in the previous chapter showed that the end-of-range damage population decreased in conjunction with an increase in misfit dislocations. Therefore, it is plausible that the proximity of the damage to the interface in fluences both trapped interstitial concentration as well as the degree of relaxation. However, future studies below 800 C are recommended to explore the

PAGE 111

111 end-of-range evolution and trapped interstitial concentration as a function of damage proximity to the interface to fully understa nd the role of excess interstitial s to the relaxation process. Conclusions The results in this chapter have show n that the implant damage proximity to the Si/SiGe interface impacts the degree of relaxation and misfit dislocati on density. After a constant anneal time and temperature, 800 C for 30 minutes, and varying implant energy (proximity) the misfit dislocation spacing is observed to decrease with increasing proxi mity. Therefore, exhibiting a higher density and observed relaxation as c onfirmed by HRXRD measurements. Further analysis using isothermal anneals and PTEM show ed that the misfit dislocations are observed to increase in density for the two lower implant energy cases. The highest energy case, already observed complete relaxation at 30 minutes and exhi bits constant misfit di slocation density with further annealing. The PTEM and HRXRD result s are consistent and confirm data observed. In summary, this experiment has determined that misfit dislocati on density and strain relaxation depends significantly on the proximity of the initial a-c interface. The closer proximity of the damage to the Si/SiGe interf ace, the higher the degree of relaxation and higher misfit dislocation density is observed. Additionall y, the results in the previous chapter showed that the end-of-range damage popul ation decreased in conjuncti on with an increase in misfit dislocations. These results indicate that close proximity of the damage to the interface could influence both trapped interstitial concen tration as well as the degree of relaxation. Therefore, from these results it can be infe rred that the epilayer interface may be acting as a sink for point defects when the implant damage is placed in close proximity. To confirm this claim, future studies below 800 C are recommended to expl ore the end-of-range e volution and trapped interstitial concentration as a function of damage proximity to the interface to fully understand the role of excess interstitials to the relaxation process.

PAGE 112

112 Figure 4-1. PTEM images for (a) Si control sample and Si on Si0.7Ge0.3 samples implanted with (b) 5 keV Si+ (c) 12 keV Si+ and (d) 18 keV Si+ and annealed for 30 minutes at 800 C. (a) (d) (c) (b) 100 nm

PAGE 113

113 0 5 1041 1051.5 1052 1052.5 1050 20 40 60 80 100 05101520Linear Defect Density (#/cm) % Strain RelaxationImplant Energy (keV) Figure 4-2. Strain relaxation (cir cles) and linear defect density (s quares) as a function of implant energy after 800 C 30 minute anneal. Zero implant energy indicates data for asgrown plus anneal case.

PAGE 114

114 Figure 4-3. PTEM images for strained Si on Si0.7Ge0.3 samples implanted with 5, 12, and 18 keV Si+ and annealed for 5, 30, and 300 minutes at 800 C. 18 keV 12 keV 5 keV 5 min 30 min 300 min

PAGE 115

115 0 5 1041 1051.5 1052 1052.5 1053 1051 10 100 103 5 keV 12 keV 18 keVLinear Dislocation Density (#/cm)Anneal time (min) Figure 4-4. Linear dislocati on density quantified using PTEM for strained Si on Si0.7Ge0.3 samples implanted with 5, 12, and 18 ke V Si+ and annealed for 5, 30, and 300 minutes at 800 C.

PAGE 116

116 CHAPTER 5 COMPRESSIVE STRAIN: DEFECT NUCLEATION AND STRAIN RELAXATION W ith the introduction of strain technology in CMOS devices the effect of strain on the individual fabrication processes has become increasingly important Specifically, understanding the upper strain limit for successful regrowth under compression is critical to device processing. This chapter outlines experiments to determine th e magnitude of compressive strain that may be used in conjunction with amorphizing impl ants without inducing strain relaxation. In these experiments, epitaxially strained silicon germanium layers are grown on silicon substrates as test structures to study the effect of compressive strain on Solid Phase Epitaxial Regrowth (SPER). The first e xperiment is designed to study the mechanism by which defects nucleate and cause strain relaxation during SPER. The defect nucleation study was performed using lower temperature anneals in order to obser ve the a-c interface prior to complete growth. The second experiment was designed to investigate the interaction of excess interstitials, a result of the implant, with the rela xation process. This study was performed using an isochronal experiment in which the relaxation and defects microstructure is mon itored during the early stages of nucleation and growth and the latter stages of coarsening as seen in Si (if present in SiGe). In summary, by monitoring the quantity of dislocations in strained vs. unstrained samples and studying the kinetics of the dislocation grow th process the relationship between the strain relaxation process and SPER can be further clarified. Experimental Design The experim ental structures were grown using Reduced Pressure Chemical Vapor Deposition (RPCVD). Pseodomor phically strained 50 nm Si1-xGex was deposited on Si (001) substrate with alloy compositions of 0, 16, 22, and 26 % germanium. The as-grown structures were characterized using TEM and XRD techniques in order to ensure proper growth of strained

PAGE 117

117 films. All Post growth, the st ructures were implanted with Si+ with an energy of 12 keV at a fluence of 1x1015 atoms/cm2 to generate a 30 nm continuous amorphous layer confined within the strained layer. Anneals were performed in a quartz-tube furnace under an inert N2 ambient environment. The experiments were designed to study specific factors that may affect the relaxation process. The first experiment consisted of an isothermal study at 500 C performed to investigate defect nucleation and propagati on in the strain ed film. Anneal times were 15, 30, and 45 minutes. With these times and temperature, th e regrowth velocity is slow enough that the amorphous-crystalline (a-c) interfac e will be captured before complete regrowth has taken place. This allows observation of defects as they nucl eate and grow with the us e of XTEM and PTEM. Only the highest and lowest strained samples, im planted with an energy of 12keV, were used for this experiment. Additionally, a preanneal of 400C for 60 minutes was conducted prior to the 500C anneal to relax and planari ze the a-c interface before initiati on of SPER. Analysis of the preanneal, shown in Appendix B, planarized the a-c interface prior to regrowth. Similar low temperature preanneals, also have been conducte d to planarize the a-c interface in prior work [Kin03, Ban00]. The second experiment consisted of a 30 minute isochronal study to investigate the interaction of extended defects and their evolution with the relaxation process. Temperatures of 500, 575, 650, 725 and 800 C were used. In implanted Si, the supersaturation of interstitials lead to evolution and agglomera tion of extended defects when ann ealed above 650 C. If present in strained SiGe, the effect of the evolution of these extended defects will be studied. The highest Ge, 26%, sample implanted with 12keV Si+ was used in this isochronal study.

PAGE 118

118 A Pananalytical MRD XPert was used to obtain XRD Rocking Curves (RC) and Reciprocal Space Maps (RSM) to obtain the degr ee of relaxation observed post anneal. A JEOL 2010F high-resolution transmi ssion electron microscope (HRTEM) and a JEOL 200CX TEM were used to study cross-sections (XTEM) of the regrown layer. Samples were prepared using a Dual-Beam FEI Strata DB 235 Focused Ion Beam. XTEM measurements were used to confirm the amorphous layer depth and layer thicknesse s. Plan-view (PTEM) samples were also prepared and imaged using a JEOL 200CX operating at 200kV using g220 weak-beam dark-field imaging to quantify dislocation density. Results and Discussion As-Grown Sample Analysis The XRD (004) RC for the as-grown structures are shown in Figure 5-1. The position of the diffraction peak depends on Ge composition and strain relaxation in the layer. The oscillation near the SiGe Bragg peak is due to different scattering factors along the Qz axis which arise from the difference in atomic speci es across the boundary la yer. The thickness of the SiGe layer can be calculated using the peak-topeak distance of the oscillations provided that the material is pseudomorphically strained [Bir03]. Therefore, PTEM (not shown) was used to determine if the strained layers were grown pseudomorphically. The 16 and 22% Ge samples showed no misfit dislocation formation and were completely pseudomorphic. The 26% Ge sample, however, was partially pseudomorphic an d exhibited a 0.77% relaxa tion calculated using misfit spacing [Mat74]. XRD RC analysis, however, indicated 0% relaxation for the same sample due to sensitivity limitations of the X -ray technique [Bir03]. The pseudomorphic/nonpseudomorphic conditions of these samples ar e in good agreement with the observations of People and Bean [Peo85].

PAGE 119

119 All as-grown samples were analyzed usi ng XRD RC, X-Ray Reflectivity (XRR) and TEM cross-section to verify Ge concentrations and la yer depth. Simulation of the experimental x-ray spectra from RC and XRR confi gurations yields Ge concentration, thickness, and roughness measurements of the SiGe layers. XTEM analys is of the samples was also used to directly measure the thickness of the layer. A summary of these results are in Table 5-1. The Ge concentrations were within 1 at% of nominal for each wafer. The thickness measurements obtained from the X-Ray simulations and the XTEM results agree well when the strained layers are completely pseudomorphic (16 and 22% Ge). The 26% Ge sample, however, is partially pseudomorphic as determined by PTEM. The disc repancy between the layer thickness measured via X-ray techniques and the XTEM could indicate higher error in X-ray measurement due to scattering from defects at the in terface. To accurately model the parameters for partially relaxed layers, the degree of relaxation must be known a nd measured using an a lternate technique. As-Implanted Sample Analysis All sam ples discussed in this chapte r were implanted with a 12 keV Si+ implant at a fluence of 1x1015 atoms/cm2. The implants were tailored to generate a continuous amorphous region confined within the strained layer as sh own in Figure 5-2. XTEM of as-implanted SiGe and Si samples confirmed that amorphous dept h did not vary with %Ge for this implant condition. The amorphous de pths were within nm for all sa mples this is within the error of XTEM measurement. Defect Nucleation/Interface Roughness Study All strain conditions form ed defects and observed strain relaxation post SPER. The mechanism of defect introduction and the mo rphology of the amorphous-crystalline interface were studied using an isothermal study performed at 500 C. At this temperature, the regrowth velocity is slow enough such that images of the a-c interface at va rious stages prior to complete

PAGE 120

120 regrowth can be captured. Samples were annealed for 15, 30, and 45 minutes and XTEM was used to observe the progression of SPER. Figure 5-3 shows XTEM images, taken under on-axis bright field conditions, of the advancing a-c front transitions from a planar in terface (as-implanted) to one that is rough as SPER progresses. The roughening process is also dependent on the degree of strain; a higher peak to valley amplitude is observed with increas ing magnitude of initial strain. This roughening effect has been previously observed in compressive ly strained Si films [Bar04] and in metastable strained SiGe films [Ang07, Cor96] during SPER and in high dose Ge+ implants into Si to form SiGe [Cri96, Eli96]. A defect-f ree regrown layer and a planar advancing a-c interface, however, has been shown in relaxed SiGe films with up to a 38% Ge concentration [Kri95a]. Additionally, Antonell et al. [A nt96] showed that incorporat ing C into SiGe prior to amorphization delayed the onset of dislocation formation and promoted planar a-c interface growth due to strain compensation effects. Th ese results indicate that the roughening of the advancing a-c front is a strain effect and not a Ge chemical effect. As the a-c front advances towards the surface under compressive strain, an increase in roughness is observed due to compe ting elastic and strain energies, similar to the perturbation observed in epitaxial growth vi a MBE or CVD. Balancing th ese energies yield a minimum wavelength of the perturbation as determined by pr evious research [Sro89]. It is also well known that SPER velocity is dependent on growth direction. The [ 110] and [111] growth velocities are 3 times and 25 times slower, respec tively, than [001] as determined by Csepregi [Cse78]. Thus, once the a-c inte rface has roughened significantly, the slower growing [110] and [111] fronts retard th e overall regrowth velocity as the initial magnitude of strain is increased. This effect can be observed by comparing Figur e 5-3 (a) for 16% Ge and (b) for 26% Ge.

