Interfacial phenomena in ion implanted silicon-on-insulator materials

MISSING IMAGE

Material Information

Title:
Interfacial phenomena in ion implanted silicon-on-insulator materials
Physical Description:
xxvi, 306 leaves : ill. ; 29 cm.
Language:
English
Creator:
Saavedra, Antonio Fernando
Publication Date:

Subjects

Subjects / Keywords:
Materials Science and Engineering thesis, Ph. D   ( lcsh )
Dissertations, Academic -- Materials Science and Engineering -- UF   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 2004.
Bibliography:
Includes bibliographical references.
Statement of Responsibility:
by Antonio Fernando Saavedra.
General Note:
Printout.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 022829398
System ID:
AA00014269:00001


This item is only available as the following downloads:


Full Text










INTERFACIAL PHENOMENA IN ION IMPLANTED
SILICON-ON-INSULATOR MATERIALS















By

ANTONIO FERNANDO SAAVEDRA


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

UNIVERSITY OF FLORIDA

2004































Copyright 2004

by

Antonio Femando Saavedra






























This dissertation is dedicated my family, as well as the mentors who supported and
influenced me throughout my educational endeavors.














ACKNOWLEDGMENTS

Henry David Thoreau said, "I would rather sit on a pumpkin and have it all to myself than

be crowded on a velvet cushion." Such would describe how I have tried to live and work over

the years. Anyone who truly knows me recognizes this sense of independence in my being.

Nonetheless, a study so vast can only be undertaken with contributions from several sources.

I must begin by recognizing those individuals that had the most profound influence on my

character. My parents, Antonio and Monica, instilled in me the moral and ethical values which I

try, often failing, to live by. My father immigrated to the United States from Bolivia in 1969

without the ability to formulate sentences in English. Diligent work allowed him to become a

respected scientist in aluminum casting technology, which he passed on to me. Oddly enough, I

became the third generation metallurgist in the family, all having the same first name, Antonio.

Maybe he planned it, maybe he did not; I have no complaints. Monica has always provided a

source of support over the years that I can never repay or fully appreciate. She believed in me

even when I felt no one else did. My mother's parents, Howard and Gladys, were pleased to

provide a safe haven from home any time we chose. I will never forget the wonderful, relaxing

times spent on their farm. Jose and Ruth, on my father's side, were the grandparents to the

south that I hope to see more of in the future. Although oral communication is difficult due to

my broken Spanish, our love is not bound. Vacation to their home in Bolivia made me

appreciate the opportunities I had, yet taught me respect for all cultures no matter the

mainstream perception. My brother, Luie, remains a formidable opponent in professional-style

wrestling every time we meet, but knows to avoid the "Power of Tone" whenever possible. His








appreciation for "B" movies (e.g., American Ninja, various Chuck Norris titles, etc.) is rivaled

only by mine. Ximena and Tania, my sisters, were always there to keep me in line when the

parents were not around. I could rely on them to snitch me out whenever I chose to go into a

creative dialogue of four-letter words. However, their love and kindness are unmatched by any

other duo. My godson, Luke, has been an inspiration since his birth and makes me want to go

back to wearing diapers. I must also recognize my sister-in-law, Sharon, and brothers-in-law,

Ronnie and Dewey, for putting up with me. Although too numerous to include, I must

acknowledge the aunts, uncles, and cousins in both hemispheres. They all hold special places in

my heart.

The congregation at First United Methodist Church in Sheffield, Alabama, made me feel

welcome every Sunday. My youth and choir director, Oliver Brazealle, exhibited patience on

more than one occasion as we prepared for choir tour. The fondest memories of my life were

the times spent touring around the country without any responsibility. Oliver's sacrifices can

never be put into context. The members of the youth group were always ajoy to be around and

will remain friends forever. I single out Ged Grimmitt, Brad Bernard, Adam Littrell, Tyler and

Matt Jones, for their continual friendship and look forward to our travels in the future.

I must also recognize the mentors I had in the Muscle Shoals City School System,

particularly Robert Young for helping me grasp the fundamentals of algebra. By doing this he

put me on the path to becoming an engineer. I cannot begin to thank the faculty in the

Department of Metallurgical and Materials Engineering at the University of Alabama including

Drs. Richard Bratt, Viola Acoff, Mark Weaver, John Barnard, Tom Piwonka, Nagy El-Kaddah,

Garry Warren, Ramana Reddy, Doru Stefanescu, and Giovanni Zangari. They provided me

with what I consider to be a true materials engineering education. Rather than sticking strictly








to a structure-properties curriculum, they taught me to think like an engineer by instilling the

fundamentals of mass and energy balances, process controls, and transport phenomena.

Unfortunately, current materials science and engineering education seems to only give the

students the facts they should memorize rather than pushing an analytical thought process. One

man, Dr. James Weston, showed me "the ropes" of engineering when I was an undergraduate.

His patience was greatly appreciated and the knowledge I gained by looking over his shoulder is

irreplaceable. I also acknowledge my officemates Dr. Tom Scharf, Dr. Kevin Minor, Dr.

Jonathan Grant, Chris Hale, and Feng Huang, for providing an enjoyable work environment.

Classmates Brett Standifer, Bill Lapp, Brian Floyd, Bobby Roberts, and Kelly Crawford, made

MTE classes one of a kind.

Numerous people have been directly involved with my research at the University of

Florida. I begin by acknowledging my committee members Drs. Paul Holloway and Cammy

Abernathy for taking the time to follow my progress and providing guidance. The help of Dr.

Michael Kaufman for providing TEM knowledge and suggestions is also appreciated. Dr.

Valentin Craciun was indispensable for teaching me much of what I know about HRXRD, and

his service on my committee is recognized. My advisors, Drs. Kevin Jones and Mark Law, have

been instrumental in my scientific growth. They allowed me to work independently and skip

weekly meetings without consequence; what else can I ask for? I also acknowledge the friends

and colleagues of the SWAMP Center. Dr. Lahir Adam became a great friend as soon as I

arrived in Gainesville and I miss his driving and goofiness. Dr. Ibrahim Avci, Russ Robison,

Ljubo Radic, and Dr. Susan Earles provided help with modeling as well as good laughs. Dr.

Patrick Keys instructed me in the art of TEM, for which I am very grateful. Dena Wrigley,

Jackie Frazier, and Andrew King helped with some of the PTEM sample preparation. Robert








Crosby, Dr. Aaron Lilak, Dr. Chad Lindfors, Heather Randall, Renata Camillo-Castillo, Diane

Hickey, Michelle Phen, and others helped balance the work environment. I am grateful for the

secretarial staff including Sharon Carter and Teresa Stevens, who sent countless FedEx

shipments and handled many travel reimbursements with a smile.

Finally, I must thank IBM and the Semiconductor Research Corporation for providing

funding for my project. Dr. Kevin Chan was as an excellent mentor during my summer at

Yorktown and his help in material processing is appreciated. Drs. Erin Jones and Omer

Dokumaci are acknowledged for serving as my SRC liaisons and also providing material

support. The work of the ASTL South laboratory including Chris D'Emic, Ray Sicina, Phil

Saunders, Ed Sikorski, and Joe Newberry also aided with processing. Lastly, 1 recognize

Mikhail Klimov from the University of Central Florida for providing SIMS analysis.















TABLE OF CONTENTS
Page

ACKN O W LEDGM EN TS ........................................................... ....................... iv

LIST O F TA BLES .................................................................. .......... ............. xii

LIST O F FIG U RES ..................................................... ................................... xiii

A B ST RA C T ....................................................................... ...................................xxv

CHAPTER

1 M O TIVATION ...............................................................1

1.1 Scaling of Planar CMOS..........................................................
1.1.1 The Silicon Age and Moore's Law....................... ......................
1.1.2 Short Channel Effects ................................... ........................2
1.1.3 O their Problem s ................................................. .......... ................. 2
1.2 Silicon-on-Insulator (SOI) for CMOS ......................... ........... ..............3
1.2.1 Advantages over Bulk Silicon.......................... .........................3
1.2.2 Challenges for SO I..................... .... ... ... .......................... 4
1.3 Objectives and Statement of Thesis................................... ........................ 5


2 LITERATURE REVIEW .................................................................. 13

2.1 Ion Implantation and Damage Recovery.................................................. 13
2.1.1 Ion Stopping and Primary Defects..................... ........... .............. 14
2.1.2 Secondary Defects and Their Structure...................................... ............. 15
2.1.2.1 Submicroscopic interstitial clusters (SMICs)............................... 17
2.1.2.2 {311 defects ........................................ ........... ............. 19
2.1.2.3 Dislocation loops.............................. ... ...................21
2.1.3 Models for Defect Evolution....................... .............................23
2.2 Dopant Diffusion in Bulk Silicon........................ .............................25
2.2.1 Mechanisms of Dopant Diffusion...........................................................25
2.2.2 Equilibrium Diffusion ................................... ........................27
2.2.3 Non-equilibrium/Enhanced Diffusion......................................................28
2.2.4 TED of Boron ................................... ................ ........................ 28
2.2.5 Boron Interstitial Clusters (BICs).......................... .....................29
2.3 Silicon-On-Insulator (SOI) Materials ....................... ..........................31
2.3.1 SIM OX ................................ ..............................31








2.3.2 SO ITEC .............................................................. ........................ 33
2.4 Interstitial and Dopant Interactions at Si/SiO2 Interfaces ................................34
2.4.1 Si (001) Free Surface and Si(001)/SiO2 Interface Structure ....................34
2.4.1.1 Si(001) .....................................................35
2.4.1.2 SiO 2 ..................... ......................................36
2.4.1.3 Si/SiO2 interface..................................... ......................... 37
2.4.2 Point Defect Interactions at the Si/SiO2 Interface ...................................39
2.4.2.1 Interface effects on interstitial kinetics under oxidizing conditions40
2.4.2.2 Interface effects on interstitial kinetics under non-oxidizing conditions and
due to ion implantation ............................................................... ............46
2.4.3 Models for Interstitial Interactions at Si/SiO2 Interfaces .......................49
2.4.4 Dopant Segregation in the Proximity of Si/SiO2 Interfaces...................51
2.4.4.1 Thermodynamic considerations.................................................51
2.4.4.2 Dynamic boundary conditions................................................52
2.4.4.3 Static boundary conditions ...................... ......................54
2.4.4.4 Consequences of dopant segregation ............................................56
2.4.4.5 Models for dopant segregation..................................................57
2.5 Dopant Diffusion in SOI ...........................................................59
2.5.1 Boron Diffusion in SOI......................... ................................59
2.5.2 Donor Diffusion in SOI.................................. ......................62
2.6 Sum m ary .................................................................... .........................62


3 EXPERIMENTAL METHODOLOGY........................ .............. ............ 113

3.1 D esign of Experim ents......................................... ............. 113
3.1.1 Self-Interstitial Experiments....................... ....................... 114
3.1.2 Boron Activation Experiments ............................................................ 115
3.2 Analytical and Simulation Techniques ................................................ 116
3.2.1 Transmission Electron Microscopy (TEM).........................................116
3.2.2 H all Effect ................................................... .................................. 118
3.2.3 Four-Point Probe....................................... .......................... 121
3.2.4 Secondary Ion Mass Spectrometry (SIMS)......................................... 122
3.2.5 High Resolution X-Ray Diffraction (HRXRD)................................... 123
3.2.6 U T-M arlow e ....................................................................................... 125
3.2.7 Florida Object Oriented Process Simulator (FLOOPS)....................... 126


4 SELF-INTERSTITIAL EXPERIMENTS IN SOI................... .......... .......... 139

4.1 Introduction ............................. ..... ..... ................................... 139
4.2 Non-Amorphizing Implants ........................................................................ 139
4.2.1 Interface Effects on {311} Defect Evolution ......................................139
4.2.1.1 Experimental ....................................... ................ 140
4.2.1.2 Results ............................................................... 141
4.2.1.3 D iscussion................................ ........ .......... .. 143
4.2.1.4 Conclusions........................................... ........... 146









4.2.2 Kinetics of {311} Defect Evolution in SOI........................................... 146
4.2.2.1 Experim ental ........................................ ........... 146
4.2.2.2 Results ........................................ ....................... 148
4.2.2.3 D iscussion.............................. ....... ............ 149
4.2.2.4 Conclusions............................. 151
4.2.3 Interface Effects on Dislocation Loop Evolution ................................. 151
4.2.3.1 Experim ental....................................................................... ....... 151
4.2.3.2 Results ............................................................. 152
4.2.3.3 D discussion ........................................................ ....... ...... .... 154
4.2.3.4 Conclusions..................... ......... .................... 156
4.3 Amorphizing Implants ......................................... 157
4.3.1 Experimental....................... .......... ........................ 157
4.3.2 R results ............................................................ ......................... 157
4.3.3 Discussion......................... ........ ........................159
4.3.4 Conclusions ............................................................... 160
4.4 Sum m ary ................................................................... ........................ 161


5 MODELING EXTENDED DEFECT EVOLUTION IN SOI ................................... 203

5.1 Introduction ..................................................... ..... .. ...................... 203
5.1.1 Model Background......................... .... .....................203
5.1.2 M odeling Results ......................................................... 204
5.2 Sum m ary ................................................................. .......................... 205


6 INVESTIGATION OF BORON INTERSTITIAL CLUSTERING IN SOI..............211

6.1 Introduction ................................................................211
6.2 TEM Analysis of Boron Implanted SOI .........................................................212
6.2.1 Experimental....................... ........... ....................... 212
6.2.2 Results ......................................................... .. .......................... 213
6.2.3 D iscussion............................... ........................ ......................... 214
6.2.4 C conclusions ................................................................................... 215
6.3 Time and Temperature Dependence of Boron Activation in SOI ..................215
6.3.1 Experimental....................... .......... ........................ 215
6.3.2 Results ............................... ...... .... ................. .........217
6.3.3 Discussion.......................... ..........................219
6.3.4 C conclusions .................................................... .. .......................225
6.4 Concentration Dependence of Boron Activation in SOI ...............................226
6.4.1 Experimental....................... .. ....... .......................226
6.4.2 Results and Discussion.............................. ...................227
6.4.3 Conclusions ........................................... .........................229
6.5 Role of Strain on Boron Activation............................ ..........................230
6.5.1 Experimental................................................. 230
6.5.2 Results and Discusssion ................................. .. ......................231
6.5.3 Conclusions ......................... ..... ....................235








6.6 Relationship Between Boron Segregation and TED......................................235
6.6.1 Experimental................................................. 235
6.6.2 Results and Discussion............................ .......................236
6.6.3 C conclusions ..................................................... ........................ 238
6.7 Sum m ary ................................................................ ........................... 238


7 SUMMARY AND FUTURE WORK ............................... .........................278

7.1 Sum m ary .......................................................................................... 278
7.2 Future W ork............................................................................................... 280
7.2.1 Local Electrode Atom Probe (LEAP) for Monitoring Dopant Segregation in SOI
................ ............................................. ..............................................2 80
7.2.2 Modification of Surface Potential Using a MOS Capacitor Structure....281
7.2.3 Critical Amorphization Depth in SO.................................................281
7.2.4 Concentration Threshold for BIC Formation in SOI ...........................282
7.2.5 N-Type Dopants in SOI.................................................... 282


APPENDIX

A QUANTITATIVE TEM FOR MEASURING TRAPPED INTERSTITIAL POPULATIONS
........................................................ ................. .................................................2 85

B FLOOPS CODE FOR SIMULATION OF {311} DEFECTS IN SOI ....................288

LIST OF REFERENCES...........................................................................292

BIOGRAPHICAL SKETCH .....................................................................306














LIST OF TABLES

Table page

1-1. Advantages and disadvantages of SOI devices over bulk silicon [IB03]...................... 10

2-1. Advantages and disadvantages of ion implantation compared to gas source and solid
source diffusion. ....................... .......................................................... ................. 64

2-2. Approximate fractional interstitial and vacancy components for various dopants in Si... 74

2-3. Advantages and disadvantages of the SIMOX process.................................................. 78

2-4. Advantages and disadvantages of using Smart-cut process for fabrication of SOI
substrates ................................................................................ 79

2-5. Segregation coefficients determined for various impurities in Si during oxidation.......... 99

3-1. Implant conditions and anticipated defect microstructures of interest in current self-
interstitial experiments. Shaded boxes indicate implant conditions that were actually
studied. Note: microstructures are only valid for bulk Si..........................................128

3-2. Insight gained from analytical techniques with regards to the BIC evolutionary process
and boron segregation in SOI. .......................................................... 132

4-1. Equivalent annealing times assuming 3.7 eV activation energy for {311} defects in bulk
Si. .............................................................................. 175

4-2. Ion range statistics determined using UT-Marlowe and SRIM simulations..................176

4-3. Dose loss for 750 A and 1450 A SOI determined using UT-Marlowe.........................177

4-4. Extracted activation energies from Figure 4-18 for SIMOX, SOITEC and bulk Si........183













LIST OF FIGURES


Figure page

1-1. Schematic of bulk silicon MOSFET device showing different ion implanted areas
w within device [JO N 98]. .......................... .................. ........................... .....7

1-2. Moore's Law describing scaling of the number of transistors on a chip [M0065]...8

1-3. Applications for varying surface Si and BOX thickness [SIG03]. ...........................9

1-4. Comparison of operating characteristics for SOI and bulk CMOS [IBM03]........... 11

1-5. Schematic of CMOS cross section illustrating elimination of latchup path in SOI
[C O L 97]........................................ ........................... ............................. 12

2-1. Schematic of collision cascade produced by light ions (e.g. atomic weight less than
Si) and heavy ions (e.g. atomic weight greater than Si).....................................65

2-2. Evolutionary path for point defects produced by ion implantation........................66

