Determination and modeling of the interaction between arsenic and silicon interstitials in silicon

MISSING IMAGE

Material Information

Title:
Determination and modeling of the interaction between arsenic and silicon interstitials in silicon
Physical Description:
Unknown
Creator:
Brindos, Richard E., 1971-
Publication Date:

Record Information

Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 26983761
oclc - 47260216
System ID:
AA00020446:00001

Full Text









DETERMINATION AND MODELING OF THE INTERACTION BETWEEN ARSENIC
AND SILICON INTERSTITIALS IN SILICON












By

RICHARD E. BRINDOS


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
2001

























This work is dedicated to the people who had faith in me to complete it.













ACKNOWLEDGMENTS

I would like to first acknowledge my research advisor, Dr. Kevin S.

Jones for giving me the chance to study and to succeed at the University

of Florida. Dr. Jones made the experience interesting and fun and

provided the necessary funding for completion of my projects. I would

also like to acknowledge Dr. Mark E. Law for his help with the experimental

and modeling efforts in this dissertation and all of the Gator sports talks

throughout the last 5 years. I further acknowledge Guna Selvaduray at

San Jose State University for giving me the courage and motivation to

apply for graduate schools and the continual encouragement to complete

my journey into doctor-hood. I would also like to thank the rest of the

committee, Dr. Paul Holloway, Dr. Cammy Abernathy, Dr. Kenneth 0 and

Fred Stevie for serving on my committee. I express additional thanks to

Fred Stevie for increasing my knowledge and interest in SIMS and Agere

Systems (Formerly Lucent Tech.). Without Fred it would have been

impossible to get all the SIMS information necessary for this project and

others. I also extend special thanks to the administrative assistants

(Cindy, Carrie, Michelle, Lauren and Kelly) who have gone through our

group, throughout my 5 years, for putting up with me and my rudeness. I
iii








want to thank the industrial folks who made my two internships possible

and successful; Bob Ogle and Emi Ishida from AMD; and from Bell Labs-

Lucent Technologies, Conor Rafferty, Hong-Ha Vuong, Tony Fiory, Hans

Gossman, George Celler, Janet Benton, John Grazul, Ken Short and the

party room (Vinnie Venezia, Lourdes Pelaz, Ramki Kalyanaraman, Peter

O'Sullivan, Emiliano Rubio, Jacques Dalla Torre). I would also like to

acknowledge the industry sponsors from Intel: Paul Packan, Steve Cea and

Hal Kennel, who provided research and made all conferences and contact

reviews pleasant and memorable. I also thank Eb Andideh from Intel for

the growing of the boron spike material.

I would like to extend a very special thanks to Jamie Rhodes and Jay

Lewis for being two of the best roommates a guy could have and for

putting up with the fur of my cat, Whitey. They are probably still getting

the hair out. I extend special thanks to the members of the cabbage

organization for their impeccable sports and beer drinking abilities. I never

got to win that T-shirt but I had a great deal of fun trying.

I thank all of the SWAMP members past and present for sometimes

intellectual conversation, but mostly not. I would like to acknowledge the

lunch group of Patrick, Ibo, Chad and Luba for all of the spectacular

lunches at Shands Hospital. I thank Aaron Lilak for the use of his carpet

steam cleaner and all of the stock-tip advice. I also thank Sushil Bharatan

iv








for his initial research directions and the mountain bike rides (which I

haven't gone on since). I would also like to acknowledge Hernan Rueda for

all the soccer wounds. I thank Wish for all the TEM advice and research

directions.

I thank all the Gator sports coaches for providing championship-

winning teams. I give special thanks to Steve Spurrier and Billy Donovan

for providing truly great football and basketball teams that were exciting

to watch. I would like to acknowledge Ananth Naman for making football

tickets a top priority for me to receive upon first arrival. I would also like

to thank everyone who was associated with the Salty Dog. This includes

Pam, Chuck and the rest of the gang, thank you for everything.

I would like to finally acknowledge the people most responsible for

my ability to complete this work. I thank my parents, Helen Rachfal and

Raymond Brindos, for their love and financial support. I thank my dad for

the wonderful Volkswagen Jetta and for all of the advice on how to fix it. I

thank my mom for always being there when needed and for continuing to

call weekly throughout my tenure. I also thank my brother for his love and

support of my venture. Finally I give thank to my significant other, Monica

Taylor, for her patience in the final leg of this journey and for being able to

start our ADULT life together. I give thanks again to all. I apologize to

and thank all those I left out.













TABLE OF CONTENTS

Page
ACKNOW LEDGMENTS................................................................................ iii
LIST OF TABLES ........................................................................................ ix
LIST OF FIGURES ....................................................................................... x

CHAPTERS

1 INTRODUCTION ..................................................................................... 1
1.1 Motivation and Objective................................................................ 1
1.2 Ion Im plantation and Defect Generation......................................... 3
1.2.1 Ion Im plantation ....................................................................... 3
1.2.2 Im planted Ions.......................................................................... 5
1.3 Im plant Damage Characteristics..................................................... 8
1.3.1 Ion Collisions.............................................................................8
1.3.2 Im plantation Related Defects .................................................. 9
1.4 Point Defect Diffusion .................................................................. 11
1.4.2 Point Defect Diffusion Model ................................................. 11
1.4.3 Transient Enhanced Diffusion ................................................ 15
1.5 Precipitation and Clustering ......................................................... 17
1.6 Arsenic Background...................................................................... 19
1.6.1 Arsenic Overview ................................................................... 19
1.6.2 Solid/Electrical Solubility........................................................ 20
1.6.3 Below Electrical Solubility ...................................................... 28
1.7 Thesis Statem ent ......................................................................... 30

2 EFFECT OF ARSENIC ON {311} FORMATION AND DISSOLUTION.........35
2.1 Introduction .................................................................................. 35
2.2 Experimental Overview ................................................................. 37
2.2.1 Arsenic W ell Formation .......................................................... 37
2.2.2 End of Range Loops from Preamorphization ......................... 41
2.3 Defect Analysis............................................................................. 42
2.3.1 Im age Analysis ....................................................................... 42
2.3.2 Effect of Arsenic on Trapped Interstitial Values................... 43








2.3.3 Effect of Arsenic on {311} Defect Dissolution...................... 44
2.3.4 Effect of Arsenic on the "Plus One Model"............................ 46
2.3.5 Activation Energy Calculations.............................................. 48
2.3.6 Effect of Arsenic on Defect Size and Density....................... 49
2.3.7 Possible Arsenic Complexes................................................... 51
2.4 Arsenic Effect on {311} Formation and Dissolution Summary .... 53

3 RELEASE OF SILICON INTERSTITIALS AND VACANCIES FROM DOPED
ARSENIC LAYERS................................................................................. 84
3.1 Boron Marker Layers..................................................................... 84
3.1.1 Overview ................................................................................ 84
3.1.2 Boron Marker Layer Setup ..................................................... 85
3.1.3 Special Considerations for Implant Conditions....................... 86
3.2 Experimental Overview ................................................................. 88
3.3 Results and Discussion .................................................................90
3.4 Boron Marker Layer Summary ......................................................92
3.5 Antimony Doped Superlattices..................................................... 93
3.5.1 Antimony Experimental Setup .............................................. 95
3.5.2 Vacancy Injection From Arsenic Doped Layers .................... 96
3.5.3 Antimony Doped Superlattice Summary ........................... 100

4 MODELING ....................................................................................... 109
4.1 {311} Defect Dissolution Model................................................ 109
4.2 Process Simulation..................................................................... 112
4.2.1 Arsenic Interstitial Binding Energy Determination.............. 112
4.2.2 Temperature Dependence of Simulations........................... 113
4.2.3 Simulated Results................................................................ 114
4.3 Using the As-I Pair Model for Testing Ramp Rate Effects......... 115
4.4 Modeling of High Concentration, Low Energy Arsenic.............. 116

5 SUMMARY AND FUTURE WORK......................................................... 137
5.1 S um m ary .................................................................................... 137
5.2 Future Work............................................................................... 145

APPENDICES

A PEAK ADJUST MACRO ...................................................................... 147
A.1 Boron Marker Layer Adjustments.............................................. 147
A.2 Peak Adjust Macro..................................................................... 148








UST OF REFERENCES ........................................................................... 153
BIOGRAPHICAL SKETCH ....................................................................... 163













LIST OF TABLES

Table 2.1. Experimental Matrix for arsenic well study ................................... 41

Table 3.1. Different process steps used in the boron marker layer study.....89

Table 4.1. Parameters used in As-I simulations............................................ 119












UST OF FIGURES

Figure 1.1. Typical CMOS transistor ............................................................... 31

Figure 1.2. Semiconductor Industry Association roadmap information..........31

Figure 1.3. Typical commercial ion-implantation machine ............................. 32

Figure 1.4. Energy loss mechanisms ............................................................... 32

Figure 1.5. Critical ion implantation parameters.......................................... ..33

Figure 1.6. Disorder produced from light and heavy ions.............................. 34

Figure 1.7. Possible dopant diffusion mechanisms ........................................ 34

Figure 2.1. Process steps used in the design of an experiment.................... 54

Figure 2.2. Secondary Ion Mass Spectroscopy plots of the arsenic wells .....55

Figure 2.3. Cross-sectional TEM image of the processed region..................56

Figure 2.4. {311} defects formation: 700C 45min .....................................57

Figure 2.5. {311} defects formation: 700C 275min ...................................58

Figure 2.6. {311} defects formation: 700C 720min ...................................59

Figure 2.7. {311} defects formation: 750C 15min .....................................60

Figure 2.8. {311} defects formation: 750C 33min .....................................61

Figure 2.9. {311} defects formation: 750C 45min .....................................62

Figure 2.10. {311} defects formation: 750C 90min ....................................63

Figure 2.11. {311} defects formation: 800C 5min ....................................... 64

Figure 2.12. {311} defects formation: 800C 10min ..................................... 65

Figure 2.13. {311} defects formation: 800C 15min .................................... 66