PAGE 121

121 Additionally, it is can be firmly concluded that roughness is enhanced at higher strain magnitudes. The roughening, in turn, causes defect forma tion as the misoriented facets meet and nucleate defects during SPER. PTEM analysis, using g220 WBDF conditions, shown in Figure 54, was used in order to determine the Burgers vect or of the defect. The dislocations are perfect dislocations of a/2<110> type which have been reported in pr ior works [Cri96, San84] and are termed hairpin dislocations. The mechanism of the nucleation and gr owth of these defects have been postulated by T. Sands [San84]. Typica lly, these dislocations were so great in number that it does not allow imaging of underlying misfit di slocations (cross-hatch pattern), a result of strain relaxation, located at th e epitaxial interface. Of the large quantity of PTEM samples analyzed, one sample happened to etch down to the interface and revealed the presence of misfit dislocations. The PTEM image of this sample is shown in Figure 5-5. By examining this image, it can be observed that the hairpin defects are much greater in number and closer in spacing than the misfit dislocations. Thus, it is important to note that the hairpin defects may obstruct the imaging and quantification of misfit dislocations. In summary, this experiment showed that th e a-c interface roughens due to strain and that the roughness increases as the magnitude of strain is increased. Also, defects are nucleated during SPER through the meeting of growth fronts of different orientations. These defects were determined to be hairpin disloca tions and are so vast in number that they can obstruct imaging of misfit dislocations. Temperature Dependence Study The relaxation behavior of 26% Ge, 12keV sa m ples was studied using a series of 30 minute isochronal anneals performed at 500, 57 5, 650, 725 and 800 C. This experiment was designed to study the evolution and agglomerati on of supersaturated in terstitials and their

PAGE 122

122 interaction with the relaxation pr ocess. Crosby et al. [Cro05] studied the evolution of these extended defects as a function of Ge in rela xed SiGe alloys. The samples studied were implanted with Si+ below the amorphization threshold. In low Ge concentration samples, the dissolutions rates were similar to that of Si. At intermediate concentrations, 25% and 35% Ge, the dislocation loops appear relatively stable. Above 50% Ge, the dislocation loops become unstable. This study aims to determine if the disl ocation loops are stable in strained SiGe films and how they influence relaxation. XRD RCs were obtained to determine strain relaxation post implantation and anneal. Additionally, PTEM and XTEM were used to study the defect microstructure and depth. PTEM quantification was used to obtain disl ocation density. All samples displayed an extended dislocation network in the form of a large connected array rather than individual dislocations. A linear (rather than area) qua ntity was therefore measured for more accurate quantification. Figure 5-6 shows a series of PTEM images as a function of anneal temperature. The spacing of the hairpin disloc ations, and thus the linear di slocation density quantified in Figure 5-7, do not depend on the anneal temperature. Additionally, the sample in Figure 5-6(a) shows hairpin dislocation presen t when regrowth has not been completed. This sample was partially regrown yet hairpins are observed. These results suggest that the defects nucleate as the SPER process progresses and remain stable at higher temperatures. Additionally, unlike the results from Crosby et al. [Cro05] no extended defects were obse rved in any samples of this study. This suggests that the roughness of the ac interface, not point de fects, is a dominant factor in defect formation in strained SiGe alloys. The hairpin dislocations are nucleated wh en the misoriented facets of the rough a-c interface meet and propagate as the a-c interface propagates towards the surface. It is also

PAGE 123

123 observed, using XTEM, (not show n) that these defects do not glide down to the epitaxial interface. The defective layer is contained between the depth of defect nucleation and the surface for lower strain cases and ex tends to the entire layer in the highest strain case. This will be further discussed in the next chapter. XRD RC for the isochronal study is shown in Figure 5-8. The SiGe peak is observed to broaden and the interference fringes are absent af ter higher temperature a nneals. This is an indication of larger variation of SiGe latti ce parameter and a disturbance of the epitaxial interface. Thus, the film is undergoing strain re laxation. Further investig ations of the highest temperature anneal, 800 C for 30 minutes, using RSM shows that th e strain state of the SiGe film is gradient in nature. The RSM spectrum, in Figure 5-9, shows that the SiGe peak is broadened and extends from full relaxed to fully st rained state. Also, the elongation of the peak shape in the direction of the relaxation line (refe r to Figure 2-5 for clarification) indicates the layer is mosaic in nature. For all samples exhib iting this gradient, the hi ghest intensity peak was used to quantify strain relaxation. When this av erage strain relaxation pl otted in Figure 5-10 is compared to the defect density shown in Figur e 5-7, the results do not similar trends. This discrepancy between the XRD and TEM data is discussed further in the next chapter. This can be further understood by considering tw o mechanisms of relaxation. The first is that the relaxation occurring at the Si/SiGe interface is due to the lattice mismatch not being accommodated elastically and nucleating misfit dislocations [Mat74, Peo85]. Second, the relaxation observed in the SiGe layer is due to the defects nucleated via roughening of the a-c interface during regrowth and when the facets of the interface meet observed in this work and others [Ang07, Hon92, Lee93]. XRD measures th e first type of relaxa tion and the second is observed using XTEM and PTEM. This fact ca uses some discrepancy in comparing the XRD

PAGE 124

124 and TEM results for these samples. More specifi cally, this discrepancy is due to the localized defect-free region and highly defective re gion, causing a strain gradient. This experiment has shown the defect nucleation is dominated by the a-c roughness. The experiment has also shown that hairpin dislo cation density is stable and not a function of temperature. It does no t correlate, therefore, with the stra in relaxation observed via XRD which increases as anneal temperature is raised. Conclusion The results in this ch apter have shown four main points. First, the a-c interface roughens due to compressive strain and this roughness increases when the magnitude of strain is increased. The roughness is a strain effect pr eviously reported in both comp ressively strained Si [Bar04] and in metastable SiGe films [Ang07, Lee93]. It is also similar to the effect seen in compressively strained epitaxial layers gr own via deposition techniques [Sro89]. With increasing initial strain in the film, there is a more strain energy to compensate by decreasing elastic energy. The bala nce of these energies yields a mi nimum wavelength of the roughness for the a-c interface. Second, defects are nucleated during SPER by the competition of the different direction growth fronts. Using g b analysis, these defects were determin ed to be hairpin dislocations with Burgers vector of a/2 <110> which were first ob served by T. Sands [San84]. These defects were later observed in metastable SiGe SPER [Ang07] and in Si samples implanted with high dose Ge+ to form SiGe via SPER [Cri96]. The a-c inte rface in these other structures also exhibited roughness similar to these observations. Third, in relation to the tensile case discussed in previous chapters, the critical strain for SPER breakdown is much lower in compression. Th e a-c interface roughne ss in compressions is the main cause for both defect nucleation and a lower critical strain in compression versus

PAGE 125

125 tension. The rough a-c interface allows defect nucle ation at much lower compressive strain than defect nucleation by yielding in tension. This is also the main cause for the difference in defect type observed. The regrowth related defect s observed in compression were a/2<110> type dislocation network, while the de fects in tension were stacking faults. The roughening a-c interface in SiGe films does not allow the film to reach its yield point. Instead, defects are introduced which relax the strain and a driving force towards further dislocation nucleation no longer exists. Fourth, the experiment has also shown that ha irpin dislocation density is stable and not a function of temperature over the ra nge studied. It does not correlate therefore, with the strain relaxation observed via XRD which increases as the anneal temperature is raised. The defects nucleate at the meeting of the facets of the r ough a-c interface and propa gate towards the surface with the advancing front. It is also observed th at these defects do not gl ide down to the epitaxial interface. The defective layer is contained between the depth at which defect nucleation occurred and the surface. At lower strain, there lies a defect-free layer which retained its initial strain with a highly defective laye r near the surface which is fully relaxed. The experiment in the next chapter will discuss how this defect-free la yer is influenced by both initial magnitude of strain and the proximity of the a-c interface to the epitaxial interface.

PAGE 126

126 Sample Conditions XRR RC XTEM Thickness %Ge X^2 Thickness (nm) Roughness (A) Density (g/cc) Thickness (nm) % Ge Thickness (nm) 50 16 1.60E-02 49.3 7 2.91 49.9 15.6 50.0 50 22 1.58E-02 48.4 10 3.08 47.0 21.6 47.0 50 26 4.20E-02 57.7 9 3.21 58.0 26.0 51.0 Table 5-1. Summary of XRR a nd RC spectra simulations compared to XTEM measurements.

PAGE 127

127 10110210310410510610733.433.633.83434.234.434.634.835 50 nm Si84Ge16 47 nm Si78Ge22 51 nm Si74Ge26Intensity (counts/sec)Omega (deg) Figure 5-1. XRD (004) rocking curves for a ll as-grown strained SiGe on Si samples.

PAGE 128

128 Figure 5-2. Cross-section of as-impla nted strained SiGe on Si samples. Surface Amorphous Region 10nm SiGe Si

PAGE 129

129 Figure 5-3. XTEM of strained Si1-xGex on Si samples implanted with 12keV Si+ annealed at 500 C for 45 minutes (a) x= 16 (b) x= 26. Arrows indicate a-c interface and hatched white lines indicate epitaxial interface. (a) 50 nm (b)

PAGE 130

130 Figure 5-4. PTEM image of strained Si0.74Ge0.26 on Si samples implanted with 12 keV Si+ and annealed for 45 minutes at 500 C.

PAGE 131

131 Figure 5-5. PTEM image of strained Si0.74Ge0.26 on Si samples implanted with 12 keV Si+ and annealed showing region etched down from surface to the SiGe/Si interface, reveling the misfit dislocations that contribute to strain relaxation.