2-3. Formation energy as a function of cluster size for self-interstitial defects in Si.
Closed diamonds represent formation energy for a compact cluster, while the open
triangles are for an elongated cluster [KIMOO]............................................67

2-4. Formation energy as a function of cluster size as determined by Cower et al.,
[C O W 99a]. ........................................................... ......... ........................68

2-5. 3D representation of {113} defect in Si lattice. Light gray balls show interstitial
chains along <110> direction [TAK91]. ....................... ..........................69

2-6. Atomic structure of planar {113} defect. Numbers represent rings different from
those in a perfect crystal [TAK94]........................... .............................70

2-7. Plan view TEM (PTEM) weak beam dark field (WBDF) micrograph of {311} defects
in Si.. ...... ................... ....... ............ ......................................... 71

2-8. Formation criteria for extended defects in Si [JON88]. ..........................................72

2-9. Mechanisms of dopant diffusion in the Si lattice [CR095]................................ 73








2-10. Enhanced and retarded diffusion of dopants under nitridation ambient conditions
[FA H 89]. ..................................................................................................... 75

2-11. Example of TED of boron due to the presence of {311 } defects. Sample was
implanted with boron at 19 keV, 3x1014 cm-2 and annealed at 7500C. The {311}s
regulate the release of the excess interstitials until their eventual dissolution at
longer tim es .......................................................................... .........................76

2-12. Energetics of evolutionary pathways for BIC formation [PEL99a].....................77

2-13. Schematic of the SOITEC process [SOI03]..................................... ...........80

2-14. Surface reconstruction of Si (001) (lxl) to (2x1). Larger circles represent surface
atoms, while smaller ones are one layer below [BAL88]..................................81

2-15. Schematic of vicinal Si(001) surface illustrating orientation of dimers on adjacent
steps. Note existence of two different single-layer steps SA and SB [ZAN00].....82

2-16. STM microscope image of vicinal Si(001) surface misoriented 0.50 along the [110]
direction. Alternating SA and SB steps are shown. Note smoothness of SA steps
and high density of kinks along SB steps [ZAN00]. ..........................................83

2-17. Structural basis of silicon dioxide the SiO4 tetrahedron. Note the constant
tetrahedral angle and varying Si-O-Si bond angle [BAL88]................................84

2-18. Phase diagram of crystalline forms of SiO2 [BAL88].........................................85

2-19. Continuous random network of A2B3 glass, similar to that present in fused silica
[CH I97]. .........................................................................................................86

2-20. Schematic illustration of Si/SiO2 interface with crystalline, cristobalite, form of
SiO2. Note presence of transition region consisting ofnon-stoichiometric SiOx
[BAL88]. ........................ .................... ............................................... 87

2-21. Structure of broken bond defects at the Si/SiO2 interface. (a) The E' defect consists
of a broken bond between two tetrahedra that would otherwise be bonded to O. (b)
The Pb center consists of a broken bond in Si that would be bonded to another Si
atom [BA L88].................................................................................................88

2-22. Schematic illustration of generation, diffusion, and recombination sources in SOI
and bulk structures [CR095]........................................................................ 89

2-23. Schematic of surface retrogrowth process proposed by Hu at an inert Si/SiO2
interface [HUS74, HUS94] ................................................ .........................90








2-24. Schematic cross sections of test structures used by Ahn et al., for determining (a) 1-
D and (b) 2-D interstitial kinetics [AHN87].............................................91

2-25. SOI structure used by Tsoukalas to study interstitial kinetics through an oxide (a)
before backside etching of W1 and (b) after etching [TS093]............................92

2-26. Difference in OSF length between thinned structure and control as a function of
oxidation time for range of temperatures [TS093]................................. ....93

2-27. Test structures used by Tsamis to monitor (a) depth dependence of interstitial
behavior and (b) lateral diffusion of interstitials under an oxidizing ambient
[T SA 95] .................................................................................................. 94

2-28. OSF length as function of distance from mask edge for SOI with varying surface Si
thickness after (a) dry and (b) wet oxidation at 11000C [TSA95]. Simulations were
performed using data from Taniguchi [TAN85]. ..............................................95

2-29. HRTEM micrograph showing zig-zag {311} defects produced by low energy Si'
implantation at 5 keV, 3x10 cm"2. Sample was annealed at 8100C for 10 minutes
[AGA97a] ................................................................................................... 96

2-30. PTEM micrographs showing defect evolution after Ge+ implantation at 5 keV and
10 keV, lxl015 cm2. (a) 10 keV unlapped, (b) 10 keV lapped, and (c) 5 keV.
Samples were annealed at 7500C for 60 minutes [KIN03]. ...............................97

2-31. Effect of impurity redistribution on segregation coefficient and diffusivity in SiO2
[G R O 64a]. .........................................................................................................98

2-32. SIMS profiles of dopant segregation for (a) arsenic, (b) phosphorus, and (c) boron
after oxidation at 1100 C for 30 minutes [SAK87]........................................ 100

2-33. Temperature dependence of the segregation coefficient during oxidation for B, P,
and A s [SA K 87].............................................................................................. 101

2-34. Test structure used by Charitat and Martinez to investigate boron segregation at a
static Si/SiO2 interface [CHA84]............................ ......................... 102

2-35. Boron segregation coefficient as a function of temperature for <100> orientation
under neutral annealing ambient. Stars indicate segregation coefficients for a Si
surface covered with a pad oxide only. Circles indicate segregation coefficients for
a Si surface covered with a pad oxide and nitride as shown in Fig. 2-34 [CHA84].
........................................................... ........................................................... 10 3










2-36. Boron segregation coefficient as a function of temperature for <111> orientation
under neutral annealing ambient. Stars indicate segregation coefficients for a Si
surface covered with a pad oxide only. Circles indicate segregation coefficients for
a Si surface covered with a pad oxide and nitride as shown in Fig. 2-34 [CHA84].
....................................................... ....................................... 104

2-37. SIMS profiles of B, 0.5 keV, lx1015 cm-2 annealed at 700 OC for 2 hours. Sample
was preamorphized with Ge, 15 keV, 1.2x1015 cm'2. Note uphill diffusion of boron
near surface, as well as gettering of boron to EOR damage around 34 nm [DUF03].
........................................... ............................................................................ 10 5

2-38. Potential energy diagram for electrons in p-type Si following Ar+ implantation.
Interfacial defects absorb positive charge from the bulk and create a space charge
region (SCR) and electric field pointing back into the bulk. Positively charged
interstitials are repelled back into the bulk by the field. Interstitials close to the
surface are able to recombine since the Fermi level approaches midgap [JUN04].
........................................................................................................................ 10 6

2-39. Simulation of boron TED experiment using model of Jung et al. Note incorporation
of surface band bending results in best fit to overall profile [JUN04].............. 107

2-40. SIMS profiles of boron after BF2 implantation at 40 keV, 1x1014 cm-2 into (a) 1986
and (b) 1988 SIMOX material. Anneals were 880 C for 100 minutes in a nitrogen
ambient. Note difference in pileup of boron at native oxide/surface Si interface
[N O R90] ................................................................................................. 108

2-41. SIMS profiles ofB implanted at 10 keV, lxl013 cm"2 into SOITEC substrates with
surface Si thickness of 60-70 nm and BOX thickness of 200 nm. Top curves show
Si signals obtained from SIMS. Following the B implant a 1050 OC, 60 second
anneal was performed. Solid curves had an additional Si+ implant at 40 keV,
5x1013 cm"2. Both samples were then annealed at 800 OC for 30 minutes. Note
increased pileup at both interfaces with addition of Si+ implant [VU099]........ 109

2-42. SIMS profiles ofB marker layers grown on SOITEC substrates using MBE. Solid
curves also had a Si+ implant at 25 keV, 1x1014 cm-2. Note enhanced segregation
of B to surface Si/BOX interface with addition Si+ implant [VU099] ............ 110

2-43. SIMS profiles of B from BF2 implanted at 60 keV, 7x1013 cm2 and annealed at
1000C for 5 sec in nitrogen. Surface Si thickness used were 530 A, 1050 A, and
1550 A [PAR99]........ ........................................... .......................... 111








2-44. SIMS profiles of P implanted at 36 keV, 7x1013 cm"2 and annealed at 10000C for 5
sec in nitrogen. Surface Si thickness used were 530 A, 1050 A, and 1550 A
[PA R 99]...................................................................................................... 112

3-1. Logic behind design of self-interstitial experiments. Implants are designated as
amorphizing or non-amorphizing. Appropriately altering implant energy and dose
allows observance of a particular extended defect. ......................................... 127

3-2. Experimental methodology invoked in first self-interstitial study of {311} defects and
dislocation loops for non-amorphizing implants. Samples were implanted at a
fixed dose with energies varying from 5 to 40 keV. Anneals were performed at
7500C for times ranging from 5 minutes to 2 hours. Note: all substrates were
fabricated using the SOITEC method .............................................. 129

3-3. Design of experiment 2 for determining kinetics of {311} defect evolution in SIMOX
and SOITEC substrates. ............................................................. 130

3-4. Experiment 3 methodology for determining effect of surface Si/BOX interface on
EOR dislocation loop formation ..................................... ............ 131

3-5. Experimental design for first study of boron activation in SOI......................... 133

3-6. Design of second experiment for investigating kinetics of BIC dissolution in SOI and
bulk Si. .............................. ..... ......................................................... 134

3-7. Principle behind weak beam dark field imaging in TEM for a edge dislocation. High
intensity occurs close to dislocation core because planes are bent back to Bragg
condition. From [WIL96] ................... ........................ 135

3-8. Illustration of Hall effect occurring in a p-type specimen. From [SCH98]........... 136

3-9. Primary components of Philips X'Pert System with 6-axis goniometer. From
[PAN 04] ................................................................................................... 137

3-10. Optics setup for rocking curve analysis using Philips X'Pert system. (a) Primary
optics hybrid mirror and (b) secondary optics triple axis detector. From
[L IN 04]............................... ................................ .. ... ............................ 138

4-1. UT-Marlowe ion profile simulations for Si implants into (a) 750 A and (b) 1450 A
SOI at 15, 30 and 48.5 keV, lxl104 cm 2 ....................................... ........... 162

4-2. Dose loss calculated from UT-Marlowe simulations for implant energies used in the
study. ..... ........ ............................................... 163








4-3. Weak beam dark field images ofSOI and bulk silicon for Si+, 15 keV, lxl014 cm2
implants after annealing at 750 OC for 5 and 15 minutes................................. 164

4-4. Concentration of trapped interstitials in all extended defects for Si, 15 keV, 1xl014
cm"2 annealed at 750 C. ..............................................................165

4-5. Weak beam dark field images of SOI and bulk silicon for Si+, 30 keV, lxl104 cm-2
implants after annealing at 750 OC for 5 and 15 minutes................................. 166

4-6. Concentration of trapped interstitials in all extended defects for Sit, 30 keV, lxl014
cm"2 annealed at 750 C. ............................................ 167

4-7. Concentration of trapped interstitials in only {311} defects for Si', 30 keV, 1xl014
cm-2 annealed at 750 OC. ....................................................... 168

4-8. Average size of {311} defects in SOI and bulk for Sit, 30 keV, xl 014 cm2 annealed
at 750 C...... .............. ..................................................................... 169

4-9. Weak beam dark field images of SOI and bulk silicon for Si+, 48.5 keV, lxl014 cm2
implants after annealing at 750 C for 5 and 30 minutes................................. 170

4-10. Concentration of trapped interstitials in all extended defects for Si, 48.5 keV,
1xl014 cm -2 annealed at 750 OC. ................................................................. 171

4-11. Concentration of trapped interstitials in only {311} defects for Si+, 48.5 keV, lxl014
cm "2 annealed at 750 C ............................................................................ 172

4-12. Average size of {311} defects in SOI and bulk for Si+, 48.5 keV, lxlO14 cm2
annealed at 750 OC....................................... ......................... 173

4-13. "+1" value as a function of surface Si thickness for different implant energies used
in the study. ............................................................ ................................ 174

4-14. Weak beam dark field micrographs of 750 A SIMOX and bulk silicon for Sit, 30
keV, lxl014 cm"2 implants after annealing at 700 C for 40 and 122 minutes.... 178

4-15. Concentration of trapped interstitials in {311} defects for Si+, 15 keV, lxl104 cm"2
annealed at (a) 700 C, (b) 750 C, and (c) 825 C. (Note: Sil = 6x109 cm-2 is TEM
detection lim it)........................................................................................... 179

4-16. Concentration of trapped interstitials in {311} defects for Si+, 30 keV, 1xl014 cm'2
annealed at (a) 700 C, (b) 750 C, and (c) 825 C......................................... 180

4-17. Concentration of trapped interstitials in {311} defects for Sit, 48.5 keV, 1x1014 cm-2
annealed at (a) 700 C, (b) 750 C, and (c) 825 C.........................................181








4-18. Plot of time constant as function of 1/kT for (a) 15 keV, (b) 30 keV, and (c) 48.5
keV ............................... .............................. ................................... 182

4-19. Ion profiles from UT-Marlowe for (a) 300 A, (b) 700 A, (c) 1600 A, and (d) bulk Si
after Si implantation from 5 -40 keV 2x101 cm2........................................184

4-20. Percentage of dose retained in surface Si layer of SOI for Si+ implants from 5 keV to
40 keV 2x1014 cm "2 ............................................ ...................................... 185

4-21. Plan-view TEM micrographs illustrating defect evolution in 300 A SOI and bulk Si
for 5 keV, 2x1014 cm after annealing at 7500 C. ........................................... 186

4-22. Concentration of trapped interstitials (Sil) in extended defects for 5 keV, 2x1014 cmn2
after annealing at 750 C. ............ ....................................................... 187

4-23. Plan-view TEM micrographs illustrating defect evolution in 300 A SOI and bulk Si
for 10 keV, 2x1014 cm-2 after annealing at 7500 C.......................................... 188

4-24. Concentration of trapped interstitials in extended defects for 10 keV, 2x1014 cm'2
after annealing at 7500 C .................................................... ........................ 189

4-25. Plan-view TEM micrographs illustrating defect evolution in 700 A SOI and bulk Si
for 20 keV, 2x1014 cm-2 after annealing at 7500 C.......................................... 190

4-26. Concentration of trapped interstitials in extended defects for 20 keV, 2x1014 cm-2
after annealing at 7500 C.................................................... ........................ 191

4-27. Plan-view TEM micrographs illustrating defect evolution in 700 A SOI and bulk Si
for 40 keV, 2x1014 cm-2 after annealing at 7500 C.......................................... 192

4-28. Concentration of trapped interstitials in extended defects for 40 keV, 2x1014 cm-2
after annealing at 7500 C ....................... ............ ............................. 193

4-29. UT-Marlowe RBS profile showing percent amorphization versus depth. Amorphous
layer is approximately 15 nm thick.... ......................................................... 194

4-30. PTEM micrographs of EOR loops in SOITEC and bulk Si after annealing at 835 OC
in nitrogen. Implant was Si, 5 keV, 1x1015 cm2 .................................. 195

4-31. QTEM data for EOR loops annealed at 8350C including (a) concentration of trapped
interstitials, (b) defect density, and (c) defect size. ................................. 196

4-32. PTEM micrographs of oxidation stacking faults (OSFs) that appeared to nucleate off
EOR loops upon further annealing at 835 C. Note significant difference in aspect
ratio between 300 A SOI and the other materials. Implant was Si, 5 keV, 1xl0"1
cm 2 .................................................. 197
cm................................................................. I................. 197








4-33. QTEM data for OSFs at 8350C including (a) concentration of trapped interstitials,
(b) defect density, (c) major axis length, and (d) aspect ratio.......................... 198

4-34. Concentration of trapped interstitials for both loops and OSFs at 8350C. Note
increase in concentration as OSFs begin to nucleate offEOR loops................ 199

4-35. QTEM data for EOR loops annealed at 9000C including (a) concentration of trapped
interstitials, (b) defect density, and (c) defect size. .........................................200

4-36. QTEM data for OSFs at 9000C including (a) concentration of trapped interstitials,
(b) defect density, (c) major axis length, and (d) aspect ratio..........................201

4-37. QTEM data for OSFs at 10000C including (a) concentration of trapped interstitials,
(b) defect density, (c) major axis length, and (d) aspect ratio..........................202

5-1. Trapped interstitial dose in SOI and bulk Si for Si', 15 keV, xl1014 cm"2 annealed at
750 C. Data points are from QTEM data, lines are FLOOPS simulations. Note
reduction in Sir in 750 A SOI as annealing proceeds.......................................206

5-2. Defect density for Si+, 15 keV, 1x1014 cm-2 annealed at 750 OC. Defects dissolve
faster in 750 A SOI as anneal time proceeds. Model predicts large decrease in
initial defect density in 750 A SOI. ............................................................207

5-3. Trapped interstitial dose in SOI and bulk Si for Si, 30 keV, Ixl014 cm-2 annealed at
750 OC. Note reduction in Sil in 750 A and 1450 A SOI as annealing proceeds.
Model predicts no {311} formation in 750 A SOI ..........................................208

5-4. Defect density for Si, 30 keV, 1x1014 cm"2 annealed at 750 C. A significant
enhancement in defect decay rate occurs in 750 A SOI. Model predicts no defects
form in 750 A SO ...................................................................... 209

5-5. FLOOPS simulation of defect size for 1450 A and bulk Si implanted at Si +, 30 keV,
lx1014 cm"2. Model overestimates differences in defects size between SOI and
bulk Si ......................................................................................... ..... 210

6-1. PTEM WBDF micrographs of defect evolution in SOI and bulk for B1, 6.5 keV,
3x1014 cm -2 ........................................................................................ 240

6-2. Trapped interstitial concentration (Sil) as a function of annealing time for B1, 6.5
keV, 3x1014 cm2 specimens.............................................................241