Figure 2.14. Number of trapped interstitials in {311} defects:700C .......... 67

Figure 2.15. Number of trapped interstitials in {311} defects:750C ......... 68

Figure 2.16. Number of trapped interstitials in {311} defects:800C ......... 69

Figure 2.17. {311} defect dissolution after furnace anneals at 700C ......... 70

Figure 2.18. {311} defect dissolution after furnace anneals at 750C ......... 71

Figure 2.19. {311} defect dissolution after furnace anneals at 800C ..........72

Figure 2.20. Plot of dissolution rate constants at various temperatures ......73

Figure 2.21. Size distribution data at various times at 700C ....................... 74

Figure 2.22. Size distribution data at various times at 750C ..................... 76

Figure 2.23. Size distribution data at various times at 800C ...................... 78

Figure 2.24. Total defect densities:700C ..................................................... 80

Figure 2.25. Total defect densities:750C ..................................................... 81

Figure 2.26. Total defect densities:800C ..................................................... 82

Figure 2.27. Missing interstitial dose ............................................................. 83

Figure 3.1. Inert and enhanced diffusion at 800C 15 min ........................... 102

Figure 3.2. Fabrication process for boron marker layer experiment............. 103

Figure 3.3. Secondary Ion Mass Spectroscopy (SIMS) plots......................... 104

Figure 3.4. Time averaged enhancements versus time................................ 105

Figure 3.5. SIMS arsenic concentration profiles ........................................... 106

Figure 3.6. Location of the peak profiles ..................................................... 107

Figure 3.7. SIMS antimony concentration profiles ....................................... 108

Figure 4.1. Difference in FLOOPS simulations and experimental data ......... 120

Figure 4.2. Scatter of various {311} dissolution data ................................. 121

Figure 4.3. Dissolution curves at 750C ....................................................... 122

xi








Figure 4.4. Effect of changing the binding energy on the dissolution ......... 123

Figure 4.5. Effect of changing the binding energy ...................................... 124

Figure 4.6. Effect of changing the binding energy (percentage)................. 125

Figure 4.7. Effect of temperature on the number of trapped interstitials .. 126

Figure 4.8. Number of trapped interstitials at zero time ............................. 127

Figure 4.9. Simulation runs versus experimental data:3x1017 cm-3.............. 128

Figure 4.10. Simulation runs versus experimental data:3x10'8 cm-3............ 129

Figure 4.11. Simulation runs versus experimental:3x109 cm-3.................... 130

Figure 4.12. Differences in activation energy .............................................. 131

Figure 4.13. Effect of changing the ramp rate in the simulations................ 132

Figure 4.14. Defect formation with no arsenic present ............................... 133

Figure 4.15. Diffusion simulations of 3 keV lx10J15 cm-2 arsenic implants... 134

Figure 4.16. Diffusion simulations of 3 keV 1x1015 cm-2 arsenic implants .. 135

Figure 4.17. Simulation of 3 keV 5x1015 cm-2 arsenic implant .................... 136













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

DETERMINATION AND MODELING OF THE INTERACTION BETWEEN ARSENIC
AND SILICON INTERSTITIALS IN SILICON

By

Richard E. Brindos

May 2001


Chairman: Dr. Kevin S. Jones
Major Department: Department of Materials Science and Engineering



Arsenic has evolved as the premier dopant for N+ source and drain

structures in current generation processors. To further use arsenic in

future devices, research is needed on the transient enhanced diffusion

and activation properties. Until now, researchers have concentrated on

higher dose arsenic implants. The results from the higher dose implants

become difficult to interpret because of effects from implant damage,

solubility, and cluster formation. The goal of this research is to use lower

arsenic concentrations to determine the basic interaction between arsenic

atoms and silicon interstitials. The arsenic-silicon interaction is then








expanded to include the high-dose effects. Arsenic wells of varying

concentrations were formulated and silicon implants known to cause

{311} defects were added. The structures were then annealed and the

nucleation, growth and dissolution of the {311} defects were monitored.

Arsenic had a distinct effect on the nucleation but little to no effect on

the dissolution. As the arsenic concentration increased, the number of

interstitials in the defects at time zero decreased. The same decrease

was realized at each temperature studied. The results show that arsenic

is pairing with interstitials during the initial stages of the annealing cycle.

The pair formation decreases the interstitial population and with fewer

interstitials present, fewer defects are able to nucleate. This result also

leads to a smaller number of trapped interstitials in {311}'s at higher

arsenic concentrations. Using Florida's Object Oriented Process Simulator

(FLOOPS) and a simple pair model the experiment was modeled and the

binding energy between an arsenic atom and an interstitial was

determined to be 0.95 eV.

Boron marker layers were used to monitor the release of

interstitials from an arsenic-implanted region. It was found that at the

initial stages of annealing, the enhanced motion of the boron marker layer

was reduced in comparison to a control wafer with only a silicon implant.








It was also determined that an enhanced diffusion was seen in the arsenic

only samples because of an injection of interstitials from cluster

formations.













CHAPTER 1
INTRODUCTION

1.1 Motivation and Objective

To comply with the speed and power demands of today's

computers, transistor performance must continually increase. A

schematic diagram of a typical CMOS transistor used in current generation

microprocessors is featured in Figure 1.1. Referring to this figure, the

terms used to describe a transistor can be defined. The SDE is the

source/drain extension, which is a shallow diffusion that connects the

channel or metallurgical spacing with the deep source and drain. The

overlap is the distance the SDE extends under the gate and the

metallurgical spacing is the channel between the SDEs of the source and

drain, where the electron flow is regulated. The junction depth refers to

the depth of the SDEs after diffusion.

Gate length, gate dielectric thickness, and junction depths are the

primary parameters that control transistor performance. In Figure 1.2,

the Semiconductor Industry Association (SIA) roadmap shows that

junction depths and gate lengths will continue to decrease in the years to








come.1 In order to continue this downward trend it becomes increasingly

important to understand and control dopant diffusion in silicon.

The manufacturing of shallow junctions mandates the introduction

and activation of dopants such that a minimal amount of diffusion occurs.

The primary method used to introduce dopants is ion implantation.

Activation is achieved through subsequent annealing steps.2 Ion

implantation is known to cause large amounts of lattice damage that

must be annealed out in order to restore device performance.3 During

this annealing step, a large supersaturation of point defects is available to

enhance the diffusivity of dopants in the area of the implanted region. In

some instances the equilibrium diffusivities of the dopants can be

increased by three orders of magnitude or more, thereby driving the

junction depths to unacceptable values. The enhancement lasts only a

short time until the local concentration of point defects returns to the

equilibrium value and equilibrium diffusivities are restored. Therefore, the

enhanced diffusion is transient in nature and is known as transient

enhanced diffusion or TED.

Of the dopants used in today's transistors, arsenic has emerged as

the most common n-type dopant. Like other dopants, implanted arsenic

in silicon has been shown to exhibit TED. Many studies have concentrated

on arsenic TED and the fundamentals associated with it.4-19 These








studies, however, have concentrated on arsenic concentrations near or

above the solid solubility limit and most of the focus of these studies is

on electrical activation. Results from these studies often become difficult

to interpret because they contain effects from solid-phase epitaxial

regrowth, arsenic precipitation, implant damage, and the formation of

dislocations. The goal of this work is to take an extensive look at the

more fundamental issue of how arsenic and silicon interstitials interact

and to provide a model of their behavior.


1.2 Ion Implantation and Defect Generation


1.2.1 Ion Implantation

The ability to change the conduction properties of a semiconductor

is the main attribute that makes semiconductors useful for electronic

devices. The way to change the conduction properties is to introduce

elements, known as dopants, into the semiconductor material such that a

high number of charge carriers are generated. This procedure of

introducing dopants can be accomplished in a variety of ways. Whichever

method is used to introduce dopants must be controllable, reproducible,

and free from undesirable side effects. In the past, dopants were

introduced by indiffusion. Dopants were diffused in from a surface source

such as a doped glass or by holding a constant atmosphere of a dopant-








containing gas over the surface. These were viable methods. However,

the amount of dopant able to be introduced was limited by solid solubility

and it became difficult to incorporate a sufficient amount of dopant. It

was also difficult to laterally diffuse the dopant under surface structures

such as gate stacks. Because diffusion from solid sources doesn't supply

all the necessary parameters, the technique of ion implantation was

developed.2,20-22

During ion implantation, a liquid or solid source containing the

desired dopant material is heated and emits vapors of the dopant atoms.

A cloud of electrons emitted by a heated filament then ionizes the atoms.

The ions then pass by an analyzing magnet and unwanted species are

filtered out. Correctly filtered ions are then accelerated toward a silicon

target and rastered over the surface. A schematic diagram of a typical

ion implanter is shown in Figure 1.3.

Monitoring the ion current can control the dopant dose and

adjusting the ion acceleration energy can control the average depth of

the ions. Ion implantation therefore satisfies the needed parameters for a

general doping process. The main disadvantage is the damage done to

the silicon lattice due to ion collisions. This damage may be removed by

subsequent heat treatments.







1.2.2 Implanted Ions

As energetic ions penetrate a solid target material, they lose

energy because of collisions with atomic nuclei and electrons in the target

material and the ions eventually come to rest. The ions are stopped in a

solid by two processes; nuclear and electronic stopping. These processes

are shown in Figure 1.4.