PAGE 132

132 Figure 5-6. PTEM images of strained Si0.74Ge0.26 on Si samples implanted with 12 keV Si+ and annealed at (a) 500C (b) 575C (c) 650C (d) 725C (e ) 800C all for 30 minutes. (a) (e) (d) (c) (b)

PAGE 133

133 0 5000 10000 15000 20000 25000 450500550600650700750800850Linear dislocation density (#/cm)Temperature (oC) Figure 5-7. Linear dislo cation density quantified us ing PTEM for strained Si0.74Ge0.26 on Si samples implanted with 12 keV Si+ and annealed for 30 mi nutes at temperatures ranging from 500 to 800 C.

PAGE 134

134 10 100 1000 10000 10510610733.633.83434.234.434.634.8Omega (o) As-grown As-grown and annealed 800 oC 30 min 12 keV and annealed 575oC 30 min 12 keV and annealed 650oC 30 min 12 keV and annealed 725oC 30 min 12 keV and annealed 800oC 30 minIntensity (counts/sec) Figure 5-8. XRD RC spect ra for strained Si0.74Ge0.26 on Si samples implanted with 12 keV Si+ and annealed for 30 minutes at temper atures ranging from 500 to 800 C.

PAGE 135

135 Figure 5-9. HRXRD (113) reciprocal space map for strained Si0.74Ge0.26 on Si samples implanted with 12 keV Si+ and annealed at 800 C for 30 minutes.

PAGE 136

136 -5 0 5 10 15 20 450500550600650700750800850% Strain RelaxtionTemperature (oC) Figure 5-10. Average strain relaxa tion from RC data for strained Si0.74Ge0.26 on Si samples implanted with 12 keV Si+ and annealed for 30 minutes at temperatures ranging from 500 to 800 C.

PAGE 137

137 CHAPTER 6 COMPRESSIVE STRAIN: PROXIMITY EFFECT The epitax ial interface of the strained silicon germanium structures used in this work may act as a sink for excess intersti tials generated during implantati on and thus influence strain relaxation during SPER. It is imperative, theref ore, to understand if the proximity of the a-c interface to the epitaxia l interface impacts defect nucleation and thus strain relaxation. To determine if such an effect is occurring, this experiment varies the depth of the amorphous layer by varying the implant energy. The quantity of dislocations and degree of relaxation will be monitored to determine if a ny relationship is present. Experimental Design Structures for this experim ent were grown using Reduced Pressure Chemical Vapor Deposition (RPCVD) at Texas In struments, Inc. Pseodomor phically strained 50 nm Si1-xGex was deposited on Si (001) substrate with alloy compositions of 0, 16, 22, and 26 at% germanium. Post growth, the structures were Si+ ion implanted with varying en ergies of 5, 12, and 18 keV at a fluence of 1x1015 atoms/cm2 to generate continuous amorphous layers with varying proximity to the epitaxial interface while confining them within the stra ined layer. Once implanted, samples were annealed in a quartz-tube furnace under an inert N2 ambient environment. An isothermal study at 800C for 5, 30, and 300 minutes was carried out to study the effect of the proximity of implantation damage and its evolution to the epitaxial interface on strain relaxation. Ion implantation is a non-conservativ e process which introduces excess point defects in the system. At temperatures between 600 a nd 800 C, the excess point defects recombine and the remaining excess inters titials agglomerate and become 311 de fects. On further annealing, the 311 defects dissolve and contribute to the coarsening of disloca tion loops. These defects are termed End-Of-Range (EOR) ]. At temperatures above

PAGE 138

138 800C, dislocations loops in the EOR coarsen at the expense of smaller loops and 311s. Since it is likely that strain relieving misfit dislocations may be nucleated at or near the EOR damage created by ion implantation, the proximity of the EOR damage to the Si/SiGe interface may influence the number and type of defects formed a nd hence affect the degree of strain relaxation. Thus, this experiment varied both strain and implant energy in orde r to investigate the effect of the proximity of the EOR to the epitaxial inte rface on strain relaxation as a function of initial strain. Once annealed, the strain state a nd crystalline quality will be characterized. A Pananalytical MRD XPert was used to obtain XRD Rocking Curves (RC) and Reciprocal Space Maps (RSM) to obtain the degr ee of relaxation observed post anneal. A JEOL 2010F high-resolution transmi ssion electron microscope (HRTEM) and a JEOL 200CX TEM were used to study cross-sections (XTEM) of the regrown layer. Samples were prepared using a Dual-Beam FEI Strata DB 235 Focused Ion Beam. XTEM measurements were used to confirm the amorphous layer depth and layer thicknesses. Plan-view (PTEM) samples were also prepared and imaged using a JEOL 200CX operating at 200kV using g220 weak-beam dark-field (WBDF) imaging to quantify dislocation density. Results As-Implanted Sample Analysis Analysis of the as-grown sam ples can be found in the previous chapter. The experiment in this chapter varies the implant en ergy in order to change the proxi mity of the a-c interface to the epitaxial interface. All implants were tailored to be confined within the strained layer. Previous work suggests that the amorphization threshold d ecreases with increasing Ge concentration due to a decrease in binding energy [Hay92, Lie93, O Ra96]. The amorphous depths of the implants in this work were measured using XTEM a nd are presented in Figur e 6-1. XTEM of asimplanted SiGe and Si samples confirmed that the amorphous depths varied with higher Ge

PAGE 139

139 concentrations and higher implants energies, in agreement with prior findings. The varying initial amorphous depth complicates the experiment by altering the final in terstitial profile and thus will alter the quantity of exce ss interstitials. If this is the case, and if the defect nucleation and density is dependent on the number of interstitials present, the effect will be observed in the quantitative PTEM study. Proximity Effect Experiment Two m ajor variables were varied in this experiment: magnitude of strain and implant proximity. First, the result will be presented in terms of strain. All Ge compositions were implanted with 12 keV Si+ and annealed at 800 C for 30 mi nutes. The XRD (left column) and XTEM (right column) data for all concentrations of Ge is presen ted in Figure 6-2. XRD (004) RC of the as-grown and as-grown annealed ca ses are accompanied with the implanted and annealed case to confirm that any enhanced relaxa tion is due to the SPER process alone. The asgrown and annealed sample shows no change in peak position but has a slight decay in peak amplitude due to the formation of misfit dislocat ions as seen in PTEM (not shown). With increasing Ge, this effect is accompanied with a slight peak shift indicating a slight relaxation due to anneal alone. After implantation and anneal of the 16% Ge sa mple, thickness fringes are still present in the XRD RC spectra indicating that the epitax ial interface is still ac commodated elastically, presented in Figure 6-2 (a). As the Ge com position is increased, Figure 6-2 (c) and (e), an absence of these fringes is observed indicating that the interface has b een compromised. Also, with increasing strain, the defects increase in density and the degree of strain relaxation increases. The percent strain relaxation i ndicated above the XRD spectra is the average relaxation of the implanted and annealed case. Th e average degree of strain relaxation increases with increasing strain as shown by the peak broa dening and peak shift in the RC spectra. This

PAGE 140

140 increased relaxation and decreased crystalline qua lity is confirmed using XTEM images. For the 16% Ge sample, Figure 6-2 (b), the defects are ne ar the surface and do not propagate down to the interface. Thus, negligible strain relaxation wa s observed. In the case of the highest strained sample, 26% Ge, the defects propagate down towards the interface, as seen in Figure 6-2 (f) and is accompanied with a higher magnitude of strain relaxation. The quantity and depth of the regrowth related defects also increases with increasing Ge, as observed by comparing Figure 6-2 (b,d, and f). This is observed with the increasi ng relaxation and decreased crystalline quality in the XRD data as a function of Ge content. A similar comparison is made in terms of im plant proximity. The XRD and XTEM data of the 16% Ge samples implanted with 5, 12, and 18 keV Si+ and annealed for 30 minutes at 800 C are presented in Figure 6-3. For the 5 and 12 keV im planted samples, all of the peak oscillations are still present and indicate th at the interface is s till intact though these samples show an additional decay in amplitude due to defects presen t in the regrown layer. A slight peak shift is also observed and indicates strain relaxation less than 1%. The 18 keV implanted case, with the closest proximity to the interface, shows a larger peak shift indicating greater relaxation and an absence of the fringes indicating that the interface uniformity has been disrupted. In Figure 6-3 (b-d), XTEM corroborates the XRD results by show ing an increase in both depth and density of regrowth related defects with in creasing interface proximity (increasing implant energy). These results suggest that the degree of relaxation is largely dependent on proximity of the a-c interface to the epitaxial interface. Further examination of all samples using PTEM elucidates the defect microstructure and density. All samples in the experimental matrix were annealed at 800 C for 5, 30, and 300 minutes. The PTEM for the 30 minutes anneal ed samples are shown in Figure 6-4. PTEM

PAGE 141

141 images provide a two dimensional representation of the defects; defects that extend deeper into the substrate will appear to have longer segments. This is observed with increasing proximity (implant energy). The results observed in XTEM are further confirmed in these PTEM images. For the 16% Ge case, the quantity of defects appear s smaller in the lower implant energies due to the dependence of defect formation on the depth of the implant. The same can be said for the higher Ge samples, the defect segments appear to be longer with increasing implant energy, therefore, the defects extend deeper into the layer. One advantage of PTEM is the ability to obs erve the microstructure and to quantify the defects. A linear dislocation density is quantifie d perpendicular to the dire ction of the g vector. The defect density calculated in this manner is presented in Figure 6-5 and is accompanied with the relaxation data from the XR D RC. The strain relaxation and proximity both have a linear relationship to initial strain (% Ge) and show the effect of proxi mity on the relaxation process. The stability of these defects was examined with isothermal anneals. The defect density for the 18keV case, annealed at 800 C for 5, 30, and 300 mi nutes, is presented in Figure 6-5. This plot shows that neither the microstructure nor the de fect density changes w ith further annealing. These results indicate that the defects are quite stable at 800 C Finally, the average degree of relaxation for the proximity experiment is summari zed in Figure 6-7. The increase in relaxation is observed as the a-c interface is placed closer to the interface and as th e initial strain level (% Ge) is increased. In summary, the results show that the rela xation and subsequent defects are dependent on both initial strain magnitude and proximity of the implant to the epitaxial interface. The implications of these results will be discussed in the next section.