6-3. PTEM WBDF micrographs of defect evolution in SOI and bulk for B1, 19 keV,
3x10 cm 2 ............................................................................. ............ 242








6-4. Concentration of trapped interstitials as a function of time for B+, 19 keV, 3x10'4 cm
2 specim ens. ............................................................. ..........................243

6-5. UT-Marlowe ion profile simulations for B+ implants at 1 keV, 3.5 keV, 6.5 keV,
3x1014 cm"2. Note location of surface Si/BOX interface for 300 A, 700 A and 1600
A SO ......................................................................... ............................. 244

6-6. Percent retained dose of boron in surface Si layer as function of implant energy for
300 A, 700 A and 1600 A SOI. Calculated using UT-Marlowe ion profiles.....245

6-7. Isothermal Hall data for B+, 1 keV, 3x1014 cm-2 at 750 OC including (a) active dose,
(b) hole mobility, and (c) sheet resistance. Note that the active dose contribution
from the background (-5xl013 cm"2 to 6x1013 cm"2) must also be subtracted from
the bulk Si numbers.......... ....... ........................... .......................... 246

6-8. Isothermal Hall data for B+, 6.5 keV, 3x10'4 cm"2 at 750 C including (a) active dose,
(b) hole mobility, and (c) sheet resistance. Note that the active dose contribution
from the background (-5x1013 cm"2 to 6x1013 cm"2) must also be subtracted from
the bulk Si num bers.................................................................................... 247

6-9. Comparison of sheet resistance data measured by four point probe and Hall Effect for
(a) 1 keV, (b) 3.5 keV, and (c) 6.5 keV, 3x1014 cm"2 annealed at 750 OC. Solid
symbols and lines represent four point probe measurements and open symbols
represent Hall measurements............................................. .........................248

6-10. Isochronal Hall data for B, 1 keV, 3x1014 cm"2 after annealing 30 minutes showing
(a) active dose, (b) hole mobility, and (c) sheet resistance. Note that the active dose
contribution from the background (-5x1013 cm"2 to 6x1013 cm"2) must also be
subtracted from the bulk Si numbers. ........................ .........................249

6-11. Isochronal Hall data for B+, 6.5 keV, 3x1014 cm-2 after annealing 30 minutes
showing (a) active dose, (b) hole mobility, and (c) sheet resistance. Note that the
active dose contribution from the background (-5x1013 cm2 to 6x1013 cm"2) must
also be subtracted from the bulk Si numbers....................................................250

6-12. Boron concentration profiles from SIMS for 300 A SOI implanted at 6.5 keV,
3x1014 cm-2 then annealed at 750 C. Note segregation of boron into buried oxide.
........... .................... .........................................................................................2 5 1

6-13. Clustered dose in SOI and bulk Si for B+, 6.5 keV, 3x1014 cm"2 annealed at 750 OC.
Dose was obtained by integrating the B concentration profiles that lie above a level
of lx l019 cm 3 .......................................... ..................................... .. 252

6-14. Active fraction of boron in SOI and bulk for 6.5 keV 3x1014 cm"2 annealed for 30
minutes............... ...................................................................... 253








6-15. Carbon and oxygen SIMS profiles for 1600 A SOI and bulk Si implanted with B+,
6.5 keV 3x10 cm2, and annealed at 6000 C for 30 minutes..........................254

6-16. (a) Boron ion profiles from UT-Marlowe for 15 keV, 3xl0'4 cm"2 and 1xl0'5 cm2.
Vertical lines indicate the location of the surface Si/BOX interface. (b) Percent
retained dose for 750 A, 1450 A, and bulk Si. ......................... ....................... 255

6-17. (a) Sheet number versus time for 15 keV, 3x1014 cm'2 annealed at 825 C. Open
symbols indicate RTA and closed circles indicate furnace anneals (FA). (b)
Fractional active dose for 750 A, 1450 A, and bulk Si under same anneal
conditions. Note that the active dose contribution from the background (-5xl013
cm"2 to 6xl013 cm-2) must also be subtracted from the bulk Si numbers............256

6-18. (a) Hole mobility versus time for 15 keV, 3x1014 cm-2 annealed at 825 C. Open
symbols indicate RTA and closed circles indicate furnace anneals (FA). (b) Sheet
resistance for 750 A, 1450 A, and bulk Si under same anneal conditions. Note that
the active dose contribution from the background (-5x1013 cm"2 to 6x1013 cm'2)
must also be subtracted from the bulk Si numbers..........................................257

6-19. (a) Sheet number versus time for 15 keV, IxlO15 cm"2 annealed at 825 OC. Open
symbols indicate RTA and closed circles indicate furnace anneals (FA). (b)
Fractional active dose for 750 A, 1450 A, and bulk Si under same anneal
conditions. Note that the active dose contribution from the background (-5x1013
cm-2 to 6x1013 cm2) must also be subtracted from the bulk Si numbers............258

6-20. Sheet number versus time for 15 keV, 3x10'4 cm'2 annealed at 750 OC for 30
seconds. Note insignificant difference in activation between SIMOX and SOITEC
m materials. ............... ............................................................................259

6-21. Schematic of (a) lattice/surface miscut and (b) bonding misorientation present in
SOITEC materials. White lines in (b) show planes corresponding to wafer B, while
black lines correspond to wafer A ..................................... ........................260

6-22. Erroneous (004) w-20 rocking curves for 1600 A SOITEC wafers implanted with
B+, 3.5 keV, 3x1014 cm"2. Anneals were performed for 30 minutes at 750C.
Samples were aligned to substrate, rather than surface Si layer.........................261

6-23. (004) w rocking curves for 1600 A SOITEC wafers as-implanted with B+, 3.5 keV,
3x1014 cm-2. Misalignment between the surface Si Bragg peak and substrate peak
is shown........................................................................................... 262

6-24. (004) w-20 rocking curves for unimplanted 1600 A SOITEC wafers, illustrating
presence of Pendellosung fringes. Samples were aligned to surface Si layer.....263








6-25. (004) o-20 rocking curves for unimplanted, as-implanted, and annealed 1600 A SOI
implanted with B+, 3.5 keV, 3x1014 cm"2. Anneal was 9000C for 30 minutes...264

6-26. (004) o-20 rocking curves for unimplanted and as-implanted bulk Si implanted with
B, 3.5 keV 3x1014 cm -2. ..................................................... .......................... 265

6-27. (004) 0-26 rocking curves for bulk Si implanted with B+, 3.5 keV, 3x1014 cm-2.
Anneals were performed at 6000C, 7500C, 9000C, and 10000C for 30 minutes. 266

6-28. Comparison of (004) 0o-20 rocking curves for 1600 A SOI and bulk Si as-implanted
with B+, 3.5 keV, 3x1014 cm -2....................................... ........................... 267

6-29. Comparison of (004) w-28 rocking curves for 1600 A SOI and bulk Si implanted
with B+, 3.5 keV, 3x10'4 cm'2. Anneals were performed at 9000C for 30 minutes.
Oscillatory pattern belongs to SOI, while the high intensity peak is bulk Si......268

6-30. SIMS profiles ofSOITEC and bulk Si materials implanted with B, 1 keV, 3x1014
cm2. Anneals were 750C for 30 minutes. Note depletion of B on surface Si side
of interface, as well as pileup on BOX side................................................269

6-31. SIMS profiles of SOITEC and bulk Si materials implanted with B+, 6.5 keV, 3x1014
cm-2. Anneals were 7500C for 30 minutes.................... ......................270

6-32. QTEM data for B+, 3.5 keV, 3x1014 cm-2 annealed at 7500C for various times. Note
little difference in dissolution between SOI and bulk Si...................................271

6-33. Comparison of TED behavior in 1600 A SOI and bulk Si after implantation at B+,
6.5 keV, 3x1014 cm-2. Anneals were performed at 7500C for 120 minutes. Note
slight tail enhancement in 1600 A SOI compared to bulk Si...........................272

6-34. SIMS profiles of SOITEC and bulk Si materials implanted with B+, 19 keV, 3x1014
cm'2. Anneals were 7500C for 30 minutes....................................... ...........273

6-35. SIMS profiles of SOITEC and bulk Si materials implanted with B+, 1 keV, 3x1014
cm2. Anneals were 10500C for 30 minutes. Note box-shape profile after
annealing SOI materials. ................................... ............................274

6-36. SIMS profiles ofSOITEC and bulk Si materials implanted with Bt, 6.5 keV, 3x1014
cm"2. Anneals were 10500C for 30 minutes...............................................275

6-37. Retained concentration versus surface Si layer thickness for B 1, 1 19 keV, 3x1014
cm"2. Anneals were performed at 10500C for 30 minutes. Concentrations were
taken as the average concentration across the surface Si layer, while avoiding
transient effects near interfaces. ........................ ...............................276








6-38. Segregation coefficients versus surface Si thickness after annealing at 10500C for
various times. Implants were B+, 1 keV, 6.5 keV, and 19 keV, with a dose of
3x1014 cm 2. ............................................................... .......................... 277

7-1. Schematic of LEAP constituents near a surface containing microtips for analysis.
From [KELOO] ..........................................................................................283

7-2. Schematic of MOS capacitor structure that could be used to alter surface potential at
Si/SiO2 interface. ................ ............................................................. 284














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

INTERFACIAL PHENOMENA IN ION IMPLANTED SILICON-ON-INSULATOR
MATERIALS

By

Antonio Fernando Saavedra

August 2004

Chair: Kevin S. Jones
Major Department: Materials Science and Engineering

Silicon-on-insulator (SOI) allows for many short channel effects to be overcome,

leading to faster operating speeds and lower power dissipation in metal-oxide-

semiconductor (MOS) devices. Integration of SOI requires an understanding of

interstitial mediated processes, such as extended defect evolution and dopant activation.

It has been shown recently that ion implantation of dopants in SOI results in anomalous

diffusion profiles leading to dopant loss to the buried oxide (BOX), and preventing that

dopant from providing electrical conduction within the surface silicon film.

The first goal of this study was to investigate the role of the surface silicon/BOX

interface on {311} defect, dislocation loop, and oxidation stacking fault (OSF), evolution

in partially-depleted (PD) and fully-depleted (FD) SOI. This helps to elucidate the

degree to which the interface acts as a sink for excess interstitials produced by ion

implantation. Energies ranging from 5 keV to 48.5 keV and doses from lxl04 cm" to

Ix 105 cm-2 were used to understand the effect of the interface on non-amorphizing and


xxV








amorphizing Si' implants. Nucleation of {311 }s and loops was shown to be strongly

dependent on the implant energy and surface silicon thickness. The FLOOPS process

simulator was used to simulate the {311} defect evolution in SOI. The model

overestimates the effect of the surface Si/BOX interface on interstitial recombination

indicating that it is not as strong a sink for interstitials as previously thought.

Investigation of boron interstitial clustering (BICs) in SOI comprised the second

part of the study. Boron implants ranging from 1 keV to 19 keV, 3x10'4 cm2 were used

to provide varying amounts of dose loss to the BOX. Hall effect data determined the

time and temperature dependence of boron activation in SOI. SIMS, TEM, and HRXRD

were used investigate the roles of strain and dopant segregation on boron activation. At

temperatures less than 750 "C, the active dose of boron in SOI was slightly less than bulk

Si. However, above 750 *C the activation in SOI began to approach bulk Si. Truncation

of the boron implant profile by the interface was found to reduce BIC formation within

the surface Si layer.













CHAPTER 1
MOTIVATION

A number of alternative materials are being considered to replace the traditional bulk

silicon substrate. One of the most promising candidates is silicon-on-insulator (SOI). This

chapter focuses on the roadblocks to continued scaling of planar bulk complementary metal-

oxide-semiconductor (CMOS) devices, as well as the advantages of moving to SOI. Lastly, the

goals for the current investigation are stated.

1.1 Scaling of Planar CMOS

The continued drive for computers that will perform greater numbers of operations per

second requires the shrinking of the components of the metal-oxide-semiconductor field effect

transistor (Figure 1-1). Moore's Law provides guidance so that the demand for computational

power can be supplied based on the economy. Scaling requires higher doping levels, shallower

junctions and narrower gate lengths. Unfortunately, these requirements lead to a number of

parasitic effects, such as latchup, short channel effects, junction capacitance, and power

dissipation.

1.1.1 The Silicon Age and Moore's Law

Different stages of development within a society are often classified according to

advances in materials science. Examples of these include the Stone Age, Bronze Age and Iron

Age. Since the invention of the transistor in 1948 [BAR01] and the integrated circuit in 1959

[KILOI], the Silicon Age has transformed our world in ways that could not have been imagined

50 years ago.








In 1965, Gordon Moore put into context the trend (Figure 1-2) that has been the driving

force of the semiconductor industry [M0065]. Although Moore's prediction that the number of

transistors on a chip would double every year overestimated the actual industry performance, it

laid a foundation. Moore's Law, as it became known, predicts that the number of transistors on

a chip doubles every 18 to 24 months. In essence, it requires continued miniaturization of the

devices on an integrated circuit chip.

1.1.2 Short Channel Effects

The reduction in dimensions of the MOSFET leads to undesirable device characteristics,

known as short channel effects. This is primarily a result of a decrease in the threshold voltage

as the channel length is reduced [TAU98]. Some of these effects include charge sharing, drain-

induced barrier lowering (DIBL), punchthrough, hot electron effects, and increased leakage

current. In general, as the junction depth, xj, and gate length, L, are reduced the short channel

effects (SCE) become more pronounced. This is one reason significant research is now being

conducted in ultra shallow junction formation technologies [JON98].

In order to curb SCE and tailor threshold voltage, a number of different ion implantation

steps are necessary as illustrated in Figure 1-1. A threshold voltage adjust channel implant is

used to modify threshold voltage for the NMOS and PMOS devices. The source/drain extension

implants provide a concentration gradient from the deep/source drain to the conducting channel

in order to reduce the maximum electric field. Halo/punchthrough implants prevent the

intersection of source and drain depletion regions when the device is turned on, thus preventing

alternate paths for current flow [ZIEOO].

1.1.3 Other Problems

Latchup occurs when a parasitic NPNP thyristor is triggered between adjacent NMOS and

PMOS devices. This can result in abnormally large currents, as well as taking the supply








voltage straight to the ground potential, causing failure of the devices. Latchup is accounted for

by building CMOS devices in a lightly doped epitaxial Si layer deposited on a highly doped

substrate. Alternatively, high energy ion implantation can also be used to create deep, highly

doped wells below the devices [PLUOO].

Another constraint being placed on bulk silicon is the increased power used by chips as

the transistors are made smaller [FRAO2]. One method for reducing power consumption in a

device is to use a lower operating or supply voltage. Portable electronic systems rely heavily on

devices that operate under a low supply voltage. It is not without coincidence that alternative

materials are now being considered to help simplify processing [COL97], reduce SCE [VEE89],

and control power consumption [FRA02].

1.2 Silicon-on-Insulator (SOI) for CMOS

By incorporating a thin layer of silicon on top of an insulating material, e.g., silicon

dioxide, many of the problems plaguing the scaling of bulk Si CMOS can be remedied.

Altering the thickness of the layers allows for SOI to be tailored towards a variety of device

applications. This is illustrated in Figure 1-3. However, there are several challenges to scaling

and integration of SOI; these are addressed below.

1.2.1 Advantages over Bulk Silicon

SOI is an ideal candidate for low voltage/low power, as well as radiation hard electronic

applications [COL98, PLO00, ADA98]. It allows for increased chip speed, lower operating

voltage, reduced parasitic capacitance, elimination of latchup, and reduced susceptibility to

interference from outside radiation sources [COL97]. SOI is also able to operate over a wider

temperature range making it desirable for high temperature operation.[FLA95] Table 1-1

summarizes the advantages and disadvantages of using SOI over bulk silicon for CMOS

applications.








Figure 1-4 compares operating characteristics for SOI and bulk silicon CMOS. This

shows that the threshold voltage can be tailored depending on the surface Si thickness. Another

advantage of SOI is that the presence of the BOX prevents interference from cosmic radiation

(a.k.a soft error), making SOI a popular material for use in satellite communication and deep

space systems.[HIR99] An SOI substrate may also be used to completely eliminate latchup

formation. This is illustrated in Figure 1-5.

Perhaps the most important reason for switching to an SOI substrate is for low

voltage/low power operating conditions. Using SOI allows for a 2X-3X reduction in power

consumption compared to bulk CMOS. Another advantage of using SOI is a simplification in

manufacturing by reduction of processing steps. It eliminates the need for a deep well

formation step, which is required in bulk CMOS. Also, use of FD SOI does away with the need

for an anti-punchthrough implant since the depletion region extends the entire thickness of the

channel region.[COL97] It is evident that using SOI provides for numerous advantages over

bulk silicon for fabrication of future CMOS integrated circuits.

1.2.2 Challenges for SOI

Any fair discussion should also mention the disadvantages and challenges that are posed

by SOI. Most of these challenges relate to the kink effect and a number of floating body effects

that occur during device operation.[COL97] The kink effect shows up when high potentials are

applied to SOI MOSFETs after saturation has been reached. This creates electron hole pairs due

to impact ionization within the channel of the device and leads to unreliability in device

performance.

SOI MOSFETs also require the fabrication of two gates rather than the singular gate

used in bulk MOSFETs. The front gate is used to control the conduction within the channel,

while the back gate is typically in a grounded configuration. This helps prevent an inversion








layer from forming at the surface Si/BOX interface. [COL97] However, this adds to the

complexity of fabricating an SOI MOSFET.