The dominant stopping mechanism depends on the atomic weight and

implant energy of the implanted species. At lower energies, a process of

nuclear stopping stops ions, while at higher energies ions come to rest by

the process of electronic stopping.23,24 The stopping power of the

target is the loss of energy per unit distance, -dE/ds, which is defined as

d= N[Se (E) + SN (E)] (1.1)

where, Eis the ion energy, s is the coordinate along the path whose

direction changes as a result of binary nuclear collisions, N is the density

of atoms in the target material, Se is the electronic stopping power and SN

is the nuclear stopping power. The total distance that an ion travels in a

solid is known as the range, and is defined as,

R=fds SEdEE (1.2)
-"f s fNf Se(E).I+ SN (E)





6

Equation 1.1 was reformulated by Lindhard et aL.25(LSS theory) for

implantation into amorphous material in terms of the reduced parameters,

e and p as,

RNM1M2 47ra2
p = (1.3)
M1 +M2

and,

EoaM2 (1.4)
ZZ2q 2 (M, +2)



where, M, and Z, are the mass and atomic number of the incident ion,

respectively; Ms and Z2 are the mass and atomic number of the target

atoms, respectively; N is the number of atoms per unit volume; and a is

the screening length, equal to

a = 0.88ao
(Z71/3 2 /3)1/2 (1,5)


where ao is the Bohr radius. (Calculations for e and p can be found in the

paper by Lindhard et a/.25). With Equation 1.3 the value of p can be

converted to the projected range, Rp, which is the average depth below

the surface an ion penetrates, using the expression








R (R
P 1+ (M2 (1.6)



Assuming that the depth profile of the implanted ions in amorphous

materials could be described by a symmetrical Gaussian curve, then the

ion concentration, n, as a function of depth, x, is given by

n(x) = -R exp[ ( P)2(1.7)
42;AR? 2AR?2{

where 0 is the ion dose in cm-2, x is the junction depth, Rp is the projected

range, and ARp is the projected straggle. The standard deviation of the

Gaussian distribution ARp is given by

AR= 2R1 (1.8)
3 LM, + M2

The peak concentration occurs when x=Rp, which leads to

nmax(Rp) = -@_ 0.4@ (1.9).
;irARP AR-p

For a given implant the predominantly desired values are the projected

range and projected straggle. These parameters are detailed

schematically in Figure 1.5. The projected range and projected straggle

for boron, phosphorous, and arsenic in silicon and silicon dioxide (Si02) for

various implant energies can be found in Smith et al.26








1.3 Implant Damage Characteristics


1.3.1 Ion Collisions

As stated in the previous section, energetic ions are involved in

several collisions with lattice atoms before coming to rest in the target

material. A collision of -15 eV is all that is needed to knock a silicon host

atom from a lattice site.21 If adequate energy is transferred in the first

collision the displaced atom may collide with other atoms to cause more

displacements and collisions. The process continues until the energy of

the collisions drops below the lattice displacement energy. Such a

process results in a collision cascade.

The damage path created will depend on the relative masses

between the dopant atom and the lattice atoms. Figure 1.6 shows a

schematic of the lattice disorder that may be created for both light and

heavy ions and with enough collision cascades the formation of an

amorphous layer will result.22 An amorphous layer is defined as a layer

that exhibits no long-range order. A light ion transfers small amounts of

energy during each collision and with a small enough transfer, few

additional displacements are created. In the case of a heavy ion, the

energy transfer is enough to cause additional collisions. The range of the

ions is generally small since the energy transfer is mostly by nuclear








stopping. The small range and nuclear collisions create localized pockets

of damage regions. When the dose of the implanted ions is increased to

a high enough point, the damage regions begin to overlap and an

amorphous region results.21

1.3.2 Implantation Related Defects

Ion implantation induced damage to the silicon lattice that must be

corrected so that the implanted dopants become electrically active. Upon

annealing, the energetically favorable position for an excess ion may not

be a lattice site. It is possible that the excess ions cluster together to

form extended defects that are more energetically favorable positions for

ions to occupy. The type of defect that forms depends on the conditions

of the implant (energy, dose, temperature, etc.) and the post-

implantation annealing conditions. In order to separate the different

types of damage, Jones et al.27 formulated a classification system in

which the various types of defects were separated according to the type

of damage from which they originated.

Type I damage is formed when the implant conditions are below

that of amorphization. In this case point defects cluster around the

projected range of the implant where the supersaturation is the highest.

Upon annealing, {311} defects or dislocation loops will form due to the

clustering of these point defects. Type II damage, also known as end-of-





10

range (EOR) damage, forms if the implant dose is high enough to cause

amorphization. Upon annealing, EOR damage will manifest itself as

dislocation loops that form just below the amorphous/crystalline

interface. Type III damage results from the regrowth of the amorphous

layer and generates several types of defects such as hairpins or

microtwins. These types of defects do not seem to affect the diffusivity

of dopants, but they can act as gettering sites for impurities such as

carbon or oxygen. Type IV damage occurs when buried amorphous layers

are created. During regrowth, the regrowing amorphous layer and

crystalline layers meet and "clamshell" or "zipper" defects are generated.

Type V damage forms when the implant is of a sufficient concentration to

reach the solid solubility limit of the dopant in silicon. Upon annealing,

precipitates of the impurity atom can form. This type of damage is

commonly found in arsenic implanted samples. If the pathway for the

formation of this precipitate is through an arsenic-vacancy cluster, then

the formation can generate high levels of vacancies which, in turn, can

lead to the dissolution of dislocation loops (however, this has never been

proven). Type I and II damage are the relevant defect classifications in

the portion of this work that involves extended defects. In other portions

of this work, implants of arsenic will be chosen carefully such that Type V

defects are avoided.










1.4 Point Defect Diffusion

In early devices, critical dimensions were large enough that the

amount of diffusion was not a significant factor. However, with the

scaling of devices to smaller dimensions, designers are forced to

reevaluate the amount of diffusion after annealing. This is complicated

due to implant-induced defects that may cause the equilibrium diffusion

to be enhanced.

Point defects such as interstitials and vacancies, created during the

implant process, are known to cause several problems during processing.

For instance, interstitials released by extended defects have been shown

to enhance diffusivities of dopants and decrease the activation

percentage.28-31 The main focus of this work is to establish how

interstitials interact with arsenic atoms. This information will provide

valuable insight in eliminating problems such as TED and low activation in

arsenic implanted regions.

1.4.2 Point Defect Diffusion Model

Ion implantation damage is removed through an annealing process.

Annealing causes dopants to redistribute in the silicon lattice through the

process of diffusion. In early diffusion studies, the diffusivity of impurities

was modeled using an Arrhenius equation of the form








D(T) =Doexp( (1.10)


where, D is the diffusion coefficient, Do is a temperature-independent pre-

exponential, E, is the activation energy for the diffusion of the impurity, k

is Boltzmanns constant, and T is the temperature. The magnitude of the

diffusion coefficient, D, indicates the rate at which atoms or impurities

diffuse. This equation, however, only describes the influence of

temperature on the diffusion of an impurity and has no terms for the

mechanisms responsible for the diffusion in silicon.

In silicon, dopant atoms are known to diffuse through interactions

with silicon self-interstitials and vacancies.29 Equation 1.10 does not

take this into account and therefore is an unreliable model for dopant

diffusion in silicon. In lieu of this, a more complicated model for dopant

diffusion in silicon must be constructed.

Dopants diffuse in silicon by combining with either an interstitial or

vacancy.29 The diffusion of a dopant, A, must then be made up of the

sum of the diffusion mechanisms or

DA =D, +DAv (1.11)

where, DA, and DAv are the interstitial and vacancy components for the

dopant diffusion. If DA is defined as the dopant diffusion under






13

nonequilibrium conditions, then DA is the intrinsic diffusion under

equilibrium conditions. Dividing Equation 1.11 through by DA* leads to

DA = DA-L DAv (1.12).
D DA DA

To account for the fraction of diffusion due to interstitials and

vacancies, Equation 1.12 can be rewritten as

DA= D DAI + DAv DAV (1.13)
D; DA, D*A DAv DA

The ratio Dx/D* (X signifies an interstitial or vacancy) is the fraction of

diffusion that occurs through an interstitial or vacancy mechanism.

D'xID/ can be defined as fx and because the interstitials and vacancies

are the only mechanisms through which dopant diffusion can happen,

then

fA,,+ =1 (1.14).

Using this definition, Equation 1.15 can be rewritten as

DA=f DA, ) DAV
A ,; + D-A) (1.15).
DA ^ AV

DAI/DA, and DAv/DAv are proportional to the relative

concentrations of interstitials, CI/C*, and vacancies, Cv/Cv, and Equation

1.15 can be written in its final form as
DA = fAI + -fA )- (1.16).
DA CI LA/






14

Equation 1.16 shows how point defect concentrations and dopant

diffusion are related and shows that by changing the defect

concentrations the dopant diffusivities can be adjusted. The important

point of this development is that to fully understand dopant diffusion in

silicon, it is critical to have an understanding of how the point defect

concentrations are affected by different processing steps.

The type of point defects that dominate the diffusion process

determines the type of diffusion mechanism, vacancy, interstitial or

interstitialcy. In the vacancy mechanism, the substitutional dopant atom

migrates through the lattice by moving on to an adjacent lattice site that

is vacant. In the interstitial mechanism, the dopant atom is kicked out of

a lattice site by a silicon self-interstitial and migrates through the

interstices as a dopant interstitial until it returns to a lattice site as a

substitutional atom. In the interstitialcy mechanism, the silicon self-

interstitial and the dopant atom form a diffusion pair.29 These processes

are shown in Figure 1.7. There is generally no distinction made between

the interstitial and interstitialcy mechanisms because they cannot be

differentiated from each other.

Studies have been done to determine what fraction of dopant

diffusion is interstitial and what fraction is vacancy, for a variety of

common dopants. It is generally agreed that boron and phosphorus are








pure interstitial diffusers and therefore, f, = 1.29 On the other hand,

antimony has been shown to exhibit pure vacancy diffusion with fAV 1.29

It has also been shown that arsenic exhibits both interstitial and vacancy

diffusion with fA, 0.5.29

Dopants that exhibit pure interstitial or vacancy diffusion can be

used as effective marker layers for studying the release of point defects

in silicon. Buried marker layers of boron or antimony can lead to essential

data on the release of interstitials or vacancies from a specific area of

interest. A discussion of how this is accomplished is presented in a future

chapter.


1.4.3 Transient Enhanced Diffusion

As device dimensions continue to shrink, source and drain regions

become closer together, causing unwanted interactions. Under

equilibrium diffusion, annealing should cause minimal motion in the source

and drain profiles. However, the ion implantation process introduces

interstitials and vacancies that lead to more motion of the profiles than is

predicted by equilibrium diffusion. Studies of the enhanced diffusion have

shown that the diffusion enhancement decays back to the equilibrium

diffusion value over time, therefore the enhancement is transient in

nature and classified as transient enhanced diffusion or TED.29,31-34 If the





16

dimensions of the device are such that little diffusion is warranted, then

the extra diffusion cannot be tolerated. In recent years many research

groups have dedicated their studies to the relationship between TED and

implant-related defects to understand and control the effects of

enhanced diffusion.28-31,33-38

The exact nature of TED has been thoroughly studied and debated

among many researchers without a general consensus. Some believe that

TED is due to the dissolution of extended defects, others feel sub-

microscopic clusters are responsible, and a portion feel that a

combination of clusters and extended defects are the cause.32,39-41

Eaglesham et al.39 compared the dissolution of extended defects with the

enhanced motion of boron marker layers. It was concluded that the

transient duration of TED, and thus the reaction rate constant, increased

very rapidly with increasing temperature. By correlating the number of

interstitials released during {311} dissolution with the TED duration, they

found that the {311}'s could account for the entire enhancement in

diffusion rates in the temperature regime 670 to 815C. Cowern et al.42

did a study of the interaction between interstitials and {311} defects and

found two distinct periods of enhancement. The initial period was due to

silicon interstitials created by collisions during the implant. The

interstitials would either enhance diffusion for low dose implants or drive








the nucleation of {311 }'s at higher damage densities. The later period of

enhancement was a much slower diffusion transient that was similar to

the mechanism described by Eaglesham et aL.39 Zhang et aL.41 came upon

conditions where no defects formed but TED was still present. From this

they implied that there may be more than one source of interstitials

available for TED. Liu et al.40 found that {311} defects could not account

for all of the excess interstitials and suggested that a combination of

{311} and cluster dissolution drive TED simultaneously.