PAGE 142

142 Discussion Most of previous SPER experim ents have b een conducted using highly metastable SiGe layers of more than 200 nm thick and am orphizing implants passing through the Si/SiGe interface [Ang07, Hon92, Lee93]. The results from these experiments showed defect nucleation once the a-c interface passed th e Matthews-Blakeslee (M-B) criti cal thickness [Mat74]. The authors of these works attributed the defect nuc leation and subsequent st rain relaxation to be related to the theoretical critical thickness. Additionally, si milar work was done in thinner layers, around 30nm, with amorphous layers al so passing through the S i/SiGe interface [Chi89, Rod97]. The results from these experiments also s howed a relationship to th e theoretical critical thickness. Rodriguez et al. conducted an e xperiment with both doped and undoped samples and found that the doped samples regr ew past the M-B critical th ickness before defects were nucleated. The authors, therefor e, concluded that the critical thickness was altered due to a decrease in strain with dopant incorporation [Rod97]. Accordi ng to M-B criteria, all of the experimental structures in this work are grown in a metastable st ate with the strained SiGe layer thicker than the critical value. If a similar relationship to M-B critical thickness is observed upon regrowth, the strained samples in this study should regrow with no defects or loss of strain within the following distances from the Si/SiGe in terface: 16% 20nm, 22% 13 nm, and 26% 10 nm [Mat74]. Above these values, defects should nuc leate and extend to the surface of the wafer. The proximity experiment results, however, are contrary to the previous findings. The defect-free thickness (critical th ickness) for all samples in this study is plotted against M-B critical thickness in Figure 6-8. The M-B criterion is observed fo r samples with lower strain and implant energy combination. However, at a cr itical strain and implant energy combination, defects are observed below the th eoretical calculations. The di fference in relaxation mechanism is key to the understanding of th is discrepancy. M-B criterion is based on an energy balance in

PAGE 143

143 which the stability is described by relative energies of the two competing interfacial structures. The energy to generate a disloca tion and the elastic energy for stra in compensation is balanced to determine the critical thickness [Mat74, Van63]. The assumption is that the strain is accommodated by generation of 60 misfit dislocation. However, as seen in the results from the prior chapter, the relaxation is accompanied by defect nucleation due to the roughening of the a-c interface. Furthermore, these defects are regrow th-related. The nucleation of defect introduction is this work is not similar to the kink model and generation of misfits as described by M-B and thus cannot be used to determine the critic al thickness for SPER breakdown observed in these samples. As predicted by Srolovitz and discussed in Chapter 5, the a-c in terface increases in roughness until an equilibrium perturbation is reache d [Sro89]. Hairpin dislocations are formed when the competing growth fronts of the rough a-c interface meet during regrowth. Therefore, shallower implants like the 5 keV do not have time to reach this equilibrium before regrowth is complete, resulting in a lower wavelength than the deeper implanted samples. Since the equilibrium wavelength dictates when the defect s are nucleated, shallower implants will nucleate a lower number of defects. W ith the deeper 12 and 18 keV impl ants the layers are deep enough to reach their equilibrium wavelength before regr owth is complete. This explains why a higher number of defects are observed when deeper implants are performed. Also, by comparing the PTEM in Figure 6-3, it can be seen that the de fect density reaches satu ration with a deep enough implant. This corresponds to the equilibrium value of the roughness wavelength. Furthermore, the stability of the def ect density at longer anneal times, pres ented in Figure 6-5, indicates that once the defects are nucleated as the a-c pert urbation reaches equilibr ium and SPER commences, no further growth or evolution of the defects is observed. More importantly, there is no evidence

PAGE 144

144 of the regrowth defects gliding down to the in terface upon further annealing as observed in the tensile strain case and discussed in Chapters Thre e and Four. Instead, below the critical strain and proximity the defects are contained within the layer. These observations support the conclusion that the relaxation behavior is different when implants are contained within the layer than when they are performed though the layer. According to the critical thickne ss criteria, the proximity of the implant should not play a major rule in the relaxation or depth of defect nucleation. Regrowth should always be defect-free below the critical thickness. This difference is likely due to the localized strain increase due to implantation and the regrowth taking place within the strained layer itself. Specifically, the impact of the implant to the strain state must be considered. Placing the implant within the layer will impact the strain magnitude of the la yer [Gla97, Kri95b, Lie95]. The point defects generated by the implant add to the overall magnit ude of strain and place it in a higher state of compression. Thus, the proximity of the projected ra nge in relation to the ep itaxial interface will impact the amount of strain in the material. This, in turn, impacts the location of defect nucleation as observed. Another variable to consider with va rying implant energy is the net interstitial population contributing to de fect nucleation. Prior wo rk by Gutierrez [Gut01b] and King [Kin03] have shown that the net interstitials do no change as a function of energy at these low energies and the proximity of the damage to the surface does not have an effect on the trapped interstitial population. Si nce the interstitials in the st rained SiGe sample did not agglomerate as observed in Si, the trapped in terstitial population coul d not be determined. Another important observation is that the re growth-related hairpin defects are seen to extend deeper into the film than the original a-c interface. This has not been reported in prior work. As the anneals were done at a higher temperature, a higher thermal budget was seen by

PAGE 145

145 the samples after regrowth was complete. It is thus possible that another mechanism was being activated once regrowth was complete, namely di slocation climb. Climb generally occurs at temperatures greater than half of the materials me lting point [Hir68, Rea53] which is true for this case. Climb is a point defect diffusion process and occurs by diffusion of either vacancies or interstitials. It is assumed th at a large number of free interstitials exist within the SiGe layer since no agglomeration of point defects is observe d. These free intersti tials are thus readily available for mediating motion and elongation of the hairpin de fects below the original a-c interface. This reasoning could explain the disc repancy in the XRD and PTEM data presented in Figures 5-7 and 5-10, where the dislocation density remains constant as a function of anneal temperature but the XRD data shows further relaxation after annea ling above 700 C. To confirm this hypothesis, further experiments are needed that alter the point defect population while keeping the proximity constant. Conclusion The relaxation of the regrown layer is quite di fferent when the strain is in a state of com pression versus tension. For the compressive state, results have shown that defects were nucleated when the competing growth fronts of the rough a-c interface meet during SPER. The a-c interface roughness is shown to increase with increasing initial strain. The defect density also shows a similar trend. Ther efore, it is evident that ther e is a correlation between a-c interface roughness and defect density. These defects are also st able and reach saturation beyond a critical strain and implant depth. The defect nucleation is highly depe ndent on both strain and proximity of the implant. A dditionally, lower strain and sha llower implant depths cause the formation of a localized layer of defects that do not extend down to the epitaxial interface even after further annealing. While this region above the a-c interface regrows with poor crystalline quality and an abundance of hairpi n dislocations, these dislocations remain confined within the

PAGE 146

146 regown region and do not affect the strain at the Si/SiG e interface. The hairpin dislocations do, however, cause some localized strain relaxation as observed with XRD measurements. At higher strain (above 22% Ge) and implant energies (above 12 keV), the en tire SiGe layer is defective and further strain relaxation is observed. Overall, this is contrary to the tensile case where the regrowth defects always extend down to the epit axial interface after regr owth regardless of the initial a-c interface position. Th ese regrowth related defects we re observed to nucleate misfit dislocations and cause further relaxati on than annealing alone (no implants). It is therefore plausible to conc lude that the critical thickness is varied when the implant is contained within the layer versus one that is performed though the laye r. This is likely due to the localized strain increase due to implantation a nd the regrowth taking place within the strained layer itself. Additionally, when im plants are performed within the layer, the proximity of the a-c interface placement with re spect to the epitax ial interface plays an importa nt role in both defect nucleation and strain relaxa tion. Furthermore, the assu mptions upon which the MathewsBlakeslee critical thickness was cal culated cannot be applied to th e present case. The relaxation mechanism itself is quite different as discussed. In summary, prior works have extensivel y studied strain relaxation and SPER in supercritical SiGe films where implants were conducted through the Si/SiGe interface. However, the effect of implant proximity to the Si/SiGe interface and how it affects defect nucleation and relaxation for implants within the layer was not well understood. This work investigates this effect and has deemed it inappropriate to apply Matthews-Blakeslees criteria to these sample conditions. Matthew-Blakeslee s criteria are calculated under equilibrium conditions. However, once the implant is conduct ed within the strained layer, it is a nonequilibrium process and MatthewsBlakeslee criteria cannot be a pplied. Also, prior work has

PAGE 147

147 been deficient in determining the relationship betw een defect microstructure and strain relaxation upon SPER. This work has shown a correlation between defect density a nd strain relaxation. Additionally, the results in this study conclude that the a-c interface roughening is the dominant factor in the defect nucleati on and that the defects were a/ 2<110> type dislocations.

PAGE 148

148 10 15 20 25 30 35 40 45 50 -5051015202530 5 keV 12 keV 18 keVAmorphous-crystalline interface depth (nm)at.% Ge Figure 6-1. Amorphous depth measuremen ts using XTEM for all strained Si1-xGex on Si samples implanted with 5, 12, and 18 keV Si+ shown as a function of Ge concentrations.

PAGE 149

149 Figure 6-2. All 12keV Si+ implanted with varying Ge con centration (a) 16% (b) 22% (c) 26% XRD RC on the left with corre sponding XTEM on the right. 16 % Ge 50nm 22 % Ge 26 % Ge 10110210310410510610733.433.633.83434.234.434.634.835 As-grown As-grown 800oC 30 min 12keV 800oC 30 minIntensity (counts/sec)Omega (deg) 10110210310410510610733.433.633.83434.234.434.634.835 As-grown As-grown 800oC 30 min 12keV 800oC 30 minIntensity (counts/sec)Omega (deg) 10110210310410510610733.433.633.83434.234.434.634.835 As-grown As-grown 800oC 30 min 12keV 800oC 30 minIntensity (counts/sec)Omega (deg) 0.5% relaxation 7% relaxation 10% relaxation (a) (c) (e) (b) (d) (f)

PAGE 150

150 10 100 1000 10410510610733.5 34 34.5 35 As-grown As-grown 800oC 30 min 5keV Si 800oC 30 min 12keV 800oC 30 min 18keV 800oC 30 minIntensity (counts/sec)Omega (deg) Figure 6-3. (a) XRD (004) RC of 5, 12, and 18keV 16% Ge samples. PTEM of 16% Ge sample implanted with (b) 5keV (c) 12keV and (d) 18keV Si+ implants. All samples annealed for 30 minutes at 800 C. 18 keV 12 keV 50 nm 5 keV (c) (d) (b) (a)

PAGE 151

151 Figure 6-4. PTEM of 5, 12, and 18keV Si+ implanted samples annealed at 800 C for 30 minutes as a function of germanium concentration. 26% Ge 22% Ge 16% Ge 5 keV 12 keV 18 keV

PAGE 152

152 0 5 1041 1051.5 1052 1052.5 1050 20 40 60 80 100 1416182022242628 5 keV 12 keV 18 keVLinear Dislocation Density (#/cm) % Strain Relaxation% Ge Figure 6-5. Linear dislocation density (fille d markers) and %Strain Relaxation (unfilled markers) for 5, 12, and 18keV Si+ implanted samples annealed at 800 C for 30 minutes as a function of germanium concentration.

PAGE 153

153 0 5 1041 1051.5 1052 1052.5 1051 10 100 1000 16% Ge 22% Ge 26% GeLinear Dislocation Density (#/cm)Anneal time (min) Figure 6-6. Linear dislo cation density of 18keV Si+ implanted samples annealed at 800 C for all germanium concentration.

