<%BANNER%>

Influence of Germanium Concentration and Homogeneous Boron Doping on Microstructure, Kinetics, and Sheet Resistance of N...

Permanent Link: http://ufdc.ufl.edu/UFE0022417/00001

Material Information

Title: Influence of Germanium Concentration and Homogeneous Boron Doping on Microstructure, Kinetics, and Sheet Resistance of Nickel Germanosilicide Thin Films
Physical Description: 1 online resource (205 p.)
Language: english
Creator: Moore, John
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: agglomeration, germanosilicide, kinetics, microstructure, nickel
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: This work studied the influence of Ge concentration and homogeneous B doping on the microstructure, kinetics, and sheet resistance of nickel germanosilicide thin films. Experimental structures consisted of 150 nm of single crystal relaxed SiGe (at% Ge = 15 or 25) layers with and without in-situ homogeneous B doping of 4.5E19 atoms/cm^2 and capped by 10 nm of sputtered Ni metal. Samples were furnace annealed under a nitrogen atmosphere at temperatures between 450 and 800 degrees Celsius in 50 degree increments for times of 10, 30, 90, 270, and 1020 minutes. Microstructure analysis showed that increasing Ge content increased the amount of film transformation while the addition of B doping had no effect on morphology. Distinct reactions were observed before and after full consumption of the parent Ni(SiGe) film during kinetic analysis of isothermal transformation curves. The first reaction, in which Si-rich Ni(SiGe) and Ge-rich SiGe were found to precipitate from the Ni(SiGe) film, was determined to have concentration dependant activation energies of 1.96 and 0.76 eV for the 15 and 25% Ge samples, respectively. In this stage, the sheet resistance was found to linearly increase with increasing area fraction of Ge-rich SiGe for all samples. The cause for the increase was determined to be related to the conductive path tortuosity. Increasing Ge content did not affect the structure/property relationship, but the addition of B caused a decrease in sheet resistance. The second reaction, in which Si-rich Ni(SiGe) grains were found to agglomerate after the initial Ni(SiGe) layer was fully consumed, was determined to have an activation energy of 0.125 eV for both the 15 and 25% Ge samples. The undoped samples in this stage were uniformly found to have very high sheet resistance values. This result was attributed to the lack of a conduction path between isolated Ni(SiGe) grains. A strong linear relationship, however, was determined for the doped samples, and increasing Ge content had a small effect on the relationship for these samples. It was determined that the stabilizing influence of B doping was caused by the availability of a conduction path through the unreacted doped SiGe layer.
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.
Statement of Responsibility: by John Moore.
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 2009-02-28

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022417:00001

Permanent Link: http://ufdc.ufl.edu/UFE0022417/00001

Material Information

Title: Influence of Germanium Concentration and Homogeneous Boron Doping on Microstructure, Kinetics, and Sheet Resistance of Nickel Germanosilicide Thin Films
Physical Description: 1 online resource (205 p.)
Language: english
Creator: Moore, John
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: agglomeration, germanosilicide, kinetics, microstructure, nickel
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: This work studied the influence of Ge concentration and homogeneous B doping on the microstructure, kinetics, and sheet resistance of nickel germanosilicide thin films. Experimental structures consisted of 150 nm of single crystal relaxed SiGe (at% Ge = 15 or 25) layers with and without in-situ homogeneous B doping of 4.5E19 atoms/cm^2 and capped by 10 nm of sputtered Ni metal. Samples were furnace annealed under a nitrogen atmosphere at temperatures between 450 and 800 degrees Celsius in 50 degree increments for times of 10, 30, 90, 270, and 1020 minutes. Microstructure analysis showed that increasing Ge content increased the amount of film transformation while the addition of B doping had no effect on morphology. Distinct reactions were observed before and after full consumption of the parent Ni(SiGe) film during kinetic analysis of isothermal transformation curves. The first reaction, in which Si-rich Ni(SiGe) and Ge-rich SiGe were found to precipitate from the Ni(SiGe) film, was determined to have concentration dependant activation energies of 1.96 and 0.76 eV for the 15 and 25% Ge samples, respectively. In this stage, the sheet resistance was found to linearly increase with increasing area fraction of Ge-rich SiGe for all samples. The cause for the increase was determined to be related to the conductive path tortuosity. Increasing Ge content did not affect the structure/property relationship, but the addition of B caused a decrease in sheet resistance. The second reaction, in which Si-rich Ni(SiGe) grains were found to agglomerate after the initial Ni(SiGe) layer was fully consumed, was determined to have an activation energy of 0.125 eV for both the 15 and 25% Ge samples. The undoped samples in this stage were uniformly found to have very high sheet resistance values. This result was attributed to the lack of a conduction path between isolated Ni(SiGe) grains. A strong linear relationship, however, was determined for the doped samples, and increasing Ge content had a small effect on the relationship for these samples. It was determined that the stabilizing influence of B doping was caused by the availability of a conduction path through the unreacted doped SiGe layer.
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.
Statement of Responsibility: by John Moore.
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 2009-02-28

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022417:00001


This item has the following downloads:


Full Text

PAGE 1

1 INFLUENCE OF GERMANIUM CONCEN TRATION AND HOMOGENEOUS BORON DOPING ON MICROSTRUCTURE, KINETI CS, AND SHEET RESISTANCE OF NICKEL GERMANOSILICIDE THIN FILMS By JOHN SAMUEL MOORE 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 John Samuel Moore

PAGE 3

3 In memory of Dr. Richard Connell. May all who teach aspire to his example.

PAGE 4

4 ACKNOWLEDGMENTS Life is a long journey, and one not traveled al one. On a personal level, I thank my family first and foremost. Without their love and s upport, this document would never have been possible. I also express gratit ude to the innumerable teachers who have helped me through my 23 years of education. I particularly recogni ze some of my high school Advanced Placement English, science, and math teachers, Ms. Mi les, Mrs. OConnor, Mr s. Slotnick, and Mr. Lederberg, who gave me the tools and foundatio n for a successful undergraduate and graduate education. My thanks are also extended to the late Dr. Richard Connell, for whom this work is dedicated, for being an outst anding professor and role m odel throughout my undergraduate degree. My dear friend and sta unch supporter, Dr. Michelle Phen, is also acknowledged for her unending assistance with the developm ent and editing of this work. I love you!!!! On a professional level, I thank Hans Weijtmans and Dr. Mark Visokay, formerly of Texas Instruments, Inc., for growing the experimental structures used in th is work. I also thank the staff, especially Kerry Siebein and Dr. Jerry Bourne, of the Major Anal ytical Instrumentation Center (MAIC) at the Univers ity of Florida for their assi stance in TEM and SEM sample analysis. I also thank Mikhail Klimov of AMPAC at the University of Central Florida for his assistance with the SIMS analysis presented in this work. I also acknowledge the staff and students of the Software and Analysis of A dvanced Materials Processing (SWAMP) research group and my advisory committee for their assistance with this work.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4 LIST OF TABLES ...........................................................................................................................8 LIST OF FIGURES .........................................................................................................................9 ABSTRACT ...................................................................................................................... .............13 CHAP TER 1 INTRODUCTION .................................................................................................................. ....15 1.1 Device Interconnections and Silicides ..............................................................................15 1.2 Silicon-Germanium in Se m iconductor Technologies ....................................................... 17 1.2.1 Low Resistivity Junctions .......................................................................................17 1.2.2 Uniaxially Strained Devices ...................................................................................18 1.3 Nickel as a Silicidation Metal ...........................................................................................19 1.4 Motivation of This Work ..................................................................................................20 2 LITERATURE SURVEY ...........................................................................................................25 2.1 Binary Systems ............................................................................................................ .....25 2.1.1 Silicon-Germanium Binary System ........................................................................25 2.1.2 Nickel-Silicon Binary System ................................................................................ 25 2.1.3 Nickel-Germanium Binary System ........................................................................ 26 2.2 Ternary System: Ni-Si-Ge ................................................................................................ 28 2.2.1 Physical Properties ................................................................................................. 28 2.2.2 Phase Diagram ........................................................................................................28 2.2.3 Microstructure ........................................................................................................29 2.2.3.1 Initial film formation .................................................................................... 30 2.2.3.2 Film agglomeration ...................................................................................... 31 2.2.4 Sheet Resistance ..................................................................................................... 34 2.2.5 Influential Variables ............................................................................................... 34 2.2.5.1 Silicon-germanium layer strain and thickness. ............................................ 35 2.2.5.2 Germanium content of Si1-xGex layer. ..........................................................36 2.2.5.3 Crystalline quality of the Si1-xGex layer. ......................................................37 2.2.5.4 Nickel layer thickness. .................................................................................38 2.2.5.5 Implanted Dopants. ......................................................................................39 2.2.5.6 In-situ Doping. .............................................................................................40 2.3 Outstanding Issues ............................................................................................................42 3 DESIGN OF EXPERIMENTS ................................................................................................... 68 3.1 Research Objectives ..........................................................................................................68

PAGE 6

6 3.2 Factors and Levels ........................................................................................................ ....69 3.2.1 Factors Altered .......................................................................................................69 3.2.2 Factors Held Constant ............................................................................................70 3.3 Experimental Structure .....................................................................................................71 3.4 Characterization Techniques ............................................................................................71 3.4.1 Cross-Section Transmission Electron Microscopy ................................................ 72 3.4.2 Scanning Electron Microscopy ............................................................................... 73 3.4.3 Electron Dispersive Spectroscopy ..........................................................................75 3.4.4 Secondary Ion Mass Spectroscopy .........................................................................76 3.4.5 Four Point Probe .....................................................................................................77 3.4.6 Image Processing and Quantification .....................................................................79 3.4.7 Statistical Analysis .................................................................................................81 3.5 Analysis of As-Grown Samples ........................................................................................83 4 INFLUENCE OF GE, B ON MICROSTRUCTURE AND KINETICS .................................... 91 4.1 Analysis using XTEM/EDS ..............................................................................................92 4.2 Plan View SEM Analysis ................................................................................................. 94 4.2.1 Phase Identification using EDS ..............................................................................95 4.2.2 Imaging and Phase Quantification using SEM/BSE .............................................. 95 4.2.2.1 Influence of Ge content ................................................................................97 4.2.2.2 Influence of B content ..................................................................................98 4.3 Reaction Kinetics ............................................................................................................101 4.3.1 Reaction Order ......................................................................................................103 4.3.1.1 Precipitation reaction ..................................................................................105 4.3.1.2 Agglomeration reaction ..............................................................................106 4.3.2 Activation Energy .................................................................................................106 4.3.2.1 Precipitation reaction ..................................................................................107 4.3.2.2 Agglomeration reaction ..............................................................................109 4.4 Summary .........................................................................................................................110 5 RELATIONSHIP OF SHEET RESISTANC E AND MICROSTRUCT URE AND THE INFLUENCE OF GE AND B .............................................................................................. 134 5.1 Sheet Resistance Analysis ..............................................................................................135 5.1.1 Influence of Ge .....................................................................................................136 5.1.2 Influence of B ....................................................................................................... 137 5.2 Sheet Resistance/Microstructure Relationship ............................................................... 138 5.2.1 Stage I (Precipitation) ...........................................................................................141 5.2.1.1 Influence of Ge ...........................................................................................141 5.2.1.2 Influence of B .............................................................................................143 5.2.1.3 Tortuosity Analysis ....................................................................................144 5.2.2 Stage II (Agglomeration) ...................................................................................... 147 5.2.2.1 Influence of Ge ...........................................................................................147 5.2.2.2 Influence of B .............................................................................................148

PAGE 7

7 5.2.3 Boron Conduction Path ........................................................................................ 150 5.2.3.1 Potential Conduction Paths ........................................................................151 5.2.3.2 Evaluation of Potential Paths .....................................................................153 5.3 Summary .........................................................................................................................155 6 CONCLUSIONS AND FUTURE WORK ............................................................................... 183 6.1 Conclusions .....................................................................................................................183 6.1.1 Microstructure and Kinetics ................................................................................. 183 6.1.2 Structure/Property Relationship ...........................................................................185 6.1.3 Influence of Homogeneous B Doping ..................................................................186 6.2 Future Work ............................................................................................................... .....187 APPENDIX A ERROR IN SEM/BSE IMAGE QUANTIFICATION ............................................................189 B GAUGE REPEATABILITY AND REPRODUCIBILITY ANALYSIS OF 4PP MEASUREMENT ................................................................................................................ 193 C AVRAMI PLOT ERROR ........................................................................................................198 LIST OF REFERENCES .............................................................................................................200 BIOGRAPHICAL SKETCH .......................................................................................................205

PAGE 8

8 LIST OF TABLES Table page 1-1: Properties of common silicide s used in salicide processing ................................................... 24 2-1: Measurement of average Ni(Si1-xGex)/Si1-xGex interface roughness for undoped and Bdoped samples calculated via AFM after annealing at 400 and 500 oC for 30 seconds .... 67 3-2: Results from SIMS analysis of as-grown samples ................................................................ 90 4-1: Linear regression results for isothermal transfor mation curves plotted on a graphs of log(log(1/(1-AF))) vs. log(t) fo r the precipitation reaction. ............................................. 130 4-2: Values of n (reaction order) for reactions ob eying Avrami kinetics .................................... 131 4-3: Linear regression results for isothermal transfor mation curves plotted on a graphs of log(log(1/(1-AF))) vs. log(t) fo r the agglomeration reaction ........................................... 132 4-4: Activation energies derive d from linear regression of seri es on the plot of ln(k) vs. (1/KT) ........................................................................................................................ ......133 5-1: Linear regression results for best fit lines sho wn for Stage I (precipitation reaction) in Figure 5-9. ........................................................................................................................179 5-2: Stage I (precipitation reaction) linear regression resu lts f or the compiled undoped and doped data shown in Figure 5-11. ....................................................................................180 5-3: Measurements of Si1-zGez area fraction, tortuosity, and sheet resistance for selected undoped and doped 25% Ge samples. ............................................................................. 181 5-4: Regression results for Stage II (agglomerat ion reaction) best fit lines shown in Figure 515......................................................................................................................................182

PAGE 9

9 LIST OF FIGURES Figure page 1-1: Diagram and SEM image of Si diffusion into Al ................................................................... 21 1-2: Salicide process flow ..............................................................................................................22 1-3: Cross-section of uniaxially strained de vice showing Si recess etch, SiGe epitax ial growth, and image of actual device ................................................................................... 23 2-1: Binary phase diagram of Si-Ge system s howing com plete solid solubility of Si and Ge in the solid phase ............................................................................................................ ....43 2-2: Binary phase diag ram of Ni-Si System ................................................................................. 44 2-3: Schematic of the evolution of Ni-Si binary system on thermal annealing ............................. 45 2-4: XRD results plotting the squared normalized intens ity of characteristic diffraction peaks for Ni, Ni2Si, and NiSi as a function of anneal time .......................................................... 46 2-5: Sheet resistance of nickel silicide samp les as a function of a nnealing temperature and initial Ni layer thickness .................................................................................................... 47 2-6: Binary phase diagra m of the Ni-Ge system ............................................................................48 2-7: XRD results plotting the squared norma lized inten sity of characteristic diffraction peaks for Ni, Ni5Ge3, and NiGe as a function of anneal time ............................................ 49 2-8: Sheet resistance of Ni-Ge system as a function of annea ling tem perature for two initial Ni layer thicknesses .......................................................................................................... .50 2-9: Free energy for a system of one gram -atom of NiSi0.5Ge0.5 in contact with one gramatom of Si0.5Ge0.5 at 600 oC as a function of w (fraction of Ge atoms transferred from NiSi1-xGex to Si1-xGex) ....................................................................................................... 51 2-10: Partial isotherms calculated fo r the Ni-Si-Ge ternary system ..............................................52 2-11: Partial isotherms calculated for the Ni-SiGe ternary system without inclusion of the NiSi2 phase .........................................................................................................................53 2-12: Partial calculated isotherm at 600 oC showing the initial and final equilibrium state for one gram-atom of NiSi0.5Ge0.5 in contact with one gram-atom of Si0.5Ge0.5. ....................54 2-13: XRD -2 scans showing evolution of Ni-SiGe system for 60 second RTAs fr om 300-500 oC ......................................................................................................................... 55 2-14: Cross-section TEM images of Ni-silicided Si1-xGex films annealed at 400 oC for 60 seconds ....................................................................................................................... ........56

PAGE 10

10 2-15: Analysis of film s annealed at 500 oC including HRXTEM ................................................. 57 2-16: XTEM images of Ni(Si0.75Ge0.25) films annealed for 60 seconds ....................................... 58 2-17: Plan-view SEM analysis of Ni(Si0.9Ge0.1) grains annealed for 30 seconds at temperatures ranging from 650 to 750 oC .......................................................................... 59 2-18: AES maps of Ge, Ni, and Si for Ni(Si0.75Ge0.25) samples annealed at 500, 700, and 900 oC for 60 seconds ........................................................................................................60 2-19: Sheet resistance of single crystal Si1-xGex (x = 0, 10, 20 at %Ge) layers silicided with 25 nm of Ni for 30 seconds as a f unction of annealing temperatures ................................ 61 2-20: Sheet resistance of nickel germanosilic id e films as a function of anneal temperatures for samples with varying initial Si1-xGex film strain .......................................................... 62 2-21: Plan-view SEM images of nickel germ anosilicide samples with varying concentrations of Ge annealed at 650, 700 and 750 oC for 60 seconds .............................63 2-23: Plan-view SEM images of nickel ge rm anosilicide films on As and B implanted single-crystal Si1-xGex ........................................................................................................65 2-24: Sheet resistance of As and B doped nick el germ anosilicide films as a function of anneal temperature ............................................................................................................ .66 3-1: SEM/BSE image from this work ............................................................................................ 85 3-2: Examples of images from this work before and after Im age J threshold operation .............. 86 3-3: Example tortuos ity m easurement ...........................................................................................87 3-4: XTEM image of sample from as-grown wafer with x = 25 at% Ge and no B doping. ......... 88 4-1: On-axis <110> XTEM images of 25% Ge samples annealed at 450 oC for 10 minutes ...... 112 4-2: On-axis <110> images of 25% Ge sa m ples with and without B doping annealed at550 oC for 10 or 1020 minutes. ............................................................................................... 113 4-3: XTEM images taken at 10o tilt from <110> zone ax is of undoped and doped 25% Ge samples RTA annealed at 600 oC for 60 seconds ............................................................ 114 4-4: SEM images of undoped 25% Ge sam ple after 1020 minute anneal at 550 oC in SE and BSE modes ..................................................................................................................... ..115 4-5: SEM/BSE images of samples annealed at 450 oC ................................................................ 116 4-6: SEM/BSE images of samples annealed at 550 oC ................................................................ 117 4-7: SEM/BSE images of samples annealed at 650 oC ................................................................ 118

PAGE 11

11 4-8: SEM/BSE images of samples annealed at 750 oC ................................................................ 119 4-9: Area fraction of Ge-rich Si1-zGez phase as a function of anneal time at temperatures of 450, 500, and 650 oC ........................................................................................................ 120 4-10: Paired t-test results (n =36) com paring the Ge-rich Si1-zGez area fraction of 15 and 25% Ge samples at all points on the thermal matrix imaged with SEM/BSE. ........................ 121 4-11: Area fraction of Ge-rich Si1-zGez phase as a function of ann eal time at temperatures of 450, 500, and 650 oC ........................................................................................................ 122 4-12: Paired t-test results (n=36) comparing the u ndoped and doped samples at all conditions imaged with SEM/BSE .................................................................................................... 123 4-13: Plot of log(log(1/(1-AF))) vs. log(t) for samples containing 15% Ge ............................... 124 4-14: Plot of log(log(1/(1-AF))) vs. log(t) for samples containing 25% Ge ............................... 125 4-15: Plot of area fraction of initial Ni(Si1-xGex) phase present as a function of Ge-rich Si1zGez area fraction. ............................................................................................................126 4-16: Plot of log(log(1/(1-AF))) vs. log(t) for Ge-rich Si1-zGez area fractions less than 60% ..... 127 4-17: Plot of log(log(1/(1-AF))) vs. log(t) for Ge-rich Si1-zGez area fractions greater than 60% ........................................................................................................................... .......128 4-18: Plots of ln(k) vs. 1/(kT) .............................................................................................. .......129 5-1: Isochronal anneal se ries of sheet resistance of undoped 15 and 25% Ge sam ples .............. 158 5-2: Isochronal anneal se ries of sheet resistance of doped 15 and 25% Ge sam ples .................. 159 5-3: Isochronal anneal se ries of sheet resistance of undoped and doped 15% Ge sam ples ........160 5-4: Isochronal anneal se ries of sheet resistance of undoped and doped 25% Ge sam ples ........161 5-5: Plot of sample sheet res istance vs. Si1-zGez grain size. .......................................................162 5-6: Plot of sample sheet re sistance vs. area fraction of Ni(Si1-uGeu) or Si1-zGez.......................163 5-7: Plot of sample sheet res istance vs. area fraction of Si1-zGez ...............................................164 5-8: Plot of sample sheet res istance vs. area fraction of Si1-zGez for the precipitation stage (Stage I) of the transformation. ........................................................................................165 5-9: Plot of Stage I sample sh eet res istance vs. area fraction of Si1-zGez for undoped and doped samples ................................................................................................................. .166 5-10: Plot of Stage I linear regression intercep t and slope of the results given in Table 5-1 ..... 167

PAGE 12

12 5-11: Plot of Stage I sample sh eet resistance vs. area fraction of Si1-zGez for all samples ......... 168 5-12: Plot of linear regressi on intercept and slop e of the results given in Table 5-2 ................. 169 5-13: Image of an undoped 25% Ge sample annealed at 450 oC for 1020 minutes before and after ImageJ processing .................................................................................................... 170 5-14: Plots of conductive path tortuosity vs. area fraction Si1-zGez and sample sheet resistance vs. tortuosity for select ed undoped and doped 25% Ge samples .................... 171 5-15: Sheet resistance of undoped and doped 15 and 25% Ge sam ples as a function of Si1zGez area fraction for the agglomeration stage (Stage II) of the film transformation ...... 172 5-16: Sheet resistance as a function of Si1-zGez area fraction for all samples in Stage II (agglomeration) of the film transformation. .................................................................... 173 5-17: SEM/BSE images of undoped 25% Ge samples annealed at 650 oC or 750 oC for 10 minutes before and after ImageJ processing ....................................................................174 5-18: SIMS analysis results of B doping distribution af ter silicidation with Ni m etal showing silicidation induced dopant segregation ........................................................................... 175 5-19: XTEM and SIMS analysis results fo r doped 25% Ge sam ple annealed at 450 oC for 10 minutes ....................................................................................................................... ......176 5-20: Cartoon schematics of nickel germanosilicide samples exhibiting segregation induced dopant segregation ...........................................................................................................177 5-21: Isochronal anneal series of doped 15% Ge sam ples with calculated sheet resistance values for the potential alternative conduction paths. ......................................................178 A-1: Individual plot results divided by sam ple and region for area fraction of Ni-rich phase and area fraction of Ge-rich SiGe phase ..........................................................................191 A-2: Minitab results for fully nested ANOVA results for area fraction of Ni-rich phase and area fraction of Ge-rich SiGe phase ................................................................................. 192 B-1: Gauge R&R ANOVA analysis of sheet re s istance measurements for the five samples which all operators measured successfully in all trials. ................................................... 195 B-2: Gauge R&R ANOVA analysis of sheet re s istance measurements including data from all 10 samples for the operator who successfu lly measured all samples for all trials. .... 196 B-3: Gauge R&R ANOVA analysis of sheet re s istance measurements, excluding data from sample 9, for the operator w ho successfully measured all samples for all trials. ............ 197 C-1: Transformed error range for a constant +/3.35% range of area fraction as a function of nom inal area fraction. .................................................................................................. 199

PAGE 13

13 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 INFLUENCE OF GERMANIUM CONCEN TRATION AND HOMOGENEOUS BORON DOPING ON MICROSTRUCTURE, KINETI CS, AND SHEET RESISTANCE OF NICKEL GERMANOSILICIDE THIN FILMS By John Samuel Moore August 2008 Chair: Kevin Jones Major: Materials Scie nce and Engineering This work studied the influence of Ge concentration and homogeneous B doping on the microstructure, kinetics, and sheet resistance of nickel germanosilici de thin films. Experimental structures consisted of 150 nm of single crystal relaxed Si1-xGex (x = 15 or 25 at%) layers with and without in-situ homogeneous B doping of 4.5E19 atoms/cm2 and capped by 10 nm of sputtered Ni metal. Samples were furnace annealed under a N2 atmosphere at temperatures between 450 and 800 degrees Celsius in 50 degr ee increments for times of 10, 30, 90, 270, and 1020 minutes. Microstructure analysis show ed that increasing Ge content increased the amount of film transformation while the addition of B doping ha d no effect on morphology. Distinct reactions were observed before and after full consumption of the parent Ni(Si1-xGex) film during kinetic analysis of isothermal transformation curves The first reaction, in which Si-rich Ni(Si1-uGeu) and Ge-rich Si1-zGez were found to precipitate from the Ni(Si1-xGex) film, was determined to have concentration dependant activation energies of 1.96 and 0.76 eV for the 15 and 25% Ge samples, respectively. In this stage, the shee t resistance was found to linearly increase with increasing area fraction of Si1-zGez for all samples. The cause for the increase was determined to

PAGE 14

14 be related to the conductive path tortuosity. Increasing Ge content did not affect the structure/property relationship, but the addition of B caused a decrease in sheet resistance. The second reaction, in which Ni(Si1-uGeu) grains were found to agglomerate after the Ni(Si1-xGex) layer was fully consumed, was determined to have an activation energy of 0.125 eV for both the 15 and 25% Ge samples. The undoped samples in th is stage were found to have uniformly very high sheet resistance values. This result was attributed to the l ack of a conduction path between isolated Ni(Si1-uGeu) grains. A strong linear relationship, however, was determined for the doped samples, and increasing Ge content had a small e ffect on the relationship for these samples. It was determined that the stabilizing influence of B doping was caused by the availability of a conduction path through the unreacted doped Si1-xGex layer.