Recent work by Cowern et aL.36 has shown that there is a

nucleation threshold for {311} defects. Therefore, if a sub-microscopic

dopant interstitial cluster it might be possible to trap enough interstitials

in the clusters to avoid the formation of extended defects. Upon

annealing, the clusters may break up and release the interstitials causing

enhanced diffusion. Boron, phosphorus, and arsenic have all been shown

to exhibit cluster formations that affect the formation of extended

defects.37,43-46



1.5 Precipitation and Clustering

In the development of new device structures the electrically active

fraction of dopants is an important quantity. Through many experiments

it is known that above a certain concentration dopants become








electrically inactive.29,46-51 The concentration of atoms that can become

electrically active at a given temperature is generally controlled by the

solubility limit. The solubility limit is given by the concentration of

dopants that will dissolve substitutionally in the silicon matrix at a given

temperature. Concentrations above this level will lead to precipitation of

a second phase in the matrix. Precipitation related defects have been

considered to be responsible for most of the electrical inactivity in boron,

phosphorus, and to some extent antimony.29

Like other dopants, arsenic exhibits precipitation if a high enough

concentration is obtained. However, electrical inactivity starts well before

the solubility limit for formation of a macroscopic second phase is

reached. Theories of clustering have been proposed to account for the

electrical inactivity, where multiple arsenic atoms form some new

configuration with an interstitial or vacancy, which is electrically inactive

at room temperature. The difference between precipitates and clusters is

that precipitates are a macroscopic second phase that may contain

thousands of dopant atoms where the size distribution is a function of

the dopant concentration above the solubility limit and the thermal

treatment that follows. Clusters, on the other, hand are composed of a

few dopant atoms in specific configurations and their formation may be

enhanced by excess silicon interstitials. Clusters exist in equilibrium with






19

isolated dopant atoms just as AX defects coexist with an isolated A,

where precipitates are regions of the crystal that have formed a second

phase of the solute and solvent constituents. Dopant will eventually

precipitate at high enough concentrations in silicon.



1.6 Arsenic Background


1.6.1 Arsenic Overview

Arsenic is the most common n-type dopant used in silicon based

microelectronic device fabrication. High Mass, high solubility, high

electrical activation, and low diffusivity are all properties that make

arsenic an attractive dopant to the device industry. Although arsenic

displays all these desired qualities, transient enhanced diffusion (TED) and

electrical activation are still concerns.

TED and electrical activation studies of arsenic implanted samples

have lead to the conclusion that there is both an electrical and solid

solubility limit associated with arsenic. From this conclusion, it is realized

that there are distinct regimes of concentration when dealing with arsenic

in terms of TED, activation, and point defects.

At low concentrations (below electrical solubility limit), but above

amorphization, TED is dominated by end of range damage and surface








effects. Increasing the arsenic concentration above the electrical

solubility limit, leads to arsenic clustering at which point deactivation of

the arsenic begins. The clustering reaction is believed to lead to

interstitial injection through the reaction AsnSi = AsnV+/.50'51 It has

been shown that if the interstitial injection reaches a sufficient level then

the formation of dislocation loops is possible.52,53 As the arsenic

concentration increases further and the physical solid solubility limit is

reached, a monoclinic AsSi phase forms.16 It has been suggested that

the formation of the AsSi phase injects vacancies and causes the

dissolution of dislocation loops.54,55 The following few sections

investigate studies in the different regimes in more depth.


1.6.2 Solid / Electrical Solubility

Over the years there have been many studies that have

concentrated on the diffusion of arsenic. In these studies, there have

been many debates as to whether or not arsenic exhibits transient

enhanced diffusion. Some authors present evidence that the diffusion

difference is within the error of the measurement techniques used, some

claim that there is TED, while others claim the TED seen in arsenic has

been confused with the standard concentration dependent diffusion

effect.56







Hoyt et al.57 found that they could model their own experimental

data along with data from the literature with an effective diffusivity that

included arsenic diffusing with a neutral, a single negatively charged

vacancy, and a doubly charged vacancy. They found that for

concentrations below 2x1020 cm-3 that they only needed to include the

neutral and the singly charged vacancy components in the effective

diffusivity equation or that

Def =Do +D-IJn (1.17)

in order to fit the data well. When the arsenic concentration was

increased to greater than 2x1020 cm-3 then the addition of a second term

was needed so that

Def, Do +D-( n+D= ( (1.18).
Sni n,-

All of the profiles fit were for rapid thermal annealing data at

temperatures greater than 1000C.

Another primary focus of research has been on the activation and

deactivation processes. Studies of other dopants have shown that the

dopant can become fully active up to the solid solubility limit where

precipitation begins and deactivation follows. Arsenic has been shown to





22

start the deactivation process well before the solid solubility limit is

reached.58

Nobili et al.58 performed experiments that compared the electrical

deactivation of arsenic with the formation of arsenic precipitates. Silicon

samples were implanted with 100 keV arsenic at doses ranging from

5x1015 up to 1x1017 cm-2. The samples were next laser annealed at

energies sufficient to melt the implanted region. By melting the

implanted region, the arsenic atoms are able to go into solution, which

allowed for complete activation of the dopant atoms. TEM shows no

crystalline defects after laser annealing. The fully activated samples were

then annealed in temperature regimes where the solid solubility is

exceeded and precipitation occurs. Using electrical measurements it was

determined that only a two-phase equilibrium, that is the formation of

precipitates, is compatible with the results. Results from channeling

studies showed that the precipitates had to be coherent with the silicon

matrix. Because the precipitates had to be coherent, little to no strain

field is associated with the precipitates and therefore TEM techniques for

viewing are hindered. Small angle x-ray scattering (SAXS) was used to

verify the existence of the precipitates and moreover that they were in

the shape of thin platelets.






23

Using samples processed in a similar manner to Nobili et aL.58,

Armigliato et aL.59 were able to view some form of precipitate using TEM.

TEM observations of samples annealed at 450C for 4 hrs showed that

small defects were visible upon annealing. The small defects were

depicted as precipitates on the basis of the large amount of arsenic that

was deactivated. In addition, the defects were determined to be in the

shape of platelets, which was in agreement with the findings of Nobili et

aL58 {311} defects were also visible, however no dislocation loops

formed at this condition. Samples annealed at 900C for 30 min showed

precipitates in platelet form as well as dislocations and loops. Further

investigation showed that assuming an SiAs composition for the

precipitate could not fully explain the amount of deactivation seen in the

electrical measurements. The remaining arsenic content was said to lie in

particles that were of a smaller size than could be imaged by the TEM.

Wu et aL.60 looked into the formation of dislocation loops in more

detail. Using 100 keV, 5x1015 cm-2 arsenic implants followed by anneals

at 600C, they showed that two distinct layers of loops form. A layer of

loops formed at the projected range of the implant and another formed at

the end or range. The results showed that the projected range loops

formed due to exceeding the solubility limit. The projected range loops

grew rapidly in size and were said to glide to the surface. The presence






24

of oxygen was also shown to have a large effect on the pinning of the

loops.

Jones et aL.61 did a study similar to that of Wu et aL.60 and found

similar results. They showed that the dissolution of arsenic precipitates

lead to the growth of half loops. It was shown that arsenic doses above

the solubility limit produced loops and half loops at the projected range

upon annealing. Further annealing caused the loops to dissolve and the

half loops to grow. In difference to Wu et al. it was suggested that the

loops were not gliding to the surface, rather they were just dissolving via

climb. It was also shown that the number of atoms bound by the

projected range loops was insufficient to account for the entire growth of

the half loops. Continued analysis lead to the result that dissolution of

arsenic clusters could provide a sufficient number of interstitials to

account for the half loop growth.

Hsu et aL.54 also saw a two-layered structure after an arsenic

implant and anneal. They concluded that the dissolution of loops at the

projected range for high dose implants was due to the injection of

vacancies from the precipitates. To this day it is still unsettled as to the

exact nature of the dissolution of the loops at the projected range. No

experiments have been performed to solely determine if the dissolution is

due to injection of vacancies from the arsenic precipitates or are the








precipitates a strong sink for interstitials that are supplied by the

dissolution of the loops.

Nobili et aL.62 has suggested that arsenic displays both an electrical

and solid solubility limit, with electrical inactive clusters being responsible

for the difference. The electrical solubility limit has been shown to be

dependent on the equilibrium carrier densities and have an exponential

dependence on the annealing temperature given by:

n,(A) = 2.2x1 0 exp .7)j (1.19)


where kT is in eV.

To further understand arsenic deactivation Rousseau et a/.49-51,63-

65 investigated arsenic concentrations around the electrical solubility limit

but below the precipitation limit, to determine the deactivation reaction.

They showed that with laser annealing very high activation levels could be

achieved. However, when subsequent annealing was performed around

750C, a significant amount of deactivation was observed. This is an

important finding because in the processing steps that go into building a

device, the wafer may see many annealing steps of a similar temperature,

which can lead to device degradation. Rousseau postulated that the

deactivation process was due to the formation of a vacancy cluster via an








interstitial kick-out mechanism first described by Fair et al66,67 Fair et

aL66,67 described the kick-out mechanism as

AsnSi 4 AsnV + I (1.20)

where AsnSi represents n (integer 1-4) arsenic atoms around a silicon

lattice site, AsnV represents a deactivated cluster with a vacancy, and I

represents an interstitial. A vacancy complex was further confirmed by

Subrahmanyan et aL.68 They showed the importance of vacancies in the

deactivation process by injecting either interstitials or vacancies from the

surface and noting a retarded or enhanced deactivation rate, respectively.