PAGE 154

154 0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00 40.00 45.00 50.00 as-grownas-grown anneal 5keV implant + anneal 12keV implant + anneal 18keV implant + annealA-C Interface Proximity% Strain Relaxation 26% Ge 22% Ge 16% Ge Figure 6-7. Summary of %Strain relaxation obtained from XRD RC data for all samples annealed at 800 C 30 minutes.

PAGE 155

155 0 10 20 30 40 50 1416182022242628 5 keV 12 keV 18 keV M-B [Mat74]Critical thickness (nm)% Ge Figure 6-8. Critical thickness, measured using XTEM, for all sa mples conditions annealed at 800 C for 30 minutes.

PAGE 156

156 CHAPTER 7 SUMMARY AND COMPARISION Overview The effect of an a morphizing implant containe d within a strained la yer has not yet been studied, nor has the crystalline quality or therma l stability of such an amorphized region. These effects may be important for future device st ructures, as arsenic a nd phosphorus, both selfamorphizing implants, are often used to create channel extensions in NMOS devices. Additionally, Solid Phase Epitaxi al Regrowth (SPER) is carri ed out under strain in stress memorization techniques. The purpose of this work is to study the eff ect of these processing conditions as a function of strai n, especially concerni ng the degree of relaxa tion, stability after amorphization and regrowth, crystalline quality of the regrown layer, and proximity of implant to the heterostructure interface. Si/SiGe heterostructures were used to study the effect of strain post amorphization and regrowth. All tensile strained layers in this study are in a metastable state because of the low temperature growth procedure. Additionally, all compressively strained layers are also metastable according to Matth ews-Blakeslee criteria. Summary Tensile Strain Case Tensile strained Si structur es were grown on relaxed Si1-xGex (Ge fractions of 0, 10, 20, and 30) virtual substrates via Molecular Beam Epitaxy (MBE). A 630 nm thick fully relaxed Si1xGex layer of corresponding composition was then grown on top of the buffer layer, followed by a 50 nm strained silicon capping layer [Gai00]. The strain of the structures was experimentally determined using lattice parameters, obtained fr om HRXRD. The results from this analysis

PAGE 157

157 show that the cappi ng layers grown on Si1-xGex with Ge fractions of 0, 10, 20 and 30 correspond to a Si layer strain of 0, 0.37, 0.74, and 1.1% strain, respectively. The structures were then implanted with Si+ at an energy of 5, 12, or 18 keV and a fluence of 1x1015 atoms/cm2 to generate continuous amorphous layers confined within the strained layer. Next, isothermal and isochronal anneals were perf ormed in a quartz-tube furnace under an inert N2 ambient environment. These experiments were carried out to determine specific factors that affect the relaxation process. The results for the tensile case study are disc ussed in Chapter Three and Four and are summarized here. Under biaxial tension there is a critical st rain for SPER breakdown that lies between 0.74 and 1.1% strain. T hus, strained silicon can be amorphized and regrown without strain relaxation for all Si1-xGex compositions up to Si0.8Ge0.2 (0.74% strain). For the 1.1% strained (Si0.7Ge0.3) sample, the solid phase regrowth of the amorphous layer breaks down and results in the formation of regrowth related de fects within the amorphous region. These defects then extend back down to the strained Si/ Si1-xGex interface upon completion of regrowth. This is the first observation of regrowth related defects extending below the amorphous/crystalline interface. Additionally, the EOR ev olution and population seems to be altered in the presence of tensile strain; the evolved disloc ation loops are fewer in number es pecially in the relaxed Si on Si0.7Ge0.3 sample. This indicates that, in conjunction with a certain level of initial strain, excess interstitials are providing a source for the generation of extrinsic stacking faults which then nucleate misfit dislocation. Thus implantation increases misfit di slocation density and promotes further relaxation than annealing alone. The possi bility of excess interstitials acting as a source for misfit dislocation nucl eation was further explored in the proximity experiment. In summary,

PAGE 158

158 these results indicate that a critical misfit st rain between 0.74% and 1.1% results in a breakdown of the SPER process and the formation of extended regrowth related defects. These defects nucleated as the a-c interface progresses towards the interface during SPER, forming regrowth related defects. Once fully regrown, these defects propagate down to the interface, promoting enhanced stra in relaxation of the layer than annealing alone. The regrowth related defects are primarily stacking faults ra ther than the traditionally observed hairpin dislocations. Since strain rela xation was only observed when these regrowth related defects were present, the results suggest that these def ects are contributing signi ficantly to enhanced relaxation of the strained layer. Furthermore, the relaxation process is a ther mally activated glide process with activation energy of 0.7 0.2 eV. This value is significan tly less than the 2.2 eV activation energy for dislocation glide observed in bulk Si. The differen ce is theorized to be due to the formation term in this work being negligible since the defects are formed as a result of SPER breakdown. Thus, the activation energy in this work is dominated by the migration term. Overall, the relaxation process is concluded to be a re sult of dislocation nuc leation and propagation via dislocation glide initiated by the SPER breakdown. Finally, the results in Chapter Four have show n that the implant damage proximity to the Si/SiGe interface impacts the de gree of relaxation and misfit dislocation density. After a constant anneal time and temperature, 800 C for 30 minutes, and varying implant energy (proximity) the misfit dislocation spacing is obs erved to decrease with increasing proximity. Therefore, exhibiting a higher density and observed relaxation as confirmed by HRXRD measurements. Further analysis using isotherm al anneals and PTEM showed that the misfit dislocations are observed to in crease in density for the two lower implant energy cases. The

PAGE 159

159 highest energy case, already observed complete re laxation at 30 minutes a nd exhibits constant misfit dislocation density with further annealing. The PTEM and HRXRD results are consistent and confirm data observed. In summary, this experiment has determined that the proximity of the initial a-c interface depends significantly on misfit di slocation density and strain rela xation. The closer proximity of the damage to the Si/SiGe interface, the hi gher the degree of relaxation and higher misfit dislocation density is observed. Additionally, th e results in the Chapter Three showed that the end-of-range damage population decr eased in conjunction with an incr ease in misfit dislocations. These results indicate that close proximity of the damage to the interface could influence both trapped interstitial concentration as well as the degree of relaxation. Therefore, it can be inferred that the epilayer interface may be acting as a sink for point defects when the implant damage is placed in close proximity. To confirm this clai m, future studies below 800 C are recommended to explore the end-of-range evolution and trappe d interstitial concentration as a function of damage proximity to the interface to fully unders tand the role of excess interstitials to the relaxation process. Compressive Strain Case Com pressive strain experimental structures were grown using Reduced Pressure Chemical Vapor Deposition (RPCVD). Pseodomo rphically strained 50 nm Si1-xGex was deposited on Si (001) substrate with alloy compositions of 0, 16, 22, and 26 % germanium. The structures were then implanted with Si+ at an energy of 5, 12, or 18 keV and a fluence of 1x1015 atoms/cm2 to generate continuous amorphous layers confined within the strained layer. Next, isothermal and isochronal anneals were performed in a quartz-tube furnace under an inert N2 ambient environment. These experiments were carried ou t to determine specific fa ctors that affect the relaxation process.

PAGE 160

160 The results for the compressive case study are discussed in Chapter Five and Six and are summarized here. First, the a-c interface roughens due to compressive strain and this roughness increases when the magnitude of strain is increa sed. The roughness is a stra in effect previously reported in both compressively strained Si [Bar04] and in metastable SiGe films [Ang07, Lee93]. It is also similar to the effect seen in compressively strained ep itaxial layers grown via deposition techniques [Sro89]. With increasi ng initial strain in the film, th ere is more strain energy to compensate by decreasing elastic energy. The balance of these energies yields a minimum wavelength of the roughness for the a-c interface. This roughness wa s not observed in the tensile case. Defects are nucleated during SPER by the compe tition of the different direction growth fronts. Using g b analysis, these defects were determin ed to be hairpin dislocations with Burgers vector of a/2 <110> which were first ob served by T. Sands [San84]. These defects were later observed in metastable SiGe SPER [Ang07] and in Si samples implanted with high dose Ge+ to form SiGe via SPER [Cri96]. The a-c inte rface in these other structures also exhibited roughness similar to these observations. In relation to the tensile case discussed in pr evious chapters, the critical strain for SPER breakdown is much lower in compression. And the mechanism by which defects are nucleated is altered. Additionally, no stacking faults were generated in the compressively strained SiGe films unlike the tensile Si counterpart. However, the dominant factor in the defect nucleation in the strained SiGe samples is the roughening of the a-c interface nucleating defects as the facets meet. Experimental results have also shown hairpi n dislocation density is stable and not a function of temperature over the ra nge studied. It does not correlate therefore, with the strain relaxation observed via XRD which increases as the anneal temperature is raised. The defects

PAGE 161

161 nucleate at the meeting of the facets of the r ough a-c interface and propa gate towards the surface with the advancing front. It is also observed th at these defects do not gl ide down to the epitaxial interface. The defective layer is contained between the depth at which defect nucleation occurred and the surface. At lower strain, there lies a defect-free layer which retained its initial strain with a highly defective la yer near the surface which is fully relaxed. The proximity experiment results discussed how this defect-free layer is influenced by both initial magnitude of strain and the proximity of the a-c interface to the epitaxial interface. The relaxation of the regrown layer is quite di fferent when the strain is in a state of compression versus tension. For the compressive state, results have shown that defects were nucleated when the competing growth fronts of the rough a-c interface meet during SPER. The a-c interface roughness is shown to increase with increasing initial strain. The defect density also shows a similar trend. Ther efore, it is evident that ther e is a correlation between a-c interface roughness and defect density. These defects are also st able and reach saturation beyond a critical strain and implant depth. The proximity experiment shows that the defect nucleation is highly dependent on both strain and proximity of the implant to th e epilayer interface. Additionally, lower strain and shal lower implant depths cause the formation of a localized layer of defects that do not extend down to the epitax ial interface even after further annealing. While this region above the a-c interface regrows with poor crystalline quality and an abundance of hairpin dislocations, these dislocations remain confined within the regown region and do not affect the strain at the Si/S iGe interface. The hairpin dislocations do, however, cause some localized strain relaxation as observed with XRD measurements. At higher strain (above 22% Ge) and implant energies (above 12 keV), the entire SiGe layer is defective and further strain relaxation is observed. Overall, this is contrary to the tensil e case where the regrowth defects

PAGE 162

162 always extend down to the epitaxia l interface after regrow th regardless of the initial a-c interface position. These regrowth related defects were obse rved to nucleate misfit dislocations and cause further relaxation than annealing alone (no implants). It is plausible to conclude that the critical thickness is varied when the implant is contained within the layer versus one that is performed though the layer. Th is is likely due to the localized strain increase due to implantation and the regrowth taking place within the strained layer itself. Additionally, when implants are pe rformed within the layer, the proximity of the a-c interface placement with respect to the epitaxial interface plays an important role in both defect nucleation and strain relaxation. Furthermore, the assump tions upon which the M-B critical thickness was calculated cannot be applied to the present case. The relaxation mechanism itself is quite different as discussed. In summary, prior works have extensivel y studied strain relaxation and SPER in supercritical SiGe films where implants were conducted through the Si/SiGe interface. However, the effect of implant proximity to the Si/SiGe interface and how it affects defect nucleation and relaxation for implants within the layer was not well understood. This work investigates this effect and has deemed it inappropriate to apply Matthews-Blakeslees criteria to these sample conditions. Matthew-Blakeslee s criteria are calculated under equilibrium conditions. However, once the implant is conduct ed within the strained layer, it is a nonequilibrium process and MatthewsBlakeslee criteria cannot be a pplied. Also, prior work has been deficient in determining the relationship betw een defect microstructure and strain relaxation upon SPER. This work has shown a correlation between defect density a nd strain relaxation. Additionally, the results in this study conclude that the a-c interface roughening is the dominant factor in the defect nucleati on and that the defects are a/2< 110> type dislocations.