Perhaps the single greatest challenge to the implementation of SOI is its lack of

knowledge base. Bulk Si has been the standard substrate for over three decades and a plethora

of information is available. For example, very little is known about how the presence of an

additional silicon/silicon dioxide interface affects dopant diffusion and activation. Anomalous

diffusion profiles have been observed in SOI materials recently, adding to the complexity of

developing models to design advanced devices. [PAR99] It is also not known how the surface

Si/BOX interface affects interstitial release from extended defects such as (311) defects and

dislocation loops. Understanding this is critical to understanding why dopants diffuse the way

they do in SOI.

1.3 Objectives and Statement of Thesis

The goal of this research is to understand the role of the surface Si/BOX interface on

interstitial mediated processes in SOI materials. This is done by first determining how self-

interstitial populations are affected by the interface. Once this is understood, it can be applied to

the problem of dopant diffusion and activation. This thesis provides significant scientific

contribution in the following areas:

-Proximity of damage to the surface Si/BOX interface on {311} defect and dislocation
loop evolution for non-amorphizing implantation in fully depleted and partially
depleted SOL.
-Kinetics of the {311 dissolution reaction for non-amorphizing implants in SOI.
-Effect of proximity of damage to the surface Si/BOX interface on {311} defect and
dislocation loop evolution for amorphizing implantation in FDSOI and PDSOI.
-Development of a physically based model for predicting interstitial evolution in
FDSOI and PDSOI.
-Investigation of time, temperature, energy, and concentration dependence of boron
activation in FDSOI and PDSOI.








-Determination of mechanism for low electrical activation of boron in SOI.





























Figure 1-1. Schematic of bulk silicon MOSFET device showing different ion implanted areas
within device [JON98].













Processor 5 a--- t i g B a
Generation g 2






-Moore's Data
-Moore's Prediction

S--Actual Industry


1.E+C

1.E+0

SI.E+O

a I.E--

01.E+
U I.E+O

. 1.E+O


I.E+O

I.E+O
1.E+0


Year


Figure 1-2. Moore's Law describing scaling of the number of transistors on a chip [M0065].


)8

7

6

15

4

13


P performance
2

1

10
1950 1960 1970 1980 1990 2000 2010































Figure 1-3. Applications for varying surface Si and BOX thickness [SIG03].






10

Table 1-1. Advantages and disadvantages of SOI devices over bulk silicon [1BI03].


FD SOI PD SOI Bulk
Junction capacitance Small Small Large
Body effect Small Large Large
Floating body effect Small Large None
Vth control Difficult Easy Easy
Subthreshold Small Larger Large
Leakage
S/D silicidation Difficult Easy Easy
Lay-out area Small Small
Circuit design Easy Difficult
Scalability Difficult Easy Difficult
Manufacturability Difficult Easy






11



SHOiW CANNL IWP ~CT IN THIN IM
'Dopingl =E7l,to*= 7nM


Channel luoIlh lin


Figure 1-4. Comparison of operating characteristics for SOI and bulk CMOS [IBM03].


0A







02


at 01=-


a 0
D 0.1 U2 U U 0.5 8as 0


I






12



GND our Vdd







P-type substrate
IN

GND OUT Vdd

P N



S Silicon substrate






Figure 1-5. Schematic of CMOS cross section illustrating elimination of latchup path in SOI
[COL97].














CHAPTER 2
LITERATURE REVIEW

The need for a faster and more efficient CMOS device has led to the insurgence of SOI.

This chapter reviews the status of SOI technology to date. Scaling of integrated circuits has

relied heavily on the ability of ion beams to create shallow, abrupt as-implanted doping profiles

in bulk Si. Unfortunately, post-implant thermal processing must be performed in order for the

dopant atoms to occupy substitutional lattice sites and contribute electrically. This can result in

unacceptable junction depths and sheet resistance due to dopant interactions with point defects.

For this reason, fundamentals of ion implantation, atomistic diffusion, and segregation in the

proximity of interfaces are also reviewed in the context of Si microelectronic processing.

Emphasis is placed on the Si/SiO, interface since it is of most interest in the investigations of

later chapters.

2.1 Ion Implantation and Damage Recovery

The discovery that atoms in the gas phase could be ionized by their bombardment with

electrons is credited to the German physicist, Eugen Goldstein. [GOL86] The work of Wien

[WIE98], Thomson [THO12], Rutherford, and Bohr [BOH13] was also critical to understanding

the behavior of ionized particles in electromagnetic fields [MOF95, WIE99]. Ion implantation

was first proposed by Shockley as a method for fabricating doped regions within semiconductor

devices. [SH003] Ever since, it has been the preferred technique for introducing dopants into

silicon in controlled amounts. It offers a number of advantages over gas and solid source

diffusion, shown in Table 2-1. Ion implantation is a very versatile process that has also been








used for introducing gettering layers, synthesis of compounds, and surface modification of

metals and polymers. Unfortunately, the main drawback of using ion implantation is resulting

damage to the target, consisting mainly of point defects. For silicon, this damage can lead to the

deleterious effects of transient enhanced diffusion, dopant-defect clustering, as well as leakage

current within the depletion region of transistors.

2.1.1 Ion Stopping and Primary Defects

The process of ion implantation is a highly non-equilibrium one. Physical, chemical and

structural changes may occur when the ions become embedded in the target material. A number

of elastic and inelastic effects also result, leaving the silicon in a damaged, metastable state.

Lattice atoms may be displaced from their equilibrium positions when the displacement energy

(15 eV for Si) is exceeded. As an energetic ion comes to rest, it may undergo a number of

collisions with atoms in the Si lattice, which leads to the production of a damage or collision

cascade. This damage cascade, illustrated in Fig. 2-1, consists of interstitials, vacancies,

amorphous regions, ionized atoms, etc. These defects produced in the "as-implanted" state are

typically referred to as primary defects [ZIEOO].

The density of a damage cascade depends significantly on the ion mass and stopping or

energy-loss mechanism. The two primary stopping mechanisms are nuclear and electronic

stopping. Nuclear stopping is characterized by a significant transfer of energy between the

nuclei of the energetic ion and the nuclei of a lattice atom. This results in a very dense damage

cascade. An ion that undergoes electronic stopping is decelerated by interacting with the

electron cloud surrounding a lattice atom. Thus, the energy losses in electronic stopping are

much less per stopping event and the collision cascade is less dense. In general, nuclear








stopping is observed at lower implant energies and heavier ions (e.g., As, Sb), while electronic

stopping occurs at higher energies and for low mass ions (e.g., B, H).

Damage cascades are also affected by the ability of the ion to channel, which is related

to the ion mass, as well as the crystal orientation relative to the incident ion beam. Channeling

refers to the phenomenon whereby an ion is able to traverse great distances into the crystal by

moving through the interstices present in the lattice. For this reason, {100} Si wafers are

typically oriented relative to the beam direction with a tilt of 7* in the [1101 direction followed

by a rotation of 22* around the [100] direction. This aids in reducing channeling and improving

the reproducibility of implant profiles. Thin screen oxides and pre-amorphization are also

common procedures for controlling channeling. [ZIEOO]

2.1.2 Secondary Defects and Their Structure

A large increase in the excess interstitial population occurs after ion implantation.

Secondary defects are those that form during subsequent thermal processing and are nearly

exclusively extrinsic, or interstitial, in nature. Post-implant annealing is always required since

the majority of implanted dopant ions are not on substitutional lattice sites. These implanted

dopants do not contribute electrically and are considered inactive. Upon annealing, the

damaged silicon lattice tends toward a more equilibrium state. Excess interstitials undergo a

number of evolutionary processes, shown in Fig. 2-2, in order to reduce the free energy

associated with the silicon lattice. These processes may be broadly classified into

recombination and clustering. Recombination occurs when the strain field surrounding an

interstitial interacts with that of a vacancy resulting in a mutual attraction and annihilation.

Frenkel pairs, a interstitial and vacancy pair around a host lattice site, may recombine during

implantation or annealing at temperatures <600"C. Clustering of excess interstitials often








results in the formation of submicroscopic interstitial clusters (SMICs), {311} defects, and

dislocation loops in order to reduce the Gibbs free energy of the system. While the final

annealed state may not be an entirely equilibrium one, it is desired that the final state does not

significantly change with time at the temperature required for device operation.

The location of the majority of excess interstitials depends on whether the implanted

dose is sufficient to produce enough disorder in the silicon lattice (~10%) to create an

amorphous layer. This dictates where the secondary defects form relative to the surface. For

non-amorphizing implants the majority of excess interstitials lie around the projected range of

the implant. In the case of amorphization, the excess interstitials are found just beyond the

amorphous-crystalline interface after regrowth of the amorphous layer has commenced. Based

on dislocation loop analysis of non-amorphizing implants, the number of excess interstitials was

found to be approximately equal to the implanted dose [JON88]. This was later termed the "plus

one" model [GIL91]. However, it has been shown that the "plus one" model varies depending

on the ion mass [HER98, PEL98].

Secondary defects are particularly important because they are believed to drive the

phenomenon of transient enhanced diffusion (TED). They do this by maintaining an interstitial

supersaturation until their eventual dissolution. [EAG94] Extended defects, such as {311}s and

loops, may also act as sources of leakage current in devices [LAN86]. This occurs because the

strain field of the dislocation behaves as a mid-band gap recombination center. Jones et al.,

provided a classification scheme for extended defects produced upon thermal annealing of

silicon [JON88]. {311} defects and dislocation loops are observed for non-amorphizing

implants (Type I) as well as continuous amorphous layers (Type II). Hairpin defects (Type III)

result from imperfect regrowth of a continuous amorphous layer. Clamshell defects (Type IV),








also associated with solid phase epitaxial regrowth, occur whenever a buried amorphous layer is

formed. These defects form at the intersection of the two amorphous-crystalline interfaces upon

regrowth. Precipitation related defects (Type V) are observed when an impurity is implanted to

very high concentrations above the solid solubility of the impurity in the matrix. {311} defects

and dislocation loops are discussed in detail below, since they are the two extended defects of

most interest in the present study.

2.1.2.1 Submicroscopic interstitial clusters (SMICs)

As their name suggests, submicroscopic interstitial clusters (SMICs) are secondary

defects that cannot be resolved optically using current microscopy techniques. SMICs, after the

interstitial point defect, are believed to be the basis from which the microscopic extended

defects evolve, although a structural transformation appears necessary [COF99]. Unfortunately,

very little is known about how SMICs evolve, as well as their influence on TED. It has been

shown that TED can occur in the absence of extended defects, providing a basis for the

existence of SMICs [ZHA95]. Due to their small size, detection is commonly done using deep

level transient spectroscopy (DLTS) [BEN97, BEN98], photoluminescence (PL) [COF99,

LIB01], and electron paramagnetic resonance (EPR). Ab initio calculations also provide insight

into the energetically favorable SMIC configurations and their evolutionary pathways.

Benton et al., [BEN97] implanted p-type Czochralski (CZ) and epitaxially grown

substrates with Si' at energies of 145 keV 2 MeV and doses of lx10' cm2 5x10' cm2.

DLTS was used to monitor the low temperature (100 680 "C) evolution of point defects into

SMICs and (311} defects. At doses less than lx10' cm"2 nearly all Frenkel pairs recombined

and SMICs were not observed to form. For higher doses (lxl102 7x0"3 cm"2) annealed at

temperatures above 600 C, two self-interstitial type defects were observed at E,+0.29 eV and








E,+0.48 eV in the absence of {311) defects. Annealing of the 5x1013 cm-2 above 680 *C

resulted in the formation of {31 l}s and a DLTS signal at E,+0.50 eV. The subsequent decrease

in the two DLTS signals at Ev+0.29 eV and E,+0.48 eV led the authors to conclude that the

SMICs are either the precursors of the {311} or they compete as sinks for the self-interstitials.

In a similar experiment for n-type material, Benton et al., [BEN98] identified 5 DLTS signals

associated with the interstitial type defects: E,-0.14 eV, E,-0.29 eV, E,-0.37 eV, E,-0.50 eV, and

E,-0.58 eV. They indicate that the defects at E,+0.29 eV and E,+0.48 eV are related to those at

E,-0.29 eV and E,-0.50 eV since they show similar annealing characteristics. The defect at E,-

0.58 eV appeared only in the presence of a higher oxygen concentration. They conclude that the

thermal stability of interstitial clusters was enhanced due to an increase in the interstitial

concentration as the dose increased.

The appearance of a sharp peak in the PL spectrum at 1376 nm was observed by Coffa et

al., to indicate a structural transformation from SMICs to {311 }s [COF99]. However, they were

unable to identify the size or configuration of the SMICs present in their specimens. This

problem has led to a number of studies using ab initio total-energy calculations [PAY92] and

inverse modeling techniques.

Kim et al., used tight-binding and ab initio local density approximation simulations to

determine the formation energy, Er, of interstitial clusters ranging from the di-interstitial (n=2)

to the {311} (n=co), where n denotes the number of interstitials in the cluster [KIMOO]. Figure

2-3 shows the formation energy as a function of the number of interstitials. E, can be seen to

decrease as the cluster size increases until it approaches the most stable configuration of the

{311}. On the other hand, Aral et al., found the most stable configuration to occur when n=4

[ARA97, KOH99]. This was further supported by Cowem et al., which found local minima in








the formation energy at n=4 and n=8 (Fig. 2-4) [COW99a]. This has since been used to model a

number of boron TED experiments by coupling of a SMIC model with a diffusion model

[COW99a, LAM03].

2.1.2.2 {311} defects

The most studied of all the extended defects in Si is the {311} (a.k.a. {113}) defect, due to

its direct link to TED [EAG94]. Since then, considerable effort has been undertaken to better

understand their characteristics. Hundreds of experiments have been conducted to determine

their affect on the diffusion of every technologically important dopant in Si. These experiments

are made easier since the {311} is microscopic and can be easily monitored by use of the TEM.

This allows the defect microstructure to be correlated with dopant diffusion, commonly

measured using secondary ion mass spectrometry (SIMS) or Rutherford backscatter

spectrometry (RBS). It should be mentioned that {311} defects are not a requirement in order

for TED to be observed, both SMICs and dislocation loops also drive TED.

Figure 2-5 shows a 3D representation of a {311} defect within the Si lattice. Two types of

{311} defects may be observed depending on whether or not they are elongated/rod-like or

planar. Both defects lie on the {311} habit plane and extend in <110> directions. A burger's

vector of b = a/25 <116> has been measured using high resolution TEM (HRTEM) by Takeda

et al., [TAK94] Planar {311} defects are less studied, and may be produced using irradiation

with high energy electrons [TAK951. They have a stacking periodicity of a/4 <110> and consist

of structural units of 5-, 6-, 7- (I units), and 8- (O units), membered rings, as shown in Fig. 2-6.

The interstitial atoms are found within the 6-membered ring. Note also that no dangling bonds

are left in the {110} cross section. However, since the I and O units are not necessarily periodic

the interstitial density around planar {311}s varies by 5 nm2 [TAK92]. It should be mentioned








that the planar defects are not true stacking faults although a small fringe contrast may be

observed in the HRTEM [CLA03].

For ion-implanted Si, the rod-like {311} defect is commonly observed, shown in Fig. 2-7.

Their elongation results due to the fact that self-interstitials can be added along the {110} cross

section without introducing dangling bonds. For this reason, these defects do not significantly

change in width ~4 nm [EAG94]. Using the structure of Takeda [TAK94], which found an area

density of interstitials of 5.1 5.5 nm2 along the {113} cross section, the total interstitials

within a rod-like {311} can be estimated to be approximately 26 nm '. This estimate commonly

serves as a basis for quantitative TEM studies to determine interstitial populations within

extended defects.

It should be mentioned that below a threshold dose, {311} defects do not form and only

SMICs might be present. This threshold is debatable, but appears to be between 7x10'2 [ZIEOO]

and Ix1013 cm2 [LIB01] Si' doses. These doses are easily reached in modem IC fabrication, so

{311} defects will nearly always form. However, the misconception that {311} defects are the

sole source of interstitials should be avoided.

{311} defects can be made to dissolve upon annealing above 700 *C. Release of

interstitials from {311} defects has been shown to follow an exponential relationship according

to

Si, Si(O)e"-' (2.1)

where Si, is the planar density of interstitials trapped in {311}s, Si(0) is the pre-exponential, t is

the anneal time, and T is the time constant for dissolution. This time constant obeys first order

reaction kinetics to yield an activation energy for dissolution via the Arrhenius relationship

S= r(0)e -E' T (2.2)








where t(0) is the pre-exponential, E, the activation energy for {311) dissolution, k Boltzmann's

constant, and T the temperature in Kelvin. The value of E, for {311 } defects has been

determined to be approximately 3.7 eV [SOL91].

{311} defects do not necessarily have to dissolve, though. Li and Jones have qualitatively

shown that {311 } defects are the source of dislocation loops for non-amorphizing implants.

They showed, via in-situ HRTEM, that {311} defects can either dissolve or undergo an

unfaulting reaction to form dislocation loops [LIJ98]. These defects are the subject of the next

section.

2.1.2.3 Dislocation loops

Somewhat less studied in comparison to the {311} defect is the dislocation loop, although

it can also drive TED long after {311} defects have dissolved [ZIEOO, NODOO]. Dislocation

loops can be stable at moderately high temperatures (750*C 850*C) for hours. Another

drawback of these defects is their ability to provide leakage current paths and degrade carrier

lifetimes when lying across a junction. This is due to the introduction of localized energy levels

that sit near the middle of the band gap in Si [MIY97, BUL78]. On the other hand, a great

advantage can be gained by using dislocation loops to getter out metallic impurities. By

introducing a dislocation loop band well below the device junctions, any metallic impurities

lying in the active regions will be attracted to the strain field introduced by the loop band

[CHA97].