PAGE 15

15 CHAPTER 1 INTRODUCTION Nickel germanosilicides ar e formed when Ni metal and SiGe compounds react during thermal processing. They are predominantly of interest to semiconductor technology where they are used as an intermediate layer between devi ces and metal device interconnections. This work investigates the influence of Ge content and in -situ B doping on the thermal stability of nickel germanosilicides. The following sections in this chapter discuss w hy interconnections and intermediate layers (silicides) are needed, why Si Ge and in-situ B doping of the SiGe are used in modern device designs, and why Ni metal is of interest as a silicidation metal. 1.1 Device Interconnections and Silicides For modern microprocessor designs, tens of millions of individual transistors must be linked together to form a single processor. The interconnections are made using complex, interconnected, multi-layered patterns of metal. The metal must also be connected to the source, gate, and drain contacts of each transistor. Dire ct contact, however, between the semiconductor and the metalsuch as that between Al and Si can lead to the forma tion of Schottky barriers which impede the flow of charge carriers. The height of the barrier depends upon the work function difference between the semiconductor and th e metal. Thus, barrier height is a function of the materials selected and will vary from sy stem to system. Regardless, Schottky barriers degrade device performance by requiring electrons or holes to overcome an energy barrier as they pass across the interface, raising overall circ uit resistance and decreasing the time constant (switching speed) of the device. The thermal stability of a metal-semiconductor interface may also be poor and lead to failure of the device. In the Al-Si system at 450-500 C, the solubility of Si in Al is between 0.5 and 1.0 at% [Mey90]. As shown in Figure 1-1(a) Si therefore diffuses

PAGE 16

16 from the substrate into the Al layer, creating pits in the silicon substrate. These pits, shown in a Scanning Electron Microscopy (SEM) image in Fi gure 1-1(b), can disrupt device function. Alternatively, an intermediate layer may be used to form low-resistance Ohmic contacts (or Schottky contacts with known, lower, barrier heights) between the substrate, silicide, and metal. Depending upon the materials selected, use of an intermediate layer can also result in improved interface stability. To form the intermedia te layer, a layer of metal is deposited onto a semiconductor substrate where electrical contact is to be made. The metal, which is often different that that used for interconnection, is then diffused into the substrate using a thermal anneal. This results in the formation of a me tal-semiconductor compound called, in the case of a Si substrate, a silicide (altern atively, for a Ge substrate, a germanicide). Depending on the metal and substrate, one or more stoichiometric compounds are possible, and more than one stoichiometric compound may be initially formed. One method of forming silicides is called a s alicide (self-aligned silicide) process. Figure 1-2 presents a simplified salicide proces s flow. The process flow begins once the devices gate, source, drain, and SiO2 isolation has been fabricate d. First, a metal layer is uniformly deposited over entire structure, usua lly by a sputtering technique. Second, a thermal anneal produces silicides at any metal/silicon in terface. Metal not in contact with Si does not react (hence the self-aligned nature of the tec hnique). Once the silicide has been formed, the remaining unreacted metal is removed using a selective etch. Once formed, however, silicide layers (or germ anicide, etc.) are s ubjected to additional thermal processing during the remaining fabrication of the device. It is important, therefore, to understand how the silicide may react or evolve due to subsequent thermal processing. For example, when NiSi is annealed to temperat ures above ~700 C a phase transformation from

PAGE 17

17 NiSi to NiSi2 occurs and increases the re sistance of the silicide [Gam 98]. For other silicides, such as TiSi2, additional thermal processing can cause th e agglomeration of silicide grains that also results in higher resistivity contacts [Gam98] Thus, knowledge of the thermal stability of a silicide is very importa nt to process design. 1.2 Silicon-Germanium in Semiconductor Technologies Silicon-Germanium (SiGe or, more specifically, Si1-xGex where x = at% Ge) are alloys of varying atomic ratios of Si and Ge which are used in some semiconductor applications. The properties of these alloys will be discussed in fu rther depth in the following chapter. SiGe is used for a variety of applications including photodiodes [Hua95], work-function-tunable gates [Hel97, Pon00], and SiGe-channel heterojunc tion MOSFETs [Pea86, Ver94]. This work, however, will focus on the use of SiGe in p-MOSFET logic technologies. 1.2.1 Low Resistivity Junctions As critical device dimensions for MOSFET logic technology decreased below 130 nm, new methods of forming shallow, low-resi stance source/drain regions became necessary [NTR97]. To overcome this problem, Raaijmak ers et al. suggested the use of elevated source/drain regions that woul d provide shallow junction depth and low sheet and contact resistance [Raa99]. The raised regions, formed by selective epitaxial growth of in-situ B doped Si above the contact region of the device, would al so act as a sacrificial layer during silicidation. Raaijmakers et al. also suggested that Ge should be added to th e elevated source/drain region for several reasons. First, the amount of electrically active B in SiGe alloys can be larger than that in pure Si [Hel97b, Sal97, Man98]; higher concentrations of active dopant leads to lower contact resistance to the silicide. In situ doping of the SiGe alloys al so allows the elimination of an activation anneal as most of the dopant is inco rporated in substitutional positions (electrically active) [Hel97b]. This effect, wh ich is stable, is due to subs titutional B compensating lattice

PAGE 18

18 strain in the SiGe alloy. Sec ond, the addition of Ge has been s hown to shift the valence band of the region to higher energy, d ecreasing Schottky barrier he ight and improving device performance [Hel97b]. Further work, by Gannavaram et al., suggested us ing an isotropic Si plasma etch to define the extension junction recess [Gan00]. The reces s would then be filled using in situ B doped epitaxial deposition of SiGe, maintaining the previously discussed benefits. Ozturk et al. suggested that this technology c ould meet the demands of future technology nodes as small as 30 nm [Ozt01]. Isotropic plasma etching, however, can damage the Si substrate and result in defect formation at the Si/SiGe interface. Loo et al. pr oposed an alternative etch technique using HCL chemical vapor etching which pr oduces defect-free epitaxial in situ doped SiGe regions [Loo04]. This work also confirmed the benefits of Ge addition and in situ B doping on source/drain contact resistance. 1.2.2 Uniaxially Strained Devices The possible benefits of the embedded process proposed by Gannavaram et al. led Intel Corporation to evaluate the technology for use in commercia l process flows [Tho06]. The technology, however, resulted in larg er than expected performance enhancement. The additional enhancement was attributed to uniaxial compre ssive channel stress induced by the SiGe wells [Tho02]. Uniaxial stress causes improved hole m obility at both low strain and high vertical electric fields due to the reduc tion in effective mass from the warping of the Si valence band under strain [Tho04, Uch06]. Improved mobility, in turn, improves the switching speed of the device [Mei04, Won04, Aub05, Lee05, Web05]. A more in-depth discussion of strained-Si technology can be found in a review by Thompson et al. [Tho06]. Uniaxial strained silicon tec hnology, therefore, has been intr oduced into commercial use for 90nm logic technologies [Tho02]. A cross-secti on of a strained-Si device is shown in Figure

PAGE 19

19 1-3. While first-generation devices used ~ 17 at% Ge in the source/drain region, future generation designs will likely increase the Ge con centration and bring the recessed region closer to the channel [Bai04, Chi04, Tho06]. Additionally, by in situ doping of the SiGe region with B, the benefits discussed in the previ ous section may be maintained. 1.3 Nickel as a Silicidation Metal A number of metals have been researched as candidates for use as a silicidation metal for both salicide and other processes. Table 1-1 pr esents important properties of silicides commonly used in salicide processes. Both TiSi2 and CoSi2 have been used in commercially produced products. TiSi2 has several benefits including low resist ivity, relatively high thermal stability, and the ability to reduce native oxides due to the high solubili ty of oxygen in Ti [Iwa85, Bar87 Mass90]. High temperature anneals (>800 C), however, are required to form the lowest conductivity C54 phase. Even higher formation te mperatures are reported to be required as silicide thickness decreases, as linewidth decr eases, and as the concentration on n-type dopants increases [Las91, Gan93, Mae93]. An alternative silicide is CoSi2, which shares a similar resist ivity and thermal stability to TiSi2. The sheet resistance of CoSi2, however, is relatively insens itive to decreasing linewidth [Las91, Mae93] and CoSi2 can be used as a dopant diffusion source to form shallow junctions [Las91, Jia92a, Jia92b, Jia92c]. Disadvantages of CoSi2 in relation to TiSi2 include consumption of more Si to produce an equivalent sheet resi stance and the need for better surface preparation, as CoSi2 does not reduce interfacial oxides. Though lower than that required for TiSi2, a relatively high temperature anneal (600-800 C) is al so required to form the lowest resistance phase [Gam98]. A third option for a silicidation metal is Ni, forming NiSi. NiSi has a low resistivity, with sheet resistance comparable to both of the pr eviously discussed silicides. As with CoSi2, the

PAGE 20

20 sheet resistance of the silicide is also relatively insensitive to linewidth. Furthermore, the formation of NiSi requires less Si consumption than either TiSi2 or CoSi2 which allows the formation of much shallower cont acts [Gam98]. Finally, NiSi can be formed in a one-step anneal at much lower temperatur es (400-600 C) than either TiSi2 or CoSi2. Disadvantages of NiSi include poor thermal stabil ity including transformation to relatively high resistivity NiSi2 at temperatures around 700 C. Further comparisons of NiSi, TiSi2, and CoSi2 are available in the review articles by Gambino et al. [G am98] and Iwai et al. [Iwa02]. 1.4 Motivation of This Work Individual transistors (and other devices) must be linked together to form a working unit through the use of metal interconne ctions. The interconnections and the semiconductor substrate can react, however, to form barriers to charge carrier flow and may also have poor thermal stability. To prevent this, intermediate layers call ed silicides are used to decrease (or eliminate) barrier height and improve inte rface stability. Most commonly, silicides are formed on the source, drain, and gate regions of a semiconductor device through the use of a salicide process. Recent developments in semiconductor technolo gy suggest that the addition of Ge to the source/drain contact regions of th e device can provide increased dopa nt solubility and activation. The use of embedded SiGe source/drain contacts also strains the channel of the device, leading to additional performance gains. Due to a large number of benefits, there is al so interest in the use of nickel as a silicidation mate rial. Thus, it is of particular interest to determine how the introduction of Ge into the source/drain well will a ffect the Ni silicidation of these regions and to determine if in situ B doping of the regions imp acts the silicidation process. While investigation of this topic has already begun (and will be discussed in the next chapter), questions remain about the behavior of nickel germanosilicides. The motivation of this work, therefore, is to answer some of these outstanding questions.

PAGE 21

21 Figure 1-1: (a) Diagram showing di ffusion of Si into Al and susbsequent pit formation in the Si substrate (b) SEM image of pits in Si subs trate after removal of Al layer. [Mey90]

PAGE 22

22 Figure 1-2: Salicide process flow (a) device with ga te, source, drain, and SiO2 isolation fabricated (b) deposition of Ti layer over entire structure (c) thermal anneal under N2 ambient producing silicides at metal/silic on interfaces (cross-hatched regions) (d) device after selective etch to remove unreacted metal. [Gam98]

PAGE 23

23 Figure 1-3: Cross-section of uniaxially strained device showing Si recess etch, SiGe epitaxial growth, and image of actual device [Tho06]

PAGE 24

24 Table 1-1: Properties of common silicides used in salicide processing [Gam98].

PAGE 25

25 CHAPTER 2 LITERATURE SURVEY The ternary Ni-Si-Ge system includes three binary systems: Si-Ge, Ni-Si, and Ni-Ge. Knowledge of the properties of the three binary systems will allow a deeper understanding of the properties seen in the ternary. Thus, each of these four systems will be discussed in this chapter, with emphasis on the microstructural evolution and electrical properties of the ternary system. 2.1 Binary Systems 2.1.1 Silicon-Germanium Binary System The binary phase diagram of the Si-Ge system is shown in Figure 2-1. It can be seen from this diagram that Si and Ge form a complete solid solution across all compositions of the alloy. This is due to the satisfaction of the Hume-Rothery rules: th e two elements share the same diamond cubic crystal structure a nd valence state and have simila r values of electronegativity and atomic diameter. The properties of SiGe vary between those of Si and Ge. For some properties, such as lattice parameter, the values vary linearly between that of Si and Ge in accordance with Vegards Law [Dis64]. For other properties, such as the melting point seen in Figure 2-3, the relations hip is nonlinear. 2.1.2 Nickel-Silicon Binary System The binary phase diagram of the Ni-Si system is shown in Figure 2-2. Six intermediate compounds are evident in the diagram ranging in composition from Ni-rich Ni3Si to Si-rich NiSi2 [Mey90]. Kinetically, however, only three compounds are of interest when Si is present in excess amounts (as is the case duri ng silicidation reactions): Ni2Si, NiSi, and NiSi2 [Gas98, Nem06]. The evolution of these phases from the initial layers is shown schematically in Figure 2-3. Generally speaking, when a thin Ni layer is annealed in contact with excess Si the Ni2Si phase forms until all Ni metal is cons umed. This reaction begins around 210 oC [Nem06]. Once

PAGE 26

26 all Ni metal has been consumed, the Ni2Si phase reacts with the Si substrate to produce NiSi. These reactions are sequential; NiSi does not fo rm until all Ni metal has reacted to form Ni2Si. Experimental observation of the transitions are reported by Nemouchi et al. [Nem06] and shown in Figure 2-4. Upon annealing above ~700 oC, NiSi transforms to NiSi2 [dHe82, dHe89]. While other intermediate phases may appear during the reaction, the presence of these phases is transient and usually neglected [Lav03, Ger 04, Nem06]. The reaction rates of these transformations (measured as a function of layer thickness) have been found to be proportional to linear time (surface reaction cont rolled) or the square root of time (diffusion controlled) [Mey90]. The formation of NiSi2 has also been found to be nuc leation controlled [dHe84] and Ni to be the dominant diffusing speci es for all phases [Fin81, dHe82, Mey90]. The electronic properties of the Ni-Si system are also of inte rest. Figure 2-5 presents the sheet resistance of the system as a function of annealing temperature for three initial Ni layer thicknesses as measured by Chen et al. [Che97]. All samples were annealed for 40 seconds. Two observations can be determined from this plot First, increasing the initial thickness of the Ni layer increases the sheet resi stance of the system for all te mperatures. Second, three regimes of sheet resistance are seen. These correspond to each of the three phases present in the evolution of the Ni-Si system. Ni2Si, present at low temperature, and NiSi2, present at high temperatures, both have sheet re sistances higher than that of the intermediate NiSi phase [Nem06]. For this reason, NiSi is the desi red phase for semiconducto r technology applications. 2.1.3 Nickel-Germanium Binary System The phase diagram of the Ni-Ge system is shown in Figure 2-6; seven intermediate compounds are evident in the diag ram. As with the Ni-Si syst em, all of the phases in the diagram are not seen experimentally. Unlike the Ni-Si system, however, some controversy exists as to which phase initially form s during the reaction of Ni metal with excess Ge. Reported initial

PAGE 27

27 phases include orthorhombic Ni2Ge [Hsi88, Li89], monoclinic Ni5Ge3 [Patt94], and hexagonal Ni3Ge2 [Nem06]. Nemouchi et al. suggest that this confusion arises from the fact that the major XRD diffraction lines are identical for all three of these phases [Nem06]. TEM/EELS analysis in the same work showed the Ni5Ge3 phase to form first at a temperature around 140 oC. Regardless, all of these authors agree that NiGe is the second and final phase formed in the evolution of the films. Also unlike the Ni-Si system, the phases present in the thermal evolution of the Ni-Ge system do not form sequentially [N em06]. Instead, the init ial N-rich phase (most likely Ni5Ge3) and NiGe form simultaneously with th e Ni-rich phase between that of the unreacted Ni and NiGe. Once the Ni metal has been consumed, the Ni-rich phase transforms into NiGe. This evolution is shown in Figure 2-7. It has been experimentally identified that Ni is the diffusing species during the evolution of all phases [Tho88]. Figure 2-8 plots the sheet resistance of the Ni-Ge system as a function of annealing temperature for two Ni layer th icknesses [Zha05]. A sample of Ni on Si was also included for comparison. All samples were RTA processed for 30 seconds. It can be observed from this plot that, as with the Ni-Si system, three regimes are present. At low temperature, the Ni-rich germanicide has a higher sheet resistance than the NiGe phase present at intermediate temperatures. At higher temperatures (~500-600 oC), a rapid increase in sheet resistance is apparent. Unlike the Ni-Si system, this increase is not attributed to phase transformation as NiGe2 has only been reported to form at high temp eratures and pressures [Tak00]. Instead, this rapid increase in sheet resistance is due to the agglomeration of NiGe which has been shown to begin at temperatur es as low as 400 oC [Zha05].

PAGE 28

28 2.2 Ternary System: Ni-Si-Ge 2.2.1 Physical Properties When Ni is reacted with SiGe, a ternary all oy of Ni, Si, and Ge is formed. While the composition of the Ni-rich initial phase is a matter of some debate [Cha04, Zha02], it is agreed that the silicidation process produces Ni(Si1-xGex) with the ratio of Si to Ge identical to that of the silicided Si1-xGex alloy at temperatures around 400 oC [Zha02, Cha04, Liu05]. Since both NiSi and NiGe have the same space group (pnma) and prototype crystal (MnP) [Mas90] and Si and Ge form a complete solid so lution, it is thought that Ni(Si1-xGex) is a complete solid solution of NiSi and NiGe in the appropria te ratio [Seg02]. It has also b een shown that the formation of NiGe2 is limited to high temperatures and pre ssures not seen during silicidation processing [Tak00]. Exhaustive experimental evidence has fu rthermore shown that small quantities of Ge (~1 at% +) prevents the formation of NiSi2 at temperatures up to 800 oC or more [Jar02, Seg02, Ish03, Seg03], most likely due to th e lack of the corresponding NiGe2 phase. Several of these authors have therefore proposed that Ni(Si1-xGex) will not transform to Ni(Si1-xGex)2 and at this date no experimental evidence has been shown to the contrary for annealing temperatures up to 850 oC for 30 seconds [Jar02]. 2.2.2 Phase Diagram S.-L. Zhang performed a series of th ermodynamic calculations treating Ni(Si1-xGex) as a complete solid solution of NiSi and NiGe [Zha03]. If the Ni(Si1-xGex) layer is in contact with Si1-xGex, as is usually the case in silicidation, the equilibrium minimum free energy of the Ni(Si1xGex)-Si1-xGex system can be calculated w ith this method. The graphi cal results for an example calculation with one gram-atom of Ni(Si0.5Ge0.5) in contact with one gram-atom of Si0.5Ge0.5 at 600 oC is presented in Figure 2-9. From this grap h, it can be determined that the system will reach equilibrium with one gram-atom of Ni(Si0.90Ge0.10) and one gram-atom of Si0.30Ge0.70

PAGE 29

29 [Zha03]. The driving force for the rejection of Ge from the nickel germanosilicide is the difference in the heats of formation for NiSi (-45 kJ/gram-atom) and NiGe (-32 kJ/gram-atom) [45 from Zha06]. Using similar methodology, partial isotherms for the ternary phase diagram can be calculated. Examples of calculated isotherms are shown in Figure 2-10. The horizontal line between NiSi and NiGe represents all possible compositions of the complete solid solution (any composition of Ni(Si1-xGex). It should be noted that the pr esence and behavior of Ni-rich phases (e.g. Ni2Si and Ni5Ge3) are neglected and not shown. As me ntioned previously, the presence of Ge has been shown to severely inhibit the formation of NiSi2. Eliminating NiSi2 from the calculations produces the isothe rms shown in Figure 2-11. This change affects only nickel germanosilicides with very small concentrations of Ge (x< ~0.05) and it has been suggested that confirmation of this change may be difficult [Jar02]. For the previous example of NiSi0.5Ge0.5 in contact with Si0.5Ge0.5, use of either 600 oC isotherm, shown in Figure 2-12 for the modified calculations, indicates the resulting equilibrium states of NiSi0.9Ge0.1 and Si0.3Ge0.7. 2.2.3 Microstructure Knowledge of the micr ostructure of Ni(Si1-xGex) and its thermal stability is important to determine the possible processi ng window of the material. Wh ile no transformation to a highresistance digermanosilicide analogous to NiSi2 has been reported in literature, nickel germanosilicide films have relatively poor ther mal stability due to the thermodynamic impetus for Ge rejection as calculated by Zhang. T hus, many researchers ha ve investigated the microstructure of Ni(Si1-xGex) and its thermal evolution over a range of processing conditions. The results from these investigations will be discussed in the following sections.

PAGE 30

30 2.2.3.1 Initial film formation The initial reaction of a Ni metal layer with a Si1-xGex alloy produces a Ni-rich phase analogous to the Ni-rich Ni2Si and Ni5Ge3 phases reported for the Ni-Si and Ni-Ge systems, respectively. For a 325 oC 30 second anneal, this phase has been reported to be Ni2(Si1-xGex) by Chamirian et al. [Cha04]. For a 300 oC 60 second anneal the presence of both Ni2(Si1-xGex) and Ni3(Si1-xGex)2 has been reported by Zhao et al.[Zha02]. The existence of Ni2Si, Ni2Ge, Ni3Si2 and Ni3Ge2 are all reported on the respective phase diagrams and so solid solutions of either pair seem reasonable. Further information on Ni-rich pha se formation is not avai lable in literature at this time; one or both phases may form sequen tially or simultaneously or other intermediate phases may also occur. Regardless of the composition and evolution of the initial Ni-rich phase, on annealing to temperatures at 450 oC for 30 seconds, and as high as 700 oC for 30 seconds, Chamirian et al. and Zhao et al. both report that the Ni-rich phase disappears and Ni(Si1-xGex) predominates for all samples. XRD spectra confirming the sole presence (disregarding SiGe and Si peaks) of Ni(Si1-xGex) at temperatures over 400 oC and ranging as high as 800 oC have been reported by several additional authors [Liu04, Ok04, Pey02, C ho06, He05, Yao07]. One set of XRD spectra containing this evolution is incl uded in Figure 2-13. Since Ni(Si1-xGex) is the phase of interest for use as an intermediate silicide layer in semiconductor applications, most work on nickel germanosilicides has therefore focused on therma l anneals at temperatures higher than 400-500 oC where no Ni-rich nickel germanosilicide forms. Ni(Si1-xGex) films formed by annealing at temperatures from 400-450 oC and times of 3060 seconds on crystalline Si1-xGex substrates (with x ranging from 10 to 20 at % Ge) have been analyzed with XTEM/EDS by several auth ors [Zha02, Jon04, Ok04, Ko06, Yao07]. These authors all agree that Ni(Si1-xGex) films produced in this annealing range have homogeneous

PAGE 31

31 compositions with Si to Ge ratio s identical to that of the Si1-xGex substrate. The films are also shown in these works to be granular, continuous and of uniform thickness. XTEM images of representative films formed within this range are shown in Figure 2-14. 2.2.3.2 Film agglomeration As shown by the previously discussed ther modynamic calculations of Zhang [Zha06], the uniform Ni(Si1-xGex) films produced at temperature ranges from 400-450 oC are not thermodynamically stable when in contact with unreacted Si1-xGex. Instead, at longer times or elevated temperatures, nickel germanosilici de films undergo Ge rejection and form Si1-zGez grains that are Ge-rich relative to the Si1-xGex substrate (z > x) interspersed between remaining Si-rich Ni(Si1-uGeu) grains (u < x). This process has b een experimentally observed by a large number of researchers [Pey02, Zha02, Pey04, Ji n05, Yao07]. Furthermore, the equilibrium concentrations of the Ni(Si1-uGeu) and Si1-zGez grains predicted by Zhangs ternary isotherms have been shown to be in good agreement with observed experimental values [Pey04]. While the overall process of the rejection of Ge and the formation of Si1-zGez is commonly referred to in literature as agglomeration of the germanos ilicides, it would be more accurate to label the process as an eutectoid-type so lid state phase transformation. The initial stages of the tran sformation have been observed by Yao et al. for samples with 20 at% Ge annealed at 500 oC for 20 seconds [Yao07]. Using HRXTEM, the researchers observed sharp v-shaped grooves form ing at the inters ection of Ni(Si1-xGex) grain boundaries and the Ni(Si1-xGex)/ Si1-xGex. One such groove is shown in Figur e 2-15(a). Similar grooving is also observed in NiSi and NiGe samples; examples of these grooves are shown in Figure 2-15 (b) and (c), respectively. The grain boundary angles, indicated as in the figures, was found to be smaller in the germanosilicide than in the silic ide or germanide. The overall shape of the interfaces in the germanosilicide was also found to be more planar. These findings led Yao et al.

PAGE 32

32 to suggest that grooving occurs much more rapidly in the germanosilicide and sufficient time is not available to diffusively smooth the interf aces. Yao et al. also performed EDS on the HRXTEM samples to determine the distributi on of Ge in the vicinity of the grain boundary/groove. The results, shown in Figure 215(d), indicate that th e grain boundary above the groove is depleted of Ge wh ile the region immediately below th e groove is enriched in Ge. These findings suggest that Ge rejection occurs along the grain boundaries of the Ni(Si1-xGex) grains. Later stages of the transformation have been observed by a number of researchers [Zha02, Pey02, Ok03, Pey04, Jin05]. In these studies, with Ge compositions ranging from 20 to 25 at %, anneal temperatures ra nging from 500 to 800 oC, and anneal times ranging from 30 to 60 seconds, XTEM analysis shows Ge-rich Si1-zGez grains have formed between grains of Si-rich Ni(Si1-uGeu) [Ok03, Pey04]. One representa tive set of XTEM images, from Pey et al. [Pey04] is shown in Figure 2-16. At lower temperatures (500 to 700 oC, Figure 2-16 (a)-(c)), distinct Gerich Si1-zGez grains are distinguishable in the film. The number and spacing of Si1-zGez grains is also observed to increase in this range. While no t explicitly discussed in any of the works, misfit dislocations can be noted between the Si1-zGez grains and the Si1-xGex layer. These results, when combined with the observations of Yao et al., su ggest that in this annealing range the Ge-rich regions below the grooved gr ain boundary form Ge-rich Si1-zGez grains which are distinct from the Si1-xGex layer. At higher temperatures (~800 oC, Figure 2-16(d)), the misfit dislocations between the grains and the layer are not observed, suggesting that the Ge concentration gradient may have lessened due to diffusion. The spacing between Ni(Si1-uGeu) grains is also larger in this temperature range.

PAGE 33

33 The transformation has also been studie d by plan-view techniques including SEM and AES. Several researchers have used plan-view SEM to observe th e extent of the transformation as a function of temperature [Ok03, Cha04, Pey04, Ko06, Yao07]. The samples observed in these works ranged from 10 to 25 at% Ge and were annealed at temperatures from 500 to 800 oC for times ranging from 20 to 60 seconds. As w ith the XTEM work discussed previously, the results from these studies generally agree. A re presentative set of images, from Ok et al. [Ok03], is shown in Figure 2-17. Evidence of the beginni ngs of the phase transformation (in the form of phase contrast between grains in the image) is present within all images at all times and temperatures. These observations are also seen a ll of the images from all of the other works. The degree of agglomeration, as judged from the size of the grains in the images, is seen to increase with increasing temperature. It is im portant to note, however, that the magnification of the images in all studies is relatively low, w ith scale bars in the 1 to 2 m range. Also, no researcher attempted to quantify the size of the grains observed in the SEM images. The importance of these matters will be discussed in later sections. Plan-view AES elemental mapping was performe d by Pey et al [Pey02]. In this study, mapping was performed on samples of 25 at% Ge a nnealed for 60 seconds at temperatures from 500 to 900 oC in 100 oC increments. The results from the mapping, some of which are shown in Figure 2-18, show non-uniform distributions of Ni and Ge which become coarser with increasing temperature. These images correspond generally well with the plan-view SEM images previously discussed and confirm the contrast seen in the previous images is due to phase contrast and not sample topogra phy. While Pey et al. conclude th at agglomeration begins around 700 oC, it is apparent from the maps that the tr ansformation has begun prior to this point (the distribution at 600 oC is clearly not uniform).

PAGE 34

34 2.2.4 Sheet Resistance One commonly studied electrical property of nickel germanosilicide films is sheet resistance, likely due to the ease of measurement via four point probe anal ysis. Sheet resistance is measured in Ohms/sq and, given the thickness of the film, the resistivity of the film can be extracted from the value. Ok et al. performed an experiment where the sheet resistance of nickel germanosilicide films was measured as a f unction of annealing te mperature [Ok03]. The structures used in the study were co mprised of 25 nm of Ni deposited on Si1-xGex films with 0, 10, and 20 at% Ge and samples were annealed for 30 seconds. The measurements from the study are presented in Figure 2-19. Below 600 oC, the sheet resistance of each film is stable. Above this temperature, the shee t resistance of all sample cond itions increases with increasing temperature (the difference in values between th e samples as a function of Ge content will be discussed later). For the sample s with 0 at% Ge (Ni on Si), the increase is attributed to the previously discussed transformation of NiSi to higher resistance NiSi2 around 700 oC. The NiSi-Ge system, however, was previously show n to lack a corresponding high resistance digermanosilicide. Thus, the increase of sheet resistance for the nickel germanosilicide structures was attributed to the agglomeration (decomposition) of the film. Similar results and conclusions have been presented by a number of researchers [Pey02, Zha02, Cha04, Cha04b, Zha04, Seg04, Liu05, Cho06, Ko06, Lau06, Yao06]. 2.2.5 Influential Variables Several variables have been determined to be influential in the stab ility and evolution of nickel germanosilicide thin films. These variable s include the strain state, thickness, Ge content, thickness, and crystall ine quality of the Si1-xGex layer and the thickness of the Ni layer. The impact of doping via both implantation and in-sit u methods has also been studied. The following

PAGE 35

35 sections briefly discuss the most important variab les and their impact on the films, both in terms of microstructure a nd sheet resistance. 2.2.5.1 Silicon-germanium layer st rain and thickness. Zhao et al. performed a study using XTEM analysis to determine if the st rain state (prior to silicidation) of the Si1-xGex layer impacted the stability of the films [Zha04]. The study used structures with a Si0.73Ge0.27 layer which was partially relaxed (50 or 75%) for some samples prior to deposition of 20 nm of Ni metal. Samples were then a nnealed at temperatures from 400650 oC for 60 seconds. The authors concluded that increasing layer strain enhanced agglomeration due to an increase in interface energy. This conclusion, however, is not wellsupported by the XTEM images presented by the authors as no distinct morphological difference is evident in the images. Plan-view analysis of the samples was not performed. The authors also noted that, for fully-strained layers annealed at 600 oC for 60 seconds, misfit and threading dislocations formed at the Si1-xGex/Si interface in locations wh ere germanosilicide grains approached the interface. These re sults suggest that the depth of th e silicidation may affect strain relaxation of the Si1-xGex layer. In conjunction with the aut hors conclusion that increasing layer strain increases agglomeration, this suggests that the proximity of the germanosilicide layer to the Si1-xGex/Si may affect agglomeration behavior. Zhao et al. also studied the sheet resistan ce of the film as a function of initial Si1-xGex layer strain. The sheet resistance measurements, s hown in Figure 2-20, are stable (~7Ohm/sq) for temperatures below 550 oC. At higher temperatures the sheet resistance of the films increased with increasing temperature. The fully strain ed samples were reported to show the highest increase (to around 47 Ohm/sq) and the 50 and 75% samples both increased to similar values (around 15 Ohm/sq). The increase in sheet resistan ce was attributed to the agglomeration of the layers, with the increased agglomeration of the fully strained layer responsible for causing the

PAGE 36

36 highest increase of sheet resist ance. No direct re lationship between agglomeration and sheet resistance was established. 2.2.5.2 Germanium content of Si1-xGex layer. Ok et al. used plan view SEM to establish th e effect of increasing th e Ge content in the Si1xGex film on germanosilicide formation and stab ility [Ok03]. Images fr om the study are shown in Figure 2-21 for samples with Si1-xGex compositions of 10 and 20 at % Ge annealed at 650, 700, and 750 oC for 30 seconds. The authors noted that the surface of the samples with increased Ge content showed a qualitatively greater number of dark regions (corresponding to areas of Si1xGex uncovered by the phase transformation) when compared to the images of samples with lower Ge content and identical anneals. Fr om these observations, it was concluded that increasing Ge content caused the germanosilicide to become degraded at lower temperatures. No quantitative study of the images, however, was pe rformed. Careful examination of the images also reveals evidence of non-uniform contest for both samples annealed at 650 oC, the lowest temperature shown. This observation suggests that the reaction has already begun for these samples. As previously shown in Figure 2-19, Ok et al also measured the sheet resistance of the samples used in the study. The measurements show ed that the magnitude of the sheet resistance in the stable region (anneal temperature less than 700 oC) increased with increasing Ge content in the initial Si1-xGex layer. No explanation for this behavior was pres ented in the work. The measurements also showed that, once sheet resi stance began to increase at higher temperatures, the sample with larger Ge c ontent increased more rapidly. Though no direct relationship was established, the increases were again attribut ed to agglomeration of the films at higher temperatures; the greater increase in sheet resist ance of the higher concentration Ge sample was ascribed to the increased agglomer ation seen in those samples.