PAGE 163

163 Comparison The relaxation of the regrown layer is quite di fferent when the strain is in a state of com pression versus tension. The main reasoning be hind the difference is in the manner in which the defects are nucleated. The advancing a-c inte rface is planar under tension and is observed to roughen under compression. This roughness was obs erved to increase with increasing initial strain value. Thus, the mechan ism of defect nucleation differed for each case. In tension, it was observed to be caused by the formation of regrow th related defects. Th ese defects then glide down to the Si/SiGe inte rface and promote further growth of misfit dislocations. In compression, the defects nucleated when differently oriented growth fronts meet. These defects are perfect dislocations and are termed ha irpins and have been observed previously. In both cases, the defects were observed to propagate below the origin al a-c interface; the first observations of such an event. In tension, the defect s are concluded to be mediated vi a dislocation glide mechanism. In compression, it is hypothesized to be dislocation climb. Further analysis is needed to confirm this claim. Future Work Dissolution rate of {311 } defects under tension. In this work, defect evolution of the EOR under tensile strain was observed to be altere d. Under tensile strain, {311}s are stable after 45 m inutes at 800 C while they are only stab le in unstrained Si for 5 minutes at this temperature. Thus, the dissolu tion rate of {311} disloc ation seems to have been affected by tensile strain. It is known that interstitials are more stable in tension than in compression. Antonell et al. showed that incr easing biaxial tensile strain significantly reduces the activation energy of interstitial formation [Ant90]. In orde r to study the true effect of tensile strain on {311} dissolution, utilization of a st rained structure without the close proximity of a sink, as exists in this work, is needed.

PAGE 164

164 Confirmation of climb mechanism in compression. The results from the samples in compression suggest that the hairpin dislocations are propagating toward s the epilayer interface via dislocation climb motion. Climb is mediated though diffusion of point defects. One method to confirm both the dislocation climb mechanism and the role of excess interstitials to the relaxation process under compression is to design an experiment such that the net interstitial population is altered yet the implant species and a-c interface depth is kept constant. This experiment can be carried out by the use of low temperature implantation. Lowering the implant temperature increases the amorphization effici ency thereby producing a deeper a-c interface depth with less energy compared to an equivalent implant carried out at room temperature. Designing the experiment in this manner will not increase complexity of the experiment by introducing additional variables wh ose effect cannot be determined. This was found to be the case with the B18H22 cluster implant experiment discussed in Appendix C.

PAGE 165

165 APPENDIX A STRESS-STRAIN CONVERSION The experim ental results in this work may be di scussed in terms of either stress or strain. The selection of strain as the metric of choice is preferred since it can be di rectly calculated from XRD results and using Braggs law to obtain lattice parameters of th e layer and substrate. In the event the reader would like to compare the results he rein to a work that presents results in terms of stress, the equivalent stress va lues in GPa for both the strained Si and strained SiGe structures are presented in Tables A-1 and A-2, respectively. Table A-1. Stress-strain conve rsions for strained Si on Si1-xGex. Strain Stress Si on Si0.9Ge0.1 0.0037 0.62 GPa Si on Si0.8Ge0.2 0.0074 1.25 GPa Si on Si0.7Ge0.3 0.0110 1.86 GPa Table A-2. Stress-strain c onversions for strained Si1-xGex on Si. Strain Stress Si0.84Ge0.16 on Si -0.0065 -1.10 GPa Si0.78Ge0.22 on Si -0.0090 -1.52 GPa Si0.74Ge0.26 on Si -0.0107 -1.81 GPa

PAGE 166

166 APPENDIX B AMORPHOUS-CRYSTALLINE INTERFACE RELAXATI ON Strained 50 nm Si1-xGex was deposited on Si (001) substr ate with alloy compositions 26% Ge and strained Si was deposited on Si1-xGex virtual substrate with al loy composition of 30% Ge. The as-grown structures were characterized usi ng TEM and XRD techniques in order to ensure proper growth of strained films. Post gr owth, the structures were implanted with Si+ with an energy of 12 keV at a fluence of 1x1015 atoms/cm2 to generate a 30 nm continuous amorphous layer confined within th e strained layer. Anneals were pe rformed in a quartz-tube furnace under an inert N2 ambient environment. The experiment consisted of an isothermal st udy at 500 C performed to investigate defect nucleation and propagation in the strained film Anneal times were 15, 30, and 45 minutes. With these times and temperature, the regrowth velocity is slow enough that the amorphouscrystalline (a-c) interface will be captured before complete regrowth has taken place. This allows observation of defects as they nucleate an d grow with the use of XTEM. The result from this experiment is discussed in Chapter five. Additionally, a preanneal of 400C for 60 minutes was conducted prior to the 500C ann eal to relax and planarize the a-c interface before initiation of SPER, this is also called a relaxation anneal. XTEM microgr aphs of the preanneal, shown in Figure B-1, planarized the a-c in terface prior to regrowth. Sim ilar low temperature preanneals, also have been conducted to planarize the a-c interface in prior work [Kin03, Ban00]. The roughness of the a-c interface post re growth is similar to samples annealed without the relaxation anneal. These results conclude that the relaxation anneal prior to SPER does not change the a-c interface morphology.

PAGE 167

167 Figure B-1. XTEM of samples implanted with B18H22 cluster (a) 26% Ge annealed for 400 C 60 minutes (b) 26% Ge annealed for 400 C 60 minutes followed by 500 C 30 minutes (c) 30% Ge annealed for 400 C 60 minutes (d) 30% Ge annealed for 400 C 60 minutes followed by 500 C 30 minutes. Arrows indicate a-c interface and hatched white lines indicate epitaxial interface. (a) (c) (d) (b) 50 nm

PAGE 168

168 APPENDIX C BORON CLUSTER EXPERIMENT: EFFECT OF EOR POPULATION P-type devices are traditiona lly im planted first with a Pre-Amorphizing Implant (PAI) prior to boron implantation to avoid channeli ng of the smaller boron atom through the silicon lattice thus achieving shallowe r junctions. Octadecaborane (B18H22) cluster ion enables the elimination of the PAI step due to its ability to amorphize the implanted region while achieving shallower junctions [Heo06, Kru06, Sek07]. The goal of using this cluster ion is to reduce the End of Range (EOR) damage from the PAI wh ich causes leakage currents and enhanced diffusion. The octadecaborane was implanted into the hi ghest tensile (Si) and highest compression (SiGe) samples. The implant was carried out at a fluence of 1 x 10 15/cm2 at an equivalent energy of 4 keV. This implant generated an amorphous layer ~ 15 nm in depth, XTEM shown in Figure C-1, the same depth as the 5 keV conventional Si+ implant as discussed in the experimental chapters of this work. The as -implanted amorphous-crystalline (a-c) interface is not as planar as a conventional im plant. It is observed to be quite non-uniform as shown in a high resolution XTEM image, Figure C-2. R ough initial a-c interf ace has been shown by previous work to nucleate defect s upon regrowth [San84]. Since the strain samples in this work already nucleate regrowth related defects, it is difficult to sepa rate out the rough a-c interface versus the strain contribution. PTEM images, taken under g220 WBDF conditions, are presented in Figure C-3 for the highest tensile and highest compressive case implanted with either 5 keV Si+ or B18H22 cluster implants and annealed at 800 C for 30 minutes The misfit disloca tion spacing (or density) difference between the implant conditions is not significant enough to determine whether the boron cluster case observed a reducti on. This is true for both stra in cases. Analysis using XRD

PAGE 169

169 is used to further study this difference. The degree of relaxation obtained from lattice parameters measured via XRD is shown in Figure C-4. In compression, according to XRD results, no difference is degree of relaxation is observed. However, in tension, the conventional implant undergoes a higher degree of relaxation than th e boron cluster case. Th is difference is not noticeable in the PTEM images, the difference be tween the two cases should exhibit an average misfit spacing difference of ~30 nm. Furthermore, from results presented in Chapter Four, the effect of excess interstitials to the degree of relaxation is enhanced with closer proximity to the Si/SiGe layer. To fully understand and determine a difference, the B cluster implant must be placed closer to the epilayer interface at a proximity where further relaxation was observed. In summary, the highest tens ile and highest compressive sa mples were implanted with either conventional Si+ or B+ cluster implant to yield amorphous layers of the same depth. The B cluster implant was used to reduce the EOR, t hus reducing the amount of excess interstitials available to contribute to relaxa tion. The results show inconclusi ve data, the difference in degree of relaxation and misfit density is not significant enough to dete rmine an affect. The implant needed to be generated closer to the inte rface where the relaxation and misfit density was enhanced (refer to proximity results in chap ters Four and Six). To fully understand the dependence of interstitial popula tion to the relaxation process, an ideal experiment would involve same species implantation with identi cal a-c interface depth while changing the net interstitial population. Add itionally, implanting boron into the strained samples adds an additional variable to the experiment, whose influence is difficult to separate out.

PAGE 170

170 Figure C-1. XTEM of as-implanted with B18H22 cluster at equivalent energy of 4 keV into Si (100) substrate. 20 nm Surface a cinterface

PAGE 171

171 Figure C-2. High-resolution XTEM of as-implanted B18H22 at equivalent energy of 4 keV into Si (100) substrate showi ng a rough a-c interface. 5 nm

PAGE 172

172 Figure C-3. PTEM images of (a) strained Si0.74Ge0.26 on Si implanted with 5 keV Si+ and (b) implanted with B18H22 and of (c) strained Si on Si0.7Ge0.3 implanted with 5 keV Si+ and (d) implanted with B18H22. All samples annealed at 800 C for 30 minutes. 100 nm 100 nm (a) (b) (c) (d)

PAGE 173

173 0 10 20 30 40 50 Strained Si0.74Ge0.26 on Si Strained Si on Si0.7Ge0.3 % Strain Relaxation5 keV Si 4 keV B18H22 Implant Condition Figure C-4. Strain relaxation qua ntified using XRD data for highest tensile and compressive strain samples implanted with 5 keV Si+ and 4keV equivalent B18H22 annealed for 30 minutes at 800 C.