Ab initio calculations confirm the As4V cluster to be energetically

favorable compared to isolated arsenic atoms in the lattice.69

Rousseau et al.50 went on to compare the deactivation that was

observed with the enhanced motion of a boron spike. If Equation 2.4 was

responsible for deactivation, then there should be a relationship between

the deactivation and an enhanced motion of a boron spike due to the

injection of an interstitial from the clustering reaction. It was shown that

the time transients of the enhanced diffusion of the boron spike

correlated with the deactivation process therefore supporting the

proposed reaction.








Alternative clusters to the As4V were suggested through modeling

by Berding et al.47 They reported that entropy considerations disfavor

the formation of such a large complex and proceeded to do a complete

free energy calculation to determine the role of As4V in the deactivation

process. Their findings showed that VAs3Sij and VAs2Si2 clusters could

be equally as effective at deactivation as was the neutral As4V. They

concluded that because As4V clusters are not needed to account for

deactivation, materials with similar arsenic concentrations and

deactivation fractions can have different microscopic states and

therefore behave differently in subsequent processing steps.

Dokumaci et al.52 made TEM observations that further support the

findings of Rousseau.49-51,63-65 Dokumaci et al.52,70,71 showed that a

reduced enhancement for larger concentrations of arsenic was due to the

formation of dislocation loops. It was stated that even though a greater

amount of interstitials were "kicked-out" at the higher arsenic

concentrations, the loops acted as barriers to interstitial motion and

therefore less enhancement was observed. It was also shown that there

was a strong dependence of the density of dislocation loops on the

amount of deactivation observed. Although the number of atoms bound

by the defects was insufficient to account for all the inactive arsenic at all

arsenic doses, the data still supports the idea that the loops are formed








due to the interstitials kicked-out during the deactivation process. Since

the loop layer could not contain all the interstitials released from the

clustering reaction, the remaining interstitials were available to enhance

the diffusion of the boron marker layer.


1.6.3 Below Electrical Solubility

Few experiments have been done at arsenic concentrations below

the electrical solubility limit because the effects of transient enhanced

diffusion and dopant activation are less problematic. However, to build

physical models for implantation damage, it is necessary to separate the

effects of high-concentration diffusion, extended defects, and point

defects. Therefore it is important to understand how the silicon

interstitials interact with dopants and how this affects TED and

activation. Park et al.72 did a low dose experiment in which no TED of

arsenic was detected. It was concluded that either the enhanced

diffusion was below detectable limits or that the motion of arsenic is due

to vacancies. In their study, the movement of the dopant profiles was

the only consideration and no correlation to extended defects was

mentioned.

In a separate experiment performed by Haynes et al.37, boron

doped wells of varying concentration with a single silicon implant were








used to examine how boron behaves in a supersaturation of silicon

interstitials. Haynes et a/.37 used the experiments as a novel way to

show that a BAB, pair can compete directly with {311} defects to retain

available interstitials. After an anneal, the number of interstitials trapped

in {311} defects was recorded as a function of well concentration

([13111J). For the same anneal time and temperature, it was shown that as

the boron concentration increases the number of interstitials trapped in

{311} defects decreased. A silicon control sample with no boron was

used to obtain the initial number of interstitials trapped in {311} defects

([131110) for the silicon implant condition used.

A simple model was used where the combination of a mobile B Si

pair plus a Bs leads to a BB, pair and releases the silicon interstitial (BsSi+

Bs ,- BB, + Si). According to this reaction, the missing interstitial dose,

[1311-A]1311B, is equal to [BRI] and proportional to [B]2. Using this model

a best-fit quadratic dependence on [B] was obtained. It was concluded

that one interstitial is stabilized by the formation of an interstitial-

substitutional BAB, pair.

This experiment provided a simple model for explaining the loss of

interstitials with an increase in background boron concentration.

Although a small cluster was used in the model, higher-order clusters are

not ruled out and an upper bound for the boron cluster size was








determined. The results of this experiment are in accordance with other

experiments that have shown that the BsBi cluster can exist. The BsBi

configuration has been observed by deep-level transient spectroscopy73

and ab initio calculations have indicated that the cluster is both bound

and immobile,74 as was assumed by the authors.



1.7 Thesis Statement

This work has contributed information in the following areas:

1. The study of {311} defect formation in the presence of an

arsenic background.

2. Quantitative TEM studies of the annealing kinetics of {311}

defects in an arsenic background.

3. Experimental investigation of the interaction between arsenic

atoms and silicon interstitials.

4. Experimental investigation into the dissolution of arsenic-

interstitial clusters.

5. Experimental investigation into the injection of vacancies from

arsenic doped region.

6. Modeling of {311} formation and dissolution in the presence of

an arsenic background.


















~Depth


Metallurgical
Spacing


Figure 1.1.











0.3 . .

0.25

0. ,2 ...

I 0.15 1 .....0 .

3 0.1 -. --


Typical CMOS transistor and terminology that describes the
device.75


00

80

eo

40




1995 2000 2005 2010 2015
Year of First Product 8hplmet Technology Generation


Figure 1.2. Semiconductor Industry Association (SIA) roadmap for
transistor gate length and source/drain extension junction
depth.1


0 . .
1995 2000 2005 2010 2015
Year of First Product ShIpmernet Technology Generation










Analyzing
magnet -


Vacuum pump
-/- Target chamber


....... J
Wafer load and
unload area


Source, magnet, power supply


Figure 1.3. Typical commercial ion-implantation machine.22


'- Atoms
Nuclear collisions


Figure 1.4. An incident ion can loses energy by nuclear collisions and
collisions with electrons.22















.1~


StaMSndatid
Oelaionge
otrfojlie-ed
IkwP


I
B


--.I


Inddnt
Ion Beam


-4-Tait
S rfse


\ PiR
\ Proloctad
^-lt


PPeak of Concentraion
Profile


Transvetm
Stra a


Figure 1.5. Critical ion implantation parameters


&ff?
I








Heavy
ion


Collision
cascade


Ion beam


I *


Amorphous
Amorphous


Figure 1.6.


Disorder produced from light
of an amorphous region.22


and heavy ions and the formation


Ile






>X,

Vacancy



Interstitial


Interstitialcy


Figure 1.7. Possible dopant diffusion mechanisms


Light
ion


/


AC













CHAPTER 2
EFFECT OF ARSENIC ON {311} FORMATION AND DISSOLUTION



2.1 Introduction

As devices continue to be scaled to smaller and smaller dimensions,

the dopant diffusion begins to control the depth of the electrical junction.

As discussed previously, Transient Enhanced Diffusion (TED) adversely

affects the diffusion of dopants and becomes an important parameter to

consider in the process design of future devices. TED from self-implants

has been a heavily studied area for many groups and a correlation has

been made between TED and extended defect formation and dissolution.

Eaglesham et al.39 has suggested that extended defects serve as storage

sites for excess interstitials and that during the dissolution of the defects

interstitials are released. Once released, the interstitials are free to

interact with any dopant atoms present. Common dopant atoms are

known to fully or partially diffuse via interstitials and therefore the release

of interstitials propels the enhanced diffusion of the dopant atoms.

The addition of excess interstitials to regions doped with either

boron or phosphorus has been shown to have a measurable influence on

the nucleation, growth and dissolution of extended defects. To
35





36

understand the influence the dopant has on the defect processes, studies

were brought about to examine the interstitial trapping by impurity

dopants. Haynes et al37 executed a study of boron interstitial trapping

using boron-doped wells. In their experiment they formed various

concentration, boron well structures and added excess interstitials by way

of silicon self-implants. Subsequent anneals were done to nucleate, grow

and eventually dissolve {311} defects. It was determined that for boron

concentrations above 1x1018 cm-3 the boron traps the interstitials and

causes a reduction in the {311} formation. It was also found that once

the defects formed the boron concentration did not affect the dissolution

process. Similar results were found for phosphorus by Keys et al.43

Unlike boron and phosphorus, which are known to be pure

interstitial diffusers, arsenic is known to diffuse by both interstitial and

vacancy mechanisms. Only being a partial interstitial diffuser, arsenic was

likely to have a lesser effect on the {311} defect processes. However,

even with being a partial diffuser, there should still be sufficient reduction

in the nucleated {311}'s. To gain knowledge on how arsenic effects the

{311} nucleation, growth and dissolution process, similar experiments to

those of Haynes et al.37 and Keys et al.43 were devised. To cover the a

range of concentrations below the clustering limit, doped arsenic wells

ranging in concentrations between 1x1017 and 1x1020 cm-3 were created








and excess interstitials were added using a silicon self-implant known to

cause {311} defects in undoped silicon. To study the effects of the

arsenic, the formation and dissolution of {311} defects that formed upon

low temperature annealing, was monitored as a function of arsenic

concentration. The ensuing sections will report the results from such an

experiment and discussions of the implications of the data are presented.



2.2 Experimental Overview


2.2.1 Arsenic Well Formation

Arsenic wells were fabricated using the following process.

Presented in Figure 2.1 is the schematic view of the process that follows.

Six p-type epi-silicon wafers were pre-amorphized with silicon such that

the damage from following arsenic implants will be equalized and

channeling reduced. A double implant was used for the pre-amorphiztion

step to ensure a deep amorphous region. The preamorphization implant

was done on a Varian El1000 with a beam current of around 4 mA. The

temperature of the wafer was held to about 90C. A deep pre-

amorphization was formed using silicon implanted at an energy of 200

keV and dose of 2x1015 cm-2. A subsequent shallow pre-amorphiztion

was done using a silicon implant at an energy of 70 keV and a dose of

1 x1015 cm-2. Following the pre-amorphization step, Arsenic was








implanted into the amorphized region. Implant energies of arsenic were

chosen such that the implant profiles would be fully contained within the

amorphous region before annealing. Also, a double implant was used at

each well condition such that after annealing a constant concentration of

arsenic would be established. Deep arsenic implants were incorporated at

an energy of 200 keV and at doses of 1.6x1012, 4.8x1012, 1.6x1013,

4.8x1013, 1.6x1014, 4.8x1014 cm-2. Following the deep implants, shallow

arsenic was implanted at an energy of 70 keV and at doses of 8.5x1011,

2.5x1012, 8.5x1012, 2.5x1013, 8.5x1013 and 2.5x1014 cm-2, respectively.