PAGE 37

37 While the previously discussed work by Zhao et al. [Zha04] only studied the microstructure evolution for Ni-silicided Si0.73Ge0.27 layers with various levels of initial strain, the study also measured the sheet resistance of identically processed samples with Si0.81Ge0.19 layers. All samples showed an increase in sheet resistance beginning around 550 oC, once more attributed to film agglomeration. The sheet re sistance measurements of the films with higher initial Ge content, however, were shown to increa se more than those with lower concentrations for each strain condition. The difference was cred ited to increased agglomeration in the samples with higher initial Ge content in the Si1-xGex film. This conclusion, while in agreement with the conclusions of Ok et al., was not supported by micr ostructure analysis of the lower concentration samples. Additionally, no direct relationship be tween microstructure and sheet resistance was determined. 2.2.5.3 Crystalline quality of the Si1-xGex layer. The works previously discussed in this litera ture survey have all used single-crystal Si1xGex layers as the basis for germanosilicide laye r formation. The formation of germanosilicide layers on poly-crystalline Si1-xGex layers, however, has also been studied and found to exhibit significantly different behavior. Using XTEM/EDS analysis, Jarmar et al. have shown that, as with nickel germanosilicide la yers formed on single-crystal Si1-xGex layers, those formed on poly-crystalline Si1-xGex layers also initially form a continuous Ni(Si1-xGex) layer that rejects Ge and agglomerates [Jar02]. A notable diffe rence, however, is that the Ni-rich Ni(Si1-uGeu) grains were shown to be much more mobile and migrated to the wafer surface and Si1-xGex/Si interface. Thus, it can be concluded that the quality of the Si1-xGex layer can clearly aff ect the evolution of the nickel germanosilicide layer. No sheet resistance measurements were included in the study.

PAGE 38

38 2.2.5.4 Nickel layer thickness. The influence of the thickness of the initial Ni layer on germ anosilicide film evolution has been studied by Ko et al. [Ko06]. In the wor k, two thicknesses of Ni metal (11 and 21 nm) were deposited on samples of relaxed Si0.75Ge0.25 and germanosilicides were formed using a two-stage anneal. Samples were first annealed at 400 oC for 60 seconds to form a homogeneous Ni(Si0.75Ge0.25) layer. The samples we re then annealed at 750 oC for an unreported time. One annealed, the samples were images with FE-SEM in BSE mode. The images, not included here due to their poor quality, show that the grain size of the samples with th e thicker Ni layer was qualitatively larger than those of with the thinner Ni layer. Since a thicker initial Ni layer will produce a thicker nickel germanosili cide film (as more Ni is avai lable to react), these results suggest that the thickness of the germanosilicide may affect the stability and evolution of the film. No quantificati on of the different morphologies was attempted. The study by Ko et al. also investigated the e ffect of varying the initi al Ni layer thickness on the sheet resistance of the layers as a func tion of annealing temperature. The samples and thermal processing used in this portion of the study were identical to those previously discussed with the exception of either 11 or 21 nm of Ni being deposit ed prior to si licidation. The measurements show that increasi ng the thickness of the Ni layer did not affect sheet resistance (stable at ~ 5-10 Ohm/sq) for any sample at temperatures below 600 oC. Between 600 to 650 oC, samples with the thinner initial Ni layer showed an abrupt rise in sheet resistance to around 75 Ohm/sq. The sheet resistance of these samples wa s then stable (at ~75 Ohm/sq) to temperatures as high as 800 oC. In contrast, samples with the thicker Ni layer showed only a gradual increase in sheet resistance above 600 oC, with a maximum value of ~75 Ohm/sq reached at 800 oC. The difference in behavior was attributed to two causes. First, the increased avai lability of Ni in the films formed from thicker Ni layers produces th icker germanosilicide laye rs and thus decreases

PAGE 39

39 sheet resistance. Second, the samples with thicke r initial Ni layers disp layed a larger grain size after annealing, causing an increas e in film interconnectivity a nd thus a more stable sheet resistance. More simply, incr easing Ni thickness decreased agglomeration and thus lowered sheet resistance. Again, no attempt to directly link the two properties was made. Liu and Ozturk also varied the initial Ni layer thicknesses (10 and 20 nm) in a study using fully relaxed Si1-xGex layers with Ge concentrations greater than 40% with a nd without in-situ B doping [Liu05]. The effect of B doping will be disc ussed later. No comp arative microstructure analysis of the samples with varying initial Ni layer thickness was made, however the sheet resistance of the samples was measured as a function of temperature for 30 second anneals. The results from the experiment, show n in Figure 2-22, show that the samples with thicker Ni layers had lower sheet resistance measurements at al l temperatures. For the undoped samples, the samples with the thicker initial Ni layer did no t show an increase in sheet resistance until ~450 oC while the samples with the thinner laye r started to display and increase around 350 oC. These results agree well with those found by Ko et al. and were attributed to the same factors: thicker and more stable germanosilicide layers produced by the increased initial thickness of Ni. 2.2.5.5 Implanted Dopants. Chamirian et al. studied th e effect of implanting Si1-xGex films with dopants prior to Ni deposition and germanosilicide formation [C ha04]. In the study, both singleand polycrystalline films of Si0.8Ge0.2 were ion implanted with either As (4 x 1015 atoms/cm2, 20 keV) or B (2 x 1015 atoms/cm2, 2 keV). Next, a spike anneal to 1100 oC was performed to activate the dopants prior to deposition of 10 nm of Ni metal. Samples were then annealed at 450 oC for 30 seconds to produce a Ni(Si0.8Ge0.2) film. To study the thermal stability, the samples were annealed a second time at temperatures of 450, 600, and 700 oC for 30 seconds. Plan view SEM images of the films formed on single-crystal Si1-xGex substrates are shown in Figure 2-23. No

PAGE 40

40 quantitative analysis of the images was perf ormed. Qualitatively, however, the morphology of the films does not appear to differ between the As and B implanted films (no non-implanted control samples were included in the study). Plan view SEM im ages of the films formed on polycrystalline substrates were also included in the work, though they are not reproduced here. These images also showed no qualitative difference in film morphology between the corresponding As and B implanted samples, though the overall grain size of the films appeared to be smaller than those form ed on the single-crystal layers. The sheet resistance of the samples used in the study by Chamirian et al. was also measured. These measurements, presented in Figure 2-24, show no significant difference between the As and B doped samples for the films formed on either singleor poly-crystalline Si1-xGex layers. The plateau at temperatures below ~375 oC was attributed to the Ni-rich germanosilicide phases present at these temperat ures having a higher resistivity than the Ni(Si1xGex) phase previously shown to be presen t at temperatures greater than around 400 oC. The increase in sheet resistance seen around 600 oC for all samples was attributed to the agglomeration of the film. The similarity in sheet resistance between the As and B doped samples was attributed to the prev iously discussed similarity in the morphology of the films. No quantitative relationship between sheet resistance and film morphology was established in the study. 2.2.5.6 In-situ Doping. The impact of heavily doping the Si1-xGex layer during its epitaxial growth (in-situ doping) on germanosilicide stability was studied by Liu and Ozturk [Liu05]. In the work, fully relaxed Si1-xGex layers with Ge concentrations greater th an 40% were doped with approximately 2 at% B (1x 1021 atoms/cm2). Undoped samples with identical Ge concentration were also grown for comparison. A 20 nm thick layer of Ni was then deposited and the samples were annealed at

PAGE 41

41 temperatures from 300 to 750 oC for 30 seconds to form nickel germanosilicide films. SIMS analysis of the B-doped sample annealed at 500 oC showed the formation of a shoulder and significantly longer tail for the Ni distribution when compared to the analysis of the B-doped sample annealed at 450 oC. It should be noted that the SIMS analysis was reported as a plot of analysis time vs. counts and wa s not converted to depth vs. co ncentration. Regardless, this change was attributed to signifi cant interface roughening. The Ni(Si1-xGex)/Si1-xGex interface was further studied by selectively etching the germanosilicide and performing Atomic Force Microscopy (AFM) analysis of th e exposed interface. Table 2-1 presents the average interface roughness values for doped and undoped samples annealed at 400 and 500 oC. The average roughness of the doped samples is significantly less than that of the undoped samples at each temperature. Liu and Ozturk attributed th ese results to the high levels of boron doping decreasing agglomeration through strain compensa tion. This theory is supported by the wellestablished ability of boron doping to compensate strain in Si1-xGex films at concentrations lower than 1 at% [Cho06a, Cho06b] and the previously di scussed conclusions of Zhao et al. [Zha06] who stated that increasing stra in increased the amount of film degradation. Other work has determined a similar effect in nickel germa nosilicide samples highly doped with C [Tol04]. The sheet resistance of the samples used in the experiment by Liu and Ozturk was also measured as a function of anneal temperature. The measurements, previously shown in Figure 222, indicate that the addition of the in-situ B dop ing stabilized the sheet resistance of the films above 350 and 450 oC for the samples with 10 and 20 nm thick initial Ni layers, respectively. While the sheet resistance of the samples with B-doped Si1-xGex layers continues to gradually increase with increasing temperature, a sharp rise in the property is not seen. The stabilization

PAGE 42

42 was attributed by Liu and Ozturk to their conclusion that the heavy B doping reduced the agglomeration of the films. 2.3 Outstanding Issues While a substantial amount of work has been accomplished with the intent of understanding the formation and stability of nickel germanosilicide films, a number of questions remain unanswered in literature. First, can a di rect relationship between film agglomeration and the increase in sheet resistance be established? While many works have proposed a qualitative link between the two properties, no quantitative relationship has been put forth. Without the establishment of a direct, quant itative relationship the connec tion between the two properties cannot be absolutely establis hed. Second, does the addition of high levels of B doping truly affect the amount of agglomeration in nickel germanosilicide films? The study by Liu and Ozturk utilized only indirect measurement t echniques during their investigation of the morphology of the films. If ther e is indeed an effect, does any quantitative relationship between agglomeration and sheet resistance still hold true? If there is not an effect, what is then causing the stabilization of sheet resistance in the films? Finally, the studies available in literature all utilize isochronal experiments with very short a nnealing times. What happens if the experiments are performed over a larger range of time and temperatures? Answers to these questions will further clarify the behavior of ni ckel germanosilicide thin films.

PAGE 43

43 Figure 2-1: Binary phase diagram of Si-Ge system showing comp lete solid solubility of Si and Ge in the solid phase [Mas90].

PAGE 44

44 Figure 2-2: Binary phase diag ram of Ni-Si System [Mas90].

PAGE 45

45 Figure 2-3: Schematic of the evolution of Ni-Si binary system on thermal annealing. The right hand path (excess Si, Ni layer consumed) is followed for silicidation processes [Mey90]

PAGE 46

46 Figure 2-4: XRD results plotting the squared normalized intensity of characteristic diffraction peaks for Ni, Ni2Si, and NiSi as a function of anneal time. Samples were 50 nm of Ni on a-Si substrate annealed at 230 oC. Note that the Ni signal disappears prior to the appearance of NiSi [Nem06].

PAGE 47

47 Figure 2-5: Sheet resi stance of nickel silicide samples as a function of annea ling temperature and initial Ni layer thickness. Anneals were performed for 40 seconds [Che97].

PAGE 48

48 Figure 2-6: Binary phase diagra m of the Ni-Ge system [Mas90].

PAGE 49

49 Figure 2-7: XRD results plotti ng the squared normalized intensity of characteristic diffraction peaks for Ni, Ni5Ge3, and NiGe as a function of anneal time. Samples were 50 nm of Ni on a-Ge substrate annealed at 160 oC. Note that the germanicide phases arise simultaneously [Nem06].

PAGE 50

50 Figure 2-8: Sheet resistance of Ni-Ge system as a function of annealing temperature for two initial Ni layer thicknesses. Sheet resi stance of Ni on Si is also shown for comparison. All samples were annealed for 30 seconds [Zha05].

PAGE 51

51 Figure 2-9: Free energy for a syst em of one gram-atom of NiSi0.5Ge0.5 in contact with one gramatom of Si0.5Ge0.5 at 600 oC as a function of w (fraction of Ge atoms transferred from NiSi1-xGex to Si1-xGex). Minimum free energy is ach ieved with w = 0.4, resulting in one gram-atom of NiSi0.9Ge0.1 in contact with one gram-atom of Si0.3Ge0.5 at equilibrium [Zha03]

PAGE 52

52 Figure 2-10: Partial isotherms calculated for the Ni-Si-Ge ternary system for (a) 600 oC and (b) 750 oC. Note that the presence and behavior of Ni-rich phases (e.g. Ni2Si, Ni5Ge3) are neglected [Jar02]

PAGE 53

53 Figure 2-11: Partial isotherms calculated for the Ni -Si-Ge ternary system w ithout inclusion of the NiSi2 phase for (a) 600 oC and (b) 750 oC [Jar02]

PAGE 54

54 Figure 2-12: Partial calc ulated isotherm at 600 oC showing the initial and final equilibrium state for one gram-atom of NiSi0.5Ge0.5 in contact with one gram-atom of Si0.5Ge0.5 [Zha03].

PAGE 55

55 Figure 2-13: XRD -2 scans showing evolution of Ni-Si-Ge system for 60 second RTAs from 300-500 oC. Ni2(SiGe) is shown to predominate at 300 oC while only Ni(SiGe) is apparent at higher te mperatures [Zha02].

PAGE 56

56 Figure 2-14: Cross-section TEM images of Ni-silicided Si1-xGex films annealed at 400 oC for 60 seconds with (a) x = 20 at% and (b) x = 30 at% [Zha02].

PAGE 57

57 Figure 2-15: Analysis of films annealed at 500 oC including HRXTEM of (a) Ni(Si0.8Ge0.2) (b) NiSi (c) NiGe. Grain boundary angle is indicated as for each sample. (d) EDS analysis of Ni(Si0.8Ge0.2) sample at locations as noted. [Yao07]

PAGE 58

58 Figure 2-16: XTEM images of Ni(Si0.75Ge0.25) films annealed for 60 seconds at (a) 500oC (b) 600 oC (c) 700 oC and (d) 800 oC [Pey04].

PAGE 59

59 Figure 2-17: Plan-view SEM analysis of Ni(Si0.9Ge0.1) grains annealed for 30 seconds at temperatures ranging from 650 to 750 oC [Ok03].

PAGE 60

60 Figure 2-18: AES maps of Ge, Ni, and Si for Ni(Si0.75Ge0.25) samples annealed at 500, 700, and 900 oC for 60 seconds [Pey02].

PAGE 61

61 Figure 2-19: Sheet resistance of single crystal Si1-xGex (x = 0, 10, 20 at %Ge) layers silicided with 25 nm of Ni for 30 seconds as a f unction of annealing temperatures [Ok03].

PAGE 62

62 Figure 2-20: Sheet resistance of nickel germ anosilicide films as a function of anneal temperatures for samples with varying initial Si1-xGex film strain. Samples were annealed for 60 seconds [Zha04].

PAGE 63

63 Figure 2-21: Plan-view SEM images of nickel germanosilicide samples with varying concentrations of Ge (10 and 20 at%) in the initial Si1-xGex layer annealed at 650, 700 and 750 oC for 60 seconds [Ok03].

PAGE 64

64 Figure 2-22: Sheet resistance of nickel germanos ilicide films as a function of temperature for samples with and without B doping and tw o initial Ni layer thicknesses. [Liu05]

PAGE 65

65 Figure 2-23: Plan-view SEM images of nickel germanosilicide films on As and B implanted single-crystal Si1-xGex. Samples underwent a two-stage anneal at 400 oC for 60 seconds and 450, 600, or 700 oC for 30 seconds.

PAGE 66

66 Figure 2-24: Sheet resistance of As and B doped nickel germanosilicide f ilms as a function of anneal temperature for samples with (a ) crystalline and (b) polycrystalline Si1-xGex layers. [Cha04]

PAGE 67

67 Table 2-1: Measurement of average Ni(Si1-xGex)/Si1-xGex interface roughness for undoped and B-doped samples calculated via AF M after annealing at 400 and 500 oC for 30 seconds. [Liu05] AFM Roughness TemperatureUndoped Doped 400 oC 12.0 nm 4.4 nm 500 oC 18.0 nm 5.0 nm

PAGE 68

68 CHAPTER 3 DESIGN OF EXPERIMENTS 3.1 Research Objectives The objective of this work is to clarif y certain outstanding issues regarding the morphological stability and elec tronic properties of nickel ge rmanosilicide thin films. Specifically, this work aims to clarify the rela tionship between the reje ction of Ge from the initial film which results in the formation of Ge-rich Si1-zGez grains (commonly referred to as film agglomeration) and observe d increases in the sheet resistance of the film. While prior works have proposed a qualitative link between the two properties, no qu antitative relationship has been put forth. This work also aims to confirm whether the addition of high levels of B doping can suppress the agglomeration process as suggested by Liu and Ozturk [Liu05]. Their work did not utilize direct measurement techni ques when evaluating film agglom eration and so the degree of suppression, if any, is not well quantified. Regard less if suppression is conf irmed to occur, this work also aims to determine if the quantitat ive relationship between agglomeration and sheet resistance is maintained for B doped samples. Finally, prior studies available in literature utilize isochronal experi ments with very short annealing times. This work aims to expand th e knowledge base of th e behavior of nickel germanosilicide films over a much larger range of anneal times and temperatures, including both isochronal and isothermal series. By increasi ng the thermal matrix, information regarding the kinetics of the phase tran sformation may be obtained. Achieving these research objectives will lead to greater understanding of the stability and properties of nickel germanosilicide thin films. This understanding, in turn, can be used to further evaluate the use of th e layers as intermediate cont acts in semiconductor devices.

PAGE 69

69 3.2 Factors and Levels The choice of factors and levels used in this wo rk were strictly constrained due to the lack of availability of in-house wafer processing capability. Many factors known to affect the behavior of nickel germanosilicide films, such as initial nickel layer thickness or Si1-xGex crystal quality, require separate wafers to study each le vel. Thus, only the two such factors thought to be most important were selected for this work: Ge concentration of the initial Si1-xGex layer and homogeneous boron doping of the Si1-xGex layer during growth. As each of these factors was varied at two levels, only four wafers were need ed for all experiments performed in this work. The other factors studied in this work, anneal temperature and anneal time, were able to be varied over many levels as ample ther mal processing ability was available. It would have been possible to include other factors in the study with a relatively small number of additional wafers through use of an experimental design allowing the screening of multiple factors, such as a fractional factorial design. Such a design, however, was rejected as the intent of this work was not to determine whic h factors are critical to nickel germanosilicide growth (this knowledge is availa ble in the literature) but to explore a few known factors in greater depth. 3.2.1 Factors Altered Four factors were varied in the experiment: a nneal temperature, anneal time, Ge content in the Si1-xGex film, and boron doping of the Si1-xGex film. The factors were varied over a range of levels dependent upon the availability of resour ces. For temperature, levels of 450 to 800 C in increments of 50 C were used. All thermal anneals were performed in a quartz-tube furnace under a N2 ambient atmosphere. For time, leve ls of 10, 30, 90, 270, and 1020 minutes were chosen. Anneal lengths were timed using a laboratory timer. For Ge content, levels of 15 and 25 at% were used. Higher Ge concentrations were not available due to limitations in reactor

PAGE 70

70 processing capability. It is known that Ge composition is a critical factor in nickel germanosilicide behavior as increasing Ge content will increase the th ermodynamic driving force for the rejection of Ge. This, in turn, results in increased reac tion rates and the formation of Si1zGez grains with higher Ge content. It was expected that these di fferences would also affect the sheet resistance of the films. The effect of incr easing Ge concentration is also important since, as previously discussed, it is expected that future device designs will utilize higher Ge concentrations to increase the strain state of the device structures. For boron doping, levels of undoped and homogeneously doped at a concentration of ~4.5E19 atoms/cm2 were selected. While this concentration was lower than the ~1E21 atoms/cm2 used by Liu and Ozturk [Liu95], it represents the maximum level possible with the reactor used to grow the material. Homogeneous boron doping during growth is an important factor as it is lik ely to be included in future device generations since, as previously discussed, it eliminates the need for implantation and activation of source/drain well dopants. All possible combinations of th e factors and levels were performed and analyzed in this work, resulting in each of the four experimental structures being studied over a significant thermal matrix. Thus, the overall matrix may also be treated as either fi ve isochronal or eight isothermal anneal series. Analysis of the repeat ability of the anneal process was also verified; the results of this analysis is available in Appendix A. 3.2.2 Factors Held Constant Additional factors determined by prior research to affect nickel germanosilicide formation, stability, and electronic properties were not varied in this work. The factors held constant include the initial Ni layer thickness and the crysta lline quality, strain state, and thickness of the Si1-xGex layer. While these factors are known to affect nickel germanosilicide film stability, they

PAGE 71

71 are expected to be either less infl uential than those varied in this work, less likely to be varied in future device designs, or both. 3.3 Experimental Structure The relaxed Si1-xGex layers used in this work were grown using an ASM Epsilon 3200 RPCVD tool at Texas Instruments, Inc. Prior to deposition, (001) Si substrate wafers underwent a HF clean and H2 bake at 1050 oC for 3 minutes to remove both the native Si oxide and any contaminants from the wafer surface. Once cleaned, single crystal Si1-xGex layers were grown using dichlorosilane and germane precursors in a hydrogen carrier gas with a flow of 40 standard liters per minute. Borane was also included during growth for the boron doped samples. Growth was carried out at 700 oC and a fixed pressure of 10 Torr. Final Si1xGex layer thickness was 150 nm. Subsequent to Si1-xGex layer growth, all wafers were capped with 10 nm of sputtered Ni metal and 10 nm of TiN. It should be noted th at the initial structure design called for only 0.5 nm of TiN deposition. As will be discussed in later chapters, however the presence of the excess TiN was not found to affect the behavior of the nickel ge rmanosilicide nor did it impact the analytical techniques utilized in this work. 3.4 Characterization Techniques Several types of characteriza tion techniques have been us ed to analyze the nickel germanosilicide films studied in this wor k. These techniques include Cross-Section Transmission Electron Microscopy (XTEM), S canning Electron Micros copy (SEM), Electron Dispersive Spectroscopy (EDS), Secondary Ion Mass Spectrometry (SIMS), Four Point Probe (4PP), and the ImageJ image analysis software. It is important that the abilities and limitations of each technique are understood so that the result s they provide may be correctly interpreted. Therefore, the following sections briefly descri be each technique, how it may be applied to film analysis, and information about the specific tools a nd methods used in this work. More in-depth

PAGE 72

72 discussions of major techniques fundamentals are available in other works, such as the text by Brundle, Evans, and Wilson [Bru92]. 3.4.1 Cross-Section Transmission Electron Microscopy Transmission Electron Microscopy (TEM) is a technique which can be used to obtain images, diffraction patterns, and composition info rmation from a sample. To image a sample with TEM, a beam of monochromatic electrons is focused on a very thin sample (typically less than 200 nm thick). As the beam of electrons passes through the sample, some electrons are inelastically scattered due to electron-atom interactions. Heavier atoms cause a stronger interaction, which leads to increas ed scattering. The transmitted electrons may then be imaged using a phosphorus screen or digital camera place d opposite and normal to the beam direction to provide precise images of the sample with ve ry high magnification. The differently scattered electrons produce phase contrast in the image (areas of higher mass in the sample appear darker). Electron-atom interactions can also produce x-rays characteris tic of the scattering atom. Characteristic x-rays may be captured by a dedicate d detector and analyzed separately, a process which will be discussed in a later section. Th is technique, called Energy Dispersive X-Ray Spectroscopy (EDS), provides information about the atoms present in the analyzed region. For analysis of nickel germanosilicide films, XTEM imaging is especially useful for determining the thickness of the f ilm as layers are directly measur ed (as opposed to calculated or simulated from models with fitting paramete rs). The technique can also provide limited information about the morphology of the film, such as the size and distribution of grains in the layer, though it is important to c onsider that estimating these charac teristics from cross-section is much less preferred than using a plan-view an alysis technique. When combined with EDS analysis, the composition of the sample within a region of the image defined by the spot size of the beam can also be semi-quantitatively determined.