PAGE 174

174 LIST OF REFERENCES [Alb95] M. Albrecht, S. Christiansen, H.P. Struck, Phy. Status Solidi A, 150, 453 (1995) [Ale68] H. Alexander and P. H aasen, Solid State Phys. 22, 27 (1968) [Ang07] D. D'Angelo, A. M. Pir o, A. Terrasi, M. G. Gri maldi, S. Mirabella, and C. Bongiorno, J. Appl. Phys. 101, 103508 (2007) [Ant90] A. Antonelli and J. Bernholc. Mater. Res. Soc. Symp. Proc. 163, 523 (1990) [Ant96] M. J. Antonell, K. S. Jones, and T. E. Haynes, J. Appl. Phys. 79, 7646 (1996) [Ban00] H. Banisaukas. PhD. Thes is. University of Florida (2000) [Bar83] W. J. Bartels, J. Vac. Sci. Tech. B1, 338 (1983) [Bar04] W. Barvosa-Carter, M. J. Aziz, A.-V. Phan, T. Kaplan, and L. J. Gray, J. Appl. Phys. 96, 5462 (2004) [Bea87] J. C. Bean, L. C. Feldman, A. T. Fi ory, S. Nakahara, I. K. Robinson, J. Vac. Sci. Technol. A, 2, 436 (1987) [Bir03] M. Birkholz, Thin Film Analysis by X-Ray Scattering, W iley-Vch, Germany, 2003, pp. 297-311 [Bon98] C. Bonafos, D. Mathiot, and A. Claverie. J. Appl. Phys. 83, 3008 (1998) [Bul79] C. Bull, P. Ashburn, G. R. Booker, K. H. Nickholas, Sol. State Electron 22, 95 (1979) [Cap92] M. A. Capano, Phys. Rev. B, 45, 11768 (1992) [Cha05] V. Chan, K. Rim, M. Ieong, S. Yang; R. Malik, Y. W. Teh, M. Yang, Qiqing, Custom Integrated Circuits Conference, 2005. Proceedings of the IEEE, pp. 667, 18-21 Sept. 2005 [Che74] D. Cherns, PhD dissertati on, University of Cambridge (1974) [Che04] C.-H. Chen, T. L. Lee, T. H. Hou, C. L. Chen, C. C. Chen, J. W. Hsu, K. L. Cheng, Y. H. Chiu, H.J. Tao, Y. Jin, C. H Diaz, S. C. Chen, M.-S. Liang, VLSI Technology, 2004. Digest of Technical Papers, pg. 56, 15-17 June 2004 [Chi89] B. T. Chilton, F. J. Robinson, and D. A. Thompson, Appl. Phys. Lett., 54(1) 42 (1989) [Chr81] L. A. Christel, J. F. Gibbons, and T. W. Sigmon, J. Appl. Phys., 52(11), 7143 (1981) [Cof00] S. Coffa, S. Libertino, and C. Spinella, Appl. Phys. Lett., 76(3), 321 (2000) [Cor96] F. Corni, S. Frabboni, R. Tonini, and G. Ottaviani, J. App;. Phys., 79(7) 3528 (1996)

PAGE 175

175 [Cri96] F. Cristiano, A. Nejim, B. de Mauduit, A. Claverie and P. L. F. Hemment, Nuclear Instruments and Methods in Physics Research Se ction B: Beam Interactions with Materials and AtomsVolume 120, Issues 1-4, Proceedings of the E-MRS '96 Spring Meeting Symp. I on New Trends in Ion Beam Processi ng of Materials, 2 December 1996, pg. 156 [Cri00] F. Cristiano, J. Grisolia, B. Colombeau, M. Omri, B. de Mauduit, A. Claverie, L. F. Giles, and N. E. B. Cowern. J. Appl. Phys. 87, 84 (2000) [Cro04] R. T. Crosby, K. S. Jones, M. E. Law, A. F. Saavedra, J. L. Hansen, A. N. Larsen and J. Liu, Mat. Res. Soc. Symp. Proc. 810 (2004) [Cro05] R. T. Crosby, PhD. Thesis University of Florida (2005) [Cse76] L. Csepregi, J. W. Mayer and T. W. Sigmon. Appl. Phys. Lett., 29(2), 92 (1976) [Cse78] L. Csepregi, E.F. Kennedy, J.W. Ma yer and T.W. Sigmon. J. Appl. Phys., 49(7), 3906 (1978) [Cow94] N. E. B. Cowern, G. F. A. van de Galle, P. C. Zalm and D. W. E. Vabdenhoudt. Appl. Phys. Lett. 65, 2981 (1994) [Dea73] G. Dearnaley, J. H. Freeman, R. S. Nelson and J. Stephen. Ion Implantation (1973) [Dis64] J. P. Dismukes, L. Ekstrom, and R. J. Paff, J. Phys Chem., 68(10), 3021 (1964) [Dod87] B. W. Dodson and J. Y. Ts ao, Appl. Phys. Lett., 51, 1325 (1987) [Dod88] B. W. Dodson and J. Y. Tsao, Phy. Rev. B. 38, 12383 (1988) [Dro82] R. Drosd and J. Was hburn, J. Appl. Phys. 53, 397 (1982) [Eag94] D. J. Eaglesham, P. A. Stolk, H.-J Gossmann, and J. M. Poate, Appl. Phys. Lett., 65(18), 2305 (1994) [Edi75] J. W. Edington. Practical Electron Mi croscopy in Materials Science. Ed. 4, Macmillian, London, 1975, pg. 6 [Ell95] R. G. Elliman P. Kringhoj, and J. M. Glasko, Nuclear Inst. And Methods B, 96, 276 (1995) [Ell96] R. G. Elliman, and W. C. W ong, Appl. Phys. Lett. 69 (18) 2677 (1996) [Fah89] P. M. Fahey, P. B. Griffin, and J. D. Plummer, Rev. Mod. Phys. 61, 289 (1989) [Far86] B. Y. Farver, Y. L. Iunin, and V. I. Nititenko, Phys. Status Solidi, 97, 469 (1986) [Few82] P. F. Fewster, J. Appl. Cryst. 15, 275 (1982)

PAGE 176

176 [Fit91] E. A. Fitzgerald, Y.-H. Xie, M. L. Green, D. Brasen, A. R. Kortan, J. Michel, Y.-J. Mii, and B. E. Weir, Appl. Phys. Lett. 59, 811 (1991) [Fri87] I. J. Fritz, Appl. Phys. Lett., 51(14), 1080 (1987) [Fre87] L. B. Freund: J. Appl. Mech. 54, 553 (1987) [Gai00] P. I. Gaiduk, A. Nylandsted Larsen a nd J. Lundsgaard Hansen, Thin Solid Films, 367, 120 (2000) [Gla97] J. M. Glasko, and R. G. Elliman, J. D. Fitz Gerald and P. Kringhoj, Nuclear Inst. And Methods B, 127/128, 397 (1997) [Gil99] L. F. Giles, M. Omri, B. de Mauduit, A. Claverie, D. Skarlatos, and D. Tsoukalas, Nuclear Inst. and Meth. B, 148, 273 (1999) [Gus01] E. Gusev, D. Buchanan, E. Cartier, A. Kumar, D. DiMaria, S. Guha, A. Callegari, S. Zafar, P. Jamison, D. Neumayer, M. Copel, M. Gribelyuk, H. Okorn-Schmidt, C. D'Emic, P. Kozlowski, K. Chan, N. Bojarczuk, L. -A. Rannarsson, P. Ronsheim, K. Rim, R. Fleming, A. Mocuta, and A. Ajmera, Elect ron Devices Meeting, 2001. IEDM Technical Digest, pg. 20 (2001) [Gut98] M. Yu. Gutkin, K. N. Mikaelyan, and I. A. Ovidko, Phys. Solid State, 40, 1864 (1998) [Gut01a] M. Yu. Gutkin, K. N. Mikaelyan, and I. A. Ovidko, Phys. Solid State, 43, 42 (2001) [Gut01b] A. F. Gutierrez.. M.S. Th esis. University of Florida (2001) [Hag78] W. Hagen, H. Strunk, Appl. Phys. Lett. 17, 85 (1978) [Hay92] T. E. Haynes and O. W. Holland. Appl. Phys. Lett. 61, 61 (1992) [Heo06] S. Heo, D. Lee, H. T. Cho, W. A. Krull, and H. Hwang, Ion Implantation Technology 2006, Marseille, France, 11-16, June 2006 [Hir68] J. P. Hirth and J. Lothe, Theory of Dislocations, McGraw-Hill, New York, 1968 [Hof74] W. K. Hofker, H. W. Werner, D. P. Oostoek, and N. J. Koeman. Appl. Phys. Lett. 4, 25 (1974) [Hon92] Q. Z. Hong, J. G. Zhu, and J. W. Mayer, J. Appl. Phys. 71, 1768 (1992) [Hor05] M. Horstmann, A. Wei, T. Kammler, J. Hntschel, H. Bierstedt, T. Feudel, K. Frohberg, M. Gerhardt, A. Hellmich, K. Hempel, J. Hoha ge, P. Javorka, J. Klai s, G. Koerner, M. Lenski, A. Neu, R. Otterbach, P. Press, C. Reichel, M. Trentsch, B. Trui, H. Salz, M. Schaller, H. J. Engelmann, O. Herzog, H. Ruel ke, P. Hubler, R. Stephan, D. Greenlaw, M. Raab, N. Kepler, Electron Devices Meeti ng, 2005. IEDM Technical Digest, pg. 233, 5-7 Dec. 2005