The lowest dose samples showed identical results to that of a silicon

control, thus the 3x1017cm-3 (4.8x1012 cm-2 + 2.5x1012 cm-2) sample was

used as the control from this point forward. All six wafers were then

annealed in a furnace under nitrogen ambient at 550C for 60 min to

regrow the amorphized layer. A subsequent 60 min furnace anneal at

1100C in nitrogen ambient was performed to form a constant arsenic

concentration to a depth in excess of 1600A. The total arsenic implant

doses used were 2.5x1012, 7.3x1012, 2.5x1013, 7.3x1013, 2.5x1014 and

7.3x1014 cm-2, which formed wells with concentrations of 2.0x1017,

4.0x1017, 1.1x1018, 3.5x1018, 1.lxlO'9 and 3.0x1019 cm-3, respectively.

Figure 2.2 shows the results from the Secondary Ion Mass

Spectroscopy (SIMS) analysis, following the 1100C well anneal. The SIMS








was done using a 5.5keV cesium beam and a sputter rate of 5.80.5

A/sec (200 nA). The raster area was set at 500 gim x 500 gim and the

crater depths were measured using a stylus profilometer. The atomic

concentrations of arsenic were calculated from relative sensitivity factors

determined from standard samples.

Following the well anneals the wafers were sectioned using a

diamond scribe and oxide etched to remove any oxide formation that

occurred during the annealing cycles. The oxide etch used was a 6:1 HF

buffered oxide etch. Following the oxide etch the samples were sent to

Kroko Ion Implantation Services for silicon ion implantation. There each

sample was independently implanted with silicon at an energy of 40 keV

and dose of 1x1014 cm-2. These conditions were used because they are

known to cause {311} defect formation after subsequent annealing.

After silicon implantation each sample was further cored into 3mm

diameter discs, to be used as Transmission Electron Microscopy (TEM)

specimens, using a Gatan ultra-sonic disc cutter. Each sample was then

furnace annealed under nitrogen ambient at various temperatures and

times as listed in Table 2.1.

After furnace annealing plan-view TEM specimens (PTEM) were

prepared by an HF etch process. This process consists of backside

grinding until the sample is approximately 100 im thick using aluminum








oxide powder and water. After grinding, the surface is protected using a

low melting temperature wax. After covering the surface with wax the

backside of the sample is exposed to a HF etchant (3:1 HNO3:HF) via a

drip etch system. The etchant is applied until a small hole and suitable

thin area is seen under white light. With practice and patience, the thin

area surrounding the hole will be electron-transparent. Once the etching

process is completed, the wax is removed by immersing the sample in

Heptane (C7H16) for 20 or more minutes. The samples were subsequently

air-dried and are then ready for viewing in the TEM.

After PTEM sample preparation the samples were analyzed using a

JEOL 200CX TEM. In order to increase the contrast of the defects in

relation to the background, all samples were imaged using a weak beam

dark field (WBDF) mode. In most cases the g220 reflection was used to

acquire the necessary images. A magnification of 50,OOOX was used for

all plan view analysis. Since only limited cross section samples were used

in this study, the procedure will not be discussed within. However,

presented in Figure 2.3 is a cross-section TEM image that shows the

{311} defects at a depth centered at -700A and the

amorphous/crystalline interface for the arsenic well at -3900A.








2.2.2 End of Range Loops from Preamorphization

End of range (EOR) dislocation loops from the preamorphization of

the silicon substrate remained after the 1100C 60 minute anneal at a

depth of -4000A. The size and distribution of the EOR defects was

constant for all the specimens. There was no difference in the {311}

formation or dissolution for the control sample with no preamorphization

and the control with the EOR loops from preamorphization. This implies

the EOR defects were sufficiently removed from the self implanted region

and thus had no effect on the {311} formation or dissolution process,

which is consistent with previous experiments.76 Figure 2.3 is a TEM

micrograph that shows a cross sectional view of the process area. The

{311} defects of study are at -700A below the surface and the EOR

loops are much deeper around 4000A.



Table 2.1. Experimental Matrix for arsenic well study______
Concentration 700C 750C 8000C
(cm-3)_____________________
lx1017 15, 30, 45, 90 min

3x1017 45, 275, 720 min 15, 30, 45, 90 min 5, 10, 15 min
lx1018 15, 30, 45, 90 min
3x1018 5, 275, 720 min 15, 30, 45, 90 min 5, 10, 15 min
1x1019 15, 30, 45, 90 min
3x1019 45, 275, 720 min 15, 30, 45, 90 min 5, 10, 15 min








2.3 Defect Analysis

2.3.1 Image Analysis

There are many methods to analyze the TEM images of the

different defect states. Three of the more useful methods are to

investigate the change in length, the distribution of defect sizes and the

total number of defects. The change in total length leads to the

knowledge of how many interstitials make up a series of defects. The

distribution of defect sizes relays important information about the

coarsening process of the defects. The total number of defects gives a

value of how many defects are in an anneal step.

For quantitative analysis of defect sizes and densities, the TEM

micrographs taken at 50,OOOX were enlarged by a factor of three to a

magnification of 150,OOOX. The defects were then traced on

transparency and scanned into a computer for image processing. Adobe

Photoshop and NIH image were used to scan and count the defects. This

data was further processed to determine the number of defects and the

approximate number of interstitials trapped by the defects. The process

is described in more detail in a previous publication by Bharatan et al.77

This data was then plotted versus various experimental parameters that

are discussed in the remaining sections in this chapter.








2.3.2 Effect of Arsenic on Trapped Interstitial Values

Plan-view TEM images are presented in Figures 2.4-2.13 of {311}

defects that remained in arsenic concentrations of 3x1017, 3x1018 and

lx1019 cm-3 after each annealing step listed in Table 2.1. In each of the

micrographs, the bright rod-shaped segments represent the {311}

defects. Each of the defects is made up of a number of interstitials.

Prior studies have shown that if the defects are on average 40A wide

then their structure leads to 26 silicon interstitials per nm of length under

the specified annealing conditions 39. The defects in this study average

about 40A and therefore the 26 interstitials per nm was used in all defect

calculations.

In all of the images in Figures 2.4-2.13 the length of each defect

was measured and by addition of each measurement the total length of

defects was determined. Knowing that there are 26 interstitials per nm

of length in the {311} defects, the length calculation can be converted to

the total number of trapped interstitials in each sample. Figures 2.14-

2.16 present the results at each temperature and annealing time, as plots

of the number of trapped interstitials versus arsenic concentration.

In these figures it is apparent that there is a significant effect on

the number of interstitials trapped in {311} defects at higher arsenic

concentrations. As the concentration of arsenic is raised the number of








interstitials in {311} defects is decreasing. As with the other dopants

studied in a similar manner, it appears that arsenic is pairing with

interstitials and acting as an alternative site. Eaglesham et aL.39 has

stated that the release of interstitials from the defects over time has a

direct correlation to TED. From this study it is now known that the

defect nucleation is affected by the presence of arsenic. The effect this

has on TED will be seen in how the defects that remain act upon

dissolution.


2.3.3 Effect of Arsenic on {3111 Defect Dissolution

The {311} dissolution process happens by the release of

interstitials from the defects. The net loss of interstitials from the

defects occurs as an exponential decay with time and can be expressed

as

Si (t) = Si 0(O)exp(-K311 *t) (2.1).

In equation 2.1 Si,(t) is the trapped interstitial concentration per area as a

function of time, Sio(O) is the concentration per area at time zero, K3,, is

the {311} dissolution rate constant and t is time. By carrying out several

measurements at various temperatures and arsenic concentrations a

family of decay curves may be generated. Determination of the

parameters in Equation 2.1 can be accomplished by plotting the {311}








trapped interstitial counts versus annealing time and then fitting each

series of data points with an exponential curve. The two most important

values are the slope, which represents the K311 value and the Y-intercept

that denotes the initial number of interstitials.

In Figures 2.14-2.16 it was shown that increasing the arsenic

concentration leads to a decreased number of interstitials in defects.

From the figures it is also seen that with increasing annealing time, the

number of interstitials in the defects is decreasing which shows the

defects are in a state of dissolution. To look at how the arsenic affects

the dissolution, the number of trapped interstitials was plotted as a

function of annealing time at each concentration. These plots are

presented for 700C, 750C and 800C in Figures 2.17, 2.18 and 2.19,

respectively.

The data points were fitted with exponential least-squares fits and

the dissolution rate constants were extracted. The dissolution time

constant for this experimental was calculated to be 505 min for each

concentration. The time constant obtained in this experiment is

consistent with the time constants of {311} studies previously

reported.30,36,39,43,44,78

Having a similar dissolution time constant at all temperatures

supports the idea that the {311} dissolution is not effected by the








presence of arsenic. It may also be recognized that at each arsenic

concentration the Y-intercept value is independent of annealing

temperature. However, when each concentration is compared there is a

large reduction in the Y-intercept value for increased arsenic

concentrations. This effect will be discussed in the following section.


2.3.4 Effect of Arsenic on the "Plus One Model"

In studies on {311} defect dissolution the initial number of

interstitials was independent of anneal conditions, suggesting it was a

function of the implant conditions. Previous studies on {311} defects

have shown that the Y-intercept or Sio(O) has related closely to the plus

one model. The plus one model assumes that after Frenkel pair

recombination there will be a dose of interstitials that remains that is

equal to the implanted dose. The plus one model appears to be

independent of dose or anneal conditions.

The implanted dose in this experiment was 1x1014 cm2 and in the

control samples the Sio(O) value hovers around 7X1013 cm-2, which is in

close agreement to the model. As the arsenic concentration was

increased, a reduction in the Sio (0) value resulted. The realization of this

data is that some sort of arsenic-interstitial complex is created during the

initial stages of annealing. With less interstitials available at nucleation

due to a dopant-interstitial complex formation, a decrease in the number








of trapped interstitials in {311} defects is eminent. The structure and

binding energy of the complex has not been disclosed at this time.