PAGE 73

73 In this work, two TEMs located at the Major An alytical Instrumentati on Center (MAIC) at the University of Florida were utilized for sample analysis. For low-magnification images, a JEOL 200CX was used. For high resolution imag es, a JEOL 2010F was used. XTEM samples in this work were produced using a FEI DB 235 Focused Ion Beam (FIB) mill to shape membranes approximately 8 microns long, 5 microns tall, and 150 nm thick from the parent material. The membrane was oriented such that its major plane was perpendicular to the surface of the material and oriented such that the [110] direction was norm al to the sample plane. Once milled and fully cut away from the parent sample, XTEM samples were extracted using an exsitu micromanipulator and placed on 3mm nickel grids backed with carbon films for imaging in the TEM. The XTEM samples were imaged in bright field along the <110> axis and/or in bright field using a two beam condition with <220> g vector. 3.4.2 Scanning Electron Microscopy Scanning Electron Microscopy (SEM) is a versat ile technique which can provide a variety of information about a sample. In SEM, a beam of energetic electrons is rastered (scanned) across the surface of a sample. The electrons interact with the sample and produce secondary electrons (SE), backscattered elec trons (BSE), and characteristic x-rays; the signal associated with each of the interactions can then be captu red with an appropriate detector. Secondary electrons, produced when primary electrons are in elastically scattered by atoms in the sample, provide topographical imaging of the surface of th e sample. Depressed areas, such as pits and crevices, trap SE while raised areas, such as ri dges and bumps, release more SE causing the areas to be darker and brighter, respectively, in th e image. Backscattered electrons, produced when primary electrons are elastically scattered by atoms in the sample, provide images showing phase contrast in the sample. The phase contrast occu rs due to the higher likel ihood of electrons being elastically scattered by atoms with higher atomic numbers. Thus, regions with more massive

PAGE 74

74 atoms will appear brighter in the BSE image. Fi nally, as in TEM analysis, characteristic x-rays can also be produced when the primary beam inte racts with the sample. The EDS analysis of these x-rays is identical to that of those produced in TEM analysis and will be discussed in the next section. Both SE and BSE imaging modes can be used to analyze nickel germanosilicide films. Plan-view SE imaging of the film surface a nd interfaces revealed by etching can provide information about qualitative surface roughness. Plan-view BSE imaging of the films is especially useful for monitoring the stability and size of the grains in the film as any non-uniform changes in composition difference ca n create contrast in the imag e. Plan-view measurement of grain size in either imaging mode also can provide direct measuremen ts of grain size and distribution over a relatively la rge sampling area defined by the ma gnification of the image. As with TEM, when combined with EDS analysis th e composition of the sample within a region of the image defined by the spot size of the beam can also be semi-quantitatively determined. In this work, a JEOL 6335F field emission SEM located at MAIC was used to image samples. Plan-view samples for SEM analysis were prepared by cleaving small regions from thermally processes wafer pieces. The small re gions were then mounted to an aluminum SEM puck using double sided carbon dots. The puck and samples were then coated with ~10 nm of carbon to prevent charging of the sample during imaging. Samples were imaged at 9.63E-5 Torr in BSE mode with a working distance of 9 mm, a probe current of 12 A, and accelerating voltage of 15 kV. Three images of each samp le for quantitative anal ysis were taken at a magnification of 30,000x and a scan speed of 23 seconds/frame. Image processing and quantification is discusse d in a later section of this chapter. While the reproducibility of the imaging quantification process was unable to be studied due to limitations in personnel time and

PAGE 75

75 availability, an analysis of the repeatability of the SEM imagi ng method used in this work is presented in Appendix A. 3.4.3 Electron Dispersive Spectroscopy As mentioned previously in the sections discussing TEM and SEM techniques, when a high energy primary electron inter acts with an atom a characteris tic x-ray can be produced. In this process, the primary electron causes the remo val of an inner shell electron from the atom. To compensate for the loss, an electron in a high er energy shell in the atom fills the inner vacant position and after releasing its excess energy through the formation of an x-ray. The x-rays energy is characteristic of the atom as the difference between the two energy levels (inner and outer) is, with few exceptions, different for each at omic species. If there are still higher energy shells, the process will repeat until only the outermost shell is missing an electron. Thus, a single strike may lead to the generation of a number of ch aracteristic x-rays. The intensity of the x-rays can then be detected and plotte d as a function of their energy. From this graph, the atomic species present in the sampled area can be iden tified and their relative concentrations semiquantitatively determined. It is important, however, to note that characteristic x-rays are also produced throughout an interaction volume. Thus depending on the spot size and energy of the beam, signal may also be produced by adjacent re gions such as neighboring grains or underlying layers. These possible additional contributions must be considered when analyzing and interpreting EDS results. It is also important to mention that EDS cannot provide absolute phase identification (as can XRD analysis) since its results are semi-quantitative and lack any information about the actual stru cture of the analyzed region. For the study of nickel germanosilicide films, EDS offers a local analysis method capable of determining the types and concentrations of atoms present within individual grains and, with high-resolution tools, in proximity to grain bound aries and interfaces. These results, when used

PAGE 76

76 in conjunction with SEM and/or TEM analysis, can provide greater insight into the imaged structures in both cross-s ection and plan-view. EDS analysis of XTEM samples in this wo rk was performed using an Oxford Inca EDS system on the JEOL2010F TEM located in the MAIC A 1 nm nominal spot size was used for point composition analysis, though a more accurate la teral resolution would be slightly higher as both beam drift and beam spreading effects must be taken into consideration. Semiquantitative analysis of the spectra was performed using the C liff Lorimer thin ration section. The ratio of the relevant atomic species (disregarding C, Cu signal from the TEM grid) was then determined to arrive at an approximate local co mposition. It should be noted, howev er, that this process is very approximate and significant error is likely present in the results. EDS analysis of plan view SEM samples was performed using an Oxford Inca EDS system on the JEOL SEM 6400 located in the MAIC. A nominal spot size of 5 nm was used for this analysis. The ratio method described above was also used to determine approximate local composition for the plan view SEM samples. 3.4.4 Secondary Ion Mass Spectroscopy Compositional profiling of films is also possi ble using SIMS analysis; the technique is especially appropriate for measuring smooth, c ontinuous, multilayered structures. During SIMS analysis, the sample is bombarded with a stream of ions which sputte r particles, including secondary ions, from the samples surface. A mass spectrometer then measures the mass-tocharge ratio of the secondary ions and this information is used to determine the average elemental composition of the sputtered layer. As the sputtering process continues, more material is removed and analysis moves deeper into the sa mple. In this manner, a raw SIMS profile of ion intensity per unit time is obtained. The raw data may then be converted into a quantitative composition profile as a function of depth by using a three step process. First, elemental signal

PAGE 77

77 intensity is converted into elemental concentr ation based on variables including secondary ion sputtering yield and detection efficiency. Next, sputtering time is converted into sputter depth via calculations involving subs trate sputtering rate. Finally, a depth resolution function (DRF) must be assessed and used to reconstruct the or iginal distribution of the atomic species. Thus, SIMS analysis can be used to profile the composition of nickel germanosilicide films as a function of depth. It is important to remember, however, that since SIMS analysis averages the composition of each sputtered la yer over the entire beam spot size a great deal of structural information is lost during analysis (i.e. lateral composition differences). If an interface is rough, SIMS analysis will also provide less-abrupt measurements of ch anges in concentration across the interface. Use of SIMS analysis should therefor e be confined to germanosilicide films that are homogeneous and planar to provide optimal results. SIMS analysis in this work was performed using a PHI Adept 1010 Dynamic SIMS system located at the Advanced Materials Pr ocessing and Analysis Center (AMPAC) at the University of Central Florida. Samples were analyzed using an 2 kV, 50 nA O2 + primary beam and raw data was converted into concentration vs depth plots using the sputtering rate and ion yield of undoped or B-doped SiGe standards, as appropriate. Analysis was performed over a 300 m square region with 10% detec tion area to avoid crater wall eff ects. It should be noted that with these conditions, analysis of Ti and Ni cont aining layers will be less precise (especially for thickness measurements). 3.4.5 Four Point Probe Four point probe (4PP), also know as Kelvin pr obe, analysis can provi de information about the resistivity of a sample. In this technique, f our terminals are placed in contact with a sample along a straight line. The outer terminals and the sample form a circuit through which a set current is passed. The inner terminals and the sa mple form a separate circuit which is passed

PAGE 78

78 across a voltmeter. The voltmeter measures th e decrease in potential between the two inner probes and, with the knowledge of the sample geometry and applied cu rrent, can allow the resistivity of the sample to be measured. Commonly, however, the resistivity of a thin film is not reported. Instead, the sheet resistan ce of the film is often discu ssed. The sheet resistance of a thin film is defined as the resistance measured for a thin film of equa l length and width (a square). Thus, sheet resistance is equal to the resistivity of the film divided be the films thickness and will be constant for any sized square sample. The sheet resistance of a sample can be directly calculated from 4PP data and, with th e use of geometrical correction factors, can also be determined for an irregularly sized sample It should be mentioned, however, that sheet resistance alone cannot define all of the electronic prope rties of a film (i.e. contract resistance, etc.) Nevertheless, sheet resistance does presen t a good general idea of the suitability of a film as a current conductor. Since the primary application of nickel germanos ilicide films is to provide an intermediate conduction layer between the souce/drain wells of a device and its in terconnect metallization, knowledge of a films electronic properties is of great importance. Due to its relatively easy measurement through use of 4PP analysis, sheet resistance is most often us ed to provide insight into the electronic properties of a film. While ot her analysis techniques are sometimes applied to determine additional electronic properties of the f ilms, most of them require the formation of 2or 3-D test structures whic h adds to the expense and complexity of the analysis. In this work, 4PP analysis was performed us ing rectangular wafer se ctions with nominal measurements of 10 by 15 mm. Samples were an alyzed using a Jandel Engineering multi height probe with a linear configurati on of carbide tungsten tips at a uniform spacing of 1.016 mm. Samples were measured using a 900 A current. If voltage readings could not be successfully

PAGE 79

79 obtained with that level of current, the curr ent was adjusted as necessary until successful readings could be taken or th e sample was determined to be non-conductive. Measurements were performed at room temperature with the prob e axis parallel to the long side of the wafer sections in both forward and reverse current mo de. A 10% difference in readings was allowed between the forward and reverse modes; othe rwise the probe location was adjusted until measurements fell within the allowance. The average of both voltage (V) measurements was then used in conjunction with the appropriate ge ometry correction factor (F) found in Table 3-1 and the probe current (I) to cal culate the sheet resistance (Rs) of each sample according to Ohms law, given in equation 3-1. F I V Rs Equation 3-1 A gauge repeatability and reproduc ibility analysis of the 4PP analysis technique used in this work is included in Appendix B. It was determined from this analysis that a 95% confidence interval of +/2.19 Ohm/sq can be applied to each data point in or der to capture any variation in the measurement process. 3.4.6 Image Processing and Quantification Quantification of SEM images is notably absent in prior work discussing nickel germanosilicide thin films. Accordingly, a method of processing and quantifying images was developed for this work. The majority of these tasks were performed using ImageJ v. 1.38x, a public domain, Java-based image processing prog ram developed at the National Institute of Health. The program, available at the time of wr iting at http://rsb/info/nih.gov/ij/, contains a large number of post-processing and image analysis options which may be used for a wide array of purposes. In this work, ImageJ was first used to remove the area of the raw image containing the scale bar in the SEM/BSE images by selec ting the area of interest and using the crop

PAGE 80

80 (Image>Crop) command. The noise in each im ages was also reduced though use of the despeckle (Process>Noise>Despeckle) command and three applications of the smooth(Process>Smooth) option. An example of a raw image is shown in Figure 3-1(a) and the same image after processi ng is shown in Figure 3-1(b) Once processed, a user-defined grayscale thre shold was selected to differentiate between contrasting grains, eliminating sh ades of gray and creating a bina ry image of black and white. Selecting a threshold allows the area fraction and av erage size of the black or white areas in the image to be quantified. Depending on the anneal conditions in this work, two or three phases were visible in each image. Images with two phases present had a single threshold applied to differentiate the phases. Images with three phases present had two thresholds individually applied to distinguish either the lightest and darkest phase from the rest of the image. Examples of thresholds applied to both types of images are shown in Figure 3-2. The average grain size (in arbitrary units) and area fraction (in percent) of the black regions was then calculated for each thres holded image using the analyze particles (Analyze>Analyze Particles) tool. Final quantification values were determined for each sample by averaging the results of three images taken of the same experimental sample. If present, the area fraction of the third phase (the medium sh ade) was determined by subtracting the area fractions of the other two phases from 100. Grain sizes of the thir d phase were not calculated for reasons which will be discussed later. Error ba rs for the quantified results were statistically determined to be +/3.35% for a 95% confidence interval; determination of this interval is discussed in Appendix A. For selected samples, the tortuosity of a conn ected phase was also calculated. Tortuosity, in its simplest form, is defined as the measure of the twisting of a path. The most basic method

PAGE 81

81 to calculate tortuosity ( ) is to divide the actual path leng th, L, by the minimum distance between the endpoints of the path (its chord, C). Th is calculation is presented as Equation 3-2. C L Equation 3-2 The quality of the SEM/BSE images in this work precluded the direct ap plication of computer software to determine tortuosity. Specifically, the images were t oo grainy and the contrast too shallow for sufficiently accurate definition of indi vidual grains and grain interfaces. Instead, the tortuosity was roughly calculated by hand. First, a line of given length was randomly drawn on a SEM/BSE image. The minimum length continuous path between the line endpoints within the desired phase was then estimated and its length me asured. The tortuosity was then calculated by dividing the path length by the le ngth of the initial line. An example of a SEM/BSE image with the initial path as a blue dashed line and the estimated conti nuous path as a solid red line is presented in Figure 3-3. This process was repeated three times for each image and the measurements averaged to determine the tortuosit y. The error in the tortuosity measurement will be addressed in a later section. 3.4.7 Statistical Analysis Statistical analysis of data was performed at many points in this work, including both paired t-tests and ANOVA analyses. A paired t-test (also known as a dependa nt t-test) is a basic statistical analysis technique which determines if the mean difference between two populations from which paired observations are drawn is likely different from a reference value, usually zero. For example, a paired t-test can be used to comp are the grain size of samples with high and low levels of Ge content at each point of a thermal matrix. Using a reference value of zero, the test will determine whether it is statistically likely that changing Ge content affects grain size. By

PAGE 82

82 pairing the samples, the test disregards the in fluence of anneal time and temperature and only calculates the influence of the Ge concentration factor. To perform a t-test, a t-stat istic is calculated from the data by first determining the difference (D = X2-X1) between each pair of samples acco rding to Equation 3-3. Next, a tstatistic is calculated from the values for D a nd the total number of paired samples (n) according to Equation 3-4. D = X2-X1 Equation 3-3 1 )(2 2 n DDn D t Equation 3-4 Once the t-statistic has been calcu lated, it is compared to a cr itical t-statistic found using the desired confidence level and degr ees of freedom (equal to n-1) from a standard t-distribution table to determine if a statistically significant di fference exists. Alternatively, a p-value may be calculated from the t-statistic and, with the desired confidence level, used to evaluate the data. To aid analysis, Minitab version 15.1. 1.0 was used to perform the st atistical calculations presented in this work Analysis of Variance (ANOVA) analyses are also used in this work as they offer the ability to test whether there is a statistically significan t difference between the levels of one or more factors in a single analysis. The analysis can also determine if multiple factors combine to create interactions (synergistic effect s) in the response variable. Performing the calculations to determine p-values for the factor and interact ion terms in an ANOVA analysis, however, is a complicated topic and the reader is referred to an introductory statisti cs text for a detailed discussion of the theory and met hodology of the analyses. As with the paired t-tests, Minitab version 15.1.1.0 was used to the statistical calcul ations for the ANOVA analysis presented in this

PAGE 83

83 work. Minitab analysis provides calculated p-values for each fact or and interaction term and, with the desired confidence level, these p-values may be used to determine the significance of each factor. 3.5 Analysis of As-Grown Samples Samples from the as-grown wafers used in this work were analyzed with XTEM to ensure proper growth of the e xperimental structures. A representative XTEM image is shown in Figure 3-4. Analysis of the images from the as-grown samples showed acceptable layer thicknesses of 150 +/3 nm, 10 +/1nm, and 10 +/1 nm for the Si1-xGex, Ni, and TiN layers, respectively. Interfaces in the as-grown samples were found to be smooth and abrupt. The presence of dislocations at the Si1-xGex/Si interface was noted in all samples. While XRD analysis of the Si1-xGex layer strain state was not performed on samples in this work, it is known that the thickness of the Si1-xGex layer in each sample exceeds th e critical value established for strain relaxation [Mat74, Pe o85]. It can be concluded, therefore, that the Si1-xGex layers in this work are relaxed and any influence of strain on a gglomeration will not be present in the samples. SIMS analysis was also performed to confir m the Ge concentrations and B doping levels of the as-grown wafers; the results from the analysis are presented in Table 3-2. Analysis of the results indicates correct growth of the experimental structures. While the doped structure at the lower Ge level showed a slightly lower Ge concentration than the undoped structure at the same level, the difference is only ~2 at%. This difference is unlikely to cause significant variation in the experimental results. Th e Ge distribution in the Si1-xGex layer was also found to be uniform and continuous in the plots of concentration vs depth, not included here Analysis of the B doped samples showed homogeneous, uniform do ping levels of 4.58E19 and 4.33E19 atoms/cm3 for the low and high Ge level samples, respectivel y. This doping level represents the maximum possible doping using the available reactors and pr ocessing time. With the XTEM results, these

PAGE 84

84 findings allow the conclusion that the experimental structures used in this work were correctly grown and may be directly compared.

PAGE 85

85 Figure 3-1: SEM/BSE image from this work (a ) prior to Image J processing and (b) after processing.

PAGE 86

86 Figure 3-2: Examples of images from this work before and after Image J threshold operation. Images include (a) two phase image before, (b) two phase image after, (c) three phase image before, (d) three phase image after w ith darkest regions se lected, and (e) three phase image after with lightest regions selected.

PAGE 87

87 (a) (b) Figure 3-3: Example tortuosity measurement showing (a) raw SEM/BSE image of the undoped 25% Ge sample annealed at 450 oC for 1020 minutes and (b) the image after application of a Si1-zGez threshold by ImageJ and with a randomly placed line (blue dashed line) and the estimated minimum c ontinuous path around the phase (red solid lines).

PAGE 88

88 Figure 3-4: XTEM image of sample from as-g rown wafer with x = 25 at% Ge and no B doping.

PAGE 89

89 Table 3-1: Table of sheet re sistance correction factors

PAGE 90

90 Table 3-2: Results from SIMS analysis of as-grown samples Nominal Ge Concentration (at%) B doping Actual Ge Concentration (at%) Actual B Doping (atoms/cm2) 15 No 15 0 15 Yes 13 4.58E19 25 No 25 0 25 Yes 25 4.33E19

PAGE 91

91 CHAPTER 4 INFLUENCE OF GE, B ON MICROSTRUCTURE AND KINETICS This chapter explores the influence of both Ge content and the presence of homogeneous in-situ B doping in the initial Si1-xGex layer on the microstructure and kinetics of nickel germanosilicide thin films. It has been prev iously reported in literature that increasing Ge content increases film sheet resistance by enhanc ing film agglomeration [Ok03, Zha04] and that incorporation of high levels of in-situ doped B stabilizes shee t resistance by suppressing agglomeration [Liu05]. None of these work s, however, provide thorough microstructural analysis, focusing mainly on cross section Tran smission Electron Microscopy (XTEM) analysis. XTEM analysis alone can provide good informati on concerning the depth di stribution of phases. Combined with EDS analysis, XTEM can also provide local phase composition and thus limited information about the size and area fraction of phases present. While some works did apply plan view Scanning Electron Microscopy (SEM), which provides good information about grain size and phase area fraction, the technique wa s performed at relatively high magnification and without quantification of the re sults. In the case of the work by Liu and Ozturk, direct observation of the reported stabiliza tion effect is not made at al l; the presented conclusion was based on sheet resistance measurements a nd roughness measurements of the nickel germanosilicide/Si1-xGex interface. Without quantitative anal ysis, conclusive determination of effects and relationships is not possible. Prior work has also been generally confined to single isochronal rapid thermal anneals for durations of 30 or 60 seconds. This limitation has prohibited the analysis of the reaction kinetics. By annealing the samples in this work over a large thermal matrix of both time and temperature, the kinetics of the nickel germanosilicide phase transformation from Ni(Si1-xGex) to Si-rich Ni(Si1-uGeu) and Ge-rich Si1-zGez can also be investigated.

PAGE 92

92 To accomplish these investigations Ge content in the initial Si1-xGex layer was varied at levels of 10 and 25 at% and the layers were either undoped or doped with ~4.5E19 atoms/cm3 of B during growth. Microstructure analysis was performed on samples annealed in a quartz tube furnace under N2 ambient at temperatures between 450 and 650 C in increments of 50 C for 10, 30, 90, 270, and 1020 minutes. Analysis included XTEM and SEM/BSE imaging as well as limited application of EDS. Additional information on the experimental structure, processing, and analysis methods was previously presented in Chapter 3. The follow sections discuss the results from the analysis techniques to determin e both the general microstructure evolution of the samples as well as the kinetics of the transfor mation process, including both the reaction order and activation energy. 4.1 Analysis using XTEM/EDS As XTEM analysis provides limited information on film morphology, only selected samples in this work were imaged using the tech nique. The analyzed samples were selected such that they could be compared to prior work, ensu ring that the films in this work behaved in a consistent manner. Specifically, it was of interest to verify that at low anneal times and temperatures the nickel germanosilicide film initially formed a homogeneous layer of Ni(Si1xGex) and then, at higher times and temperatures, this layer agglomerated into Si-rich Ni(Si1uGeu) and Ge-rich Si1-zGez grains. Accordingly, samples with 25% Ge content with and without B doping were XTEM analyzed after anneals of 450 oC for 10 min and 550 oC for 10 and 1020 minutes. For better comparison to literature which extensively uses 30 to 60 second RTA anneals, samples with 25% Ge content with a nd without B doping were also RTA annealed for 60 seconds at 600 oC. Samples with 15% Ge content were not selected for analysis as it has been previously shown that agglomer ation increases with increasing Ge content [Ok03]. Thus, any effect will be greatest, and most easil y observed, in the 25% Ge samples.

PAGE 93

93 Figure 4-1 presents an on-axis <110> XTEM images of 25% Ge undoped and doped samples annealed at 450 oC for 10 minutes. It is apparent from the images that the films in this work form a homogeneous nickel germanosilicide layer approximately 20 nm thick. These results agree well with prior litera ture that reports the initial formation of a granular, continuous, homogenous Ni(Si1-xGex) layer of uniform thickness [Zha02, Jon04, Ok04, Ko06, Yao07] prior to the agglomeration process. It is also evid ent from the images that the presence of the homogeneous B doping used in this work does not si gnificantly affect film be havior at this time and temperature, as the qualitative ap pearance of both images is identical. On-axis <110> images of 25% Ge samples with and without B doping annealed at 550 oC for 10 and 1020 minutes are shown in Figure 4-2. It is apparent from the phase contrast between adjacent grains that the germanosilicide layers in the samples annealed at 550 oC for 10 minutes have begun to transform into Si-rich Ni(Si1-xGex) and Ge-rich Si1-zGez grains. The transformation is much more advanced, and evident, in the films annealed for 1020 minutes. These samples show large, distinct gr ains of dark-contrast Si-rich Ni(Si1-xGex) interspaced with thin grains of light contrast Ge-rich Si1-zGez (the relationship betw een grain contrast and composition was confirmed by EDS, discussed later in this section). The features in these images agree well with the intermediate and advanced stages of the phase transformation previously reported in literature [Zha02, Pey02, Ok03, Pey04, Jin05]. As noted previously, the addition of the homogeneous B doping in this work does not appear to affect the morphology of the films in cross section. For improved comparison to the XTEM/EDS results of prior literature, additional doped and undoped 25% Ge samples were RTA annealed at 600 oC for 60 seconds. Figure 4-3 presents the XTEM images of these sa mples taken at a tilt of 10o from the <110> zone axis (necessary for

PAGE 94

94 EDS analysis on the TEM used in this work). As with the samples annealed at 550 oC for 1020 minutes, these images show a microstructure with large, distinct grains with dark contrast interspersed between thin grains with lighter contrast. EDS an alysis was performed on several locations for both samples; the results are also sh own in Figure 4-3. As previously mentioned, the results indicated that the grains w ith dark contrast were Si-rich Ni(Si1-uGeu) and the regions with light contrast were Ge-rich Si1-zGez. Additionally, the EDS analysis of the remaining Si1xGex layers agreed well with the SIMS analysis performed on the as-grown structures (22 and 23 vs. 25 and 25 at% for the undoped and doped samples, respectively). Overall, the EDS results were found to agree well with previously report ed germanosilicide film cross-section analyses [Zha02, Pey02, Ok03, Pey04, Jin05]. It was also noted in the EDS analysis that the initial TiN capping layer had oxidized to TiO2. In summary, it can be concluded from the XTEM/EDS analyses presented in this section that the evolution of the germanos ilicide films in this work gene rally agrees with the progression outlined in prior literature. In itially, at short annealing time and temperature, a homogeneous layer of nickel germanosilicide forms. At hi gher times and temperatures, the homogeneous layer transforms into grains of Si-rich Ni(Si1-uGeu) and Ge-rich Si1-zGez. 4.2 Plan View SEM Analysis While XTEM analysis can provide valuable information about the cross section of a sample, such as layer thicknesses, the technique is not well-suited to monitoring grain size or the relative area fraction of phases pr esent during a phase transformation. Hence, this work used extensive plan view SEM analysis to monitor th e microstructure of the nickel germanosilicide films as they underwent the phase transformation from a homogeneous layer to an agglomerated Si-rich Ni(Si1-uGeu) and Ge-rich Si1-zGez structure.

PAGE 95

95 4.2.1 Phase Identification using EDS As previously seen in the XTEM analysis, ni ckel germanosilicide f ilms that are undergoing a phase transformation from a homogeneous laye r to an agglomerated structure show phase contrast between the Si-rich Ni(Si1-uGeu) and Ge-rich Si1-zGez grains. Similar phase contrast was also observed in plan view SEM images taken in both SE and BSE modes. An example of the observed phase contrast is show n in Figure 4-4, where SE a nd BSE images of an undoped 25% Ge sample annealed at 550 oC for 1020 minutes are shown. To identify the phases, an EDS line scan of a undoped 25% Ge sample annealed at 550 oC for 1020 minutes was performed. The results from this line scan, also presented in Figure 4-4, indicate that light contrast regions were rich in Ni while the Si and Ge signal are relatively constant due to the influence of the underlying Si1-xGex layer. These results indicate that the ph ase contrast seen in the SEM images was the reverse of that seen in the XTEM images. The correct identification for the SEM results is that light contrast regions co rrespond to Si-rich Ni(Si1-uGeu) grains and dark regions to Ge-rich Si1zGez grains. 4.2.2 Imaging and Phase Quantification using SEM/BSE The microstructure of all samples annealed from 450 to 650 oC in 50 oC increments at times of 10, 30, 90, 270, and 1020 minutes were pl an view imaged in BSE mode according to the method outlined in Chapter 3. As will be presen ted later, these samples capture the region of the thermal matrix in which sheet resistance was de creased by the presence of B doping. It was of interest, therefore, to dete rmine if the B-doped samples in this region exhibited less agglomeration according to the theory proposed by Liu and Ozturk [Liu05]. Additionally, as will also be later discussed, this region of the thermal matrix also encompasses all successful sheet resistance measurements of undoped samp les. These samples were, therefore, of

PAGE 96

96 maximum interest as their microstructure would allow a structure/property relationship to be determined for both the doped and undoped samples (discussed in Chapter 5). Figures 4-5, 4-6, and 4-7 present BSE images of the samples annealed for 10, 90 and 1020 minutes at temperatures of 450, 550, and 650 oC, respectively. In general, the samples all show a similar morphological evolution with increasing time and temperature. At the lowest times and temperatures used in th is study, no contrast is evident in the images. This corresponds to the homogenous initial film of Ni(Si1-xGex) previously seen in cross section in Figure 4-1(a) and (b). As the time and temperature of the anne als increases, the images begin to show regions of light and dark contrast which have been s hown by Figure 4-4 to correspond to Si-rich Ni(Si1uGeu) and Ge-rich Si1-zGez grains, respectively. At yet higher times and temperatures, the medium grey regions of Ni(Si1-xGex) disappear as the amount of Si-rich Ni(Si1-uGeu) and Ge-rich Si1-zGez increases further. Finally, at the highest times and temperatures, the grains of Si-rich Ni(Si1-uGeu) appear to begin to agglomerate. The evidence of agglomeration was further investigated by imaging samples annealed at 750 oC for 10, 30, 90, 270, and 1020 minutes. The BSE images of the samples annealed for 10, 90, and 1020 minutes are presented in Figure 4-8. Distinct agglomeration is evident in these se ries of images; the samples annealed for 1020 minutes are qualitatively fewer in quantity and much more circular than those annealed at lower times and temperatures. Therefore, it is expected that the agglomeration of the grains is resulting in an increase in grain thickness. Support fo r this conclusion can be observed in the XTEM images previously presented in Figure 4-2, where the Ni(Si1-uGeu) grains are much thicker in the samples annealed for 1020 minutes than in the samples annealed for 10 minutes. While prior studies of nickel germanosilicide films have relied on the qualitative analysis of images, such as that leading to the previously discussed description of the evolution process,

PAGE 97

97 this work makes extensive use of quantitative image analysis. As one of the main goals of this work is to monitor the decomposition (usually referred to as agglomeration) of nickel germanosilicide films, a quantitative metric linked to the process must be established. For this work, that metric was selected to be the area frac tion (or, occasionally, grain size) of the Ge-rich Si1-zGez phase present in an image. As Ge-rich Si1-zGez is a product of the film transformation process, the area fraction of this phase w ill increase with increasing agglomeration. Accordingly, many of the analyses in this work will discuss the agglomeration process in terms of the amount of Ge-rich Si1-zGez present. 4.2.2.1 Influence of Ge content Prior works by Ok et al. [Ok03 ] and Zhang et al. [Zha04] ha ve qualitatively established that increasing the Ge content of the initial Si1-xGex layer increases nickel germanosilicide agglomeration. The works, however, did not qua ntify the microstructures presented in their results. While a qualitative comparison of th e SEM/BSE images presented in Figures 4-5 through 4-8 of the samples with initial Si1-xGex layer concentrations of 15 and 25% Ge supports their conclusions, the images in this wo rk were also quantitatively analyzed. Figure 4-9(a) and (b) plot a 95% confidence interval for the area fraction of Ge-rich Si1zGez as a function of annealing time for temper atures of 450, 550, and 650 for the undoped and doped samples, respectively. Calculation of the area fraction range required for the 95% confidence interval is presented in Appendix A. It should be noted, however, that while for simplicity a single error range (est ablished in Appendix A) was used for all values in these plots, the actual error decreases with increasing area fr action. The quantitative results further support the conclusion that increasing Ge content worsen s film stability. At lower temperatures (450550 oC), the area fraction of precipitated Ge-rich Si1-zGez is much lower for the 15% vs. the 25% Ge samples for both the undoped and doped structur es. At higher temperatures, however, the

PAGE 98

98 difference in Ge-rich Si1-zGez area fraction between the 15 and 25% Ge samples is much less. This trend is also consistent in the quantitative analysis of the isothermal series not shown in the figures (500, 600, 750 oC). For further analysis, a paired ttest can be used to statistica lly compare the differences in Ge-rich Si1-zGez area fraction at each point in the therma l matrix. A single paired t-test was applied to all points on the thermal matrix for both the undoped and doped samples (i.e. comparing the 15% Ge 500 oC 90 minute sample with the 25% Ge 500 oC 90 minute sample, and also comparing the 15% Ge + B 500 oC 90 minute sample with the 25% Ge + B 500 oC 90 minute sample in the same statistical test). A ccordingly, a total of 36 paired data points were used in the analysis. The results from the analysis, presented in Figur e 4-10, show that the difference in Ge-rich Si1-zGez area fraction between the samples with 15 and 25% is statistically significant at a 99.999% level (p=0 .000), with the 15% Ge samples having a generally lower area fraction (and hence less agglomerat ion). It is interesting, howe ver, to note the range of the difference in area fraction. The difference ranges from the 15% samples showing over 60% (absolute) less area fraction to th em having 3% (absolute) more. This range highlights the fact that, while increasing Ge content definitively increases agglomeration at lower times and temperatures, the effect is not consistent acro ss the entire thermal matrix. The increase in agglomeration is only present at lower times and temperatures. The cause of this discrepancy, however, will be addressed in a later section. 4.2.2.2 Influence of B content Previous work by Liu and Ozturk [Liu05] has s uggested that the presen ce of high levels of homogeneous B doping reduced nickel germanosili cide film agglomeration and led to the stabilization of film sheet resist ance. The previous work, however lacked any direct observation of the microstructure effect and instead formul ated the theory from AFM measurement of the

PAGE 99

99 germanosilicide/Si1-xGex interface and SIMS compositional depth profiles. A main objective of this work was to determine, through direct, qua ntitative microstructure observation using plan view SEM/BSE imaging, if this theory coul d be substantiated. While the homogeneous B doping level in this work was limited by the ability of the reactor used to grow the experimental structures, it will be presented in Chapter 4 th at the B doped samples in this work exhibited a similar decrease in sheet resistance when comp ared to undoped samples annealed at the same time and temperature. Accordingly, a differenc e between sample microstructures for the doped vs. undoped structures in this work was expected The SEM/BSE images presented in Figures 45 to 4-8, however, do not qualita tively support Liu and Ozturks th eory. The area fractions, grain sizes, and interconnect ivity of the phases present do not appear to differ between the undoped and doped samples at any time or temperature for either the 15 or 25% Ge samples. Quantitative analysis of the images provides further evidence that there is no difference in area fraction or grain size between the undoped a nd doped samples in this work. Figure 4-11(a) and (b) plot a 95% confidence interval for the area fraction of Ge-rich Si1-zGez as a function of annealing time for temperatures of 450, 550, and 650 oC for the 15 and 25% Ge samples, respectively. Calculation of the area fraction ra nge required for the 95% confidence interval is presented in Appendix A. It should be again no ted that while for simplic ity a single error range was used for all values in these plots, the actual error decreases with increasing area fraction. The plots for the undoped and doped samples appear to largely overlap within the experimental error at all times and temperatures for both the 15 and 25% Ge samples. In fact, it appears if the area fraction of Ge-rich Si1-zGez for the doped samples may be consistently slightly higher than that of the undoped samples, suggesting that doped samples may display more agglomeration than the equivalent undoped samples.