PAGE 177

177 [Hou91] D.C. Houghton, J. Appl. Phys. 70, 2136 (1991) [Hul89] R. Hull, J. C. Bean, D. J. Werder and R. E. Leibenguth, Phy. Rev. B. 40, 1681 (1989) [Hul93] R. Hull, J. C. Bean, L. J. Peticolas, D. Bahnck, B. E. Weir, and L. C. Feldman, Appl. Phys. Lett. 73, 7407 (1993) [Hwa91] D. M. Hwang, S. A. Schwarz, T. S. Ravi, R. Bhat, and C.Y. Chen, Phy. Rev. Lett. 66, 739 (1991) [Ima83] M. Imai, K. Sumino, Philos. Mag A, 47, 599 (1983) [Jai90] S. C. Jain, J.R. Willis, R. Bullough, Adv. in Phys. 39(2), 127 (1990) [Jon88] K. S. Jones, S. Prussin and E.R. Weber. Appl. Phys. A. (USA), 34, 1 (1988) [Jon97] K. S. Jones, K. Moller, J. Chen, a nd M. Puga-Lambers, J. Appl. Phys., 81 (9), 6051 (1997) [Kas75] E. Kasper and H.J. Herz og, Thin Solid Films, 44, 357 (1975) [Kin03] A.C. King. M.S. Thesis. University of Florida (2003). [Koe01] S.J. Koester, K. Rim, J.O. Chu, P. M. Mooney, and J.A. Ott, Appl. Phys Lett. 79, 2148 (2001) [Kri95a] P. Kringhoj, R. G. Elliman, M. Fyhn, S. Y. Shiryaev and A. Nylandsted Larsen, Nuclear Inst. And Methods B, 106, 346 (1995) [Kru06] W. A. Krull, B. Haslam, T. Ho rsky, K. Verheyden, and K. Funk, Ion Implantation Technology 2006, Marseille, France, June 2006 [Kuk83] Y. Kuk, L. C. Feldman, and P. J. Silverman, Phys. Rev. Lett. 50, 511 (1983) [Lam99] E. Lampin, V. Senez, and A. Claverie. J. Appl. Phys. 85, 8137 (1999) [Lee93] C. Lee, T. E. Haynes, and K. S. Jones, Appl. Phys. Lett. 62(5), 501 (1993) [LeG93] F. K. LeGoues, P.M. Mooney, J. Tersoff, Phys. Rev. B, 40, 1681 (1993) [Lei01] C. W. Leitz, M. T. Currie, A. Y. Kim, J. Lai, E. Robbins, E. A. Fitzgerald, and M. T. Bulsara, J. Appl. Phys. 90, 2730 (2001) [Lie93] D. Y. C. Lie, A. Vantomme, F. Eisen, T. Vreeland, Jr., M.-A. Nicolet, T.K. Carns, V. Arbet-Engels, and K. L. Lang. J. Appl. Lett. 74, 6039 (1993) [Lie95] D. Y. C. Lie, J. H. Song, A. Vantomme, F. Eisen, M.-A. Nicolet, N. D. Theodore, T. K. Carns, and K. L. Wang, J. Appl. Phys. 77, 2329 (1995)

PAGE 178

178 [Lin03] C. D. Lindfors. PhD. Th esis. University of Florida (2003) [Liu95] J. Liu, M.E. Law, and K.S. Jone s, Solid-State Electronics, 38(7), 1305 (1995) [Mar87] P. M. J. Mare, J.C. Barbour, J. F. van der Veen, K. L. Kavanagh, C. W. T. BulleLieuwma, and M. P. A. Viegers, J. Appl. Phys. 62, 4413 (1987) [Mass90] T. B. Massalski, Binary Alloy Phase Diagrams, ASM International, Materials Park, OH (1990) [Mat74] J. W. Matthrews and A. E. Bl akeslee, J. Cryst. Growth, 29, 273 (1974) [Mat91] M. Matsuura, J. M. Tonnerre, and G. S. Cargill III, Phys. Rev. B, 44, 3842 (1991) [Mau94] B. de Mauduit, L. Laanab, C. Berga ud, M.M. Faye, Nuclear Inst. And Methods B, 84, 955 (1994) [Moo65] G. E. Moore, Electronics 38, 114-117 (1965) [Moo94] P. M. Mooney, F.K. LeGoues, J. Tersoff, J. Appl. Phys. 75, 3968 (1994) [Mil05] A. Mills, III-Vs Review 18 (7), 28 (2005) [Nar82] J. Narayan, J. Appl. Phys. 53, 8607 (1982) [Nay96] D. K. Nayak, K. Goto, A. Yutani, J. Murota, and Y. Shiraki, IEEE Trans. Electron Devices 43, 1709 (1996) [Nod00] T. Noda, S. Odanaka, and H. Umimoto. J. Appl. Phys. 88, 4980 (2000) [ORa96] C. ORifeartaigh, R. C. Barklie, A. Nylandsted Larsen, F. Priolo, G. Franzo, G. Lulli, M. Bianconi, J. K. N. Lindner, F. Christiano, and P. L. F. Hemment. Nucl. Instr. Meth. Phys. Res. B. 120, 165 (1996) [Ols85] G. L. Olson, Mat. Res. Soc. Proc., 35, 25 (1985) [Ols88] G. L. Olson and J. A. Roth, Mat. Sci. Reports, 3, 1 (1988) [Pat66] J. R. Patel and A. R. Chaudhuri, Phys. Rev, 143, 601 (1966) [Pan96] G. Z. Pan, K. N. Tu, and S. Prussin, Appl. Phys. Lett., 68(12), 1654 (1996) [Peo85] R. People and J. C. B ean, Appl. Phys. Lett. 47, 322 (1985) [Pai91] D. C. Paine, N. D. Evans, and N. G. Stoffel, J. Appl. Phys, 70, 4278 (1991) [Phe04] M. S. Phen, The Effects of Stress on Dopant Activation and Defect Morphoogy, Semiconductor Research Corporation Annual Review, Raleigh, NC, 2004

PAGE 179

179 [Phe05] M. S. Phen, R. T. Crosby, K. S. Jones, M. E. Law, J. L. Hansen, A. N. Larsen, Mat. Res. Soc. Symp. Proc. 864 (2005) [Phe06] M. S. Phen, V. Craciun, K. S. Jones, J. L. Hansen and A. N. Larsen, Nuclear Inst. And Methods B, 253, 22 (2006) [Pie04] U. Pietsch, V. Holy, and T. Baumbac h, High-Resolution X-Ray S cattering: From Thin Films to Lateral Nanostructures, Sp ringer-Verlag, New York, 2nd ed., 2004, p.205 [Poa84] J. M. Poate and J.S. Williams, Ion Im plantation and Beam Processing, Academic Press, New York, 1984, p. 27 [Raa99] I. J. Raaijmakers, H. Sprey, A. Storm, T. Bergman, J. Italiano, and D. Meyer, J. Vac. Sci. Technol. B 17, 2311 (1999) [Rab00] J. Rabier and J. L. De menet, Phys. Stat. Sol., 222, 63 (2000) [Rea53] W.T. Read, Dislocation in Crystals, McGraw-Hill, New York, 1953 [Rod97] A. Rodriguez, T. Rodriguez, A. Kling, J. C. Soares, M. F. da Silva, and C. Ballesteros, J. Appl. Phys. 82, 2887 (1997) [Sam99] S. B. Samavedam, W. J. Taylor, J. M. Grant, J. A. Smith, P. J. Tobin, A. Dip, and R. Liu, J. Vac. Sci. Tech. B 17, 1424 (1999) [San84] T. Sands, J. Washburn, R. Gronsky, W. Maszara, D. K. Sadana and G. A. Rozgonyi, Proc. 13th Int. Conf. on Defects in Semi conductors, The Metallurg. Sot. of AIME, Warrendale, PA, 1984, p.531 [Sek07] K. Sekar, W. A. Krull, T. N. Ho rsky, and D. C. Jacobson, Ion Implantation Technology 2007 [Sin05] D. V. Singh, J. W. Sleight, J. M. Hergen rother, Z. Ren, K. A. Jenkins, O. Dokumaci, L. Black, J. B. Chang, H. Nakayama, D. Chidamba rrao, R. Venigalla, J. Pan, W. Natzle, B. L. Tessier, A. Nomura, J. A. Ott, M. Ieong, W. Haensch, Electron Devices Meeting, 2005. IEDM Technical Digest, pg. 505, 5-7 Dec. 2005 [Sin06] D. V. Singh, K. A. Jenki ns, J. Sleight, Z. Ren, M. Ieo ng, W. Haensch, Electron Device Letters, IEEE vol.27, no.3, pp. 191, March 2006 [Sro89] D. J. Srolovitz Acta Metl. 38, 621 (1989) [Ste07] M. Steinert, J. Acker, S. Oswald, K. Wetzig, J. Phys. Chem. C. 111, 2133 (2007) [Sto95] P. A. Stolk, H. J. Gossmann, D. J. Eaglesham, D. C. Jacobson, J. M. Poate, and H. S. Luftman. Appl. Phys. Lett. 66, 568 (1995) [Sug01] N. Sugii, J. Appl. Phys. 89, 6459 (2001)

PAGE 180

180 [Sug04] N. Sugii, S. Irieda J. Morioka, and T. Inada, J. Appl. Phys. 96, 261 (2004) [Tam96] M. Tamura, Appl. Phys. A, A63 (2), 359 (1996) [Tho04] S. E. Thompson, M. Armstrong, C. Auth, S. Cea, R. Chau, G. Glass, T. Hoffman, J. Klaus, M. Zhiyong, B. Mcintyre, A. Murthy, B. Obradovic, L. Shifren, S. Sivakumar, S. Tyagi, T. Ghani, K. Mistry, M. Bohr, Y. El-Mansy, Electron Devi ce Letters, IEEE vol.25, no.4, pg. 191, (2004) [Tho06] S. E. Thompson, G. Sun, Y. S. C hoi, and T. Nishida, Electron Devices, IEEE Transactions, vol.53, no.5, pp. 1010 (2006) [Tsa87] J.Y. Tsao, B.W. Dodson, S.T. Pi creaux, D.M. Cornelison, Phy. Rev. Lett. 59, 2455 (1987) [Tup90] C. G. Tuppen, C. J. Gibbings, M. Hock ly, S. G. Roberts, Appl. Phys. Lett. 56, 54 (1990) [Uem99] M. Uematsu. Jpn. J. Appl. Phys. 38, L1213 (1999) [Van63] Van der Merwe, J. Appl. Phys. 34(1), 117 (1963) [Van05] W. Vandervorst, T. Ja nssens, B. Brijs, R. Delhougne, R. Loo, M. Caymax, B.J. Pawlak, M. Posselt. Appl. Phys. Lett. 86, 081915 (2005) [Weg90] W. Wegscheider, K. Eberl, U. Men czigar, G. Abstreiter, Appl. Phys. Lett. 57, 875 (1990) [Wil96] D. B. Williams and C. B. Carter, Transmission Electron Microscopy, Plenum Press, New York, NY, 1996, pp. 423 [Yon96] I. Yonenaga, K. Sumino, Appl. Phys. Lett. 69, 1264 (1996) [Yon99] I. Yonenaga, Phys. St atus Solidi A. 171, 41 (1999)

PAGE 181

181 BIOGRAPHICAL SKETCH Michelle S. Phen is the daughter of Shin thong Phen and Ramona Phen. Michelle w as raised in Miami, Florida. She attended Barry Un iversity in Miami, Florida before transferring to the University in Florida in 2000. In May 2003, she was awarded a B.S. in Materials Science and Engineering Cum Laude with a specialty in Polymers. Upon graduation, she enrolled in graduate school to pursue a PhD. During her gra duate studies, she interned at Texas Instruments, Inc (Dallas, Texas) during two se parate fall semesters. Her first internship was with the High Performance Analog (HPA) package development group where she worked on failure analysis and development of new package technology. He r second internship was with the Silicon Technology Development group where she worked as a process development engineer on the SiGe and SALICIDE process line. Upon r eceiving her doctoral degree, she will begin employment at Intel Corporation.