These defects always form under amorphizing implant conditions, assuming the implant

energy is not ultra low energy, as well as non-amorphizing conditions when the dose is

marginally high. Figure 2-8 shows the conditions under which extended defects, particularly

loops, are expected to form for varying doses and ion mass [JON88]. Faulted Frank loops and








perfect elongated loops are the two common dislocation loops observed in ion implanted Si.

They are both two-dimensional precipitates placed in between adjacent {111} planes of Si, but

have different burger's vectors. The faulted loops have a b = a/3 <111>, while the perfect loops

have a b = a/2 <110>. The planar density of interstitials is believed to be approximately the

same, 1.566x101 cm2. Loops that form as a result of amorphizing implants are termed end-of-

range (EOR), but bear no relation to the faulted or perfect loop; both types are observed as EOR

loops [CLA03].

As mentioned, compared to {311} defects, dislocation loops are much more thermally

stable. They exhibit an activation energy for dissolution in range of 5 eV. Another key

difference is their ability to Ostwald ripen, allowing a larger dislocation loop to absorb the

interstitials from a smaller one [JAI02]. Thus, the larger dislocation loops are more stable. Still

an issue of debate is whether or not {311} defects truly Ostwald ripen. Moller et al., found the

average size of {311} defects to increase as annealing proceeded [MOL98]. Others have noted

that {311 }s should acquire an equilibrium shape at long times, but the aspect ratio should be

constant during ripening [EAGOO, COWOO]. However, Law and Jones [LAWOO] have used the

experimental results of Li [L1J98] to develop a model for {311} evolution that does not depend

on Ostwald ripening. They note that dissolution of {31 l}s depends on the ability of interstitials

to hop off the ends of the defects, thus smaller {311 }s dissolve faster than larger ones, but not

necessarily because they are more stable. Similarly, interstitials can only be added to the {311}

by attaching to the ends of the defect, whereas in dislocation loops the interstitials may attach

anywhere along the edge.








2.1.3 Models for Defect Evolution

A number of models for predicting the evolution of {311} defects and dislocation loops

have been developed, but only after their link to TED. Only recently has coupling between

{311} models and loop models been achieved. Clearly, the ultimate goal is to provide a

universal model that accurately predicts all the evolutionary stages of clustering starting with the

self-interstitial and ending with its dissociation from a secondary defect.

The first {311} model to be developed was based on first order kinetics, but was only

applied to the results of one experiment [RAF96]. In it, the interstitial release rate was

determined by the hopping frequency and binding energy to the cluster, while the cluster growth

was depended on the ability of the cluster to trap diffusing interstitials. Unfortunately, this

model did not consider defect size or the dependence of the interstitial binding energy on cluster

size. This was the basis of the first model of Law and Jones [LAW96], as well as others

[HOB97, GEN97], which used a two-moment method to predict the {311} clustering and

dissolution process. However, the models of Hobler et al., [HOB97] and Gencer and Dunham

[GEN97] used parameters that allowed for an energy dependence of the defect based on their

size.

This was the motivation for the second model of Law and Jones [LAWOO] since a line

defect should not exhibit such dependence. This model incorporated three new ideas. First, that

the {311} defect size does not affect the binding energy of interstitials to it. Second, the

dissolution of the defect is governed by the ability of the interstitials to dissociate from the ends

of the defects rather than their diffusion to the surface. Third, nucleation of the defects is

heterogeneous in nature, i.e. defects form from damage created by the ion implantation process.








Based on this, the number of interstitials in {311 }s and SMICs are solved, as well as the defect

size as a function of time according to

dC, D3,(C, -C3I.) (2.3)
dt T3,,


dr = 3 (2.4)
dt 3,,, C31,,,1

dCSMc CSMc(CI CsMCE) (2.5)
dt TSMIC

where C3,, is the concentration of interstitials trapped in {311} defects, D3,, is the density of

{311} defects, T3,, is the dissolution time constant for {311 }s, C, is the total concentration of

interstitials, C3,, is the equilibrium concentration of interstitials in {311 }s, and t is time.

Notation for parameters with the subscript SMIC apply to the SMIC defects.

The coupled {311} and loop model developed by Avci etal., [AVCO4] is based partly

off the model of Law and Jones [LAWOO], with the slight addition of a nucleation rate term.

This model is reviewed in detail since it forms the basis of the model developed in the current

studies. It assumes the dislocation loop size to be governed by the interaction of the loop

boundary with point defects. It calculates the effective equilibrium concentrations of point

defects according to that given by Borucki [BOR92]

C gC, (P)e k'"r (2.6)

CVb g-C, (P)e I I T (2.7)

where Cb, is the effective equilibrium concentration of interstitials at loop boundaries, Cvb is the

effective equilibrium concentration of vacancies, g, is a geometric factor (-0.7), C,' is the

equilibrium concentration of interstitials, Cv" is the equilibrium concentration of vacancies, P is

pressure, and AE, is the change in the defect formation energy as a result of self-force of a








dislocation loop developed by Gavazza et al. [GAV76]. The interstitial and vacancy continuity

equations are then modified as

SV [D,C (P)V( C,)]-K,(C,C, C;(P)C (P)P))-K, (C, -C,b)f(R)dR (2.8)
t C: (P) 0,

-V[DC',(P)V( )]-K,(C,Cv-C (P)C (P))-K (Cv -CV)fD(R)dR (2.9)
at C(Pt)

where KR is the bulk recombination rate, D corresponds to the interstitial and vacancy

diffusivity, respectively, KI is the reaction rate constant between interstitials and loops, and KVL

is the reaction rate constant between vacancies and loops. This leads to the formulation for the

change in dislocation loop density with time according to

dD,_ N 'l 1 2DK (210)
dt (C,/C, +10) R (2.

where N,,, "" is the loop nucleation rate, D,, is the loop defect density, and R, is the average

loop radius. Avci applied the model to a variety of implant and anneal conditions with

reasonable success and details can be found elsewhere. [AVC02]

2.2 Dopant Diffusion in Bulk Silicon
A brief review of dopant diffusion mechanisms is now given due to its importance in the

current investigations. More detailed reviews can be found in a number of excellent sources and

the reader is referred to those for a comprehensive discussion of the literature [FAH89, HUS94,

CHA97, JAI02, SHA03, HAY0O]. Topics have been selected as they pertain mainly to boron

interactions with point defects and the modeling associated with it.

2.2.1 Mechanisms of Dopant Diffusion

Dopants migrate through the Si lattice by interaction with point defects via 4 main

mechanisms, shown in Fig. 2-9. These can be understood as 4 separate, reversible reactions:








A+V AV (2.11)

A+I A. (2.12)

A+ I A, (2.13)

A A, + V (2.14)

where A represents a impurity atom in a substitutional configuration, I a self-interstitial, V a

vacancy, and A, a impurity atom in a interstitial position. The forward of the first reaction

occurs when a substitutional dopant pairs with a nearby interstitial to form a dopant-vacancy

pair. This is commonly referred to as the vacancy mechanism of diffusion. In the second

reaction a substitutional dopant pairs with a self-interstitial forming a dopant-interstitial pair.

This is known as the interstitialcy mechanism, whereas Eq. 2.13 is the interstitial mechanism.

These two are different in that the interstitial mechanism requires either the self-interstitial or

substitutional dopant be completely "kicked" off the lattice site. In the interstitialcy mechanism

the AI pair sort of share a lattice site as they migrate. It should be mentioned that the distinction

between the interstitialcy/interstitial mechanisms is rarely made. The last reaction is known as

the dissociative reaction or Frank-Tumbull mechanism diffusion. This requires a substitutional

dopant to hop into an interstitial position, leaving behind a vacancy.

Most dopants are dominated by diffusion of either interstitials or vacancies. B, P, and

Ga, diffuse mainly through interaction with interstitials, while Sb is nearly a pure vacancy

diffuser. Arsenic, on the other hand, diffuses by interactions with both types of point defects.

The fractional interstitial, f,, or fractional vacancy, f,, component determines the degree to

which that dopant species prefers to diffuse via the particular point defect. These values are

shown for different dopants in Table2-2.








2.2.2 Equilibrium Diffusion

Although point defects are efficiently created at room temperature, it does not approach

the large supersaturation that can be created using ion implantation or other processes such as

oxidation or nitridation. Under this case, a near equilibrium formulation for dopant diffusion

can be introduced based on Fick's Laws of diffusion. Under these intrinsic and dilute dopant

concentrations the flux of dopant A is expressed as


-J, -d +d d(2.15)
ad Ox

where dAV and dA, are the diffusivities associated with the particular defect complex, CAV and C,

are the concentrations of the particular complex, and x is the one-dimensional distance of

interest [FAH891. This is a form of Fick's first law of diffusion which states that a flux of the

impurity will occur in the presence of a concentration gradient. It can be shown that Fick's

second law can be applied to near equilibrium such that


A C (2.16)

where CA is the concentration of the dopant, t is time, and DA* is the equilibrium diffusivity of

the dopant defined to be the sum of the equilibrium diffusivities of the AV and AI complexes

[FAH89]. Basically, the change in concentration with time within a volume element is

dependent on the difference in the flux of the impurity entering and leaving the volume element.

Unfortunately, these formulations are highly idealized situations and are not applicable to the

processes that take place during IC fabrication. This requires the use of equations that take into

account the local point defect populations.








2.2.3 Non-equilibrium/Enhanced Diffusion

Enhanced diffusion refers to the phenomenon whereby dopants diffuse rapidly under a

supersaturation of point defects. TED is a type of enhanced diffusion that proceeds for a

specified amount of time, as long as secondary defects are able to store interstitials. A number

of processes that commonly occur during IC fabrication significantly alter the point defect

populations. It is obvious from the previous discussions that damage created by ion

implantation can easily do this. Oxidation of the silicon surface is another method. Only a

submonolayer of Si is required to form a monolayer of SiO, (ratio of ~1:2.25). This results in

the injection of interstitials due to the net volume expansion difference between Si and SiO2.

Oxidation enhanced diffusion (OED) of B and P is observed, while oxidation retarded diffusion

(ORD) is observed for Sb [HUS74]. The opposite phenomena occurs as a result of nitridation

(NED) and silicidation (SED). NED is believed to occur as interstitials are swept toward the

surface, where they react to form SixNy leaving behind an excess of vacancies. In SED,

vacancies are generated at the interface as silicon atoms are removed to react at the silicide-

metal interface. Under these conditions, B and P diffusion is retarded, whereas Sb and As are

enhanced. Examples of enhanced diffusion under nitridation is illustrated in Fig. 2-10. A

number of non-equilibrium formulations are available to couple point defect and dopant

diffusion and the reader is referred to the review articles for specifics.

2.2.4 TED of Boron

As mentioned previously, TED results from dopant interactions with interstitials stored in

secondary defects. For a dopant such as B, which interacts very strongly with interstitials, this

leads to unacceptable junction depths in bulk Si [JON98]. Figure 2-11 shows an example of the

TED behavior of B as a result of the presence of {311) defects. Significant motion of the

profile is seen to occur at shorter times, while at long times the {311 s have mostly dissolved








and can no longer drive TED. At this point, the B profile is mostly stable and the diffusivity

exhibits more of the characteristics of near-equilibrium diffusion.

TED of B has a number of interesting characteristics and is affected by a number of

implant and anneal parameters, including dose, energy, time, temperature, ramp rate, etc.

Implant energy and dose tend to increase the amount of observed TED. However, saturation in

the amount of TED occurs as the dose increases above approximately Ixl0s1 cm'2. This is

attributed to stable dislocation loop formations, that act as strong sinks for excess interstitials.

Similarly, intuition tells us that as time increases so does TED; at least under isothermal

conditions. TED can be reduced by annealing at higher temperatures for shorter times, though.

[JAI02] For this reason there has been considerable effort aimed at alternative annealing

technologies such as rapid thermal annealing (RTA), laser thermal annealing (LTA), and

FLASH lamp annealing. Each of these has specific advantages and disadvantages, but in

general the goal is to obtain shallow junction depths by reducing TED, while at the same time

enabling high dopant activation.

2.2.5 Boron Interstitial Clusters (BICs)

Equilibrium solid solubility of boron is well above lxl02 cm3 at temperatures of 850 *C

and higher [TRU60]. However, these levels of dopant activation are not obtained after

annealing of ion implanted boron in silicon, even after TED has ended. The reason for this is

attributed to the formation of boron interstitial clusters (BICs) due to a high interstitial

supersaturation [STO95]. These defects consist of one or more boron atoms bound with one or

more self-interstitials and should not be confused with the SMIC discussed earlier. In general,

BICs are immobile and electrically inactive for the most part, although active BICs are believed

to exist [LIL01]. An immobile peak in Fig. 2-11 is observed near the surface, which is where








the BICs reside. Complete dopant activation cannot be obtained until they dissolve; this

requires long periods at high temperatures.

The kinetics of BIC dissolution have been investigated by a number of authors and the

most significant findings are now discussed [PEL99b, HUA98, CRI03, MANGO, MAN01,

SOLOO, SCHOO, LIL02, RAD02, MIR03]. The boron concentration threshold for BIC formation

appears to be between I xl08 cm" and I x10" cm ". The amount of clustering depends strongly

on the separation between the region of high boron concentration and peak of the interstitial

concentration [JON96J. A thermal activation energy for BIC dissolution has been

experimentally determined to be between 3.0 eV [LIL02] and 3.2 eV [MIR031. These are

significantly less than that found by Mokhberi, et al., of 4.7 eV [MOK02]. However, this could

be due to the different doses used in the studies. Lilak et al., [LIL02] used a dose of boron of

2x10" cm2, whereas Mokheri et al., [MOKO2] used Ixl0"5 cm"2. These differences in

concentration could result in BIC species with significantly different binding energies.

Ab initio and tight binding calculations have provided useful information on the relative

stability of specific BIC configurations [ZHU96, CAT98, PEL99a, LIUOO, LENOO, LUO01,

ADE03, HWA03]. The model of Pelaz et al., is shown in Fig. 2-12 illustrating the formation

energies required for evolution of BICs up to a size of B414 [PEL99a]. They came to the

conclusion that BICs with a high interstitial content (e.g., BI2, B313, B414) form at early times

when the interstitial supersaturation is greatest. As annealing proceeds, the BICs emit

interstitials that can contribute to TED leading to BICs with a lower interstitial content.

Therefore, the most stable configurations are those where m < n for a B.Im cluster. When the

BICs completely dissociate the immobile peak then dissolves out, but only long after TED has

ended.








2.3 Silicon-On-Insulator (SOI) Materials

Lilienfield first proposed the idea of a three terminal device operating with an insulated

gate in 1926 [LIL26]. His patent describes a thin layer of semiconductor deposited on an

insulating material; thus, one could argue that the idea of a field effect transistor was first

proposed as an SOI structure. SOI is most commonly described as a thin layer of silicon

(typically hundreds of A to a few microns thick) on top of an insulating material with an

underlying bulk silicon substrate. The most common insulating material currently is silicon

dioxide. To date, SOI has been incorporated into the process flow of nearly every major

semiconductor company including IBM, Motorola, AMD, Intel, Philips, Canon, etc. [HAN02].

For years the advantages of SOI devices over bulk silicon were well documented, but

SOI suffered from the inability to produce adequate, device-quality materials [HOV96]. Bulk

silicon had a huge experience base and scaling in accordance with Moore's Law was easily met.

In effect, SOI had a huge mountain to climb, in order to gain acceptance as a viable alternative.

Early SOI materials, such as silicon-on-sapphire (SOS), were mainly used in niche

markets including space exploration and high temperature environments. In the late 1970s,

separation by implanted oxygen (SIMOX) was developed and has become the most mature of

all commercially available SOI materials [IZU98]. The 1990s saw development of advanced

techniques for fabrication of SOI materials, such as the Smart-cut and Nanocleave methods.

Today, the fabrication of SOI materials is a multi-billion dollar industry and is projected to

account for 50% of the production of all wafers by 2008 [SO103].

2.3.1 SIMOX

At present, the most mature of all SOI materials is the separation by implantation oxygen

(SIMOX) process. Oxygen ion implantation was first used for the synthesis of silicon oxide in

the late 1960s [WAT66]. However, it was not until the late 1970s that the SIMOX process was








actually developed [IZU78]. The 1980s and early 1990s saw considerable interest in enhancing

the quality and throughput of SIMOX materials, as evidenced by the number of studies

conducted [HEM83, JAU85, HAY80a, HAY80a, HAY80b, CEL86, CHA87, DOU87, WH187,

HOL84, MAO86]. In order to form a buried insulator using conventional ion implantation took

over 2.5 days to implant a dose of 1.2x 018 cm"2 with a beam current of 100 (tA [IZU91]. The

development of a high current oxygen implanter by Eaton Corporation significantly reduced the

processing time devoted to the implant step. The quality of the surface Si layer was enhanced

by the use of lower doses and higher post-implant annealing temperatures [NAK90]. The

advantages and disadvantages of the SIMOX process are shown in Table 2-2. Today, the IBIS

Corporation is the largest producer of SIMOX wafers [IBI03].

The SIMOX process basically consists of three steps. The first step consists of high

dose oxygen implantation at an elevated temperature (> 5000C). This is done in order to

prevent complete amorphization of the surface Si layer. Creation of an amorphous surface Si

layer would be disastrous because it would be impossible to recrystallize off an amorphous

BOX. Typical implant energies and doses would be 180 keV, 1.8x101' cm'2 for standard

SIMOX (180 nm surface Si/400 nm BOX), or 30 keV, x1l017 cm"2 for low dose SIMOX (57 nm

surface Si/47 nm BOX). A high temperature annealing step at 13000C for 6 hours follows this.