PAGE 100

100 The validity of these conclusions may be furt her established by performing a paired t-test to statistically compare the differences in Ge-rich Si1-zGez area fraction at each point in the thermal matrix. Accordingly, a single paired t-test was applied to all SEM/BSE imaged points on the thermal matrix for the 15 and 25% Ge samples (n = 36). The results from this analysis, presented in Figure 4-12(a), show that there is no statistically significant difference in Ge-rich Si1-zGez area fraction between the undoped and dope d samples at a 95% confidence level. While area fraction provides information on the quantity of a phase present in a microstructure, the metric does not offer any information on the size or distribution of that phase. A difference in grain size (with constant area fraction) could suggest that the interc onnectivity of grains may be altered with the addition of B doping. While this prope rty appears qualitatively identical in the SEM/BSE images of the undoped and doped samples presented in Figures 4-5 to 4-7, statistical analysis was also performed to determine if a quantitative difference existed. Specifically, a single paired t-test as previous ly outlined for area fraction was used as the statistical analysis method. The results from this analysis, presented in Figure 4-12(b), again show that there is no statisti cally significant difference betw een the undoped and doped samples at a 95% confidence level. Thus, in conjuncti on with qualitative observation of the images, it can be concluded that the addition of B doping to the samples did not affect agglomeration, either through an increase in the area fraction of the precipitated phase or by varying the grain sizes of the precipitate d (or any other) phase. The quantitative results and statis tical analysis in this work, therefore, strongly contradict the conclusions drawn by Liu and Ozturk on th e influence of homogeneous B doping in the initial Si1-xGex layer. No statistically significant differen ce in either grain size or area fraction of the Ge-rich Si1-zGez phase was determined at a 95% confidence level. While their results

PAGE 101

101 indicate that the addition of B doping may d ecrease the roughness of the germanosilicide/Si1xGex interface, the decrease in roughness does not correlate to a decrease in agglomeration. Instead, an alternate mechanism must lead to the reduction of sheet resistance in the B doped samples observed in both works. This mechan ism, and the structure/property relationship between agglomeration and sheet resistance, is further addressed in Chapter 5. 4.3 Reaction Kinetics The reaction kinetics of nickel germanosilicid e agglomeration have not been established as most published work has been confined to short, isochronal anneal series of 30 or 60 seconds. Accordingly, without a full thermal matrix includ ing anneals of multiple times and temperatures, the reaction rate as a function of time and temperature cannot be calculated. This, in turn, prevents the activation energy of th e reaction to be established. Accordingly, this work utilized a large thermal matrix incorporating a large range of times and temperatures in order to be able to investigate the kinetics of the reaction process. It should be noted that published literature (and, at times, this work) generally refers to the agglomeration of nickel germanosilicide film s. As shown in the SEM/BSE images in the previous section, this term is somewhat of a misnomer. Agglomeration, according to the Merriam-Webster dictionary, refers to the actio n or process of collecting in a mass [Mer04]. In materials science terminology, agglomeration is usually applied to a process in which a precipitate, thin film, or other structure becomes more spherical in order to reduce the energy of its interface. While XTEM micr ographs of the nickel germanos ilicide films in this work do show some signs of spheroidization (and hence aggl omeration in its strictest definition), it is well established in both prior literature and this work that the film also rejects Ge and forms Ge-rich grains of Si1-zGez. The process, therefore, was not and should not be treated as a pure agglomeration process but instead be treated as (at least partially) a solid state transformation

PAGE 102

102 reaction: the initial Ni(Si1-xGex) film transforms into Si-rich Ni(Si1-uGeu) and Ge-rich Si1-zGez grains. Parallel or subse quent agglomeration of eith er phase may, in fact, need to be handled as a separate event. Transformation reactions have been widely st udied across a variety of materials systems, and a number of theories have been advanced to describe their kinetic pr ocesses. One kinetic theory of transformation for a nucleation a nd growth process was proposed by M. Avrami [Avr39, Avr40, Avr41] and well discussed in the text by J.W. Christian [Chr02]. This theory considers the case where transformation begins at several points (nuclei) in a material. As the transformed regions grow outwards from the nucle i, they eventually impinge on each other and create a common interface ove r which growth ceases (though it will continue normally elsewhere). Assuming an isotropic growth rate and a constant or decreasing nucleation rate per unit volume, Avrami proposed a general relation for a three dimensional nucleation and growth process. This relation, presented in Equati on 4-1, relates the volume fraction transformed ( ) to a rate constant (k), time (t), and the reaction order (n). = 1 exp(-ktn) Equation 4-1 An extensive discussion of the th eory behind the relationship is in cluded in the text by Christian [Chr02] and is beyond the scope of this work. It should be not ed, however, that confining the transformation within a 2-D film (and hence including the effects of a free surface) will modify the reaction order by up to 1 (i.e. decrease n = 3 for a bulk system to n = 2 for a thin film). Isothermal transformation curves of a process obeying the Avrami relationship will have a sigmoidal shape on a linear plot of vs. t. It is difficult, however, to fit linearly plotted sigmoidal data in order to determine n and k fo r a set of experimental results. Accordingly,

PAGE 103

103 Equation 4-1 may be mathematically transformed to produce the relationship given by Equation 4-2. log(log((1)-1) = n(log(t)) + log(k) Equation 4-2 It can be seen from Equation 4. 2 that a plot of log(log((1)-1) vs. log(t) will produce a straight lines with slope n and y-axis in tercept of log(k) for an isothe rmal transformation curve of a process that obeys the Avrami relationship. T hus, experimental data from a process thought to obey Avrami kinetics is often graphed on such a plot and analyzed us ing linear regression analysis after the appropriate transformations have been made. The isothermal transformation curves in Fi gures 4-9 and 4-11 do not show a classic sigmoidal shape as the curves are increasing in va lue from the earliest time. The lack of classic shape, however, does not preclude the process obeying Avrami kinetics, however. Instead, it illustrates that the initial nucleation period may be very short (i.e. the first part of the sigmoidal shape may lie below 10 minutes). It was of interest, therefore, to determine if the nickel germanosilicide transformation obeyed the Av rami theory. As the volume fraction of transformed phase was not determined in this work, the area fraction of the phase was substituted. This substitution is justified as th e reaction can be generally assumed to be confined in a thin layer (as shown by the XTEM in Fi gures 4-1 and 4-2). Plots of log(log((1-AF)-1) vs. log(t) will still produce straight lines with slope n and y-axis in tercept of log(k) for a process obeying Avrami kinetics. As mentioned previous ly, however, the reaction order will be reduced by up to 1. 4.3.1 Reaction Order To determine n, the reacti on order, the log(log((1-AF)-1) of all imaged sample conditions were plotted as a function of log(t). The plot s are presented in Figure 4-13 for the undoped and doped 15% Ge samples and in Figure 4-14 for the undoped and doped 25% Ge samples. As it

PAGE 104

104 was determined in the preceding discussion that, while Ge content influenced film microstructure, homogeneous B doping did not, the undoped and doped samples were considered and analyzed as a single case (though the provenance of each data point was maintained on the graphs). Restated, while the undoped and doped sa mples are plotted separately on the graphs, single statistical regressions c ontaining all data points (both undoped and doped) was performed for each combination of Ge composition and anneal temperature. It should also be noted that, due to the mathematical transformation, the erro r bars plotted for the transformed log(log((1AF)-1) values are not constant, as shown, but actually vary with area fraction. A more comprehensive explanation of this effect and th e reasoning behind the sele ction of the displayed error ranges is contained in Appendix C. It is evident from Figures 4-13 and 4-14 that while most temp erature series appear to be linear, all anneal temperatures do not share a cons tant slope in either pl ot. Furthermore, the results from the 600 oC anneal series for the 15% Ge samples and the 550 oC series for the 25% Ge samples do not appear to be linear. An e xplanation for this effect can be determined by analysis of a plot of the ar ea fraction of the initial Ni(Si1-xGex) phase as a function of Ge-rich Si1zGez area fraction, presented in Figure 4-15. It is evident from this figure that for all samples, regardless of Ge or B content, the initial Ni(Si1-xGex) phase disappears by ~60% Ge-rich Si1-zGez area fraction. When transformed for the Avrami plots, the log(log((1-AF)-1) of 60% area fraction corresponds to a value of -0.400. A horizontal line of this value plotted on the graphs shown in Figures 4-13 and 4-14 would divide the graphs into two regions. Above and below this line, the slopes of all temperature series appear to be qualitatively equal, including the corresponding sections of the 550 and 600 oC plots. These results, along with th e earlier observation of the film

PAGE 105

105 morphology in plan view SEM/BSE, confirm that th e nickel germanosilicide film evolution must be considered as two separate reaction s: precipitation and agglomeration. 4.3.1.1 Precipitation reaction Figure 4-16(a) and (b) present the sections of the Avrami plots lying below 60% area fraction Ge-rich Si1-zGez for the 15% and 25% samples, respectively. These regions of the plot correspond to the precipitation stage of the transformation reaction, where the Ge-rich Si1-zGez phase is being formed through Ge rejection from the initial Ni(Si1-xGex) phase and reaction with the remaining Si1-xGex layer. Linear regression analysis for every temperature series appearing in each plot was performed using the Minitab software; the results from the regression analysis are presented in Table 4-1 and the best fit equati on lines are plotted in Figure 4-16(a) and (b). The average value of n for the 15% Ge samples was determined to be 0.676 +/0.142 for a 95% confidence level. For the 25% Ge samples, n wa s determined to be 0.491 +/0.106 for the same confidence level. With the previously menti oned influence of a 2-D film on n, it can be concluded that the equivalent n for this reaction in a bulk materi al would be between ~0.5 and 1.5. This value for n can be compared to a tabl e of values from the text of Christian [Chr02], presented in Table 4-2, for diffusion controlle d reactions obeying Avra mi behavior. While several types of growth have n values between 0.5 and 1.5, the only reaction which can describe the morphology seen in Figures 4-5 to 4-7 is that of all shapes growing from small dimensions, zero nucleation rate which ha s n = 1.5. It can be concluded, th erefore, that the first stage of the nickel germanosilicide transformation is of this type. This conclusion is further supported by the prior work of Yao et al. [Yao07] who showed that Ge enrichment occurred at grooves formed between neighboring germanosilicide grains (see Figur e 2-15). These results, which suggest that grain boundary nucleation is occurring, support the conclusion that a zero nucleation rate is in

PAGE 106

106 effect, as no new grain boundaries are likely to be formed during the reaction in unreacted regions. 4.3.1.2 Agglomeration reaction Figure 4-17(a) and (b) present the sections of the Avrami plots lying above 60% area fraction Ge-rich Si1-zGez for the 15% and 25% samples, respectively. These regions of the plot correspond to the agglomeration stage of the transformation reaction, where the Ge-rich Si1-zGez area fraction increases as the Si-rich Ni(Si1-xGex) phase agglomerates. Linear regression analysis for every temperature series appearing in each plot was performed using the Minitab software; the results from the regression analysis are presen ted in Table 4-3 and the best fit equation lines are plotted in Figure 4-17(a) a nd (b). The average value of n for the 15% Ge samples was determined to be 0.067 +/0.023 for a 95% confid ence level. For the 25% Ge samples, n was determined to be 0.056 +/0.015 for the same conf idence level. With the previously mentioned influence of a 2-D film on n, it can be concluded that the equivalent n for this reaction in a bulk material would be between ~0.05 and 1.05. This value for n can be compared to the table of values from the text of Christian [Chr02], fo rmerly presented in Table 4-2, for diffusion controlled reactions obeying Avrami behavior. Th e results of this comparison suggest that the reaction is the diffusion controlled thickening of very large plates (e.g. after complete edge impingement) which has a bulk n of 0.5. Th is conclusion is supporte d by the previously presented XTEM (Figure 4-2) and SEM/BSE images (Figures 4-5 to 4-7) in this work and in prior work (Figure 2-16) which show both the Ge-rich Si1-zGez and Si-rich Ni(Si1-xGex) to have large, plate-like morphologies. 4.3.2 Activation Energy The linear regression analyses necessary to determine the order of the reactions also provided values for log(k). These values, reporte d in Tables 4-1 and 4-3 for the precipitation and

PAGE 107

107 agglomeration reactions, respectively, may be mathematically c onverted to ln(k) and used to determine the activation energy (Ea) of each reaction according to the Arrhenius equation, which is presented as Equation 4-3: ln(k) = -Ea (1/KT) + ln(A) Equation 4-3 From this equation, where K is Boltzmanns constant (8.67e-5 eV/K), it is apparent that on a plot of ln(k) vs. (1/KT), Ea will be equal to the negative slope of the line. Figure 4-18(a) and (b) present such plots for the preci pitation and agglomeration reacti ons, respectively. While each data point on these graphs has a unique error range (calculated from the standard error of log(k) in Tables 4-1 and 4-3), the error for each reactio n series was found to be relatively consistent. Hence, for simplicity, a single error range corresponding to a 95% confidence interval for the average standard error was presented for each series on the graphs in Figure 4-18. The error ranges are equal to +/0.680, 0.459 for the 15 and 25% Ge precipitation reactions and +/0.117, and 0.078 for the 15 and 25% Ge agglomeration react ions, respectively. Linear analyses were performed on the reaction series shown in Figure 18; the results from the analyses are presented in Table 4-4 and will be discussed in the following paragraphs. 4.3.2.1 Precipitation reaction For the precipitation reaction, Fi gure 4-18(a) and the results in Table 4-4 show that the activation energy for the 15% Ge samples was hi gher than for the 25% Ge samples. A twosample t-test using the appropriate data fr om Table 4-4 shows the nominal difference in activation energy to be 1.196 eV, with a 95% confidence interv al for the difference ranging between 0.595 and 1.797 eV. While this analysis is formed using somewhat limited data (four points for the 15% sample and three for the 25% sample), the difference is significant and the error range for the data relatively small. It is unlikely, therefore, that additional data points would cause enough of a shift in the linear regression to elimin ate the difference in activation

PAGE 108

108 energies. It can be concluded, accordingly, that a statistically significant difference in activation energies for the precipitation reaction ex ists when Ge content is increased. It is evident from these results that the activation energy of this reaction is concentration dependant; increasing Ge content from 15 to 25% in the initial Si1-xGex layer decreased the activation energy by ~1.2 eV. The precise cause of this decrease it not known. Many steps of the reaction, however, have been found to be co ncentration dependant in other system. Such steps include the bonding energy be tween atoms (which must be broken to switch from Ni-Ge to Ni-Si), and the diffusion of atoms in an alloy. An example of a concentration influence on bonding energy can be found in the work by J. Tersoff, where bond strength was calculated to decrease with increasing Ge content in Si1-xGex [Ter89]. Experimental evidence of this effect has also been reported [Cros03]. An example of concentration depe ndant diffusion activation energy has been shown by Zangenburg et al., w ho studied the diffusion of Ge in Si1-xGex alloys [Zan01]. Their work showed that the activation energy of Ge diffusion in silicon germanium alloys decreases with increasing Ge content. It can be extrapolated from their results that the activation energy decreases from 4.33 to 3.92 eV for a change in composition from 15 to 25% Ge, a difference of 0.42 eV. It is likely, therefore, that the influence of Ge on activa tion energy could be accounted for by similar concentration dependent effects, as the initial (metastable) Ni(Si1-xGex) films contain different compositions of Ge base d upon the concentration of the Si1-xGex layer being silicided. Additional research into the bond strengths a nd mobility of Ni, Si, and Ge in nickel germanosilicide films is needed, however, to dete rmine if a single concentration dependant effect dominates and creates the differen ce in activation energy, or if th e difference is a compilation of

PAGE 109

109 two or more lesser effects. The concentration dependency should al so be studied over a greater range of Ge compositions to determine the other composition ranges to establ ish a stronger trend. 4.3.2.2 Agglomeration reaction Figure 4-18(b) and the results in Table 4-4 show approximate ly equal activation energies for the 15 and 25% Ge samples undergoing the aggl omeration reaction. A two-sample t-test using the appropriate data from Table 4-4 show s the nominal difference in activation energy to be 0.044 eV, with a 95% confidence interval for the difference ranging between -0.0054 and 0.0931 eV. It can be concluded, therefore, that no statistical difference between the reactions exists at the 95% confidence level. While this analysis is formed using somewhat limited data (three points for the 15% sample and four for the 25% sample), the nominal difference is small in relation to the error of the data. It is unlikely, therefore, that additional data points would cause enough of a shift in the linear regression to cause significant difference in th e activation energies. Thus, unlike for the precipitation reaction, this work found no statistically significant difference in activation energy for the agglom eration reaction between the 15% and 25% Ge samples. It can be concluded, therefore, that this reaction does not have a dependency on Ge concentration. The lack of concentration dependence, however, is not unreasonable. The agglomeration reaction does not re sult in any composition changes. Instead, it is likely driven solely by the minimization of free energy th rough geometric rearrang ement of the Ni(Si1-uGeu) grains (i.e. minimizing surface area to volume ra tio); a conclusion also supported by observation of the generally circular shape of the Ni(Si1-uGeu) grains after annealing at high times and temperatures. With consideration of the fact that the compositions of the Ni(Si1-uGeu) grains for both the 15 and 25% Ge samples are very close (l ess than ~1 at%) at the completion of the precipitation reaction (and hence th e start of the agglomeration reac tion), it is possible that the rearrangement is occurring by a mechan ism which occurs within the Ni(Si1-uGeu) grains. Thus,

PAGE 110

110 the mechanism would not be influenced by the extern al variations in Ge concentration. As with the precipitation reaction, however, not enough in formation about the Ni-Si-Ge system (mobility, diffusivity) is known to draw a firm conclusion as to the exact nature of the mechanism beyond the fact that it is not very temperature dependant as indicated by the small activation energy. 4.4 Summary A number of important conclusions can be drawn from the work presented in this chapter. First, the general cross-section and plan-view morphology of the nickel germanosilicide films in this work agree well with those reported in prior literature. Second, while prior work has generally described the film evolution as aggl omeration, two distinct reactions have been captured by the anneal matrix used in this work. Initially, Ge-rich Si1-zGez grains precipitate from the parent film as Ge is rejected to th e grain boundaries. This pr ocess continues until the parent film has completely transformed in to Ge-rich Si1-zGez and Si-rich Ni(Si1-uGeu) grains. Complete transformation corresponds to ~60% area fraction Ge-rich Si1-zGez for all samples in this work. Once the transformati on is complete, the Si-rich Ni(Si1-uGeu) grains begin to agglomerate. Third, increasing Ge content in the initial Si1-xGex film from 15 to 25% increases the amount of transformation observed for equiva lent anneals, likely due to the increase in activation energy for Ge rejection with increasing Ge content. Fourth, the addition of ~4.5E19 atoms/cm2 homogeneous B doping during Si1-xGex layer growth does not affect nickel germanosilicide film microstructure for any anne al. Finally, Avrami reaction kinetics can be used to model the isothermal transformation curves of nickel germanosilici de thin films. Two separate reaction orders were observed, corresponding to the precipitation and agglomeration reactions. By analysis of the reaction order, the precipitation reaction was determined to be diffusion controlled growth with a zero nucl eation rate. The agglomeration reaction was similarly determined to be diffusion controlled thickening of very large plates. The activation

PAGE 111

111 energies of the reactions were also able to be calculated and showed that for the precipitation reaction, Ea was 1.96 and 0.76 eV for the 15 and 25% Ge samples, respectively. For the agglomeration reaction, Ea was determined to be 0.14 and 0.10 eV, respectively (though statistically these values were found to be equal).

PAGE 112

112 Figure 4-1: On-axis <110> XTEM images of 25% Ge samples annealed at 450 oC for 10 minutes with (a) no doping and (b) B doping

PAGE 113

113 Figure 4-2: On-axis <110> images of 25% Ge samples with and without B doping annealed at550 oC for 10 or 1020 minutes.

PAGE 114

114 Figure 4-3: XTEM images taken at 10o tilt from <110> zone axis of undoped and doped 25% Ge samples RTA annealed at 600 oC for 60 seconds. The locations and results of EDS analysis are indicated on the image.

PAGE 115

115 1 m 1 m BSE SE Line Scan Figure 4-4: SEM images of undoped 25% Ge sample after 1020 mi nute anneal at 550 oC in SE and BSE modes. The location of the line scan and the resulting spectra are also shown. Note that the SE and BSE images show different regions of the same sample.

PAGE 116

116 Figure 4-5: SEM/BSE images of samples annealed at 450 oC

PAGE 117

117 Figure 4-6: SEM/BSE images of samples annealed at 550 oC

PAGE 118

118 Figure 4-7: SEM/BSE images of samples annealed at 650 oC

PAGE 119

119 Figure 4-8: SEM/BSE images of samples annealed at 750 oC

PAGE 120

120 Undoped 0 10 20 30 40 50 60 70 80 90 100 020040060080010001200 time (min)AF Si(1-z)Ge(z) 15% Ge, 450 C 15% Ge, 550 C 15% Ge, 650 C 25% Ge, 450 C 25% Ge, 550 C 25% Ge, 650 C Doped 0 10 20 30 40 50 60 70 80 90 100 020040060080010001200 time (min)AF Si(1-z)Ge(z) 15% Ge + B, 450 C 15% Ge + B, 550 C 15% Ge + B, 650 C 25% Ge + B, 450 C 25% Ge + B, 550 C 25% Ge + B, 650 C(a) (b) Figure 4-9: Area fraction of Ge-rich Si1-zGez phase as a function of anneal time at temperatures of 450, 500, and 650 oC for (a) undoped samples and (b) doped samples.

PAGE 121

121 Figure 4-10: Paired t-test result s (n=36) comparing the Ge-rich Si1-zGez area fraction of 15 and 25% Ge samples at all points on the thermal matrix imaged with SEM/BSE.

PAGE 122

122 15% Ge 0 10 20 30 40 50 60 70 80 90 100 020040060080010001200 time (min)AF Si(1-z)Ge(z) 15% Ge, 450 C 15% Ge, 550 C 15% Ge, 650 C 15% Ge + B, 450 C 15% Ge + B, 550 C 15% Ge + B, 650 C 25% Ge 0 10 20 30 40 50 60 70 80 90 100 020040060080010001200 time (min)AF Si(1-z)Ge(z) 25% Ge, 450 C 25% Ge, 550 C 25% Ge, 650 C 25% Ge + B, 450 C 25% Ge + B, 550 C 25% Ge + B, 650 C(a) (b) Figure 4-11: Area fraction of Ge-rich Si1-zGez phase as a function of anneal time at temperatures of 450, 500, and 650 oC for (a) 15% Ge samples and (b) 25% Ge samples.

PAGE 123

123 Figure 4-12: Paired t-test re sults (n=36) comparing the undoped and doped samples at all conditions imaged with SEM/BSE for (a) Ge-rich Si1-zGez area fraction and (b) Gerich Si1-zGez grain size.

PAGE 124

124 15% Ge -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 0.511.522.533.5 log(t)log(log(1/(1-AF))) 15% Ge, 450 C 15% Ge, 500 C 15% Ge, 550 C 15% Ge, 600 C 15% Ge, 650 C 15% Ge, 750 C 15% Ge + B, 450 C 15% Ge + B, 500 C 15% Ge + B, 550 C 15% Ge + B, 600 C 15% Ge + B, 650 C 15% Ge + B, 750 C Figure 4-13: Plot of log(log(1/(1-AF))) vs. log(t) for samples containing 15% Ge

PAGE 125

125 25% Ge -2 -1.8 -1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.511.522.533.5 log(t)log(log(1/(1-AF))) 25% Ge, 450 C 25% Ge, 500 C 25% Ge, 550 C 25% Ge, 600 C 25% Ge, 650 C 25% Ge, 750 C 25% Ge + B, 450 C 25% Ge + B, 500 C 25% Ge + B, 550 C 25% Ge + B, 600 C 25% Ge + B, 650 C 25% Ge + B, 750 C Figure 4-14: Plot of log(log(1/(1-AF))) vs. log(t) for samples containing 25% Ge

PAGE 126

126 0 20 40 60 80 100 01020304050607080 % AF SiGe% AF Ni(SiGe) 15% Ge 25% Ge 15% Ge + B 25% Ge + B Figure 4-15: Plot of area fraction of initial Ni(Si1-xGex) phase present as a function of Ge-rich Si1-zGez area fraction.

PAGE 127

127 15% Ge -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 0.511.522.533.5 log (t)log(log(1/(1-AF)) 15% Ge, 450 C 15% Ge, 500 C 15% Ge, 550 C 15% Ge, 600 C 15% Ge + B, 450 C 15% Ge + B, 500 C 15% Ge + B, 550 C 15% Ge + B, 600 C 25% Ge -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 0.511.522.533.5 log (t)log(log(1/(1-AF)) 25% Ge, 450 C 25% Ge, 500 C 25% Ge, 550 C 25% Ge + B, 450 C 25% Ge + B, 500 C 25% Ge + B, 550 C(a) (b) Figure 4-16: Plot of log(log(1/(1-AF))) vs. log(t) for Ge-rich Si1-zGez area fractions less than 60% (corresponding to precipitation reaction) for (a) 15% Ge and (b) 25% Ge samples.

PAGE 128

128 15% Ge -0.5 -0.45 -0.4 -0.35 -0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 0.511.5 22.533.5 log (t)log(log(1/(1-AF)) 15% Ge, 600 C 15% Ge, 650 C 15% Ge, 750 C 15% Ge + B, 600 C 15% Ge + B, 650 C 15% Ge + B, 750 C 25% Ge -0.5 -0.45 -0.4 -0.35 -0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 0.511.522.533.5 log (t)log(log(1/(1-AF)) 25% Ge, 550 C 25% Ge, 600 C 25% Ge, 650 C 25% Ge, 750 C 25% Ge + B, 550 C 25% Ge + B, 600 C 25% Ge + B, 650 C 25% Ge + B, 750 C(a) (b) Figure 4-17: Plot of log(log(1/(1-AF))) vs. log(t) for Ge-rich Si1-zGez area fractions greater than 60% (corresponding to agglomeration reac tion) for (a) 15% Ge and (b) 25% Ge samples.

PAGE 129

129 Activation Energy (Precipitation) -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 12131415161718 1/kTln(k) 15% Ge 25% Ge(a)1.96 eV 0.77 eV Activation Energy (Agglomeration) -1.5 -1.4 -1.3 -1.2 -1.1 -1 -0.9 -0.8 -0.7 -0.6 -0.5 101112131415 1/kTln(k) 15% Ge 25% Ge(b)0.10 eV 0.14 eV Figure 4-18: Plots of ln(k) vs. 1/(kT) for the (a) precipitation and (b) agglomeration reactions.