Modeling efforts presented in Chapter 4 help to decipher some of this

information. However, there are a couple schools of thought on the

strength of the bond between silicon interstitials and an arsenic atom.

First, it could be that the complex is more stable than the {311} defects.

In this case the complexes would hold interstitials until an energy

sufficient was present and the interstitials would be released. This would

most likely happen after the dissolution if the {311} defects and so the

release from the pairs would not affect the dissolution. It is also possible

that the defect complex is less stable and that the dissolution of the

{311} defects is dependent only on the defect itself. In other words, the

release of interstitials from {311} defects by how fast the interstitials

that make up the defect are able to diffuse away. In this view, the

dissolution of the {311} defect is based on how fast the interstitials can

leave the ends of the defects and the release of interstitials from the As-I

complexes would not affect the dissolution rate either. Another view may

be that the {311} defects have a supersaturation of interstitials

surrounding them. When interstitials diffuse away then the defects

release the interstitials to satisfy the supersaturation requirements. In

this instance the release of interstitials from As-I complexes would affect








the dissolution. Assuming that there are enough interstitials in the

complexes to cause an effect, the release of the interstitials from the

complexes would supply the {311} with the need interstitials to keep the

supersaturation satisfied. If there were more interstitials released from

the complex formations than from {311} defects then a delay in

dissolution would occur. This is not seen experimentally and therefore is

not believed to be true. A final scenario may be that with the addition of

arsenic the Fermi level is raised to a point where the interstitials diffuse

out of the area before they are captured by the {311} defect. This is not

believed to be true since this same experiment was done with a number

of different dopants and different results were acquired in each case.

The true answer can not be obtained from this experiment alone.

Additional experiments presented in the next few chapters are presented

in hopes of obtaining the answers.


2.3.5 Activation Energy Calculations

The activation energy for the dissolution of {311} defects can be

calculated using the dissolution kinetics experimental data. The

activation energy is related to the dissolution rate constant by


K311 = K31 1(0)exp-Ea (2.2).
( k(.)








In Equation 2.2, E, is the activation energy, K is Boltzman's constant

(8.616x10-5 eV/k) and T is temperature in Kelvin. Rearranging this

equation gives the linear relation of

E.
In( K311) =In( K311(0)) + E-- (2.3).
kT

So, by plotting K3,1 versus 1/T a straight line with a slope of E/k should

result. Multiplying the slope by k results in the activation energy. The

rate constants as a function of 1/T (K/) for each arsenic concentration

are plotted in Figure 2.20. The activation energy for each concentration

was calculated to be 3.40.2 eV. This activation energy is in the range of

published values for {311} defects that range from 3.3 to 4.2 eV.32,.78,79

Again, there is no effect of arsenic on the dissolution kinetics of {311}

defects. This seems to indicate that the dissolution of the defects is

limited by either the release rate of interstitials from the ends of the

defects or by the diffusion of interstitials away from the damage region.

The addition of arsenic to the system has no effect on either of these

mechanisms.


2.3.6 Effect of Arsenic on Defect Size and Density

Presented in Figures 2.21-2.23 is the effect arsenic has on the size

distribution of {311} defects. For higher arsenic concentrations, not only

do less defect form but they tend towards smaller sizes also. The smaller






50

sizes can be seen in a shift in the histograms to only include defects in

the lower size ranges. Also, total defect density data as a function of

annealing time at different arsenic concentrations is presented in Figures

2.24-2.26 for each annealing temperature. Apparent in these figures is

the lack of change in overall defect density at each concentration.

However, when the total number of interstitials trapped in the defects

was calculated in the previous sections, the number of interstitials in the

defects was decreasing with increasing arsenic concentration. The only

way for these two to coincide would be to have smaller defects.

This may explain why the defects at higher concentrations are less

stable. In most growth phenomena it is known that smaller defects grow

at the expense of larger defects. A critical size is determined dependent

on the energy of the system. If a defect is larger than the required size it

will grow at the expense of the defects that are not of critical size. Once

the smaller defects are dissolved the larger ones may then dissolve. This

can be seen in the density plots at most of the concentrations. At small

anneal times there is a number of small defects, as time evolves less

small defects are realized and some larger defects are present. At some

point the larger defects dissolve and at long times there are small and

few defects.








2.3.7 Possible Arsenic Complexes

The data presented suggests that the introduction of arsenic is

having a significant effect on the amount of excess interstitials available

for {311} defects. There are several possible mechanisms that might

account for this effect. Two of the more likely reasons include formation

of arsenic clusters or enhanced diffusivity of charged interstitials when

the material becomes extrinsic (n>n,). First, as was suggested for boron,

the arsenic could be trapping the excess interstitials in some form of a

complex. A simple chemical equation that would relay the transfer of an

arsenic atom and silicon interstitial into a dopant-defect complex may be

developed as

ASx + Si, <* ASxI (2.4)

In order to determine what the value of x may be the concept of missing

interstitials is introduced. It is known that if there is no arsenic present

the total number of interstitials available at time zero is on the order of

7x1013 cm-2. As arsenic is added this number decreases. If it is assumed

that all of the arsenic within the silicon implant damage region is

interacting with the silicon interstitials then a missing dose of interstitials

can be calculated. In this experiment the maximum depth of the {311}

bottom of the defect layer was measured to be -800A by cross-sectional

TEM. If this value is multiplied by the concentration of arsenic, the dose








of arsenic atoms affected by the excess interstitial dose is estimated.

These values range from 2.4x1012 As/cm2 for the 3x1017 cm-3 well to

2.4x1014 As/cm2 for the 3x1019 cm-3 well. As previously stated, the dose

of the interstitials injected by the silicon implant was determined to be

-7x1013 cm-2. The missing interstitial dose is then the difference

between the y-intercept of the undoped well (-7x1013 cm-2) and the y-

intercept of the doped samples in Figures 2.17-2.19. Since the percent

change is the same at each temperature, only one temperature needs to

be evaluated. A plot of the missing interstitial dose versus arsenic dose

is presented in Figure 2.25. A few things are noticed when this graph is

reviewed. First there is not a linear relationship throughout the

concentration ranges. This is due to the fact that as the arsenic

concentration is raised there becomes a point at which there are more

arsenic atoms available than interstitials in the system. If the graph is

broken into two sections then a number can be reached. The slope

between the first two points turns out to be 0.5, which means that there

is a 2:1 arsenic to interstitial ratio before a saturation of interstitial

trapping is reached. Thus if arsenic trapping were occurring the cluster

might be an As21l. This is a crude estimate and other complexes with

higher silicon to arsenic ratios might also be possible to explain the role








off in the curve. If the whole curve were assumed to be linear then an

As1ol would be the complex of choice.



2.4 Arsenic Effect on (311} Formation and Dissolution Summary

The effect of arsenic on {311} formation and dissolution was

studied, by the formation of arsenic wells. Silicon implants known to

cause {311} defects were introduced into the wells. After various

anneals the formation and dissolution behavior of the defects was

determined. It was determined that arsenic had a strong effect of the

nucleation of the defects. With increasing arsenic concentration a

decrease in the number of interstitials at zero time was captured.

However, once formed the defects dissolved at the same rate

independent of the arsenic concentration. The activation energy for

dissolution also was not affected by the arsenic concentration. It was

determined that in the initial stages of annealing As-I pairs were formed

and provided alternate sites to {311}'s for the interstitials. These pairs

break up in time but by the time they do the {311}'s are in a dissolution

mode and since the {311} dissolution is mediated by the release rate at

the ends of the defects, the breakup does not effect dissolution.




























Figure 2.1. Process steps used in the design of an experiment to examine
the effects of arsenic on extended defect nucleation and
dissolution.






55






1x1020
"N B 7.3 x lO"/cm2

E1 \ 7.3 x 10'3/cm2
0%000l2. 100 20"00/0cm
---,k--- 7.3 x 101C2/ m
S)1012/cm2

w lx10"a
C 1X1017
0
IX1016
C
cc lxlO'6




0 1000 2000 3000 4000 5000 6000 7000 8000

Depth (A)


Figure 2.2. Secondary Ion Mass Spectroscopy (SIMS) plots of the arsenic
wells after an 1100C 60 min anneal.






56





EOR A/C {311} '
Defects Interface









700










Figure 2.3. Cross-sectional TEM image of the processed region. A band
of {311} defects is located -700A below the surface and
large EOR defects are centered about 4000A.



















3x10 17 cm-3 As+


3xl 018 cm-3 As+


3x1019 cm-3 As+

Figure 2.4. Effect of arsenic concentration on {311} defects formation:
700C 45min




















3x10 17cm-3 As+


3xl 018 cm-3 As+


3xl 019 cm-3 As+

Figure 2.5. Effect of arsenic concentration on {311} defects formation:
700C 275min




















3x1017 cm-3 As+


3x 1018 cm-3As+


3x01 cm-3 As*

Figure 2.6. Effect of arsenic concentration on {311} defects formation:
700C 720min



















3x1 017 cm-3As+


3xl 018 cm-3 As+


3xl 019 cm-3 As*


Figure 2.7. Effect of arsenic concentration on {311} defects formation:
750C 15min




















3xl 017 cm-3As+


3xl 018 cm-3 As*


3x1019 cm-3 As*


Figure 2.8. Effect of arsenic concentration on {311} defects formation:
750C 33min




















3xl 017 cm-3As+


3xl 018 cm-3 As*


3x1019 cm-3 As+

Figure 2.9. Effect of arsenic concentration on {311} defects formation:
750C 45min




















3xl 017 cm-3As+


3xl018 cm-3As+


3x1019 cm-3 As+


Figure 2.10. Effect of arsenic concentration on {311} defects formation:
750C 90min




















3x1 017 cm3 As*


3x1 0Q cm-3As+


3x1 019cm-3As+

Figure 2.11. Effect of arsenic concentration on {311} defects formation:
800C 5min




















3xl 017 cm-3 As*


3xl 018 cm-3 As+


3x1 019 cm-3As+

Figure 2.12. Effect of arsenic concentration on {311} defects formation:
800C 10min




















3xl 017 cm-3 As+


3xl 018cm-3 As+


3xl 019 cm-3 As*


Figure 2.13. Effect of arsenic concentration on {311} defects formation:
800C 15min






67












1014 11





E
U


; ~13
10
C

CL


T1 700C 45min-
a.
I-,- -'N--TI 700C 45min
-*--TI 700C 275min
-6 TI 700C 720min

10 12 a I a i a lI-I I I 1 111 1 1 a a "
1017 1018 10" 1020

Concentration (cm'3)


Figure 2.14. Number of trapped interstitials in {311} defects as a
function of arsenic concentration after a 700C anneal for
various times.





































1 018 1 019

Concentration (cm3 )


Figure 2.15.


Number of trapped interstitials in {311} defects as a
function of arsenic concentration after a 750C anneal for
various times.


1 014


1 0 13


1012 1
1 017


1 020





































1018 101" 1020

Concentration (cm3 )


Figure 2.16.


Number of trapped interstitials in {311} defects as a
function of arsenic concentration after a 800C anneal for
various times.


1 014


1 0 13


1012 _
1017














1 014


1013s


1 012


S1x1 04 2x1 04 3x1 04 4x1 04 5x104


Time (sec)


Figure 2.17.


Effect of arsenic concentration on the {311} defect
dissolution after furnace anneals at 700C. A similar slope
represents similar dissolution constants showing the arsenic
has no effect on the dissolution.


CMa
E
0




Co


C-


0.
cc
I-















1 014


00%I
CM
E
Q
%WO
mC,
(0,



4)
4W,
I-
C

4)
0.
0.
,L
cc


1 013


1 012





Figure 2.18.


0 1000 2000 3000 4000 5000 6000

Time (sec)


Effect of arsenic concentration on the {311} defect
dissolution after furnace anneals at 750C. A similar slope
represents similar dissolution constants showing the arsenic
has no effect on the dissolution.


!














1 014


1 013


1012 0 ,,Imal ,m I,,,uI, ,I ,,,,a I,,,,I,,,,I,,,u
200 300 400 500 600 700 800 900 1000


Time (sec)


Figure 2.19.


Effect of arsenic concentration on the {311} defect
dissolution after furnace anneals at 800C. A similar slope
represents similar dissolution constants showing the arsenic
has no effect on the dissolution.


N

E
0


cu
CO




a)
V



CO
t,
a.
a.

0














0.01 -






0.001

4)
0
(a)


S0.0001 -






10".5 -
9.2


9.4 9.6 9.8 10

1 04/T (K1 )


Figure 2.20.


Plot of dissolution rate constants at various temperatures
for activation energy calculations. The slope at each
concentration is the same relaying that arsenic has no
effect on the activation energy to {311} dissolution. The
activation energy was calculated to be 3.4 0.2 eV, which is
in the same range as previous studies on pure {311}
dissolution.


10.2 10.4




















E
10"
A

#
I 10


3S17_700_46
.... .... .... .... ... . . . I .5


C e ......c. .0.... ''.'i@0e"

Range (A)



301870046
1 0"






10I.
101,









10'I
o .oooo .ooooooooo.. o.o m

Rang. (A)



3S19_700_45

1 0''









10*I


10'*
o :moBewone:ee:iiiBM0"0100
-l"B**tWww .5n *.h B S*0 S


3.17_700_275
10'2 U I I lIII


o2


000000
*~~-t4


R-ng. (A)


318_700_275
















..ooaooooo.oa.o@000000.0w

Range (A)



3019_700_275
I.......I....................................I


1 0''


10"



10n



1 0 l
0s0


Hang* (A) Rang* (A)

Figure 2.21. Size distribution data at various times at 700C


I,

































mangg (Aj


1o0'

I 0"















1o'
10"




10"













lol
10 I'










1 0" LI


Range (A)


3el9_700_720
I'I'I II~l 'I~i III~~lel Mll l I I l I I


CooeeCoe-aeoeaoooCooo00e

Rang. (A)

Figure 2.21. (Continued)


3018_700_720
.. .. .. ....lI ii i .i i 1 1. I 1. ,1iil..ii 1, ,


10"









10'
I 0v



I Ole


II..........


3o17_700_720
10 Wi llIIl s i 1 1 1 1


m







76




3e17_7S0_1 5 31 7_78033
1 0'' i~,ri ~ ~~~~~~, ,11111111111 I 0''


10" 10' l
I I

I 0l 10"
S 10 1 'l










31l1i7SO_18 318_70_33





i i if

loll 10
I I ofi



10' 10'

Range (A) Range (A)




3F1- 7S0_1. 31 7850_33
10'" r1 10"r1





10"i 01 i 0i
I I












10' 1'*
o00o0oow 0ooooowo @@ @0 10*om









R*ng- (k) Range (A)

Figure 2.22. Size distribution data at various times at 750C.












3017_750_90
10e ,,, .,. .. ,,.. ..... ..,.


*II | I" '



*-


3015_750_45

S 10,






lit
| ,!
10




... 000.........@0
~oS5SS S S|Range (A)
Reng*lA)


3119_750_45
............... ...


Range (A)


301_750_90






10" -





100 000
I Op.hhhhhhhh~


^oco eo ~Ro nec5,5 g (A)*
ROnae (A)


3019_750_90
i 0II


10" F


10"0 9





Rd (
10R (A)
000000000c~00000M,.c0--0
Rang. (A)


Figure 2.22.


(Continued)


3.17_750_45


Rang. (A)


R Ln (A)
Rang (A)


















3*17_800_5




















I^^^^^^^^ i 111111 *1' *1 1.1@000 0

Range (A)


3*17800_10


10- .








10'"








101











10"




1 0'L
I 0









10'


















1 0"r




1 0''-






10'



10'
0Is


1 0"






1 0d'






0,d


30o1860010






















Range (A)


3s1_o80010
~I ITT


















iU" i'll 'l'Rian .g .|,|.(A l) .i.i.|.|.
.00 000000000000000000000
00 00000 oSoi0000onotfo)'oOO
Rang. (A)


Figure 2.23. Size distribution data at various times at 800C.


3e18_00_8
I II I I i I i Ii i .~~l i i ~. i. i i i-.. .


"'rTT'l


Range (A)


Range (A)


3U19_00_5


Range (A)








79









317_800_15



S10













Range (A)
5 10
I
C




















Range (A)
3*18_800_15
I







I'1 I"" I I l l l l l l l I I 1 1 I 1' 1" i"
1 0'






10'













Range (A)



9u13_0 0_1e
10 ,
10I













j 10''
10'












Rang. (A)



Figure 2.23 (Continued)
















1012 ~, I * ~ I U I I I I I


I I I I I I I I I I


N
E
CM
0


1 011






1 010


1 09


I I I I I I I I I I I I I I I I I I I I I I


0 1x104 2x104


3x104 4x104 5x104


Time (sec)


Figure 2.24. Total defect densities as a function of annealing times at 700C
for each arsenic concentration studied.


0 3x1017 cm-3 As
E 3x101" cm"* As
* 3x1019 cm"3 As


1012


I I I I I I I I I a














1 012


N


CM
0
*









I-
0)
I-


1 011





1 0'0


10'


9


.31.1 I 11111.11111.31 I liii,, II


0 1000 2000 3000 4000 5000


6000


Time (sec)

Figure 2.25. Total defect densities as a function of annealing times at 750C
for each arsenic concentration studied.


. . .U I . .S I S
o 3xI017 cm'3 As
o 3x10"' cm"3 As
3x10"9 cm"3 As


" I . I . I














1 012


1 011






1 010


* ~1U
o 3x10 cm-' As
'3 3x101 cm-3 As
3x1019 cm"3 As


1 I I IIIIIII' ',l,,,,1i,, ,l ,, l 1iu t m mi m I i
200 300 400 500 600 700 800 900 1000


Time (sec)


Figure 2.26. Total defect densities as a function of annealing times at 800C
for each arsenic concentration studied.


CM
E


C






0
0)


0
Q)






83




1014 1- gi I I 111111 I I I I 1111 1 I I




E
0


400 13
0 10

5 -


Si
0) /




1012 s I 1 -- a a I I I I I ,,
1 012 1013 1 014 1015

Arsenic Dose (cm"2) at 800A




Figure 2.27. Missing interstitial dose as a function of the dose of arsenic
available at 800A. The slope of the initial portion of the curves
is 0.5, which relates to a possible As21l complex.













CHAPTER 3
RELEASE OF SILICON INTERSTITIALS AND VACANCIES FROM DOPED
ARSENIC LAYERS

3.1 Boron Marker Layers

3.1.1 Overview

Gate length, gate dielectric thickness and junction depths are the

primary parameters that control transistor performance. According to

the Semiconductor Industry Association (SIA) roadmap junction depths

and gate lengths will continue to decrease at a rapid rate in the years to

come. With decreasing junction depths, the doping of the source/drain

structures needs to increase beyond current solid solubility limits. In

order to continue this decreasing trend, it becomes increasingly important

to understand and control the solid solubility and the diffusion of dopants

in silicon. The ability to accomplish this will depend on our understanding

of how dopants and silicon interstitials interact. This chapter will

approach the more fundamental issue of how arsenic and silicon

interstitials interact in reference to the release of silicon interstitials.

CVD grown boron marker layers were used to monitor the release

of silicon interstitials from an arsenic doped surface region that was

subsequently implanted with silicon. These structures were annealed for

84








various times at 750C in an ambient of nitrogen. A comparison of boron

spike enhancement and defect dissolution is made. It is shown that

enhancement values from the silicon implant were reduced at short times

for samples containing arsenic compared to samples implanted with

silicon alone or arsenic alone. The TEM results showed that defect

densities were dramatically reduced for the samples containing arsenic.

These results imply that the previously reported reduction in {311}

formation observed in As doped wells is most likely not a Fermi level

effect and is consistent with the formation of As interstitial complexes.

The data shows that As-I complexes form, and control extended defect

formation, which slows the enhanced diffusion.


3.1.2 Boron Marker Layer Setup

Boron marker layers can be used to study the release of interstitials

from an implanted region. The release of interstitials from the implanted

region will subsequently enhance the motion of a buried boron marker

layer. The motion of the marker can be used to determine an

enhancement value over that of inert diffusion and the enhancement can

be monitored over a range of times and temperatures. Figure 3.1 is an

example of a boron marker layer sample after annealing showing the

difference between inert and enhanced diffusion. TEM samples can also

be made from these samples to compare the release of the motion of the