The annealing step is necessary in order to synthesize a box shaped buried SiOx layer. The final

step may consist of polishing in order to remove oxidation at the surface as a result of the high

temperature annealing. Despite the high temperature annealing, the BOX properties vary from

thermally grown SiO2 due to the presence of silicon islands, stoichiometry, increased oxygen

concentration in surface Si, etc. For this reason, research has been done into other methods of

SOI fabrication such as the Smart-cut and Nanocleave processes. However, the recent








development of the internal oxidation of Si (ITOX) process can help improve the dielectric

properties of the surface Si/BOX interface in SIMOX materials, as well as eliminate many of

the silicon islands [COL97].

2.3.2 SOITEC

While the SOI community made huge strides with the development of the SIMOX

process, higher quality SOI materials could be made by utilizing thermal oxidation followed by

wafer bonding [CON96, STA97]. Early attempts, such as bonded and etch-back SOI (BESOI),

could not overcome the problem of material waste [COL97]. In the early 1990s, the Smart-cut

process was invented [BRU96] and resulted in the formation of the company SOITEC, now the

largest producer of SOI wafers [SOI03]. The advantages of the Smart-cut process are shown

in Table 2-3.

An illustration of the Smart-cut process is shown in Fig. 2-3. It utilizes two wafers,

but results in one final SOI wafer and another bulk silicon wafer, which may be reused. The

first step involves thermal oxidation of one wafer, which will later provide the BOX of the SOI

wafer. Next, hydrogen ion implantation is performed through the BOX to the surface silicon

thickness desired. The implant energy can be tailored to dictate the thickness of the surface Si

layer, while the dose is around 5x10'6 cm"2. The two wafers are then bonded together at room

temperature via van der Waals forces at the two surfaces. A low temperature anneal (-600-

700C) is then performed in order for the implanted hydrogen to coalesce into micro bubbles.

As the micro bubbles grow the pressure inside them increases until they fracture resulting in the

splitting of the SOI wafer from the recyclable wafer. A second anneal is performed at I100C

for two hours to strengthen the bond at the bonded interface. A final polishing step is required

in order to smooth the surface after splitting [BRU97].








2.4 Interstitial and Dopant Interactions at SI/SiO2 Interfaces

The foundation has been laid for understanding point defect-dopant interactions in bulk

Si; this is now discussed in the context of the current investigations. Oxide growth kinetics and

the Si/SiO2 interface structure are one of the most well understood phenomena related to Si

microelectronic fabrication thanks to the work of Deal and Grove [DEA65a]. Routine growth of

SiO, on Si occurs for the formation of field oxides, masking oxides, pad oxides, and gate oxides,

during IC fabrication. These are all in addition to the native oxide/Si interface that exists

anytime a clean Si surface is exposed to the atmosphere. Thus, the presence of numerous

Si/SiO, interfaces is unavoidable. The structure of these interfaces determines the manner in

which point defects interact with them. Predicting dopant diffusion in SOI is further

complicated due to the presence of a buried Si/SiO, interface, in addition to the native

oxide/surface Si interface. Significant numbers of investigators have sought a better grasp of the

interface reconstructions and electrical defects that exist at Si/SiO2 interfaces, but few of these

bothered to apply their findings to point defect interactions. It is also known that certain

dopants prefer to remain in Si, while others tend to segregate towards SiO,. However, only a

handful of these studies used SOI substrates to understand the effect of a buried interface.

These are discussed in the final section of the chapter.

2.4.1 Si (001) Free Surface and Si(001)/Si02 Interface Structure

Clearly the most studied of all condensed matter interfaces are the Si(001) surface and

Si(001)/SiO2 system. They are used in well over 95% of all semiconductor devices produced by

the microelectronics industry [ZAN0. This is due to excellent stability between Si and silicon

dioxide, both thermally and electrically, as well as the low density of structural and electrical

defects. Si(001) is the most common crystallographic orientation used for MOS fabrication due

to its low interface state density (~-xl0o' charges/cm2) [STROO].








2.4.1.1 Si(001)

The surface free energy is governed by the ability of surface atoms to rearrange

themselves into more favorable configurations. Surfaces that already possess a low surface free

energy will not tend to reconstruct (e.g., metals), while those with high energies will. A perfect

cut along the Si(001) leaves each surface atom with two dangling bonds. This causes the

surface atoms to reconstruct into a (2x1) unit cell, forming rows of dimerized atoms, shown in

Fig. 2-14. The driving force for the (2x1) reconstruction is the reduction in the number of

dangling bonds from 2 to 1. The reconstruction appears to be stable up to temperatures of 1200

"C, although (lxl) and (4x2) unit meshes have also been observed [BAL88].

Vicinal Si(001) surfaces consist of rows of dimerized atoms that are orthogonal on

adjacent terraces separated by single or odd number-layered steps, shown in Fig. 2-15. Thus,

the existence of the (1x2) reconstruction must also be conjectured. The figure also illustrates

the existence of two different types of single-layer steps, SA and SB. Dimers at the SA step run

parallel with the dimers in the upper terrace, while the SB step dimers run perpendicular those in

the upper terrace. This results in abrupt (SA) or graded (S,) steps depending on the orientation

of the step dimer. The S, step contains significant densities of kinks, while SA steps appear to

be fairly smooth. A scanning tunneling microscope (STM) image of alternating S^ and SB steps

are shown in Fig. 2-16 illustrating the difference in the two steps [ZAN00].

For the current investigations, the free Si surface primarily idealistic. Interstitials would

be able to annihilate if they were able to diffuse to surface and attach themselves to a step or

kink. Thus, many investigators have treated the surface as an infinite source of vacancies and

infinite sink for interstitials. However, the surface is practically always covered with some form

of silicon oxide that will affect this treatment. Oxidation occurs by transport of oxygen (O) or








H20 to the Si/SiO, interface where it reacts, pushing the interface deeper into the substrate. The

position of the interface is continually changing, so the existence of a free Si surface is not

critical in the present context, rather it is the arrangement of Si and O atoms at the Si/SiO2

interface.

2.4.1.2 SiO2

More than 95% of all rockforms on the earth possess silicon dioxide as the main

constituent [PAN75J. It consists of tetrahedral SiO4 structural units with an Si-O bond distance

varying between 0.152 nm and 0.169 nm, shown in Fig. 2-17. Each oxygen atom is bonded to

two silicon atoms, while each silicon is bonded to four oxygen atoms. The O-Si-O bond angle

is 109.18*, while the Si-O-Si angle varies from 120* to 1800 in crystalline SiO2. There is also a

rotational angle between tetrahedral that is either 0 or 60* for crystalline SiO2. The way in

which the SiO4 tetrahedron are arranged in 3-D determines the specific crystal structure, or if it

is random/amorphous. Figure 2-18 shows the different allotrophs of SiO2 that exist at high

pressures and elevated temperatures. Note under most oxidation temperatures the stable form is

tridymite, but a and 0 quartz may also exist at lower temperatures. The density of these phases

depends strongly on the Si-O bond length and Si-O-Si bond angle. Coesite is the densest form

of SiO2 due to its small, 1200 bond angle. Three properties distinguish crystalline SiO2 from

amorphous SiO2: (1) Rotational angle between adjacent tetrahedra may be any angle, (2)

variation in Si-O-Si bond angle of 150 15", and (3) number of tetrahedra in rings may also

vary, but 8, 10 and 12-membered rings are more energetically favorable [CHI97].

At room temperature for thermally grown oxides there is no long range order present.

Instead, fused/amorphous silica is believed to be similar to a continuous random network

(shown in Fig. 2-19) or collection of microcrystalline grains of SiO4 tetrahedron arranged in








space. The network model can be thought of consisting of 8, 10, and 12-membered rings. The

presence of voids in the structure can be seen; this accounts for the low measured density of

fused silica (~2.2 g/cm3) [CHI97]. The microcrystalline model presumes the distribution of Si-

O-Si bond angles to be due to the presence of small grains that cannot be resolved using x-ray

diffraction techniques. The grains would consist of different SiO, crystalline phases. A

combination of the two models is likely most representative of amorphous silica. Thus, there

may be short range order of the Si-O-Si bond angles, but a variation in the bond angles over the

long range [BAL88].

2.4.1.3 Si/SiO2 interface

While the growth kinetics of silicon oxidation are well understood, there remains a large

amount of debate over the precise configuration of Si and O atoms in the vicinity of the Si/SiO2

interface. The interface is not atomically abrupt, rather the transformation from Si to SiO2 takes

place over a few monolayers. Idealized models have been proposed using both crystalline and

amorphous SiO,, although both assume no broken bonds to exist at the interface [BAL88]. It is

generally accepted that the interface consists of a transition region sandwiched between the bulk

Si and amorphous silica regions, as shown in Fig. 2-20. The local bulk Si and SiO, are also

altered near the transition region.

The transition region is believed to consist of non-stoichiometric suboxides, SiO, that vary

depending on their proximity to either the bulk Si or SiO2. The suboxides can be detected using

methods such as x-ray photoelectron spectroscopy (XPS) or photoemission spectroscopy (PES),

although PES appears more sensitive [LUZ93]. The oxidation states Si'i, Si'2, and Si'3, are

associated with the Si 2p core levels and correspond to Si atoms with one, two, and three nearest

oxygen neighbors, respectively [HOL83]. The thickness of the transition layer has been








estimated using quantitative PES and XPS to be between 0.6 and 1.5 monolayers (-5-8 A) thick

[LUZ93]. The relative amounts of the particular oxidation state are sensitive to the growth

temperature of oxidation [LUZ95]. Lu et al., found the areal density of the +2 and +3 states to

increase with growth temperature, while the +1 state was fairly constant [LUZ95]. They

attributed this to strain relaxation at the interface as more oxygen reaches the interface during

growth.

Two structural defects are of primary interest to the Si/SiO2 interface and are illustrated in

Fig. 2-21. The E' defect consists of a broken bond between two tetrahedra that would otherwise

be bonded to an O atom. One of the Si atoms then becomes positively charged, while the other

keeps its unpaired electron in a dangling bond orbital. It has been hypothesized that the E'

defect is the source of fixed positive interface charge [BAL88]. The Pb center consists of a

broken bond in Si that would be bonded to another Si atom. This defect has been extensively

studied using electron spin resonance (ESR).

Electrical defects of interest to the Si/SiO, system can be divided into four categories.

Fixed oxide charge, Qf, is net positive charge existing within the transition region of the

interface with an areal density of 10 10" cm'. It is given this nomenclature because the

positive charge persists under normal device operation. This charge is believed to be due to

incompletely oxidized Si atoms having obtained a positive charge. Q,, refers to trapped interface

charge and may have a positive, negative, or neutral charge state. Similar to Q,, these defects

are believed to be due to Si atoms with unsatisfied bonds. The other two types of electrical

defects are present in the oxide, but away from the transition region. Mobile ionic charge, Qm, is

associated with cations of alkali metals such as Na* or K'. Oxide trapped charge, Qa, is due to

broken Si-O bonds present in the oxide, created during processes such as ion implantation and








reactive ion etching. Fortunately, these last two defects are less significant today because of

stringent contamination control and the ability of broken bonds to repair themselves during

thermal processing. However, charge trapping due to Q, can cause shifts in threshold voltage in

devices requiring current to be passed through the oxide (e.g., EPROM).

2.4.2 Point Defect Interactions at the Si/SiO2 Interface

For a number of years the ability of an interface to alter point defect populations has been

a concern and an issue of debate. TED is expected to be strongly affected by this, since it is

strongly dependent on the interstitial supersaturation. If an interface efficiently allows

interstitials to be trapped or annihilated it would serve to reduce TED. Thus, the ability to

model the behavior of an interface greatly increases the reliability of process simulators for

predicting dopant diffusion profiles.

Classically the silicon surface has been thought of as an infinite source of vacancies or

infinite sink for interstitials. An interstitial diffusing towards the surface will instantaneously be

annihilated. On the other hand, a vacancy diffusing to the surface acts to expand the free

surface by annihilating a surface Si atom. Thus, many authors generically speak of kinks and

traps as sites for interstitials to recombine or become immobile at the Si/SiO, interface, while

mentioning nothing regarding what a kink or trap physically is. While this rudimentary picture

is very useful for understanding processes such as epitaxy that involve atomic attachment to free

surfaces, it does not create a realistic picture of interstitial recombination at the Si/SiO,

interface. However, before delving into the intricacies of the physical processes underlying this,

it is necessary to give a general description of the parameters for determining point defect

generation and annihilation at interfaces.








The continuity equations governing point defects appear as a modification to Fick's

second law with the addition of another term, thus for the two types of native point defects


d, -- k,, (C,C, -CC) (2.17)



at dx
-- ^d ,.v (C, Cv -C;C;) (2.18)

where terms have been defined previously in Eqs. 2.8 and 2.9. These equations can be

understood to mean the change in concentration of point defects in a volume element is

determined by the flux entering and leaving minus the bulk recombination rate of both types of

point defects. Thus, those defects that enter the volume element and do not leave must

recombine with the opposite point defect within the bulk. In the case of an interface, where

point defects can be created or annihilated it becomes necessary to invoke additional

formulations


g, +d, dC --Ks,(C,-C,;) 0 (2.19)
Ox

sv + d, Ks.f,v (C, C ) 0 (2.20)
ax

where g is the flux of point defect injection into the bulk, and K,, is the recombination velocity.

These equations can be understood as point defects created at an interface that do not diffuse

into the bulk recombine at the interface. These processes are illustrated in Fig. 2-22 for the

cases of bulk Si and SOI with the obvious difference being the presence of an additional

interface for recombination in SOI.

2.4.2.1 Interface effects on interstitial kinetics under oxidizing conditions

The majority of studies for determining interface recombination velocities have involved

measurement of the growth and shrinkage of oxidation induced stacking faults (OISF) and








oxidation stacking faults (OSF). Structurally, there is no difference between OISFs and OSFs;

they are distinguished by how they are nucleated. OSFs nucleate off defects present before

oxidation begins, such as dislocation loops, while OISFs prefer a more homogeneous path.

These stacking fault defects are easily produced as interstitials are injected during high

temperature oxidation of Si and form on { 111 planes. Because of their extrinsic nature, the

OSFs can drive the phenomenon of OED. The work of Hu was highly instrumental in this field

as far as tailoring the OSF results with surface processes [HUS74, HUS75, HUS85a, HUS85b].

Since oxidation, under most conditions, follows a linear-parabolic growth law [DEA65a], Hu

proposed a similar parabolic dependence of the interstitial injection flux


g, (t) = (2.21)

where A is the oxidation growth rate constant, t is time, and to is the time constant for separating

the linear-parabolic transition [HUS85b]. This was later modified to a power law dependence

by Fahey et al., as

g, (t) A(t, + t)" (2.22)

that prevents infinite flux at time zero. [FAH89] In the absence of an energy barrier to

recombination, Hu determined the interstitial recombination velocity to be

Ksu,,. Paod, (2.23)

where p is the density of surface kinks, and ao the capture radius [HUS85b, FAH89]. This

equation assumes a constant kink density, but it is entirely possible that these sites could

become saturated if the number of interstitials attempting to recombine is greater than the kink

density. Hu also proposed the kink site density to be dependent on orientation [HUS74]. The

mechanism of interstitial recombination proposed by Hu is illustrated in Fig. 2-23 [HUS74,








HUS94]. In this process, an interstitial or di-interstitial diffuses to the interface and recombines

along a step edge or kink site near the transition region.

Taniguchi et al., used a boron implant to define the depth of OSFs followed by frontside

and backside oxidation to observe their growth and shrinkage [TAN83]. Since the boron

implant was close to the surface, backside etching had to be used to move the backside interface

closer to the damage so the OSFs could nucleate. Polysilicon and silicon nitride films were

deposited on the frontside prior to oxidation, so that only backside oxidation would take place.

They found the maximum length of OSFs to decrease exponentially as the thickness of silicon

increased. For silicon thickness 40 tmi or less, the OSF length increased with oxidation time.

The growth rate also increased as the thickness of silicon decreased. The 500 (im film did not

result in growth of OSFs, while the 80 Rm had a modest decrease at early times followed by an

increase with oxidation time. They then developed a model based on the following: oxidation

at the backside injects interstitials that diffuse to the frontside causing OSF growth. However,

the interstitial concentration at the frontside must be greater than the equilibrium interstitial

concentration in the vicinity of the fault in order for growth to occur. Using this model they

were able to develop an empirical expression for the interstitial recombination velocity at the

frontside interface, given by:

-24eV
K -4.6xx 10 ee (2.24)

where K,,,, is given in cm/s.

Ahn et al., used two different test structures to investigate 1-D and 2-D interstitial

kinetics, shown in Fig. 2-24 [AHN87]. Figure 2-24(a) shows the 1-D test structure with 4 points

of interest within the structure. Position 1 is blocked from oxidation since both the surface and

backside are protected with silicon nitride, thus the phosphorus marker should diffuse close to








equilibrium conditions. Position 2 is protected at the backside, but interstitial injection occurs at

the surface. Position 3 is protected at the surface, while backside oxidation occurs. Finally, the

fourth position is not protected at either interface, so interstitial injection occurs from two

directions. The 2-D test structure in Fig. 2-24(b) illustrates the reduction of trench width down

to 3 prm, as interstitial injection occurs at the surface. They found the recombination velocity to

vary based on the trench width, which they attribute to stress produced by bending of the silicon

membrane. For a 20 1im trench they extracted Ks,, to be 5xl0"7 cm/s, while a 104 gm trench

was found to be 2x107 cm/s. They seem to report an average recombination velocity of 3x107

cm/s at 1100'C. This is more than a factor of 2 less than that reported by Taniguchi et al., that

found Ksr,, to be 7x1l07 cm/s at the same temperature [TAN83]. However, both these values of

Ks.f are lower than that extracted by Scheid and Chenevier [SCH86]. They found Ksfj to be

25x107 cm/s. Ahn [AHN87]attempts to reconcile these differences by noting that Scheid and

Chenevier [SCH86] assumed their nitride to be perfectly reflecting, while trying to obtain the

recombination velocity at the pad-oxide interface. Differences in processing conditions could

also lead to different oxide properties that could affect Ksg,,. Taniguchi et al., [TAN83], used

OSFs to determine the recombination velocity, whereas Ahn et al., [AHN87] used dopant

diffusion of phosphorus. This difference could also explain the difference in values reported by

the two authors.