PAGE 130

130 Table 4-1: Linear regression resu lts for isothermal transformation curves plotted on a graphs of log(log(1/(1-AF))) vs. log(t) for the precipitation reaction. n standard error nlog(k) standard error log(k) 15% Ge, 450 C0.8340.093-4.0410.197 15% Ge, 500 C0.6480.106-2.8690.223 15% Ge, 550 C0.4790.072-1.8970.152 15% Ge, 600 C0.7410.020-1.8100.031 15% Ge (average)0.6760.073-2.6540.151 25% Ge, 450 C0.4260.048-2.0420.101 25% Ge, 500 C0.5090.050-1.8320.105 25% Ge, 550 C0.5370.065-1.3920.099 25% Ge (average)0.4910.054-1.7550.102

PAGE 131

131 Table 4-2: Values of n (reaction order) for reactions obeying Avrami kinetics [Chr02].

PAGE 132

132 Table 4-3: Linear regression results for isothermal transformation curves plotted on a graphs of log(log(1/(1-AF))) vs. log(t) for the agglomeration reaction n standard error nlog(k) standard error log(k) 15% Ge, 600 C0.0610.012-0.4850.030 15% Ge, 650 C0.0670.014-0.4480.029 15% Ge, 750 C0.0710.009-0.3660.020 15% Ge (average)0.0660.012-0.4330.026 25% Ge, 550 C0.0430.006-0.4280.016 25% Ge, 600 C0.0600.005-0.4420.012 25% Ge, 650 C0.0510.009-0.3770.019 25% Ge, 750 C0.0680.011-0.3230.024 25% Ge (average)0.0560.008-0.3920.017 Agglomeration

PAGE 133

133 Table 4-4: Activation energies derived from linear regression of series on the plot of ln(k) vs. (1/KT) Eas.e. Ea 15% Ge-1.9610.376 25% Ge-0.7650.186 Eas.e. Ea 15% Ge-0.1430.009 25% Ge-0.1000.029 Precipitation Agglomeration

PAGE 134

134 CHAPTER 5 RELATIONSHIP OF SHEET RESISTANCE AND MICROSTRUCTURE AND THE INFLUENCE OF GE AND B This chapter explores the influence of both Ge content and homogeneous in-situ B doping of the initial Si1-xGex layer on the sheet resistance of nickel germanosilicide thin films. Using the results from the quantitative microstructure anal ysis performed in the previous chapter, this chapter also investigates the structure/property relationship between sheet resistance and microstructure. The influence of Ge content and B doping on the relationship is also addressed. Prior work has widely utilized four point probe (4PP) sheet resistance measurements to characterize the electrical quality of nickel germanosilicide thin films. Isochronal sheet resistance measurements of undoped films, such as the one previously shown in Figure 2-19, have been reported by many researchers [Pey02, Zha02, Cha04, Cha04b, Zha04, Seg04, Liu05, Cho06, Ko06, Lau06, Yao06]. In each case, th e isochronal sheet resi stance curves show a relatively sharp increas e in value around ~700 oC. In the absence of a high-resistivity digermanosilicide (analogous to the NiSi2 phase), this increase is attributed to the agglomeration of the film. However, due to the limited th ermal matrix and common lack of quantitative microstructure analysis, any dire ct structure/property relationship between sheet resistance and the microstructure has not been established. Also of interest is determining how both Ge content and homogeneous in-situ B doping of the initial Si1-xGex layer may affect the structure/prope rty relationship. While it has been established in prior work [Ok03, Zh a04] and confirmed in the previ ous chapter of this work that increasing Ge content increases film agglomera tion, it is unknown whether Ge content would affect any microstructure/sheet resistance relationship. Prior work by Liu and Ozturk [Liu05] has shown that incorporation of high levels of in-situ doped B stabilizes sheet resistance. Their explanation for the cause of th e stabilization (a reduction in a gglomeration), however, was not

PAGE 135

135 seen in the B doped samples of this work. Thus it is of interest to first determine whether similar stabilization occurred in the samples in th is work and, if so, how any structure/property relationship between sheet resistance and microstructure may have been altered. To investigate the sheet resistance, structure/ property relationship, a nd influence of Ge and B, the nickel germanosilicide films were form ed at a range of thermal conditions. The Ge content in the initial Si1-xGex layer of the structures was varied at levels of 10 and 25 at% and the layers were undoped or inco rporated ~4.5E19 atoms/cm3 of homogeneous in-situ B doping. The samples were annealed in a quartz tube furnace under N2 ambient at temperatures between 450 and 800 C in increments of 50 C for 10, 30, 90, 270, and 1020 minutes. Once annealed, the sheet resistance of the samples was measured using a 4PP and the method previously presented in Chapter 3. It should be noted that, with the exceptions of the additional thermal series at 700 and 800 oC, the samples from the previous chapter were used for the analysis presented in this chapter. 5.1 Sheet Resistance Analysis The sheet resistance of all samples was m easured using the 4PP method outlined in Chapter 3. In keeping with prior work, the valu es of sheet resistance were plotted as a function of anneal temperature (isochronal series). It is important to not e, however, that all samples were not able to be successfully measured for all time s and temperatures. Inst ead, at higher values of temperature (generally over ~600 oC), some undoped samples were determined to be nonconductive. In this case, the plots of sheet resistance for those samples show a dashed line extrapolating to infinity after the last successful measurement of the isochronal series. The following sections discuss the sheet resistance m easurement results in terms of Ge and B content to determine the respective influences of each variable.

PAGE 136

136 5.1.1 Influence of Ge The sheet resistance measurements for the undoped 15 and 25% Ge samples are presented in Figure 5-1(a) as a function of anneal temperat ure (isochronal series). Only the 10, 90, and 1020 minute anneal series are shown for clarity. It is apparent from the plot that increasing Ge content caused earlier increases in sheet resistance for all anneal series. The difference between the sheet resistances at each point in the thermal matrix was also calculated by subtracting the sheet resistance value of the 15% Ge sample from that of the 25% Ge sample, shown in Figure 51(b) for the isochronal 10, 90, and 1020 minute se ries. Dotted lines in the figure represent extrapolation to infinite sheet resistance (i.e. the next point in the se ries was unable to be successfully measured). It is evident from this plot that increasing Ge content increased the sheet resistance of the film by 10s to 100s of Ohm/sq. The intermediate isochrona l series that were not shown in Figure 5-1 (30 and 270 min) displayed similar trends. The trend seen in these results agree well with those shown in prior work, wher e increasing Ge content was shown to result in poorer thermal stabili ty of sheet resistance [Ok03, Zha04]. Isochronal sheet resistance measurements for the doped 15 and 25% Ge samples are presented in Figure 5-2(a). As with the undope d case, only the 10, 90, and 1020 minute anneal series are shown for clarity. It is apparent fr om the plot that, similar to the undoped samples, increasing Ge content caused an increase in sheet re sistance for all anneal series. Figure 5-2(b) plots the differences between the sheet resistance values of the 25 and 15% Ge samples. It is evident from this plot that increasing Ge conten t increased the sheet resistance of the film by ~2 to ~70 Ohm/sq. In comparison with the undoped sa mples, these increases were generally smaller and more consistent across the thermal matrix. The intermediate isochronal series that were not shown in Figure 5-2 (30 and 270 min) displayed si milar trends. This experiment was the first time the effect of Ge content on samples with homogeneously B doped Si1-xGex layers was

PAGE 137

137 investigated; the work by Liu and Ozturk [Liu05] did not vary Ge c oncentration (though it did vary initial Ni layer thickness). It is important to observe, therefore, that the trend of increasing Ge content increasing sheet resistance seen in undoped samples is also present in the doped samples, though to a lesser degree. 5.1.2 Influence of B Previous work has shown that the ad dition of dopants to the initial Si1-xGex prior to silicidation may [Liu05] or may not [Cha04] affect the sheet resist ance of nickel germanosilicide thin films. Isochronal sheet resistance measurements for the undoped and doped 15% Ge samples in this work are presented in Figure 5-3(a). Only the 10, 90, and 1020 minute anneal series are shown for clarity. It is apparent fr om the plot that the B doped samples have a lower sheet resistance than the corresponding undoped samples at all times and temperatures. It is also apparent that the addition of B stabilized th e sheet resistance measurements to much higher temperatures; as previously shown, no doped samp le registered as non-conductive in this work. Figure 5-3(b) plots the differences between the sheet resist ance values of the doped and undoped 15% Ge samples. It is evident from this plot that the addition of B doping initially provides a small decrease in sheet resistance. However, once the sheet resistance of the undoped sample begins to rapidly increase the ad dition of B doping results in a larg e decrease of sheet resistance. The magnitude of the decrease was eventually equa l to infinity as the undoped samples read as open circuits and the doped samples maintained values less than ~120 Ohm/sq. The intermediate isochronal series that were not shown in Figure 5-2 (30 and 270 min) displayed similar trends. Isochronal sheet resistance measurements for the undoped and doped 25% Ge samples are presented in Figure 5-4(a) and their differences in (b). These results are very similar to those discussed for the undoped and doped 15% Ge samples with the exception that the stabilization begins earlier (due to the influe nce of Ge content). As with the 15% Ge samples, the addition of

PAGE 138

138 B stabilized the sheet resistance measurements to much higher temperatures. The difference between the sheet resistances of corresponding undoped and doped samples was also initially small and then rapidly grew, approaching infinity. These results confirm that both the 15 and 25% Ge structures used in this work show the stabilizing effect on sheet re sistance of the addition of hi gh levels of homogeneous B doping during Si1-xGex layer growth reported by Liu and Ozturk [Liu05]. While the level of B doping in this work (~4.5E19 atoms/cm3) is less than that used in their published work (~2.1E21 atoms/cm3), it is reasonable to assume that the sa me mechanism is causing the reduction and stabilization of sheet resistance for both works. Liu and Ozturk proposed that the cause for the effect was due to a reduction in agglomeration ca used by strain-reliving e ffects of the high levels of B. The previous chapter of this work, however, showed that no difference in sample microstructure or evolution was caused by the ad dition of B doping. Thus, the effects of B on sheet resistance cannot be expl ained by its influence on film microstructure. A different proposed mechanism, and the influence of B doping on any sheet resi stance/microstructure relationship, will be discussed in the following sections. 5.2 Sheet Resistance/Microstructure Relationship Apart from the specific influences of Ge and B content, Figures 5-1 to 5-4 show a general trend of increasing thermal annealing causing an increase in sheet resistance (to a greater or lesser degree. dependent on Ge and B content). Si milar trends have been extensively reported in literature and are generally ascribed to the worsening of f ilm morphology with increasing anneal time and temperature (e.g. agglomeration) [P ey02, Zha02, Cha04, Cha04b, Zha04, Seg04, Liu05, Cho06, Ko06, Lau06, Yao06]. In genera l, however, prior lite rature has lacked quantitative analysis of film morphology in plan view, as well as having been confined to a single isochronal series. These limitations ha ve precluded establishment of any direct,

PAGE 139

139 quantitative structure/property relationship between film mo rphology and sheet resistance. Instead, only a circumstantial, qualitative relationsh ip has been established. This work, however, performed quantitative microstructure analys es of a large number of anneal times and temperatures. These results, presented in the previous chapter, can be used in conjunction with the corresponding sheet resistance measurements presented in the preced ing sections of this chapter to investigate the structure/property rela tionship between film microstructure and sheet resistance in a quantitative manner. Film microstructures in Chapter 4 were quantified using three metrics: Ge-rich Si1-zGez grain size, Si-rich Ni(Si1-uGeu) area fraction, and Ge-rich Si1-zGez area fraction. Each of these three metrics was evaluated for suitability in establishing a structure/property relationship by plotting the sheet resistance of each sample as a function of the metric. Figure 5-5 presents the result of plotting sheet resist ance as a function of Ge-rich Si1-zGez grain size for all four structure types. Analysis of the plot shows that data from each sample type is widely scattered and intermixed with the other types, no discernable patterns are apparent. It is evident from this figure that Ge-rich Si1-zGez grain size is not a suitable metric for establishing a quantitative relationship. Next, sample sheet resistance was plotted as a function of Si-rich Ni(Si1-uGeu) area fraction, shown in Figure 5-6(a). While this plot shows stronger tr ends than Figure 5-5, consistent relationships are not readily observed. Finally, sa mple sheet resistance was plotted as a function of Ge-rich Si1-zGez area fraction, shown in Figure 5-6( b). This plot shows strong, consistent trends for all sample series which ar e clearly superior to t hose generated using the other metrics. It was determined, therefore, that Ge-rich Si1-zGez area fraction was the best metric for use in establishing a structure/property rela tionship. It should be noted that Figure 56(b), and the plots derived from it throughout this chapter, is a plot of two experimentally

PAGE 140

140 measured parameters. Accordingly, each data point has error in both the x (Ge-rich Si1-zGez area fraction) and y (sheet resistance) axes of the plot. For the x-axis, error remains as discussed previously and in Appendix A, with a 95% confid ence interval of 3.35%. The y-axis error, as determined in Appendix B, however, is narrow enough in range (~2.2 Ohm/sq for a 95% confidence interval) in comparison to the magnitude of the measurements that the error bars are not visible on the plot. It is clearly evident from Fi gure 5-6(b) that a structure/pr operty relationship between Gerich Si1-zGez area fraction and film sheet resistance can be established. Both area fraction and sheet resistance are area properties. Hence, it coul d be expected that the sheet resistance of an anisotropic, non-textured two-phase microstructure (such as that evident in the SEM/BSE images of Figures 4-5 to 4-8) would vary linearly betwee n the values of each phase. This relationship, however, is clearly not shown by the series in Figure 5-6(b). The discrepancy can be further illustrated by Figure 5-7(a), which shows the experimental relationship for the doped 15% Ge samples versus a linear interpolation betw een the values for the homogeneous nickel germanosilicide film (~10 Ohm/sq) and the B-doped Si1-xGex layer (~120 Ohm/sq). While, on initial observation, it may be attractive to model the microstructure/sheet resistance relationships for the sample series as exponential functions the results from the microstructure analysis performed in Chapter 4 sh ould be considered first. It was previously shown in Figure 4-15 that the ar ea fraction of the initial Ni(Si1-xGex) phase decreases with increasing Ge-rich Si1-zGez area fraction. If this same plot is added on a secondary axis to a plot of the structure/property relationships, as is shown in Figur e 5-7(b), it can be seen that the inflection point in the relationship corresponds to the disappearance of the initial Ni(Si1-xGex) phase (at ~60% area fraction of Ge-rich Si1-zGez) The analysis in Chapter 4 also showed that it

PAGE 141

141 is at this point that the morphological evolution of the film should be se parated into two stages: precipitation and agglomeration. With this consideration, it is apparent that the structure/property relationships should be modele d as two sets of linear functions, one for each of the stages of microstructure evolution. These two stages, cal led Stage I for the precipitation reaction and Stage II for the agglomeration re action, are shown in Figure 5-7(b). These structure/property relationships, and the influence of Ge and B on them, will be discussed in the following sections. 5.2.1 Stage I (Precipitation) Figure 5-8 presents the sheet resistance of th e sample series as a function of Ge-rich Si1zGez area fraction for area fractions less than 60%, corresponding to the preci pitation stage of the microstructure evolution process. It is evident from this figure th at the distribution of data points is not even, but skewed towards lower area fractions (less than ~25%). This is a result of the anneal sequences not being tailored to produce ev enly spaced steps in area fractions, as the necessary rate calculations to engineer such a spacing were not availa ble. It is also evident from the figure that there appears to be more spread in the data as the area fraction increases. Both of these observations will become important in the fo llowing sections as the in fluences of Ge and B are considered. 5.2.1.1 Influence of Ge The influence of initial Si1-xGex layer Ge content on the struct ure/property relationship of the undoped samples was determined by plot ting the area fraction of Ge-rich Si1-zGez area fraction versus sheet resistance, s hown in Figure 5-9(a). Linear regr ession analysis of the series was also performed and the best fit lines show n on the plot. The linear regression analysis provided a best fit line with an R2 value of 0.7454 and 0.8431 for the 15 and 25% Ge samples, respectively. The regression results including slope, slope standard error, intercept, and intercept

PAGE 142

142 standard error, are presented in Table 5-1. These results, in combination with the R2 values, indicate that the accuracy of the linear regressions for the undoped samples are hampered by the relatively few data points and large scatter above ~25% Si1-zGez area fraction, especially for the 15% Ge samples. Nevertheless, it can be concluded that increasing Ge-rich Si1-zGez area fraction causes an increase in sheet resistance. Figure 5-9(b) presents a plot of the area fraction of Ge-rich Si1-zGez area fraction versus sheet resistance for the doped 15 and 25% Ge samples. As with the undoped samples, linear regression analysis was performed and the best fit line shown for each series, with R2 values of 0.8663 and 0.8423 for the 15 and 25% Ge series, resp ectively. The numeric al regression results for these lines are also given in Table 5-1. Th ese results indicate that the doped samples have a stronger linear trend than their undoped counterparts. Less di sparity between the regression lines for the two series is also qualitatively apparent for the doped samples when compared to the undoped ones. As with the undoped series, however, th e accuracy of the result s is still hampered by the relative scarcity of data between ~25 and 60% Ge-rich Si1-zGez area fraction, especially for the 15% Ge samples. The values for the regression intercepts and their 95% confidence intervals are shown in Figure 5-10(a) and for each of the four sample se ries. Using 2-sample t-tests, it can be shown that no statistically significant differences in intercept value exist at the 95% confidence level between any of the sample series. This result wa s expected, as at low times and temperatures all samples exhibited a homogeneous continuous film of Ni(Si1-x)Gex. It would be expected, therefore, that the inte rcept value of each plot of area fraction Ge-rich Si1-xGex vs. sheet resistance would be equal to each other and the sheet resistance of this initial film (where there is 0% area fraction Si1-zGez). The average intercept value of ~8 +/3 Ohm/sq does in fact agree

PAGE 143

143 well with the previously reported values for a such a film which have usually averaged between 5 and 12 Ohm/sq dependent upon initial Ni laye r thickness, etc [Ok03, Cha04, Zha04, Liu05]. The values for the regression slopes and thei r 95% confidence intervals are shown in Figure 5-10(b) and for each of the four sample se ries. While 2-sample t-tests cannot statistically prove that the slopes for the 15 and 25% Ge samp les are equivalent, this is likely due to the uncertainty introduced by the afor ementioned scatter and lack of data at ar ea fractions between 25 and 60%, especially for the 15% Ge samples. Qualitatively, however, the slope regression results for the 15 and 25% Ge series appear to be fairly similar for either the undoped or doped samples. These observations are further reinfo rced by observation of Figure 5-9(a) and (b) where the data points for both the 15 and 25% Ge samples appear to be well-interspersed for both the undoped and doped samples. It can be tentatively concluded, therefore, that increasing Ge content of the initial Si1-xGex layer does not affect the struct ure/property relationship between precipitated Ge-rich Si1-zGez area fraction and sheet resistance. This conclusion, however, requires further experimentation to provide statistically significan t support; the needed experimentation will be disc ussed in a later section. 5.2.1.2 Influence of B Figure 5-11 presents a plot of film sheet resistance vs. Ge-rich Si1-zGez area fraction for all four sample series. The analysis in the previ ous section showed that it can be tentatively concluded that varying Ge content in undoped and doped nickel ge rmanosilicide thin films does not affect the relationship be tween film morphology and sheet resistance. This section, therefore, grouped together the data from th e 15 and 25% Ge samples when considering the influences of homogeneous B doping of the initial Si1-xGex layer on the structure/property relationship. Linear regression analysis wa s accordingly performed on the grouped 15 and 25% Ge samples for the undoped and doped cases and th e best fit regression lines shown on Figure 5-

PAGE 144

144 11. The numerical regression results are include d in Table 5-2 and the regression intercepts, slopes, and respective 95% confiden ce intervals of the lines are presented in Figure 5-12(a) and (b). Comparison of the regression intercept values wi th a 2-sample t-test shows that there is no difference between the undoped and doped samples at a 95% confidence level. As in the previous section, the average intercept value of 9.7 +/2 Ohm/sq agrees well with prior literature. Comparison of the regression slope va lues, however, shows that there is a significant difference between the undoped and doped sample s at a 95% confidence level with the doped samples having a slope approximately ~60% le ss than the undoped samples. It can be concluded, therefore, that th e addition of homogeneous B doping influences the fundamental structure/property relationship be tween film microstructure and sh eet resistance. The cause for this influence is likely the addition of a more conductive current path provided by the B doped samples. The specific nature and location of this conductive path will be addressed separately in a later section. 5.2.1.3 Tortuosity Analysis The previous sections have established that the sheet resistance of nickel germanosilicide thin films increases approximately linearly with increasing Si1-zGez area fraction within this stage of the overall microstructure evolution (less than 60% AF Si1-zGez). Figure 5-7(a) has previously shown, however, that the effect cannot be desc ribed by a liner interpolation between the sheet resistance of the initial film and the substrate. It is of interest, therefore, to determine the if the underlying cause for the relati onship can be determined. When considering possible mechanisms, it is important to remember that the defining characteristic of the first stage of the microstructure evolution is the presence of initial Ni(Si1xGex) phase which has not yet fully rejected Ge and transformed into Si-rich Ni(Si1-uGeu) and

PAGE 145

145 Ge-rich Si1-zGez. Thus, it can be qualitatively observed in the SEM/BSE images that within this stage, the highly conductive Ni(Si1-xGex) and Si-rich Ni(Si1-uGeu) phases form a continuous network which is disrupted by grains of relatively less conductive Ge-rich Si1-zGez. This observation leads to the suggestion that the cond uctive phases may be carrying the majority of the current and that the less conductive Ge-rich Si1-zGez grains are disrupting the current path, causing an increase in sheet resistance with incr easing area fraction (and hence interference) of Si1-zGez. It was of interest, therefore, to determin e how the conductive path tortuosity and the area fraction of Si1-zGez were related. Accordingly, the tortuosity of selected doped and undoped 25% Ge samples was calculated according to the method described in Chapter 3. Figure 5-13 shows a example image and estimated continuou s conductive path across the sample. The samples used in this analysis, and their Si1-zGez area fraction, tortuosity, and sheet resistance are presented in Table 5-3. The samples were select ed such that the covere d the entire range of Si1zGez area fractions within the first stage of microstruc ture transformation. It is important to note that, due to the requirement of manual calculation, sample to rtuosity was calculated using the average of three random measurements of a sing le image for each sample condition. The small sample size and relatively inaccurate calculati on (when compared to computer modeling), therefore, resulted in substantial error in the tortuosity measurement. The average tortuosity range for the three measurements of each image was determined to be 0.0333. The measurement error, therefore, was estimated by dividing this value by the square root of the sample size (3) according to the work of Kyker [Kyk83] which showed that for normal distributions, this calculation would estimate the true standard de viation within 10-15%. For the tortuosity

PAGE 146

146 calculations in this work, the resulting 95% confidence interval based on the error estimation proposed by Kyker was determined to be +/0.0377. Figure 5-14(a) presents a plot of conductive path tortuosity vs. Si1-zGez area fraction for the selected undoped and doped 25% Ge samples. It is evident from this plot that conductive path tortuosity increases linearly w ith increasing area fraction of Si1-zGez. It is also apparent without statistical analysis that B doping does not influence the relationship between the two properties. A single linear regression was therefore performed for all data points and the best fit regression line and equation shown in the figure. While the tortuosity of the 15% Ge samples was not studied, the same relationship is expected due to the similarity in microstructure evolution observed in the previous chapter. Additional work should be performed, however, to confirm this expectation and strengthen the observed re lationship by reducing the error in tortuosity measurement. The direct relationship between tortuosity and sheet resistance was also explored; the sheet resistance of the undoped and doped 25% Ge samples was plotted as a function of conductive path tortuosity and is presented in Figure 5-14(b). While observa tion of this plot shows that there is a distinct relationship between the two pr operties, it is not evident (due to the few data points and large error) either thr ough qualitative or statistical analysis whether the relationship is of an exponential nature or linear relationship. The R2 values for an expone ntial relationship are 0.864 and 0.874 for the undoped and doped 25% Ge samples, respectively. For a linear relationship, the values are 0.830 and 0.817, respec tively. Regardless, the stabilization caused by B doping is clearly eviden t in the figure as the doped samp les have smaller sheet resistance values than the undoped samples of equal tortuosity. While additional analysis is necessary to better establish the relationship, these results support the theory that the mechanism behind the

PAGE 147

147 structure/property relationship in this stage of the films microstructural evolution is likely related to the disrup tion of the conductive path through the Ni(Si1-xGex) and Si-rich Ni(Si1-uGeu) grains by the precipita tion of less conductive Si1-zGez. It is also probable, therefore, that the addition of B doping reduces the magnitude of this disruption by providing an alternative conduction path; as mentioned before, the nature and location of this path will be addressed in a later section. It is also im portant to note that the linear relationship between conductive path tortuosity and Si1-zGez area fraction suggests that the properties may be co-dependant variables. If so, conductive path tortuosity may be a be tter predictor of sh eet resistance than Si1-zGez area fraction. 5.2.2 Stage II (Agglomeration) The structure/property relations hip between film microstructu re and sheet resistance was also considered within the agglomeration stage of the film transformation. The samples in this stage had Si1-zGez area fractions above 60% and no sample in the study exhibited an area fraction greater than 78.16%. It is e xpected, however, that samples w ith significantly greater area fractions than this value will not occur, as while Ni(Si1-uGeu) grains are agglomerating in this stage to reduce their interface energies resul ting in the deepening of grains and through conservation of mass an associated reduction in pr ojected size, eventually an equilibrium will be reached in this process. More simply put, the area fraction of Si1-zGez will never reach 100% as long as grains of Ni(Si1-uGeu) are present in the film. Thus the analyses in the following sections, which consider the influences of Ge and B, are likely representative of the entire reaction stage and not just a single segment or subset thereof. 5.2.2.1 Influence of Ge Figure 5-15(a) presents a pl ot of the sheet resistance of the undoped 15 and 25% Ge samples as a function of Si1-xGez area fraction. Error ranges are omitted for clarity. The data in

PAGE 148

148 this plot, which is shown on a loga rithmic plot due to the substantia l disparity in sheet resistance measurements, shows a wide degree of scatter fo r the 25% Ge samples. Additionally, only two data points are included for the 15% Ge samples. The scarcity of data may be attributed to the fact that the sheet resistance of only two of the ten 15% Ge samples with Si1-xGez area fractions contained in this stage of the reaction was able to be successfully measured. Likewise, only ten of the seventeen 25% Ge samples were able to be measured. The cause for this difficulty will be addressed in the following section. It should also be noted that the analysis in Appendix B showed that 4PP measurement of sheet resist ances above ~500 Ohm/sq showed much greater variability than those at lower values. Acco rdingly, no influence of Ge content may be determined for the undoped samples. Figure 5-15(b) presents a plot of the sheet resistance of the doped 15 and 25% Ge samples as a function of Si1-xGez area faction with error bars omitted fo r clarity. It is evident from this plot that the sheet resistance of the doped sample s in this range has a linear relationship with Si1xGez area fraction. Accordingly, linear regression of the data series was performed; the best fit lines are shown in the figure and the regression results given in Table 5-4. Analysis of the regression results using a 2 sample t-test shows that the slope and intercept of the best fit lines for the doped 15 and 25% Ge are not equal at a 95 % confidence level. Th us, it can be concluded that, unlike within the first transformation stage, increasing Ge content in this stage affects the structure/property relationship. 5.2.2.2 Influence of B Figure 5-16 presents a plot of the sheet resistance of all four sample series as a function of Si1-zGez area fraction. It should be noted that a log scale was used for the y-axis of the plot (sheet resistance) to display all data points within reas onable proximity and that error bars were omitted for clarity. Observation of this figure shows that the doped samples have much smaller

PAGE 149

149 sheet resistances than thei r undoped counterparts for all Si1-zGez area fractions, as well as a much stronger structure/property relationship. Thus, the addition of B doping has a significant effect on the relationship between microstr ucture and sheet resistance. The cause for the significant stabilization of sheet resistance by the addition of B doping within this stage is clarified by considering representative images of the film. Figure 5-17(a) and (b) show an image of a doped 25% Ge sample after annealing at 650 oC for 10 minutes before and after application of a contrast threshold, respectivel y. This sample has a Si1-zGez area fraction of 64%, at the low end of the range with in this stage. Figure 5-17(b) and (c) show an image of a doped 25% Ge sample after annealing at 750 oC for 10 minutes before and after application of a contrast threshold, respectively. This sample has a Si1-zGez area fraction of 73.7%, at the high end of the range w ithin this stage. Both samples show that within this stage of the transformation process, individual grains of Si-rich Ni(Si1-uGeu) are isolated from each other by regions of Si1-zGez. For the undoped samples, no highly conductive current path is available between these islands. Hence, it is extremely di fficulty for current to pass through the film and the measured sheet resistance is very high or the sample reads as an open circuit. Conversely, for the B doped samples, a conductive current path between Si-rich Ni(Si1-uGeu) grains is likely available through a B doped region. This leads to the much lower sheet resistances measured for these samples. Unlike the precipitation stage, de termination of the conduction pa th of least resistance for the doped samples is complicated by the fact that the path will, up to a certain island separation, likely utilize the low resistivity Ni(Si1-uGeu) grains for portions of the path (i.e. jump between Ni(Si1-uGeu) grains). However, as area fraction of Si1-zGez increases with increasing agglomeration of the Ni(Si1-uGeu) grains, the distance between Ni(Si1-uGeu) grains increases.

PAGE 150

150 This effect is shown in the SEM/BSE images of the microstructure; the sample in Figure 5-17(a) with area fraction of 64% is cl early less agglomerated than the sample in Figure 5-17(c) with area fraction of 73.7%. The increasing sepa ration between highly conductive grains will therefore cause the observed increase in sheet resistance with Si1-zGez area fraction as an increasing percentage of the total conduc tion path is spent between the Ni(Si1-uGeu) grains. Eventually, this path may become so tortuous th at it becomes favorable for it to pass directly through the B doped regions and intersect only the Ni(Si1-uGeu) grains which happen to lie on a line directly along the chor d of the path (i.e. directly between the terminals of a 4PP). If this occurs, it would be expected that the plot of sheet resistance vs. Si1-zGez area fraction would show an inflection point as the relations hip would become relatively insensitive to agglomeration. While no such point is visibl e in Figure 5-15(b), it is possible that the Si1-zGez area fractions obtained in this work are not larg e enough. Additional work should be performed, therefore, to see if such a point exis ts at area fractions greater than 75%. 5.2.3 Boron Conduction Path Prior sections of this chapter have suggest ed that doped regions of the experimental structures containing homogene ous B doping are providing an alte rnative conduction path for current flow, thereby reducing the sheet resistance of the film. In some cases, such as during the precipitation reaction, the amount of reduction is relatively low. In other cases, such as during the agglomeration reaction, the amount of reducti on is large (even approaching infinity). The nature and location of the alternative path however, has not yet been addressed. Published literature has show n that full silicidation (FUS I) of doped polysilicon gates using Ni metal can lead to improved device performance through tuning of the gate work function and reduction of gate tunneling effects [Mas05]. The improvements are attributed to silicidation induced dopant segregation leading to high concentr ations of dopants in proximity to

PAGE 151

151 the gate oxide. Specifically, literature has shown that nickel silicidation of a B implanted Si sample past the initial implant depth results in th e pile-up of B at the silicide/silicon interface. Zhang et al. performed an experi ment where a 2 keV 3E15 atoms/cm2 B implant was performed into bulk silicon resulting in an initial peak con centration 10 nm below the sample surface. After deposition of excess Ni metal, the im planted samples were annealed at 500 oC for 90 seconds under a N2 atmosphere. The samples were then analyzed using SIMS; the results are shown in Figure 5-18. It is evident from this figure that silicidation induced dopant segregation has moved the peak dopant concentration from a depth of ~10 nm to ~80 nm leaving only residual B doping of the silicide. Silicidation induced dopant segreg ation has not been studied in the nickel germanosilicide system. It is plausible, however, to expect that similar segregation occurs in this system as well. To determine if any dopant segregation was obser ved in the B doped samples in this work, SIMS analysis of the doped 25% Ge sample annealed at 450 oC for 10 minutes was performed. This sample was selected for analysis as it has th e most homogeneous layer compositions (beneficial for SIMS analysis) and any sili cidation induced dopant segregati on would have already occurred (as the segregation would occur dur ing the formation of the silicide layer). An XTEM image of the sample is presented in Figure 5-19(a) and the SIMS results are s hown in Figure 5-19(b). While evidence of dopant segregation is apparent in the SIMS results, due to the interface roughness observed in the XTEM image the SIMS results are of overall poor quality (as evidenced by the large concentration tails). T hus, another technique less susceptible to interface effects should be used to confirm th e presence of dopant segregation. 5.2.3.1 Potential Conduction Paths Assuming that silicidation induced dopant segregation does occur for the B doped samples in this work, three potential conduction paths can be proposed for the structure. As shown in

PAGE 152

152 Figure 5-20, silicidation induced dopant segregation would resu lt in a residually doped nickel germanosilicide silicide layer, the first possible conduction path. Assuming that the segregated B doping is fully electrically active (which is conservative, as it the mechanism by which segregation occurs is not understood), the re gion with the segregated dopant would form a second path. Finally, the unreacted, actively doped Si1-xGex layer will offer a third conduction path. The potential sheet resistance of each of the three potential conduction paths was estimated for the doped 15% Ge sample. The sheet resistan ce of the doped 15% Ge sample annealed at 450 oC for 10 minutes was measured to be ~ 10 Ohm/sq. This value is accordingly approximately equal to the sheet resistance of a 20 nm homogeneous nickel germanosilicide layer. A wafer without deposited Ni was also available for this experimental structure; 4PP measurement of this wafers 150 nm B doped Si0.85Ge0.15 layer showed that the sheet resistance of the layer was 120 Ohm/sq. Sheet resistance (Rs) and sample resistivity ( ) are related according to Equation 5-1 when the ratio of samp le length (L) to width (W) is equal to 1. )(W L t Rs Equation 5-1 Equation 5-1 shows, therefore, th at dividing the sheet resistance of the layer by its thickness (150 nm) allowed the resistivity of the layer to be calculated as 1.8E-5 Ohm*m. As the resistivity of a semiconductor is inversely proportional to its le vel of doping, assuming that all segregated dopant from the ~20 nm germanosilicide layer is contained within a 10 nm layer below the interface, the layer containing the pileup of se gregated dopant would have three times the nominal dopant concentration and th us an estimated resistivity of 6E-6 Ohm*m (one-third of the nominal resistivity). The estimated sheet resist ance of the region with the pileup of segregated dopant can accordingly be calculated as ~600 Ohm/ sq. The estimated sheet resistance of the

PAGE 153

153 remaining 120 nm layer of unreacted, nominally B doped Si0.85Ge0.15 layer would then be equal to 150 Ohm/sq (due to its reduced thickness). It should also be noted that if the dopant pileup caused by silicidation induced dopant segregation is not electrica lly active, the expected sheet resistance of the remaining 130 nm of nominally B doped Si0.85Ge0.15 would have a sheet resistance of 140 Ohm/sq. Comparison of the calculated sheet resistance va lues for the 10 nm region with the pileup of segregated B dopant and the remaini ng 120 nm of unreacted, nominally B doped Si0.85Ge0.15 layer indicates that, while the concentration of active dopant in the region with pileup is potentially 3 times higher than that of the unreac ted, nominally doped layer, the differences in layer thicknesses cause the sheet resistance of the region with pileup to be significantly higher (600 Ohm/sq. vs. 150 Ohm/sq, respectively). When considered as resistors in parallel, however, the combined resistance of the two layers de creases to 120 Ohm/sq according to Equation 5-2: 2111 1 RR Rparallel Equation 5-2 Thus, if pileup of active dopant does occur due to silicidation i nduced dopant segregation, the measured sheet resistance of the non-germanosili cide layers would be expected to be ~120 Ohm/sq. If pileup does not resu lt in increased amounts of activ e dopant, the measured sheet resistance of the non-germanosilicide layers would be expected to be ~140 Ohm/sq (equal to the sheet resistance of 130 nm of nominally dopes Si0.85Ge0.15). 5.2.3.2 Evaluation of Potential Paths The previous section calculated the expected sheet resistance values of several potential conduction paths. As shown in Figure 5-20, it is expected that silici dation induced dopant segregation leads to the pileup of B dopant at the nickel germanosilicide/Si1-xGex interface for

PAGE 154

154 the B doped samples in this work, leaving the in itially homogeneous germ anosilicide film with low levels of residual doping. As B diffusivity has been shown to be very low at the times and temperatures used in this work [Rad06], the Si1-zGez grains which precipitate from the germanosilicide are thus likely only lightly doped, at best. It is expected, therefore, that the initial low resistance path provide d by the germanosilicide layer will be significantly degraded with increasing anneal time and temperature. Unlike the undoped samples, however, the doped samples in this work exhibited lower, more st able sheet resistance measurements. Thus, an alternative conduction path must be provided by th e B doped regions. Expected sheet resistance values for two possible conducti on paths (10 nm highly doped Si1-xGex plus 120 nm of nominally doped Sii1-xGex in parallel (120 Ohm/sq), or 130 nm of nominally doped Si1-xGex(150 Ohm/sq)) were calculated in the previous section for the doped 15% Ge samples. By comparing the estimated values to experimental results, the more likely path may be determined. Figure 5-21 presents a plot of the sheet re sistance of the doped 15% Ge samples for isochronal anneal series of 10, 90, and 1020 minut es for temperatures ra nging from 450 to 800 oC. The calculated sheet resistance values for the homogeneous nickel germanosilicide film and the two possible conduction paths are also presented in the figure. It is evid ent from the plot that the sheet resistance of the samples at low times and temperatures is equal to that of the homogeneous nickel germanosilicide film. As anneal time and temperature increases, the sheet resistance of the samples increases past 120 Oh m/sq and begins to asymptote at a value around 140 Ohm/sq. These results indicate that any pi led up dopant is not electrically active and the most likely alternative conduction path is through 130 nm of nominally doped Si0.85Ge0.15. In context of previous work, the determination that the dopant pileup is likely electrically inactive can explain why the work of Chamiria n et al. [Cha04] found no effect of implanted

PAGE 155

155 dopants on film sheet resistance. Silicidation in duced dopant segregation likely resulted in the pileup of inactive dopants at the germanosilicide/Si1-xGex interface. Unlike this work or the work of Liu and Ozturk [Liu05], the Si1-xGex layer in the study by Chamirian et al. was not doped. Accordingly, no alternate, actively dope d conduction path was ava ilable in the latter work and the sheet resistan ce of the films behaved iden tically to the undoped case. It should be noted that these conclusions sugge st that the influence of Ge content on the structure/property relationship between film microstructure a nd sheet resistance during the agglomeration reaction are most likely due to differences in th e resistivity of the underlying Si1xGex film due to differences in B dopant incorpora tion or activation with increasing Ge content. As this path is a major contributor to overall sample sheet resistance for this stage of the film transformation, a significant difference in sheet resistance with Ge content was seen. For the precipitation stage of the reaction, however, the c ontribution of the alternative conduction path to sample sheet resistance is much smaller, accord ingly any difference in the resistance of the alternative path was negligible and no influence of Ge content was seen. The resistivity of the doped Si0.75Ge0.25 layer was unable to be determined as no structure without Ni and TiN deposition was available for this sample conditions Efforts to chemically or physically remove the Ni and TiN layers without damage to the doped Si0.75Ge0.25 layer were attempted but were unsuccessful. 5.3 Summary Several important conclusions can be drawn from the work presented in this chapter. First, as previously shown in literature, incr easing the Ge content of the initial Si1-xGex layer was found to worsen sheet resistance. Second, as s hown by Liu and Ozturk [Liu05], the addition of homogeneous B doping to the initial Si1-xGex layer resulted in the decrease and stabilization of sheet resistance values for both the 15 and 25% Ge samples used in this work. Third, a

PAGE 156

156 quantitative structure/property re lationship can be esta blished between area fraction of Ge-rich Si1-zGez and sheet resistance for both the precipitati on and agglomeration stages of the films microstructure evolution. For the precipitation stage, Ge composition was not determined to influence the structure/property relationship for either the undoped or doped samples. The addition of homogeneous B doping, however, was f ound to significantly affect the relationship. Overall, the general increase in sh eet resistance for all samples in this stage of the transformation can be attributed to the in creasing tortuosity of the conduc tion path with increasing Si1-zGez precipitation. The decreased sheet resistance of the B doped samples in this stage is likely due to the samples containing an alternative conduction path that reduces the influence of the increasing tortuosity. For the agglomeration stage of the fi lm transformation, the sheet resistance of the undoped samples was found to rapidly increase to la rge (or infinite) valu es unrelated to the amount of Si1-zGez present. The influence of Ge cont ent on the structure/property relationship for these samples was therefore unable to be conclusively determined. This effect was determined to be due to the absence of a con tinuous conduction path between isolated Si-rich Ni(Si1-uGeu) grains for these sample. For the doped samples in the agglomeration stage, the sheet resistance of the samples was found to increa se linearly with increasing area fraction of Si1-zGez, resulting in a predicative stru cture/property relationship for th ese samples. These results suggested that an alterna tive conduction path between the isolated Si-rich Ni(Si1-uGeu) was provided by B doped regions. It was also determin ed that increasing init ial Ge concentration influenced the relationship, with the 25% Ge samp les having higher sheet resistance values than the corresponding 15% Ge samples. Fourth, the alternative conduction path for the B doped samples was determined to most likely be the unreacted, nominally B doped Si1-xGex layer below the nickel germanosilicide layer. It was also determined that differences in the sheet resistance

PAGE 157

157 of the Si1-xGex layer with Ge content is the most likely cause for the influence of Ge noted in the agglomeration stage of the film transformation.

PAGE 158

158 Undoped (15 and 25% Ge) 0 250 500 750 1000 1250 1500 450500550600650700750800 Temperature (C)Sheet Rho (Ohm/sq ) 15% Ge, 10 min 15% Ge, 90 min 15% Ge, 1020 min 25% Ge, 10 min 25% Ge, 90 min 25% Ge, 1020 min(a) Difference (25% Ge 15% Ge) 0 100 200 300 400 500 450500550600650700750800 Temperature (C)Sheet Rho (Ohm/sq ) 10 min 90 min 1020 min(b) Figure 5-1: Isochronal anneal series of (a) sheet resistance of undoped 15 and 25% Ge samples and (b) difference in sheet resistance betw een 25 and 15% Ge samples. Note that dotted lines are extrapolations to infinity indicating that subsequent samples in the series read as open circuits.

PAGE 159

159 Doped (15 and 25% Ge) 0 25 50 75 100 125 150 175 200 225 450500550600650700750800 Temperature (C)Sheet Rho (Ohm/sq ) 15% Ge + B, 10 min 15% Ge + B, 90 min 15% Ge + B, 1020 min 25% Ge + B, 10 min 25% Ge + B, 90 min 25% Ge + B, 1020 min Difference (25% 15% Ge) 0 20 40 60 80 450500550600650700750800 Temperature (C)Sheet Rho (Ohm/sq ) 10 min 90 min 1020 min(a) (b) Figure 5-2: Isochronal anneal series of (a) sheet resistance of doped 15 and 25% Ge samples and (b) difference in sheet resistance be tween 25 and 15% Ge samples.

PAGE 160

160 15% Ge (Undoped and Doped) 0 50 100 150 200 250 450500550600650700750800 Temperature (C)Sheet Rho (Ohm/sq ) 15% Ge, 10 min 15% Ge, 90 min 15% Ge, 1020 min 15% Ge + B, 10 min 15% Ge + B, 90 min 15% Ge + B, 1020 min 15% Ge Difference (Doped Undoped) 0 20 40 60 80 100 120 450500550600650700750800 Temperature (C)Sheet Rho (Ohm/sq ) 10 min 90 min 1020 min(b) (a) Figure 5-3: Isochronal anneal series of (a) sheet resist ance of undoped and doped 15% Ge samples and (b) difference in sheet resistance between the doped and undoped samples. Note that dotted lines are extrapolations to infinity indicating that subsequent samples in the series read as ope n circuits or that an infinite difference was calculated.

PAGE 161

161 25% Ge (Undoped and Doped) 0 50 100 150 200 250 450500550600650700750800 Temperature (C)Sheet Rho (Ohm/sq ) 25% Ge, 10 min 25% Ge, 90 min 25% Ge, 1020 min 25% Ge + B, 10 min 25% Ge + B, 90 min 25% Ge + B, 1020 min 25% Ge Difference (Doped-Undoped) 0 25 50 75 100 125 450500550600650700750800 Temperature (C)Sheet Rho (Ohm/sq ) 10 min 90 min 1020 min(b) (a) Figure 5-4: Isochronal anneal series of (a) sheet resist ance of undoped and doped 25% Ge samples and (b) difference in sheet resistance between the doped and undoped samples. Note that dotted lines are extrapolations to infinity indicating that subsequent samples in the series read as ope n circuits or that an infinite difference was calculated.

PAGE 162

162 0 50 100 150 200 0.002000.004000.006000.008000.0010000.0012000.00 Grain Size (arb units)Sheet Resistance (Ohm/sq ) 15% Ge 25% Ge 15% Ge + B 25% Ge + B Figure 5-5: Plot of samp le sheet resistance vs. Si1-zGez grain size.

PAGE 163

163 Function of Ge-rich Si(1-z)Ge(z) 0 50 100 150 200 250 02 04 06 08 01 0 0 Percent Area Fraction Si(1-z)Ge(z)Sheet Resistance (Ohm/sq ) 15% Ge (Undoped) 25% Ge (Undoped) 15% Ge + B 25% Ge + B Function of Si-rich Ni(Si(1-u)Ge(u)) 0 50 100 150 200 250 0510152025303540 Percent Area Fraction Ni(Si(1-u)Ge(u))Sheet Resistance (Ohm/sq ) 15% Ge (Undoped) 25% Ge (Undoped) 15% Ge + B 25% Ge + B(a) (b) Figure 5-6: Plot of sample sheet re sistance vs. area fraction of (a) Ni(Si1-uGeu) and (b) Si1-zGez.

PAGE 164

164 0 25 50 75 100 125 0 20 40 60 80 100 Percent Area Fraction Si(1-z)Ge(z)Sheet Resistance (Ohm/sq ) 15% Ge + B Linear Interpolation 0 50 100 150 200 250 02 04 06 08 01 0 0 Percent Area Fraction Si(1-z)Ge(z)Sheet Resistance (Ohm/sq ) 0 10 20 30 40 50 60 70 80 90 100Area Fraction Ni(Si(1-x)Ge(x) ) 15% Ge (Undoped) 25% Ge (Undoped) 15% Ge + B 25% Ge + B AF Ni(Si(1-x)Ge(x))(a) (b) Figure 5-7: Plot of sample shee t resistance vs. area fraction of Si1-zGez for (a) doped 15% Ge samples with linear interpolation of film and substrate sheet resistance values and (b) for all samples with area fraction of remaining initial Ni(Si1-xGex) layer and divisions into reaction stages.

PAGE 165

165 10 20 30 40 50 60 70 01 02 03 04 05 06 0 Percent Area Fraction Si(1-z)Ge(z)Sheet Resistance (Ohm/sq) 15% Ge (Undoped) 25% Ge (Undoped) 15% Ge + B 25% Ge + B Figure 5-8: Plot of sample shee t resistance vs. area fraction of Si1-zGez for the precipitation stage (Stage I) of the transformation.

PAGE 166

166 Undoped R2 = 0.7454 R2 = 0.8431 0 10 20 30 40 50 60 70 80 01 02 03 04 05 06 0 Percent Area Fraction Si(1-z)Ge(z)Sheet Resistance (Ohm/sq ) 15% Ge (Undoped) 25% Ge (Undoped) Doped R2 = 0.8663 R2 = 0.8423 0 10 20 30 40 50 60 70 80 01 02 03 04 05 06 0 Percent Area Fraction Si(1-z)Ge(z)Sheet Resistance (Ohm/sq ) 15% Ge + B 25% Ge + B(a) (b) Figure 5-9: Plot of Stage I sample sheet resistance vs. area fraction of Si1-zGez for (a) undoped and (b) doped samples. Best fit line ar regression lines are also shown.

PAGE 167

167 Regression Intercept-5 0 5 10 15 20Intercept Value 15% Ge (Undoped) 25% Ge (Undoped) 15% Ge + B 25% Ge + B Regression Slope 0 0.25 0.5 0.75 1 1.25 1.5 1.75Slope Value 15% Ge (Undoped) 25% Ge (Undoped) 15% Ge + B 25% Ge + B(a) (b) Figure 5-10: Plot of Stage I lin ear regression (a) intercept and (b ) slope of the results given in Table 5-1. The 95% confidence interv al for each value is also shown.

PAGE 168

168 0 10 20 30 40 50 60 70 80 01 02 03 04 05 06 0 Area Fraction Si(1-z)Ge(z)Sheet Resistance (Ohm/sq) 15% Ge (Undoped) 25% Ge (Undoped) 15% Ge + B 25% Ge + B Linear (All Undoped) Linear (All Doped) Figure 5-11: Plot of Stage I sample sheet resistance vs. area fraction of Si1-zGez for all samples. Best fit linear regression lines for th e compiled undoped and doped samples are also shown.

PAGE 169

169 Regression Intercept 0 2 4 6 8 10 12 14 16Intercept Valu e Undoped Doped Regression Slope 0.25 0.5 0.75 1 1.25Slope Value Undoped Doped(a) (b) Figure 5-12: Plot of linear regres sion (a) intercept and (b) slope of the results given in Table 5-2. The 95% confidence interval for each value is also shown.

PAGE 170

170 (a) (b) Figure 5-13: Images of (a) raw SEM/BSE im age for undoped 25% Ge sample annealed at 450 oC for 1020 minutes and (b) image of same sample after ImageJ processing and threshold application with example tortuosity calculation shown.

PAGE 171

171 y = 0.0029x + 1.0149 R2 = 0.9253 1 1.05 1.1 1.15 1.2 1.25 01 02 03 04 05 06 07 0Area Fraction Si(1-z)Ge(z)Tortuosity 25% Ge (Undoped) 25% Ge + B y = 0.0102e7.1279xR2 = 0.8636 y = 0.0264e6.0707xR2 = 0.8743 10 15 20 25 30 35 40 45 1 1.05 1.1 1.15 1.2 1.25TortuositySheet Resistace (Ohm/sq) 25% Ge (Undoped) 25% Ge + B Expon. (25% Ge (Undoped)) Expon. (25% Ge + B)(a) (b) Figure 5-14: Plots of (a) conductive path tortuosity vs. area fraction Si1-zGez and (b) sample sheet resistance vs. tortuosity for selected undoped and doped 25% Ge samples. Best fit linear regression line fo r compiled undoped and doped samples is shown in (a).

PAGE 172

172 Undoped 100 1000 10000 60 65 70 75 80 Percent Area Fraction Si(1-z)Ge(z)Sheet Resistance (Ohm/sq ) 15% Ge (Undoped) 25% Ge (Undoped) Doped 25 50 75 100 125 150 175 200 60 65 70 75 80 Area Fraction Si(1-z)Ge(z)Sheet Resistance 15% Ge + B 25% Ge + B(a) (b) Figure 5-15: Sheet resistance of (a) undoped and (b) doped 15 and 25% Ge samples as a function of Si1-zGez area fraction for the agglomerati on stage (Stage II) of the film transformation. Best fit linear regression lines for 15 and 25% Ge samples are shown in (b).

PAGE 173

173 10 100 1000 10000 6065707580 Percent Area Fraction Si(1-z)Ge(z)Sheet Resistance (Ohm/sq ) 15% Ge (Undoped) 25% Ge (Undoped) 15% Ge + B 25% Ge + B Figure 5-16: Sheet resistance as a function of Si1-zGez area fraction for all samples in Stage II (agglomeration) of the film transformation.

PAGE 174

174 Figure 5-17: Images of (a) raw SEM/BSE fo r undoped 25% Ge sample annealed at 650 oC for 10 minutes, (b) previous sample after Imag eJ processing, (c) raw SEM/BSE image annealed at 750 oC for 10 minutes, and (d) previous sa mple after ImageJ processing. Possible conduction paths vs. chord lines are shown in (b) and (d).

PAGE 175

175 Figure 5-18: SIMS analysis results of B doping distribution after silici dation with Ni metal showing silicidation induced dopant segregation [Zha06].

PAGE 176

176 Ni(Si1-xGex) TiN Figure 5-19: (a) XTEM and (b) SIMS analysis re sults for doped 25% Ge sample annealed at 450 oC for 10 minutes

PAGE 177

177 Si1-xGex(+ B) Si1-xGex(+ B) Ni0.5(Si1-xGex)0.5 B Pileup Si1-xGex(+ B) Lightly Doped Si1-zGez Si1-xGex(+ B) Si1-xGex(+ B) Lightly Doped Si1-zGez Residual Doping Figure 5-20: Cartoon schematics of nickel germ anosilicide samples exhibiting segregation induced dopant segregation of initial Si1-xGex B dopants before and after precipitation of Si1-zGez grains.

PAGE 178

178 Sheet Resistance 0 25 50 75 100 125 150 450500550600650700750800 Temperature (C)Sheet Resistance (Ohm/sq ) 10 min 90 min 1020 min Homogeneous NiSiGe 10 nm pileup and 120 nm of SiGe + B in parallel 130 nm SiGe + B (No Pileup) Figure 5-21: Isochronal anneal se ries of doped 15% Ge samples w ith calculated sheet resistance values for the potential a lternative conduction paths.

PAGE 179

179 Table 5-1: Linear regression results for best fit lines shown for St age I (precipitation reaction) in Figure 5-9. Intercept Standard Error InterceptSlope Standard Error Slope 15% Ge (Undoped)7.7883.178 1.1730.190 25% Ge (Undoped)5.2744.889 0.9540.137 15% Ge + B8.3951.278 0.6760.077 25% Ge + B10.4592.632 0.4950.071

PAGE 180

180 Table 5-2: Stage I (precipita tion reaction) linear regression re sults for the compiled undoped and doped data shown in Figure 5-11. Intercept Std. Error InterceptSlope Std. Error Slope Undoped9.4972.7870.8510.129 Doped9.9291.2310.5280.045

PAGE 181

181 Table 5-3: Measurements of Si1-zGez area fraction, tortuosity, and sheet resistance for selected undoped and doped 25% Ge samples. Temperature (C) Time (min) % AF Si(1-z)Ge(z) Tortuosity Sheet Resistance (Ohm/sq) 15% Ge (Undoped)450 302.10 1.01015.14 15% Ge (Undoped)450 9015.951.08017.69 15% Ge (Undoped)450 102048.501.15039.85 15% Ge (Undoped)500 3021.331.07020.35 15% Ge (Undoped)500 9034.731.10028.39 25% Ge (Undoped)450 9019.601.08315.79 25% Ge (Undoped)450 102057.151.16031.19 25% Ge (Undoped)500 103.63 1.01614.53 25% Ge (Undoped)550 1035.401.12021.08 25% Ge (Undoped)550 3045.171.17237.00

PAGE 182

182 Table 5-4: Regression results for Stage II (agglomer ation reaction) best fit lines shown in Figure 5-15. Intercept Std. Error InterceptSlope Std. Error Slope 15% Ge + B-211.2734.784.280.51 25% Ge + B-429.2357.907.880.84

PAGE 183

183 CHAPTER 6 CONCLUSIONS AND FUTURE WORK The stated objective of this work was to cl arify certain outstanding issues regarding the morphological stability and electr onic properties of nickel germ anosilicide thin films. The following sections discuss the results and conclusi ons determined by this work in the context of these outstanding issues. A discussion of potenti al future work which would complement the results of this work is also included. 6.1 Conclusions Three major research objectives were addressed by the experimentation in this work. The three objectives included clarification and quan tification of the structure/property relationship which published literature had proposed to ex ist between nickel germanosilicide film agglomeration and increases in film sheet resist ance, further investigatio n of the influence of high levels of homogeneous B doping on the relations hip, and expansion of the thermal matrix in which films were studied in order to gain info rmation about the kinetics of the agglomeration process. The following sections discuss the findi ngs of the experiments performed in this work in terms of these three objectives. 6.1.1 Microstructure and Kinetics Prior studies available in liter ature have utilized isochronal experiments with very short annealing times (30 to 60 seconds). This work aimed to expand the knowledge of the behavior of nickel germanosilicide films over a much larger range of anneal times and temperatures, including both isochronal and isothermal series Through analysis of samples annealed at temperatures ranging from 450 to 800 oC and times from 10 minutes to 1020 minutes, information about the kinetics of th e phase transformation was obtained.

PAGE 184

184 Results from the analysis of the annealed samples showed that the general cross-section and plan-view morphology of the nickel germanosili cide films in this work agreed well with those reported in prior literature. Also, while prior work has genera lly described the film evolution as agglomeration, two distinct reactions were captured by the anneal matrix used in this work. Initially, Ge-rich Si1-zGez grains precipitated from the parent film as Ge was rejected (presumably to the grain boundaries). This process continued until the parent film had completely transformed in to Ge-rich Si1-zGez and Si-rich Ni(Si1-uGeu) grains. Complete transformation corresponded to ~60% area fraction Ge-rich Si1-zGez for all samples in this work. Once the transformation was complete, the Si-rich Ni(Si1-uGeu) grains then began to agglomerate. It was also shown that increasing Ge content in the initial Si1-xGex film from 15 to 25% increased the amount of transformation observe d for equivalent anneals, likely due to the increase in driving force for Ge rejection with increasing Ge content. It was also determined that the reaction kinetics for both stages of the film transformation could be modeled using general Avrami isothe rmal transformation curves. Two separate reaction orders were observed, corresponding to th e precipitation and agglomeration reactions. By analysis of the reaction order, the precip itation reaction was determined to be diffusion controlled growth with a zero nucleation rate The agglomeration reaction was similarly determined to be diffusion controlled thickening of very large plates. The activation energies of the reactions were also able to be calculated and showed that for the precipitation reaction, Ea was 1.96 and 0.76 eV for the 15 and 25% Ge samp les, respectively. For the agglomeration reaction, Ea was determined to be 0.14 and 0.10 eV, respectively (though statistically these values were found to be equal).

PAGE 185

185 6.1.2 Structure/Property Relationship Prior literature has proposed a qualitative link between the agglomeration of a nickel germanosilicide thin film and increases in the film s sheet resistance. The evidence in favor of the relationship, however, was circumstantia l and no quantitative relationship had been established. Therefore, one objec tive this work was to clarify the relationship by determining if a direct, quantitative relations hip could be established betw een a metric describing film agglomeration and sheet resistance. The in fluence of increasing Ge content was on the structure/property relatio nship was also studied. The best metric for establishing a structure/ property relationship between the film quality and its sheet resistance was determined to be the area fraction of Ge-rich Si1-zGez present in the film. Use of this metric allowed a direct, quant itative relationship to be established in both the precipitation and agglomeration stag es of the film transformation. For the precipitation stage of the transformation, the sheet resistance was found to increase linearl y with increasing area fraction of Si1-zGez. Increasing Ge content in this stag e of the reaction did not affect the relationship. It was determined, therefore, that the influence of increasing Ge on sheet resistance could be attributed solely to the increased agglomeration seen in these samples and not to a fundamental change in the struct ure/property relationship. Overa ll, the general increase in sheet resistance for all samples in this stage of the transformation was found to be attributable to the increasing tortuosity of the c onduction path with increasing Si1-zGez precipitation. For the agglomeration stage of the film transformation, the sheet resistance of the undoped samples was found to rapidly increase to large (or infinite) values unr elated to the amount of Si1-zGez present and no quantitative structure/property was able to be established. The influence of Ge content was similarly unable to be determined. This effect was determined to be due to the absence of a continuous conduction path betw een isolated Si-rich Ni(Si1-uGeu) grains for these sample. The

PAGE 186

186 doped samples, however, showed a linear trend of increasing sheet resistance with increasing area fraction Si1-zGez, though with a greater slop e than seen in the prec ipitation stage of the reaction. 6.1.3 Influence of Homogeneous B Doping Prior literature has proposed that high le vels of homogeneous B doping in the initial Si1xGex layer can suppress the agglomeration process a nd thus stabilize sheet resistance. The prior work, however, did not utili ze direct measurement techni ques when evaluating film agglomeration and so the degree of suppression, if any, is not well quantified. Accordingly, this work aimed to confirm the presence of a gglomeration suppression using plan-view SEM imaging. This work also aimed to determine if the quantitative relationship between agglomeration and sheet resistance was maintained for B doped samples. Analysis of the samples in this work s howed that the addition of ~4.5E19 atoms/cm2 homogeneous B doping during Si1-xGex layer growth did not affect the morphology of the nickel germanosilicide film microstructure for any anne al time or temperature. The doped samples, however, showed a decrease and stabilization of sheet resistance values for both the 15 and 25% Ge samples used in this work in accordance with prior literature. This effect was determined to be present in both the precipitation and agglomer ation stages of the film transformation, though the magnitude of the effect was much larger in th e latter stage. Overall, these results suggested that a mechanism other than a decrease in aggl omeration, most likely an alternative conduction path in the B doped structures, was responsible for the effects. This path was also determined to most likely be the unreacted, nominally B doped Si1-xGex layer below the nickel germanosilicide layer. This conclusion also suggested that differences in the sheet resistance of the Si1-xGex layer with changing Ge content, though unconfirmed, caused the difference in structure/property relationship for the doped samples within the agglomeration stage of the transformation.

PAGE 187

187 6.2 Future Work While this work clarified many of the outstanding questions concer ning the stability and behavior of nickel germanosilicid e thin films, some questions were also raised by this work. Specifically, the cause for the increase of activ ation energy for the preci pitation reaction with increasing Ge content is not known. Also, the existence of segregation induced dopant segregation was not directly observed for the samples in this work, nor was the distributions of the B dopant determined within the structure for e ach anneal time and temperature. These points could possibly be clarified by performing additional investigations of the nickel germanosilicide system using Local Electrode Atom Probe (LEAP) analysis. LEAP analysis has been shown to be able to provide three dimensional maps of atomic distributions for semiconducting samples [Kel04,]. The technique has also been shown to be sensitive enough to investigate the lateral distribution of a gate st ructure implanted with B dopant [Moo08]. It is expected, therefore, that LEAP analysis of samples similar to those used in this work could provide information about B and Ge distributions in proximity to the grain boundaries of the germanosilicide grains. With su ch information, the mechanism by which Ge is rejected from the sample (i.e. via bulk or grai n boundary diffusion) could possibly be clarified. The cause for the concentration dependence of the activation energy might, in turn, also be clarified. The B dopant distri bution, specifically confirmation of whether silicidation induced dopant segregation is present, could definitely be determined through LEAP analysis. Information about the doping of the precipitated Si1-zGez grains could also be determined through this type of analysis. This information would, in turn, allow confirmation of the alternative conduction path for the B doped samples. Another area of potential future research w ould be the continued investigation of the relationship between area fraction of precipitated Si1-zGez, conduction path tortuosity, and

PAGE 188

188 sample sheet resistance. The tortuosity investig ations in this work were performed using manual measurement and calculations. It is expected, therefore, if computer software was used to measure tortuosity measurement error would decreas e. The more rapid speed of software-based calculations would also allow both a larger number of samples at each anneal time and temperature to be studied as well as a larger ove rall number of anneal times and temperatures. Additionally, software aided an alysis may allow the modeling of the conductive path for the doped samples in the agglomeration stage of the relationship to explain the linear dependence of sheet resistance on area fraction of Si1-zGez. Both of these analyses will require higher resolution, less noisy SEM images of the sample s, as the graininess of the images (and the resulting difficulty of defining the grain edges) was the primary reason such analyses were not attempted in this work. Performing these additional inve stigations would continue to develop the scientific knowledge of the behavior of nickel germanosilici de thin films. This knowledge, in turn, can then be used to address the use of such film s as intermediate layers between semiconductor devices and metallization layers.

PAGE 189

189 APPENDIX A ERROR IN SEM/BSE IMAGE QUANTIFICATION SEM/BSE images in this work were mainly quantified using ImageJ software. To validate the repeatability of the SEM/BSE image quantifi cation in this work, as well as explore the repeatability of the anneal process and regiona l sample variation, samples of the undoped, 15% Ge samples were annealed at 550 oC for 90 minutes on three separate occasions. Between each anneal, the furnace was set to another temp erature and then re-set back to 550 oC. While it does not represent an exact center po int in the experimental matrix, the data point (undoped, 15% Ge, 550 oC, 90 minute) selected for replication was chosen as it displays the most complex intermediate morphology with all three phases, Ni(Si1-xGex), Ni-rich Ni(Si1-uGeu), and Ge-rich Si1-zGez, present in the image. It is expected, therefor e, that the variation in this sample would be the highest in the experiment due to the difficult y in quantizing the complex microstructure. By applying the variation found at this data point to the entire matr ix, a conservative estimate of variation would be included in the data analysis. Additional data points for replication were not chosen due to time and resource constraints. Once annealed, each of the replicates was imaged at 30,000x magnification using SEM/BSE five times in two separate regions on each sample, for a grand total of 30 images. This nested design allowed both sample-to-sample and region-to-region comp arisons to be made. The area fraction of Ni-rich Ni(Si1-uGeu) and Ge-rich Si1-zGez phases present in each image was then quantified using ImageJ and the results analy zed with Minitab to determine the variation in the process. Individual value plots of the area fraction va lues divided by sample and region are shown in Figure A-1 for both response variables. The plots indicate the qualitative presence of some variation, both between and within each sample. To determine if the apparent variation was

PAGE 190

190 statistically significant, a fully nested ANOVA analysis was performed on the data for each response variable using sample and region as th e predictive factors in the model. The ANOVA results, presented in Figure A-2, show that there is no statistically si gnificant difference between samples for either response variable at a 95% co nfidence level (p-values of 0.335, 0.375). While no statistically significant difference is seen be tween regions for the area fraction of Ni-rich Ni(Si1-uGeu) (p = 0.438) at the 95% c onfidence level, a significant difference is noted for the regional variation of Ge-rich Si1-zGez (p = 0.003). It is not known why regional variation is seen for one phase and not the other; however it is un surprising that some local variations in phase quantity exist due to the relatively small sampling area contained by an image at 30,000x magnification. Nevertheless, this level of magnification is necessary to resolve the small grains involved in the phase transformation. The maximum total variance in the nested ANO VA analysis was found to be 8.761. As three images were independently analyzed and their values averaged to determine the area fraction of each phase at each data point in the experiment, the variance of the averaged data is calculated according to Equation A-1: n XVar2)( Equation A-1 The variance for the averaged data was thus found to be 2.92, corresponding to a standard deviation of 1.71%. For a 95% c onfidence interval, this equates to a range of +/3.35%. This conservative estimate takes into account both sample-to-sample and local variations. Error from image processing and selection of contrast threshold for quantifica tion analysis is also contained in this term. It should be noted, however, that experimental re sults suggest that actual error decreases with increasing area fraction and that this confidence interval results in very conservative error estimation at high area fraction contents (greater than ~ 60% AF).

PAGE 191

191 Figure A-1: Individual plot results divided by sample and re gion for (a) area fraction of Ni-rich phase and (b) area fractio n of Ge-rich SiGe phase

PAGE 192

192 Figure A-2: Minitab results fo r fully nested ANOVA results for (a) area fraction of Ni-rich phase and (b) area fractio n of Ge-rich SiGe phase

PAGE 193

193 APPENDIX B GAUGE REPEATABILITY AND REPRODUCIBILITY ANALYSIS OF 4PP MEASUREMENT A full gauge repeatability and reproducibility study (gauge R&R) was performed on the four point probe (4PP) sheet resistance measuremen t technique used in this work. For this study, 10 samples were selected at random from the fu ll anneal matrix used in this work. Three operators then measured the shee t resistance of each sample in random order. This process was then repeated twice (with the order of the samples randomly re-arranged between each repetition) for a total of 3 repetitions and a grand total of 90 measurements. Of the three operators, only one operator (the author of this work) was able to successfully measure every sample for each of the three repetit ions. In total, only 5 of the 10 samples were successfully measured for all repe titions by all operators. For th e five samples that were not completely successfully measured, some samples were successfully analyzed by an operator for none, one, or two repetitions. These results indi cate problems with both the repeatability and reproducibility of the measurement process. Data from the five samples that were completely successfully measured was analyzed using a Gauge R&R ANOVA package contained in the Minitab 15.1.1.0 software. The results, presented in Figure B-1, indicate very low contributions to th e total study deviation by both measurement repeatability and reproducibility. Th ese results are very interesting when taken in conjunction with the fact that many samples c ould not be successfully measured by some operators. Thus, it can be concluded that whil e an operator may have a difficult time obtaining a successful measurement of a sample, if a m easurement is obtained it is of good quality. While this finding is of general interest regarding the measurement process, all sheet resistance measurements presented in this work were taken by a single ope rator (the author). Hence, a deeper understanding of the repeatability of the measurement pro cess by this operator is

PAGE 194

194 of interest when analyzing the results in this work. Accordingly, a one-way ANOVA analysis was performed on the results from this operato r. As this operator successfully obtained measurements for all samples in the study, data from all 10 samples was used in the analysis. The ANOVA results from this analysis, presented in Figure B-2, indicate that both repeatability and part-to-part variation contribut e to the total overall variance in the study. It can also be observed in the results that one sample, number 9, has a sheet resistance several orders of magnitude higher than the other sample s in the study. It is also apparent that the sheet resistance measurements of this sample varied over a cons iderable range (from ~2000 to ~6500 Ohm/sq). This observation suggests that measurement rep eatability may suffer for samples with very high sheet resistances. To further explore this suppos ition, the data from sample 9 was eliminated from the study and the ANOVA analysis performed ag ain. The results from the revised analysis are presented in Figure B-3 and s how that the contributi on of repeatability to the study variation has decreased to less than 5% of the overall value. It can be concluded, th erefore, that the sheet resistance measurement process is repeatable for th is operator (the author) for most samples. If a sample presents a very high sheet resistance va lue (on the order of thousands of Ohm/sq), however, it should be cautioned th at significant variation in th e results may be present. Nevertheless, most samples in this work have sheet resistance values below ~200 Ohm/sq. For these samples, based on the results from the revi sed ANOVA analysis of the first operator (the author of this work), a 95% confidence interval of +/2.19 Ohm/sq can be applied to capture any variation in the measurement process.

PAGE 195

195 Figure B-1: Gauge R&R ANO VA analysis of sheet resistance measurements for the five samples which all operators measured successfully in all trials.

PAGE 196

196 Figure B-2: Gauge R&R ANO VA analysis of sheet resistance measurements including data from all 10 samples for the operator (the author) who successfully measured all samples for all trials.

PAGE 197

197 Figure B-3: Gauge R&R ANO VA analysis of sheet resistance measurements, excluding data from sample 9, for the operator (the aut hor) who successfully measured all samples for all trials.

PAGE 198

198 APPENDIX C AVRAMI PLOT ERROR Generating error bars for a sigmoidal func tion plotted on a graph of log(log((1-AF)-1) vs. log(t) is complicated by the transformation mathematics. While it was determined in Appendix A that a constant error of +/3.35% should be used for a 95% conf idence interval of area fraction in this work, the transformation of this interval for use on a log(log((1-AF)-1) vs. log(t) results in significantly varying error with the nominal (mean) value of area fraction. This complication is graphically shown in Figure C-1. In the figur e, the resulting upper a nd lower confidence error bounds for the transformed interval are shown as a function of nominal area fraction. For small area fractions (less than ~15%), the transformed error range is ve ry large. For moderate area fractions (~15 to 85%), the error range is relativ ely small, and for large area fractions (greater than ~85%), the error range again increases. An example of the non-linear nature of transformed error on this t ype of plot is as follows. Consider a sample with 5% AF. For a 95% confidence interval in this work, the resulting range of AF lies between 1.65% and 8.35%. The re sulting range on the transformed plot would therefore span between -2.28 and -1.39, a range of 0.88. For a sample with 50% AF, the linear range of 46.65% to 53.35% corresponds to a span of -0.57 to -0.47 on the transformed plot, a range of only 0.10. Thus, a constant error range on a linear plot of area fraction vs. time can be shown to be much more difficult to handle after mathemati cal transformation for use with a plot of log(log((1-AF)-1) vs. log(t). For simplicity, this work therefore uses a constant error term of +/0.15 for the data points that lie below 60% area fr action. For points with greater than 60% area fraction, a term of +/0.045 was used. Thes e values were selected according to their representation of the average error in th eir respective ranges (based on Figure C-1).

PAGE 199

199 Error Range ( +/3.35% AF) -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 0102030405060708090100 Area Fraction (%)log(log(1/(1-AF))) High Limit (+ 3.35%) Nominal Low Limit (3.35%) Figure C-1: Transformed error range for a c onstant +/3.35% range of area fraction as a function of nominal area fraction.

PAGE 200

200 LIST OF REFERENCES [Aub05] V. Aubry-Fortuna, P. Dollfus, S. Galdin-Retailleau, Solid-State Electron. 49, 1320 (2005) [Avr39] M. Avrami, J. Chem. Phys. 7, 1103 (1939) [Avr40] M. Avrami, J. Chem. Phys. 8, 212 (1940) [Avr41] M. Avrami, J. Chem. Phys. 9, 177 (1941) [Bai04] P. Bai et al., IEDM Tech. Dig. 2004, 657 (2004) [Bar87] J.C. Barbour, A.E.M.J. Fischer, J.F. van der Veen, J. Appl. Phys. 62, 2582 (1987) [Cha04] O. Chamirian, A. Lauwer s, J.A. Kittl, M. Van Dal, M. De Potter, R. Lindsay, K. Maex, Mic. Eng. 76, 297 (2004) [Cha04b] S. Chattopadhyay, L.D. Driscoll, K.S.K. Kwa, S.H. Olsen, A.G. ONeill, Solid State Electron. 48 1407 (2004) [Che97] J. Chen, J.P. Colinge, D. Flandre, R. Gillon, J.P. Raskin, D. Vanhoenacker, J. Electrochem. Soc. 144(7), 2437 (1997) [Chi04] P.R. Chidambaram, B.A. Smith, L.H. Hall, H. Bu, S. Chakravarthi, Y. Kimm VLSI Symp. Tech. Dig. Honolulu, HI, 48 (2004) [Cho06a] Saurabh Chopra, Mehmet C. Ozturk, Veen a Musra, Kris McGuire, Laurie E. McNeil, App. Phy. Lett. 88, 202114 (2006) [Cho06b] Saurabh Chopra, Mehmet C. Ozturk, Veen a Musra, Kris McGuire, Laurie E. McNeil, App. Phy. Lett. 89, 202118 (2006) [Chr02] J.W. Christian, The Theo ry of Transformations in Metals and Alloys, Part I, Pergamon, Oxford (2002) [Cro03] R.T. Crosby et al., Mat. Sc i. in Semicond. Proc. 6, 205 (2003) [dHe82] F. dHeurle, C.S. Pete rsson, L. Stolt, B. Stritzke r, J. Appl. Phys. 53, 5678 (1982) [dHe84] F.M. dHeurle, C.S. Petersson, J.E.E. Ba glin, S.J. LaPlaca, C. Wong, J. Appl. Phys. 55, 4208 (1984) [dHe89] F.M. dHeurle, J. Vac. Sci. Technol. A 7, 1467 (1989) [dHe86] F.M. dHeurle P. Gas, J. Mater. Res., 1(1), 205 (1986) [Dis54] J.P. Dismukes, L. Ekstrom, R.J. Paff, J. Phys. Chem. 68(10), 3021 (1964)

PAGE 201

201 [Fin81] T.G. Finstad, Phys. St atus Solidi. 63(a), 223 (1981) [Gam98] J.P. Gambino, E.G. Colgan, Mat. Chem. Phys. 52, 99 (1998) [Gan93] E. Ganin, S. Wind, P. Ronshei, A. Yapsir, K. Barmak, J. Bucchingnano, R. Assenz, Rapid Thermal and Integrated Processing II, Vo l. 303, Materials Research Soc. Puttsburgh, PA, 109 (1993) [Gan00] S. Gannavaram, N. Pe sovic, M.C. Ozturk, in: Proceedings of the IEDM 2000, p.347 (2000) [Gas98] P. Gas, F. dHeurle, Numerical Data and Functional relationships in Science and Technology Volume III, Landolt-Bo rnstein, Berlin 33A (1998) [Ger04] P. Gergaud, C. Rivero, M. Gaillanou, O. Thomas, B. Froment, H. Jaouen, Mater. Sci. Eng. B 114-115, 67 (2004) [He05] J.H. He, W.W. Wu, L.J. Chen, Nuc. Ins. Meth. Phys. Resch. B 237, 174 (2005) [Hel97a] P.E. Hellberg, S.L. Zhang, C.S. Pe terson, IEEE Elec. Dev. Lett. 18(9), 456 (1997) [Hel97b] P. E. Hellberg, A. Gagnor, S.L. Zhang, C.S. Peterson, J. Electrochem. Soc. 144, 3968 (1997) [Hsi88] Y.F. Hsieh, L.J. Chen, E.D. Marshall, S.S. Lau, Thin Solid Films 162, 287 (1988) [Hua95] F.Y. Huang, X. Zhu, M.O. Tanner, K.L. Wang, Appl. Phys. Lett. 67(4), 566 (1995) [Ish03] C. Isheden, J. Seger, H. Radamson, S. L. Zhang, M. Ostling, Mater. Res. Soc. Symp. Proc. 745, N4.9 (2003) [Iwa85] M. Iwami, A. Hiraki, J pn. J. Appl. Phys. 24, 530 (1985) [Iwa02] Hiroshi Iwai, Tatsuya Ohguro, Shunichiro Ohmi, Mic. Eng. 60, 157, (2002) [Jar02] T. Jamar, J. Seger, F. Ericson, D. Mangelick, U. Smith, S.L. Zhang, J. App. Phy. 92(12), 7193 (2002) [Jia92a] H. Jiang, C.M. Osburn, Z.G. Xiao, D. Giffis, G. McGuire, G.A. Rozgonyi, J. Electrochem. Soc. 139, 196 (1992) [Jia92b] H. Jiang, C.M. Osburn, Z.G. Xiao, G. McGuire, G.A. Rozgonyi, B. Patniak, N. Parikh, M. Swanson, J. Electrochem. Soc. 139, 206 (1992) [Jia92c] H. Jiang, C.M. Osburn, Z.G. Xiao, G. McGuire, G.A. Rozgonyi, J. Electrochem. Soc. 139, 211 (1992) [Jin04] L.J. Jin, K.L. Pey, W.K. Choi, E.A. Fitz gerald, D.A. Antoniadis, A.J. Pitera, M.L. Lee, D.Z. Chi, C.H. Tung, Thin Solid Films, 462, 151 (2004)

PAGE 202

202 [Jin05] L.J. Jin, K.L. Pey, W.K. Choi, E.A. Fitz gerald, D.A. Antoniadis, A.J. Pitera, M.L. Lee, D.Z. Chi, Md. A. Rahman, T. Osipowicz, C.H. Tung, J. Appl. Phys. 98, 033520 (2005) [Kel04] T. Kelly et al., Micr osc. Microanal. 10, 373 (2004) [Ko06] H. Ko, J. Vac. Sci. Technol. A, 24(4), 1468 (2006) [Kyk83] G.C. Kyker, Jr., Am/ H. Phys. 51, 852 (1983) [Las91] J.B. Lasky, J.S. Nakos, O.J. Cain, P.J. Geiss, IEDM Trans. El ectron. Devices, ED-3458, 262 (1991) [Lau06] A. Lauwers, M.J.H. van Dal, P. Verheyen O. Chamirian, C. Demeurisse, S. Mertens, C. Vrancken, K. Verheyden, K. Funk, J.A. Kittl, Mic. Eng. 83, 2268 (2006) [Lav03] C. Lavoie, F.M. dHeur le, C. Detavernier, C. Cabral, Micro. Eng. 70, 144 (2003) [Lee05] S.G. Lee et al. Mater. Sci .. Semicond. Process. 8, 215 (2005) [Li89] J. Li, Q.Z. Hong, J.W. Mayer, J. Appl. Phys. 67, 2506 (1989) [Liu05] Jung Liu, Mehmet C. Ozturk, IEEE Trans. Ele. Dev. 52(7), 1535 (2005) [Mae93] K. Maex, Mater. Sci. Eng. Rev. R11, 53 (1993) [Man98] D. Mangelick, P.E. Hellberg, S.L. Zha ng, F.M. dHeurle, J. Electrochem. Soc. 145, 2530 (1998) [Mas90] T.B. Massalski, Binary Alloy Phase Di agrams, ASM International, Materials Park, OH (1990) [Mas05] W.P. Maszara, J. Elect rochem. Soc. 152 (7), G550 (2005) [Mat74] J.W. Matthrews and A.E. Blakeslee, J. Cryst. Growth, 29, 273 (1974) [Mei04] D.L. Mei, J.C. Li, J. Zhang, W.J. Xu, K.Z. Tan, M.H. Yang, Microelecton. J. 35, 969 (2004) [Moo08] J.S. Moore et al., Ultramicroscopy, 108(6), 536 (2008) [Mor84] Cpt. Morgan, Diageo, (1984) [Nat04] R. Nath, M. Yeadon, Electroch em. Solid-State Lett. 7(10), G231 (2004) [Nem06] F. Nemouchi, D. Mangelinck, J.L. Labar, M. Putero, C. Bergman, P. Gas, Mic. Eng. 83, 2101 (2006) [NTR97] 1997 National Technology Roadmap for Semiconductors, Semiconductor Industry Association, San Jose, CA (1997)

PAGE 203

203 [Nyg91] S. Nygren, D. Caffin, M. Ostling, F.M. dHeurle, A pp. Surf. Sci. 53, 87 (1991) [Ok03] Young-Woo Ok, S.H. Kim, Y.J. Song, K.H. Shim, Tae-Yeon Seong, Sem. Sci. Tech. 19, 285 (2003) [Ok04] Young-Woo Ok, S.H. Kim, Y.J. Song, K.H. Shim, Tae-Yeon Seong, J. Vac. Sci. Technol. B 22(3), 1088 (2004) [Pat94] J. Ken Patterson, B.J. Park, K. Ritley, H.Z. Xiao, L.H. Allen, A. Rockett, Thin Solid Films 253, 456 (1994) [Pea86] T.P. Pearsall, J.C. Bean, IEEE Electron. Device Lett. EDL-7, 308 (1986) [Peo85] R. People and J.C. Bean Appl. Phys. Lett. 47, 322 (1985) [Pey02] K.L. Pei, W.K. Choi, S. Chattopadhyay, H.B. Zhao, E.A. Fitzgerald, D.A. Antoniadis, P.S. Lee, J. Vac. Sci. Technol. A 20(6) 1903 (2002) [Pey04] K.L. Pey, S. Chattopadhy ay, W.K. Choi, Y. Miron, E.A. Fitzgerald, D.A. Antoniadis, T. Osipowicz, J. Vac. Sci. Technol. B 22(2) 853 (2004) [Pon00] Youri V. Ponomarev, Peter A. Stolk, Cora Salm, IEEE Trans. Elect. Dev. 47(4) 848 (2000) [Raa99] Ivo J. Raaijmakers, Hessel Sprey, Ar jen Storm, Timo Bergman, Joe Italiano, Doug Meyer, J. Vac. Sci. Technol. B 17(5) 2311 (1999) [Rad06] L. Radic, PhD Dissertation fo r the University of Florida (2006) [Sal97] C. Salm, D.T. van Veen, D.J. Gravesteijin J. Holleman, P.H. Woerlee, J. Electrochem. Soc. 144, 3665 (1997) [Seg02] J. Seger, S.L. Zhang, D. Mangelick, H.H. Radamson, Appl. Phys. Lett. 81(11), 1978 (2002) [Seg03] J. Seger, S.L. Zhang, Thin Solid Films 429, 216 (2003) [Seg04] J. Seger, T. Jarmar, Z.B. Zhang, H.H. Radamson, F. Ericson, U. Smith, S.L. Zhang, J. App. Phy. 96(4), 1919 (2004) [Tak00] Hirotsugu Takizawa, Kyota Uheda, Tadashi Endo, J. All. Com 305, 306 (2000) [Ter89] J. Tersoff, Phys. Rev. B 39, 5566 (1989) [Tho88] R.D. Thompson, K.N. Tu, J. Angillelo, S. Delage, S.S. Iyer, J. Electrochem. Soc. 135, 3161 (1988) [Tho02] S. Thompson et al. IEDM Tech. Dig. 2002, 61 (2002)

PAGE 204

204 [Tho04] S.E. Thompson, G. Sun, K. Wu, J. Lim, T. Nishida, IEDM Tech. Dig. 2004, 221 (2004) [Tol04] N. Toledo, P. Leeb, K. Pey, Thin Solid Films, 462-463, 202 (2004) [Uch06] K. Uchida, T. Krishnamohan, K.C. Sa raswat, Y. Nishi, IEDM Tech. Dig. 2006: San Francisco, CA (2006) [Ver94] Sopie Verdonckt-Vandebroek, Emmanuel F. Crabbe, Bernard S. Meyerson, David L. Harame, Phillip J. Restle, Johnannes M.C. Stor k, Jeffrey B. Johnson, IEEE Trans. Elect. Dec. 41(1) 90 (1994) [Wan95] Zhihai Wang, D.B. Aldrich, Y.L. Chen, D.E. Sayers, R.J. Nemanich, Thin Solid Films, 270, 555 (1995) [Web05] J. Weber, L. Nebrick, F. Bensch, K. Ne umeier, G. Vogg, R. Wileland, D. Bonfert, P. Ramm, Microelecton. Eng. 82, 215 (2005) [Won04] L.H. Wong et al. Thin Solid Films 462-463, 76 (2004) [Yao07] H.B. Yao, M. Bouville, D.Z. Chi, H.P. Sun, X.Q. Pan, D.J. Srolovitz, D. Mangelinck, Electrochem. Solid State Lett. 10(2) H53 (2007) [Zan01] N.R> Zangenberg et al. P hys. Rev. Lett. 87(12), 125901 (2001) [Zha02] H.B. Zhao, K.L. Pey, W.K. Choi, S. Ch attopadyay, E.A. Fitzgerald, D.A. Antoniadis, P.S. Lee, J. Appl. Phys. 92(1) 214 (2002) [Zha03] S.L. Zhang, Mic. Eng. 70, 174 (2003) [Zha04] Q.T. Zhao, D. Buca, St. Lenk, R. Loo, M. Caymax, S. Mantl, Mic. Eng. 76, 285 (2004) [Zha05] Qingchun Zhang, Nan Wu, Thomas Osi powicz, Lakshmi Kanta Bera, Chunxianh Zhu, Jap. J. of Appl. Phys., 44(45), 1389, (2005)

PAGE 205

205 BIOGRAPHICAL SKETCH John Samuel Moore, who publishes as J.S. Moore, received his B.S. cum laude from the University of Florida in 2004, with a major in materials science and engineering (metals specialty) and a minor in busine ss administration. After enteri ng graduate school at UF, he received his M.S. in materials science and e ngineering in 2006 and c ontinued with doctoral studies. He graduated with a Ph.D. in 2008 in materials science and engineering (electronic materials specialty) and a minor in industrial engineering. During his years at the University of Florida, he was active in both engineering and student government societies including terms as the President of the Society of Automotiv e Engineers (Spring 2002 to Spring 2003) and Treasurer of the Benton E ngineering Council (2004-2005).