In another experiment, Ahn et al., used both phosphorus and antimony diffusion, as well

as OSF growth and shrinkage to monitor interstitial kinetics at the SiO2/Si interface [AHN89].

In this case, the post fabrication annealing was done in an inert ambient using argon. They

noted enhanced Sb diffusion, retarded P diffusion, and enhanced shrinkage of OSFs in the








proximity of the SiO2/Si interface. This is attributed to a vacancy supersaturation due to the

formation of SiO molecules as self-interstials diffuse into the oxide according to the reaction:

Si (s) + SiO2 (s) = 2 SiO (g) (2.25)

It is proposed that the fast diffusing, gaseous SiO molecules efficiently transport interstitials

from the interface. Unfortunately, the authors give no analytical evidence that such a reaction

takes place, so another vacancy mechanism cannot be ruled out. It is also unclear whether the

SiO formation only occurs under high temperature processing, since the authors only annealed

at 1100C. This has, in fact, been suggested by Celler and Trimble in a number of papers

[CEL89a, CEL89b].

Only a few studies have been aimed specifically at studying interstitial interactions with

interfaces using SOI materials. In their first study, Tsoukalas, Tsamis, and Stoemenos set out to

investigate interstitial reactions with a thin oxide film in the structure shown in Fig. 2-25

[TS093]. They oxidized a SIMOX wafer (W2) at 11000C for 2 hours in order to grow OSFs,

which they then directly bonded to another SIMOX wafer (Wl). Prior to bonding, W1 had a

thin oxide of 8 nm grown on the surface to serve as the transport site for the interstitials. After

bonding, KOH etching was used to thin through the Si bulk from the backside of WI. The

surface Si layer of W1 was then oxidized to inject interstitials that then diffused to the thin oxide

to react. The growth of the OSFs was monitored to determine the transport mechanism of the

interstitials through the oxide. Figure 2-26 shows the OSF length difference between the

thinned SOI structure and reference samples as a function of oxidation time. It illustrates that

the thinned samples tend to grow longer than the reference samples. A clear temperature

dependence of the change in OSF length can also be seen. There is also a delay time constant

associated with the OSF growth. The authors attribute this difference in growth ITS093) and in








another study of isotopic Si diffusion [TSO01] to SiO formation at the front interface of the thin

oxide given by the reaction of Eq. 2.25. Further, the SiO molecules then diffuse through the

thin oxide and react at the back interface, according to the reverse reaction of Eq. 2.25. This

effectively injects interstitials into the OSF layer leading to the growth. There are some

problems with this explanation though. Once again, no analytical evidence is provided to

suggest the reaction actually takes place. The authors also ignore the effects that oxidation of

the thin oxide could play. Although most of the oxygen is expected to react at the surface, a

small amount will also react at the other interfaces [COL97]. This could also provide additional

interstitials to create the growth of OSFs. In their second study, Tsamis et al. investigated the

lateral distribution of interstitials under a nitride mask using bonded SOI substrates [TSA95].

The test structures shown in Fig. 2-27 were used to investigate both the depth distribution of

interstitials and the lateral distribution. Oxidation of the trenches was used to inject the

interstitials and the growth of OSFs was monitored at the surface Si/BOX interface and under

the nitride/pad oxide. Figure 2-28 shows the OSF length for varying surface Si thickness under

wet and dry oxidation conditions. The data shows that thinner surface Si films result in smaller

OSFs compared to bulk Si. Also, the effect is stronger under dry oxidation conditions compared

to wet. The authors believe the reduction of OSF length in the lateral study is due to interstitials

that diffuse to the surface Si/BOX interface and recombine. However, they do not explain why

differences exist between the wet and dry oxidation. This is could be due to differences in the

roughness of the Si/SiO2 interface. Dry oxidation is known to create a rougher interface that

could, in principle, lead to a higher recombination velocity at the surface. This roughness could

be in the form of dangling bond defects or kink sites. This would explain why the OSFs are

larger in the case of wet oxidation, but cannot explain the difference that is observed between








the different materials because the surface roughness would be the same. This suggests that

there may also be differences at the surface Si/BOX interface due to the oxidation conditions.

Other studies using SOI materials were mainly aimed at reducing OISF formation for

low-energy SIMOX fabrication [GIL99a, GUI92, GIL94]. During this process, sacrificial

oxidation of the surface Si layer followed by etching in dilute HF is used to obtain thinner

surface Si films. In their paper, Giles et al., propose a model based on vacancy injection from

the surface at temperatures above 1190"C. They suggest this suppresses the interstitial

supersaturation, thus inhibiting the growth of OISFs in SOI. It is also suggested that the

interstitial supersaturation is further reduced by recombination at the surface Si/BOX interface,

as mentioned above. Giles et al., also extended their studies to included bonded SOI wafers

[GIL99b]. In this study they noted a stronger temperature dependence of OISF shrinkage in

bonded SOI versus SIMOX. Also, the length of OISFs was, on average, greater in SIMOX.

2.4.2.2 Interface effects on interstitial kinetics under non-oxidizing conditions and due to
ion implantation

Up to this point the analysis of recombination velocities dealt with injection of interstitials

due to oxidation. Ion implantation potentially allows for interstitial recombination to be

observed without injection of point defects from the surface during annealing. The implant

damage may be brought closer to the surface by reducing the implant energy or by using a

higher energy implant followed by lapping of the surface. The effect of the surface can then be

studied using TEM studies of extended defect evolution. Other studies have used Si

implantation close to a marker layer and then observed motion of the marker layer.

Unfortunately, this has led to a range of experimental interpretations and significant debate

remains over the effect of the surface on a interstitial supersaturation.








Lim et al., studied the affect of surface removal on the motion of a boron marker layer

after it had been implanted with Si' [LIM95]. They found the Dt product to decrease

monotonically with etch depth. The data was then found to fit with a recombination length of

0.1 gum. From their study, they conclude that the surface is a strong sink for interstitials. A

similar study was performed by Gossmann et al., without etching and by annealing in

vacuum[GOS95]. They claim a depletion of interstitials occurs during annealing in vacuum, but

do not propose any mechanisms for the behavior. Cowem et al., attempted to elucidate the

effect of the surface by etching an amorphous layer produced by a 150 keV, 2x10'5 cm2, Ge'

implant to varying depths [COW99b]. This was followed by a boron implant at 3 keV, lx10"

cm2 into the different layers. They noted a linear decrease in the diffusion enhancement as the

EOR loop layer was brought farther from the surface. Their conclusion is that the surface is a

great sink for interstitials, although their data seems to contradict that of Lim [LIM95]. Cowern

basically concludes that a shallower amorphous layer enhances the diffusivity of boron because

the interstitials are recombining at the surface, whereas Lim claims the recombination of

interstials reduces the boron diffusivity. It is unclear how these two authors came to the same

conclusion, although their data is contradictory.

Similar results were also reported by Agarwal et al., but instead of etching the surface they

simply reduced the Si' implant energy [AGA97b]. They found the diffusivity enhancement of

boron doping superlattices (DSL) to decrease as the Si' implant energy was reduced from 5 keV

to 1 keV [AGA97b]. From this they came up with a recombination length less than 10 nm,

which is one order of magnitude less than that reported by Lim et al. [LIM95]. In a defect study

without the boron DSL, Agarwal et al., noticed the formation of zig-zag {311} defects, shown

in Fig. 2-29, for low energy Si* implants [AGA97a, EAG96]. During growth, these particular








{311} defects appeared to unfault and then continue growing on another {311} plane. The zig-

zag defects were also found to be significantly more stable than the normal rod-like {311}s.

They also found that both types of {311} defects formed at energies as low as 1 keV, 3xl0" cm

2. This indicates that {311} defects still form before the interstitials can recombine at the

Si/SiO2 interface. This seems to contradict their other work in [AGA97b], but the authors do

not address the point.

Other authors have all but confirmed the Si/SiO2 interface is not a significant sink for

interstitials under most conditions. The effect of the surface on EOR defect evolution after Ge'

implantation has been studied by a couple of authors. Omri et al., used etching of a single

amorphous layer to bring the EOR damage closer to the surface [OMR96]. In their study, a 150

keV, 2x10" cm-2 implant was used to produce a 175 nm continuous amorphous layer. Etching

the amorphous layer down to 30 nm did not significantly change the flux of trapped interstitials

to the surface. A change in defect density was observed when the amorphous layer was thinned

to below 55 nm. They propose that the a/c interface acts as a diffusion barrier as the EOR loops

begin to nucleate. However, only after solid phase epitaxy (SPE) has occurred while the loops

are in the coarsening phase can the surface begin to have an effect. King et al., also studied

EOR dislocation loop evolution after Ge' implantation [KIN03a]. They studied the effect of

lapping on the loop evolution for 5, 10, and 30 keV, lxl0'5 cm2 implants. These implants

produced amorphous layers of varying depths, which were then lapped to bring the EOR

damage closer to the surface. For example, the 10 keV amorphous layer was lapped to less than

that of the 5 keV, yet the defect evolution did not vary between the 10 keV lapped and unlapped

samples. The defects produced from the 5 keV implant dissolved faster than either of the 10

keV samples, shown in Fig. 2-30. This indicated that the 5 keV defects did not dissolve because








they were closer to the surface, but rather because of an implant energy effect. In a follow up

study, King et al., found that the surface began to affect the defect evolution when the damage

was brought to a depth of -60 A [KIN03b].

2.4.3 Models for Interstitial Interactions at Si/SiO2 Interfaces

Physically based modeling of interstitial recombination has been the subject of at least

three groups' attention, in addition to the work of Hu discussed previously. Dunham developed

a model to explain OED and ORD behavior, which he then applied to non-oxidizing conditions

[DUN92]. He claims, based on his interpretation of other author's experiments, that the

majority of interstitials produced during oxidation flow back into the growing oxide rather than

diffuse into the bulk. The interstitial segregation coefficient is defined as the ratio of the

equilibrium concentration of interstitials between SiO, and Si:


C'
m (2.26)

Dunham's proposition would seem to be a legitimate basis for a model when the interstitial

segregation coefficient is considered [AGA95. TSA98]. This is at least three orders of

magnitude higher for interstitials in SiO, at 1100 C, according to the work of Agarwal

[AGA95] and Tsamis [TSA98]. However, as Hu points out [HUS94], a anisotropic segregation

coefficient, as suggested by Dunham could result in a perpetual motion device where two

crystallographic faces intersect.

Tsamis and Tsoukalas developed a time dependent recombination velocity for non-

oxidizing conditions, as suggested by Ahn [AHN87, TSA98]. Intuitively, this model makes

since if the number of sites that can result in recombination is assumed to be a finite number.

Thus, the interface would lose its effectiveness as a sink for interstitials with annealing time.

They make the following assumptions: there exists a fixed concentration of interstitials in SiO2,








the oxide extends infinitely in the lateral direction, and no sinks or sources of interstitials exist

in the oxide. From this, they are able to define the time dependent effect surface recombination

velocity:


Ks,,w,(t) Ks j (O)e erfc(o) (2.27)

which does not consider Si reaction at the interface. For reaction at the interface to form SiO it

takes the form:


Ks.C,ff.g(t) = 2k (S+( +l -l)e '! erfc( ( ) (2.28)
SI t0

where k, is the reaction rate constants for Si and S, is the supersaturation ratio of self interstitials

in bulk Si. Some of the limitations of the model should be pointed out. The model is not

applied to temperatures below 1050C, which is more relevant in modem IC fabrication. The

authors point out that interstitial incorporation into the oxide will decrease with temperature.

This could severely limit the applicability of this model except in the case of oxidation.

Law et al., have also investigated point defect recombination at the Si/SiO interface

[LAW91, LAW98]. They were the first to attempt to correlate TED with OED by proposing di-

interstitial recombination as a dominant mechanism [LAW98]. In the case of recombination

being dominated by single interstitials, K,, is expressed as

FsYJ Ksu.r,(C, -C) (2.29)

where F,,, is the recombination flux of silicon interstitials at a non-oxidizing interface. For di-

interstitial recombination, the recombination flux becomes proportional to the square of the

interstitial concentration according to


Fsu.a Ksj.,2(C, C; )


(2.30)








where K.,ein includes both the surface recombination velocity and a temperature dependent

factor for the dependence of di-interstitial population on the interstitial concentration.

Subsequently, the model is able to fit a variety of OED data and TED data including the

interstitial concentration as a function of lateral distance for long and short channel stripe

widths. The advantage of this model is that it can be applied at lower temperatures, where TED

and SPE take place.

2.4.4 Dopant Segregation in the Proximity of Si/SiO2 Interfaces

The following question should now be posed: Why is dopant segregation important?

First, it may lead to alterations in the channel and source/drain doping profiles. This, in turn,

affects the threshold voltage and device drive currents. Second, segregation to the oxide could

change the properties of a gate oxide and the surface state density. These are just two reasons;

others will be pointed out along the way. In this section, we begin with a thermodynamic

description of impurity redistribution in a two phase system. Next, the effect of static (e.g.,

inert) and dynamic (e.g., oxidation) interfaces on dopant segregation is described. Finally, the

specific studies aimed at understanding dopant distributions in SOI materials are reviewed.

2.4.4.1 Thermodynamic considerations

In order to understand the reasoning as to why an atom prefers to reside in one material

instead of another, one must turn to a thermodynamic approach. Our present system can be

thought of as a two phase region (Si and SiO2) with a additional component/impurity (e.g.,

dopant atom). In order for the impurity to attain equilibrium within the two phase region, the

chemical potential of the impurity on either side of the phase boundary/interface must be equal.

The chemical potential is related to the activity according to

l^ p' Ai RT na NAkT In a (2.31)








where AlLkA is the chemical potential of component A in phase k relative to a reference state, R

is the ideal gas constant, T is temperature, and ak^ is the activity of component A in phase k. In

the last formulation, NA is Avogadro's number and k is Boltzmann's constant. The activity

given in Eq. 2.31 is valid at a particular temperature, composition, and pressure. The activity

coefficient is typically used to describe the behavior of a component in a solution as is given as

a^ -yXk (2.32)

where ykA is the activity coefficient of component A in k, and XkA is the mole fraction. This

suggests that if ykA > 1, then akA > XkA and the impurity behaves as if there is actually more of it

in k. On the other hand, if ykA < then akA < XkA and the component behaves as if there is less

than suggested by the composition [DEH93]. For a dopant impurity at the Si/SiO2 interface the

concentration of the dopant on either side will be related to the activity coefficients as
A A A cA
sC ^-ysAoso, (2.33)

where CsiA and Csi0o2 are the concentrations of the dopant in Si and SiO2, respectively. We can

now redefine the segregation coefficient for dopants, along with volume and pressure, as


m e Rr X (2.34)
cAo, Y s

where the terms are the same as those defined previously [GRO64a, CHA84]. Thus, when CsiA

> CsioA the impurity is rejected from the SiO2, while CiA < CSiA means it tends to segregate to

the oxide.

2.4.4.2 Dynamic boundary conditions

The case of oxidation represents a significantly different problem compared to inert

ambient studies. This was studied extensively by Grove and others in the early 1960s [ATA60,

GRO64a, GRO64b, DEA65b, SN065]. Here the interface behaves as a moving boundary as the








SiO2 grows at the expense of Si, effectively pushing the interface deeper within the bulk. This

causes continual redistribution of the dopant during the oxidation process.

The diffusivity of the dopant in the oxide also plays a major role in the redistribution. A

high diffusivity in the oxide could mean that the dopant tends to escape from the oxide into the

gaseous ambient, leading to a dose loss effect. This will further affect the concentration of

dopant in the silicon by means of Eq. 2.33. This is illustrated in Fig. 2-31 for both m < 1 and m

> 1. A fast diffusing species in the oxide is shown to reduce both CsiA and Cso,2A for either m >

1 or m < 1. Note that when diffusion through the oxide is slow, and m < 1, a pileup results on

the Si side of the interface. This effect becomes even more pronounced during oxidation, and is

often referred to as a "snow plow" effect.

Segregation coefficients for a number of dopants are summarized in Table 2-5. Although

the precision of the numbers is low, a few general comments can be made. Boron appears to be

the only dopant that prefers to reside in the oxide (e.g., m < 1), while P, As, Sb, and Ga, prefer

Si (e.g., m > 1). The variation in m found by different authors seems to be due to their specific

experimental conditions used to study the dopant segregation. Figure 2-32 shows secondary ion

mass spectrometry (SIMS) profiles of As, P, and B, after oxidation in dry 02 at 1100 *C for 30

minutes [SAK87]. Note the snow plow effect observed in P and As as oxidation proceeds, and

its absence in B.

Several authors have proposed a temperature dependence of the segregation and mass

transfer coefficients in order to model the behavior [SAK87, ALE98]. This also accounts for

some of the discrepancies found in Table 2-5. The results of the temperature dependence of the

segregation coefficient are shown in Fig. 2-33, as determined by Sakamoto et al.[SAK87]. Note

that the B segregation coefficient shows a much more pronounced temperature dependence than








that of P or As. Aleksandrov and Afonin noted that the segregation coefficient can be a function

of temperature, yet independent of the oxidation rate, when one considers the that


m=e "r (2.35)

where (cp, and ps, are the concentration-independent portions of the chemical potential in the

oxide and silicon, respectively [ALE98].

2.4.4.3 Static boundary conditions

More recently, the segregation of dopants at stationary Si/SiO2 interfaces has taken on

increased importance. This is due to the drive towards continually shallower junctions, which

often use low energy ion implantation close to the surface to produce the shallow dopant

profiles. This can result in the phenomenon referred to as "uphill diffusion", whereby a dopant

appears to diffuse towards the surface rather than into the bulk.

Charitat and Martinez investigated boron segregation at the Si/SiO, interface using the

nitride stripe pattern shown in Fig. 2-34 [CHA84]. This allowed them to also include the role of

stress on the segregation. In addition, they also studied the orientation dependence. Figures 2-

35 and 2-36 show the segregation coefficient as a function of temperature for <100> and <111>

orientations, respectively. The nitride layer can be seen to significantly increase the segregation

coefficient for both orientations by preventing the incorporation of boron into the pad oxide. On

the contrary, the orientation does not appear to affect the segregation coefficient to any

substantial degree. In general, m appears to increase between 850 *C and 950 C, followed by a

gradual decrease as the temperature is further increased. The authors attribute this phenomenon

to viscous flow of the oxide above 1000 C. They suggest the pressure term in Eq. 2.34 plays

the most important role in affecting m. Unfortunately, they were unable to quantitatively

measure the stress to strengthen their argument. Tensile stress in the nitride film is able to








explain the increase in m. Since boron produces a tensile stress in the Si lattice, another tensile

stress at the surface will serve to repel the boron preventing further incorporation into the oxide.

Numerous authors have observed the phenomenon of uphill diffusion of B [WAN01,

KASOO, SHI01, DUF03], Sb [SA185] and As [KAS98, SA185] in recent years. Figure 2-37

illustrates this for an ultra-low energy boron implant. Note the segregation of boron near the

surface towards the native oxide. In general, the same trends are observed for the different

dopants as discussed previously: B segregates to the oxide, while As tends to pileup on the Si

side of the interface. Duffy et al., found the uphill diffusion of B to be highly sensitive to the Ge

preamorphization energy [DUFO3]. They found very little pileup of B without

preamorphization, while the pileup increased as the preamorphization energy was increased.

The amount of tail diffusion also increased with preamorphization energy, and this was

attributed to both the increased EOR damage and a "chemical pump" effect as a result of a high

substitutional concentration of B. These results were somewhat contradictory to those found by

Kasnavi et al., when comparing B and BF, implants [KASOO]. They found the dose loss to the

oxide to increase as the BF2 energy was reduced, and the loss was greater than that for B alone.

This difference can partly be attributed to the fact that the BF2 implants were amorphizing,

although it is unclear the role that F could be playing. This is supported by the fact that they

saw a reduction in dose loss for BF, implants performed at 2x10" cm-2. Kasnavi et al., also used

XPS to determine that the majority of segregated As resides within the first monolayer of Si

[KAS98]. Similarly, Sai-Halasz et al., previously found Sb and As to be confined to a single

monolayer at the interface [SAI85].








2.4.4.4 Consequences of dopant segregation

Three consequences of dopant segregation should now be emphasized: dopant trapping,

precipitation, and electrical deactivation. Clearly, the loss of dopant to an oxide will show an

electrical deactivation effect since there cannot be any donors or acceptors in such an insulating

material. It has also been reported that trapping of dopants at the interface also results in

deactivation [SA185, VUO00]. Precipitates of segregated impurities such as Pb, [HOL88] Ag,

[HOL88] As, [IAC98b] Ge, [IAC98a, RAI96] and Sb, [WIL92] have been observed to form at

the Si/Si02 interface upon annealing.

As mentioned above, the work of Sai-Halasz et al., found the majority of segregated Sb

and As to be confined to a monolayer at the interface [SAI85]. After implantation and

annealing in an inert ambient, the dopants segregated to the interface and appeared to remain

attached. The dopants remained attached until an areal density of dopants approaching 2xl0i'

cm'2 was realized. Using Van der Pauw measurements they were able to determine that the

trapped Sb dopants at the interface were electrically inactive. They also note that if a interface

sticking coefficient of unity is assumed, ~25% of the implanted As dose can effectively be lost

when the implant dose is less than 5x1014 cm2.

The importance of dopant precipitation at the interface is also of profound importance,

since it can affect the gate oxide quality. lacona et al., found SiAs precipitates to form at

relatively high doses (3x106 cm-2) during oxidation, but did not form at 3x1015 cm'2 [IAC98b].

The precipitates were determined to have a monoclinic crystal structure and lie on the {111}

plane of Si. Their presence also led to significant changes in not only the interface morphology,

but also the surface roughness.








2.4.4.5 Models for dopant segregation

In recent years, a number of authors have recognized the need for accurate modeling of

dopant-interface interactions. Vuong et al., investigated the affect of dopant trapping on device

characteristics for BF,, P, and As implants [VUO00]. They were able to successfully model the

device data for surface-channel and buried-channel devices for both NMOS and PMOS using a

trap and detrap methodology. Their model is based on the hypothesis that a flux of dopant

arrives from the bulk and becomes trapped at the Si/SiO, interface according to


rCA,( 1 ) (2.36)
Nm,

where Jp is the flux of dopants becoming trapped at the interface, r, is the dopant trapping rate,

Cs5, is the active dopant concentration, Q,, is the areal trapped dose, and N, is the maximum

number of trap sites. In Eq. 2.36, note that the trapped flux is proportional to the active dopant

concentration indicating that trapping will become increasingly important at future technology

nodes. They also developed another equation to simulate detrapping effects


J -' -Q (2.37)

where Jda,, is the detrap flux, rd the detrap rate, and t, the thickness of the trapped layer.

Ab initio methods were used by Dabrowski et al., to investigate the mechanism of dopant

segregation of P and As [DAB02]. These included substitutional donor atoms just below the

interface, bonding donor and oxygen atoms, Si dangling bonds, dopant pairs, and defects with

unoxidized Si. The last mechanism included a silicon bridge, ledge atom, or a Si vacancy at an

unoxidized but fully coordinated Si atom. They found a high energy gain for donor trapping at

electrically active defects (e.g., unoxidized dangling bonds), but this process was only efficient

at low donor concentrations. For trapping at electrically inactive defects (e.g., unoxidized step








ledges and bridges) the energy gain was lower, but still occurred frequently at high

concentrations. This mechanism relied on migration and recombination of dangling bonds,

along with reoxidation of the Si bridges. Dopant pairing trapping mechanisms were found to

dominate at very high concentrations because the interface becomes saturated with as much as a

monolayer of inactive donors.

More recently, work in the laboratory of E. G. Seebauer has sought alternative

explanations for the phenomenon of dopant segregation [DEV03, JUN04, DEV04]. Using

photoreflectance (PR), they were able to quantify the degree of band bending at the Si/SiO,

interface after Ar' implantation at 0.5 keV, lxlO' cm'2 [DEV03]. Band bending persisted at

temperatures up to 940 C for several minutes. This resulted in a change in the surface potential

of 0.4 and 0.52 eV for as-implanted with oxidation and oxide-free surfaces, respectively. The

presence of the change in surface potential results in a net electric field as shown in Fig. 2-38.

The electric field tends to repel positively charged interstitials attempting to move from the bulk

towards the surface. Interstitials closer to the surface can be negatively charged due to the

position of the Fermi level near midgap, and move uninhibited towards the surface. Jung et al.,

developed a model for predicting both TED and dopant segregation based on these results

[JUN04]. Using the general form of Eqs. 2.17 and 2.18, and incorporating an electromigration

term to Fick's first law, they were able to model a variety of TED and segregation data. Their

general flux and electric field equations are given as

J -Dj -' + zC E(x) (2.38)


E(x) (2.39)
E(x) ---- (2.39)








where z is the charge on species j, Ri the mobility, E the electric field, and W the electrostatic

potential. They included boron interstitial transitions of (+/-) and (+/0) and a (++/0) transition

for self-interstitials in the model. Their model shows an excellent fit with both the tail diffusion

and surface effects exhibited by boron in Fig. 2-39. They also indicate that their model provides

an alternative explanation for the boride-enhanced diffusion (BED), previously observed in a

number of studies [AGA99, COW99c]. Jung et al., indicates that the formation of a boride

phase will then lead to a number of interfaces with possibly larger potentials; this could make

the diffusion appear more enhanced [JUN04].

2.5 Dopant Diffusion in SOI

This section focuses specifically on diffusion studies performed using SOI substrates.

Boron is focused on because of its pertinence to the studies in Chapter 6. It should be

emphasized that as SOI material has evolved over the years, so the results of experiments may

also vary over the course of a few years. Thus, the results from earlier studies must be put into

context with more recent ones. For example, an experiment performed using early SIMOX

material will likely yield different results from one using state of the art SIMOX today.

2.5.1 Boron Diffusion in SOI

The ability of SOI material quality to affect the physics of processing was recognized in

early studies of boron diffusion by Normand et al. [NOR90]. They studied BF, implants in

SIMOX materials that had been fabricated in 1986 and 1988. Figure 2-40 shows SIMS profiles

for BF, implants at 40 keV, lxl0O4 cm' into the 1986 and 1988 specimens. Significant

differences in the pileup of boron near the surface can be seen to occur upon annealing in

nitrogen at 880 C for 100 minutes. The boron appears to pileup over a larger depth and lower

concentration in the 1986 SIMOX compared to the 1988. However, a higher concentration of








boron at the surface Si/BOX interface was found in the 1986 sample. Cross-sectional TEM

(XTEM) analysis showed a high density of SiC precipitates and threading dislocations in the

1986 specimens, while the 1988 showed no signs of process induced defects. This difference

was attributed to carbon contamination as a result of a long implant time used to fabricate the

1986 substrates. The 1988 SIMOX were fabricated using a high current implanter, similar to

that used to produce state of the art SIMOX today and allows for a much shorter implant time.

The threading dislocations may have been the reason for the enhanced diffusivity of B in the tail

region for the 1986 SIMOX. Spreading resistance profilometry (SRP) of the surface Si layer

showed that the boron piled up near the surface was inactive.

The most extensive studies of boron diffusion in SOI materials were done by Crowder et

al., in the mid 90s [CRO94a, CRO94b, CR095]. They used boron marker layers to study OED

and TED in SIMOX and bonded SOI materials. They noted an enhancement in the

recombination velocity of interstitials in single implant versus multiple implant SIMOX

materials. They attributed this to differences in the surface Si/BOX interfacial roughness.

SUPREM IV was used to simulate the OED and TED profiles. An effective recombination

velocity at the surface Si/BOX interface for bonded SOI was extracted and found to be IKD, =

4.7x10 lexp (+1.34/kT).

Vuong et al., studied the phenomena of B pileup and clustering in SOITEC materials with

surface Si thickness of 60-70 nm and BOX thickness of 200 nm [VUO99]. They used both B

marker layers and implants along with, in some cases, Si' implants. SIMS profiles of B

implanted at 10 keV, lxl0" cm2 are shown in Fig. 2-41. Top curves show Si signals obtained

from SIMS. Following the B implant a 1050 *C, 60 second anneal was performed. Solid curves

had an additional Si' implant at 40 keV, 5x10" cm-2. Both samples were then annealed at 800








*C for 30 minutes. A significant increase in the pileup at both the screen oxide/surface Si and

surface Si/BOX interface can be seen to occur with the addition of the Si' implant. Figure 2-42

shows SIMS profiles of B marker layers grown on SOITEC substrates using molecular beam

epitaxy (MBE) and doped to a concentration of lxl10" cm'3. Solid curves also had a Si' implant

at 25 keV, IxlO cm-2. Once again, enhanced segregation of B to the surface Si/BOX interface

occurred with the additional Si' implant. This allowed the authors to come to the conclusion

that self-interstitials significantly aid the pileup of boron in SOI films. However, the marker

layer was grown extremely far from the surface Si/BOX interface (~100 nm). If this were not

the case, perhaps noticeable segregation would have occurred in the unimplanted case. SIMS is

unable to resolve the low concentrations of B far from the marker layer, so it is impossible to

comment on the unimplanted marker. In their clustering study, they found little difference in

the clustering behavior between SOI and bulk Si. Less diffusion did occur in the tail region in

SOI, though. The authors believe the clusters form at early times before interstitial

recombination becomes a critical factor.

Recent work at IBM has investigated the effect of implant energy and surface Si thickness

on B diffusion [PAR99, DOK02]. In Park's experiment, using SIMOX of varying thickness

they implanted B and BF2 at equivalent energies to give the same projected range of the implant

[PAR99]. Figure 2-43 shows the B from BF2 redistribution after RTA at 1000 *C for 5 seconds

in nitrogen. A pileup of B appears on both sides of the surface Si/BOX interface in the 530 A

SOI, but the effect diminishes as the surface Si thickness is increased. The effect was also not

as pronounced at low implant energies.








2.5.2 Donor Diffusion in SOI

In their study, Park et al., also investigated P and As diffusion [PAR99]. Phosphorus

SIMS profiles are shown for each of the different thickness in Fig. 2-44. It shows a significant

reduction in the pileup of P compared to the B in Fig. 2-43. A similar result was obtained for

As. They proposed a model to try and account for differences between B, P, and As in SOI. It

incorporated stress at the surface Si/BOX interface due to compressive stress in the BOX,

gradual oxygen concentration at the interface, interface transport, and thermomigration.

Arsenic diffusion has also been studied by a few other investigators [OGU01, NOR89,

ROB90, SAT95]. In general, all these authors found As to pileup on the surface Si side of the

surface Si/BOX interface. Normand et al., also noted an enhanced diffusivity of As in their SOI

films [NOR89], but this was disputed by Robinson et al. in their experiment [ROB90].

Phosphorus diffusion was also studied extensively by Crowder in experiments similar to

those mentioned previously [CRO94a, CRO94b, CRO95). They found significantly less

diffusion of P in their OED experiments and this was confirmed by Uchida and co-workers

[UCHOO].

2.6 Summary

This chapter has reviewed the current understanding of point defect physics and how they

relate to SOI technology. After ion implantation, interstitials may undergo a number of

evolutionary stages that determine the effect they have on the diffusion of dopants. The

presence of interfaces is unavoidable in microelectronic processing and their interactions with

point defects and dopants must be accounted for in order for process simulators to provide

accurate results. Conflicting results from a number of studies cloud the true nature of how the

Si/SiO, interface interacts with interstitials. Several methods are currently available for






63

fabricating SOI substrates, but the most prevalent are the Smart-cut'" and SIMOX methods. It

is clear that the type and age of a SOI substrate may significantly alter the process physics and

must be accounted for when conducting a experiment.






64




Table 2-1. Advantages and disadvantages of ion implantation compared to gas source and solid
source diffusion.


Advantages Disadvantages

Easy introduction of desired impurity into Ion channeling
target

Good mass-charge separation Crystal damage

Accurate dose control Transient enhanced diffusion (TED) makes

ultra shallow junction formation difficult
Not confined to surface Extended defects can be source of leakage
Not limited by solid solubility current within device

Reproducibility of impurity profiles
Lower process temperature
Ability to tailor doping profiles













REGION
LIGHT
ION







UNIFORM DAMAGE
ALONG ENTIRE
HEAVY TRACK
ION

DISORDER
CLUSTER

Figure 2-1. Schematic of collision cascade produced by light ions (e.g. atomic weight less than
Si) and heavy ions (e.g. atomic weight greater than Si).





66


Ion Implantation


I I
{311) Defect Nucleation Loop Nucleation and
and Growth Growth



Defect Dissolution


Interstitial Release to bulk



Figure 2-2. Evolutionary path for point defects produced by ion implantation.





67



+-+Rp) )1.02. .+ 3.9aS


2

1-
E;

0 2 4 6 8 10
n number of interetitlals



Figure 2-3. Formation energy as a function of cluster size for self-interstitial defects in Si.
Closed diamonds represent formation energy for a compact cluster, while the open
triangles are for an elongated cluster [KIM00].













0




I"


0.1


0.0


S--7,,,3=0.65 eV


-,'2'


-- O Aluar-( 13)
- ---- PDLo


Y"=0.027 eV \


10 10 10"
Number of atoms


lo'
10'


io
10a

10


2 "
0


Figure 2-4. Formation energy as a function of cluster size as determined by Cower et al.,
[COW99a].






69




113




3j2




Figure 2-5. 3D representation of {113} defect in Si lattice. Light gray balls show interstitial
chains along <110> direction [TAK91].





70










7 Oh a s h 8 G h 6h a 6. Bh C Sh 4 Sh 7



E11l 3
t -e [53 21




Figure 2-6. Atomic structure of planar { 113) defect. Numbers represent rings different from
those in a perfect crystal [TAK94]
































Figure 2-7. Plan view TEM (PTEM) weak beam dark field (WBDF) micrograph of {311}
defects in Si.














101



E tol

0
IO

_C


1014

No extended defects

0 20 40 60 80 100 120 140
Ion mass (amu)
Figure 2-8. Formation criteria for extended defects in Si [JON88].


Categoy I defects


Category U defects




CA I dh l d


-.5-7 r o ose

egory H heshold dos












S'












+0
IntdrlUW loOeM wun(ntM























Vacancy Mechanism



Figure 2-9. Mechanisms of dopant diffusion in the Si lattice [CR095].





74

Table 2-2. Approximate fractional interstitial and vacancy components for various dopants in
Si.


Impurity Fractional Interstitial Fractional Vacancy
Boron 0.8- 1 0 0.2
Phosphorus 0.9- 1 0-0.1
Arsenic 0.4 0.6
Antimony 0.02 0.98
Silicon 0.6 0.4




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E3KTHMMUE_7X5MPH INGEST_TIME 2013-09-27T23:26:24Z PACKAGE AA00014269_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES