Citation
Defect and diffusion study in boron implanted silicon

Material Information

Title:
Defect and diffusion study in boron implanted silicon
Creator:
Liu, Jinning, 1968-
Publication Date:
Language:
English
Physical Description:
viii, 171 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Activation energy ( jstor )
Annealing ( jstor )
Atoms ( jstor )
Boron ( jstor )
Density ( jstor )
Diffusion coefficient ( jstor )
Dosage ( jstor )
Ions ( jstor )
Point defects ( jstor )
Silicon ( jstor )
Dissertations, Academic -- Materials Science and Engineering -- UF ( lcsh )
Materials Science and Engineering thesis, Ph. D ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1996.
Bibliography:
Includes bibliographical references (leaves 163-170).
Additional Physical Form:
Also available online.
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Jinning Liu.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Jinning Liu. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
36714335 ( OCLC )
026354498 ( ALEPH )

Downloads

This item has the following downloads:


Full Text










DEFECT AND DIFFUSION STUDY IN BORON IMPLANTED SILICON














By


JINNING LIU















A THESIS 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 1996






























dedicated to my parents













ACKNOWLEDGMENTS



This dissertation would not have been possible without the guidance and support of my advisor, Dr. Kevin Jones, who not only provides us illuminating advice but also creates a stress-free environment for us to work in and do our best. I would like to thank Drs. Steve Pearton, Cammy Albernathy, Rajiv Singh and Mark Law for serving on my committee, and Dr. Robert Park for sitting in my defense. Special gratitude is due to Dr. Jian Chen, who patiently helped me get into the door when I first joined the group four years ago and gave me many valuable suggestions in the years following. Special gratitude is also due to Dr. Wish Krishnamoorthy for many stimulating discussions and a thorough correction on my dissertation draft. I would like to thank all my colleagues, especially members of our silicon group, Brad Herner and Sushil Bharatan, for many helpful discussions, and my officemate as well as good friend, Ranju Datta, for keeping me company and making my time at school enjoyable.

It would have been impossible to finish this work

without the assistance of many other people. Dr. Lenny Rubin at Eaton Corporation and Craig Jasper at Motorola helped me with the implantations. Dr. Jinghong Shi and Joe Bennett from SEMATECH helped me with the SIMS analysis. Dr. Hans



iii








Gossmann kindly provided me with the boron doping superlattices as well as his PROPHET simulation results. Dr. Heemyong Park, Dr. Samir Chaudhry and Omer Dokumaci showed me how to run the simulation program FLOOPS. I thank all of them.

I am especially grateful to Joe and Alisa Zhu, John and Sherry Ding for their long-lasting friendship, which has driven away much of my lonely time and made my stay in Gainesville full of joy.

I am unable to express my gratitude to my parents Zezhan and Yafan and my sister Jinqi for the sacrifice they have made for me to make it so far.

Finally, I would like to thank my husband Jun Zhou for his love and understanding.




























iv















TABLE OF CONTENTS


ACKNOWLEDGMENTS ........................................... iii

ABSTRACT ..................................................vii

CHAPTERS

1 INTRODUCTION .............................................1

1.1 Motivation and Objective ..............................1
1.2 Background ............................................3
1.2.1 Shallow Junctions ................................3
1.2.2 Implant Damage and Types of Extended Defects ..... 7 1.2.3 {311} Rod-like Defects ..........................10
1.2.4 Dopant Diffusion in Silicon ....................18
1.3 Implantation, Characterization and Simulation
Techniques ............................................25
1.3.1 Ion Implantation ................................25
1.3.2 Transmission Electron Microscopy ................27
1.3.3 Secondary Ion Spectroscopy ......................29
1.3.4 Florida Object Oriented Process Simulator ....... 30
1.4 Thesis Statement .....................................31

2 THE FORMATION THRESHOLD OF {311} DEFECTS AND
SUB-AMORPHIZATION DISLOCATION LOOPS. ......................33

2.1 Overview .............................................33
2.2 Experimental Procedures ..............................34
2.3 Defect Microstructure.................................35
2.4 Criteria for the Formation of {311} Defects and
Sub-amorphization Loops ...............................37

3 EFFECT OF ANNEALING ON DEFECT EVOLUTION AND TRANSIENT
ENHANCED DIFFUSION .......................................48

3.1 Overview........................................... 48
3.2 Effect of Annealing on Defect Behavior ...............49
3.2.1 Experimental Procedures .........................49
3.2.2 Defect Evolution: Microstructure,
Interstitial Density and the Activation Energy
of Interstitial Decay ..............................50
3.2.3 Defect Evolution: Density, Size and Size
Distribution .......................................53
3.3 Effect of Annealing on Transient Enhanced
Diffusion .............................................58
3.3.1 Experimental Procedures .........................58


v








3.3.2 Dopant Diffusion during Annealing ...............59
3.3.2.1 Static Peak at High Concentration .......... 59
3.3.2.2 Broadened Tail at Low Concentration ........ 63
3.3.3 The Activation Energy for the Diffusion
Saturation Process .................................65

4 THE INFLUENCE OF IMPLANT ENERGY ON DEFECT BEHAVIOR
AND TRANSIENT ENHANCED DIFFUSION. .........................86

4.1 Overview.............................................86
4.2 Implant Energy Effect on Defect Behavior .............87
4.2.1 Experimental Procedures .........................87
4.2.2 Defect Microstructure and Dissolution
Process ............................................87
4.2.3 Defect Density, Size and Size Distribution
during Annealing ...................................90
4.3 Implant Energy Effect on Transient Enhanced
Diffusion ................................. ......... ..93
4.3.1 Experimental Procedures .........................93
4.3.2 Dopant Diffusion Behavior .......................94
4.3.3 Correlation Between Defects and TED .............96

5 THE INFLUENCE OF IMPLANT DOSE ON DEFECT BEHAVIOR AND
TRANSIENT ENHANCED DIFFUSION ............................120

5.1 Overview........................................... 120
5.2 Implant Dose Effect on Defect Behavior ..............122
5.2.1 Experimental Procedures ........................122
5.2.2 Defect Microstructure and Dissolution
Process ...........................................122
5.2.3 Defect Density, Size and Size Distribution
during Annealing ..................................126
5.3 Implant Dose Effect on Transient Enhanced
Diffusion ............................................129
5.3.1 Experimental Procedures ........................129
5.3.2 Dopant Diffusion Behavior ......................130
5.3.3 TED Saturation at Higher Dose ..................132

6 SUMMARY AND FUTURE WORK ................................151
6.1 Summary .............................................151
6.2 Future Work .........................................158

LIST OF REFERENCES ........................................163

BIOGRAPHICAL SKETCH .......................................171











vi












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


DEFECT AND DIFFUSION STUDY IN BORON IMPLANTED SILICON By

Jinning Liu

December, 1996




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


The aim of this work is to better understand the nature of transient enhanced diffusion (TED) of boron in silicon.

A matrix of sub-amorphization boron implants was used in this work with energies ranged from 5 to 40 keV and the doses ranged from 5x1013 to 1x1015 cm-2. The formation threshold of {311} rod-like defects and sub-amorphization dislocation loops are higher for lower implant energies. For the same energy, the threshold dose for {311) defect formation is lower than that for loop formation.

The study of annealing condition effect on {311} defect behavior and boron TED shows that {311) defects dissolve more rapidly at higher temperatures. The concentration of interstitials trapped in {311) defects was below the value predicted from previous studies of Si+ implants, indicating



vii







the presence of submicroscopic boron interstitial complexes. The interstitial emission from {311) defects during anneals shows an exponential time dependence with an activation energy of 3.8 eV. The time constant for TED saturation process is smaller at higher temperatures and the dependence shows an activation energy of 1.6 eV. Both (311} defects and boron interstitial pairs are believed to be the source of interstitials driving TED.

The implant energy effect on (311} defects and TED was

also studied. The dissolution of {311} defects is slower for higher energy implants. This may be due to either an increase in the surface recombination or an increase in the ratio of boron interstitial clusters to {311} defects with decreasing energy. Lower diffusion enhancement and shorter duration of TED were observed for the 5 keV implant than for all the higher energies at a dose of 2x1014 cm-2. The (311} defects might anneal slower than mobile boron interstitial pairs so that the increase in {311} defect concentration may account for the increase in TED duration. Surface recombination is believed to play an important role for the 5 keV implant so that the amount of TED is reduced.

The study of dose effect on defect and TED shows that (311} defects dissolve slower at higher doses. Diffusion inceases with increasing dose initially. Boron interstitial clusters and dislocation loops formed at high doses suppress any further increase in TED with increasing dose above 2x1014 cm-2 at 20 keV by acting as interstitial sink.



viii













CHAPTER 1
INTRODUCTION


1.1 Motivation and Objective


Since the development of the first integrated circuits, the minimum feature sizes of devices have been reduced dramatically. This miniaturization is expected to continue and by the year 2000, the number of transistors per chip is expected to reach one billion. The scaling of devices to smaller dimensions both vertically and horizontally implies lower voltages, thinner dielectrics and more abrupt and shallower dopant profiles.1-3 The new devices will possess the properties of higher packing density, higher speed and lower power consumption.4,5 However, before the realization of this goal, many challenges remain in processing and packaging. One of them, for example, is to scale the source/drain junction depth and the lateral penetration under the gate in order to maintain a threshold voltage for the MOSFET independent of the channel length.

Among the various methods available for forming shallow junctions,6-11 ion implantation is currently the most widely used one, due to its precise controllability, uniformity and reproducibility. However, future devices require shallow p+/n junctions with depths less than 1000A to minimize short




1





2


channel effects and to maintain low contact resistance.12 These shallow junctions will be difficult to form by ion implantation. One of the reasons is that the implantation process induces a supersaturation of silicon interstitials, which will eventually lead to anomalous dopant diffusion, known as transient enhanced diffusion (TED).13-20 TED is caused by the pairing of excess silicon interstitials with dopant atoms which reduces the energy barrier for the migration of dopant atoms through the bulk. As a result, the dopant penetration depth from diffusion is greatly increased. TED lasts for only a limited time at typical annealing temperatures so it is transient in nature.

TED will affect the source/drain dopant profiles both

vertically and laterally. The vertical diffusion results in deeper profiles, which will weaken the control of the gate voltage. The lateral diffusion reduces the effective channel length which may result in punch-through phenomenon. Besides these two effects, the enhanced diffusion of channel dopant near the gate edge from excess interstitials created during the lightly doped drain (LDD) or source/drain implant causes an increase in the threshold voltage for shorter gate length devices.21 This is the so-called reverse short channel effect, i.e., an anomalous increase in the threshold voltage with a decrease in device dimension. As device dimension shrinks further, these effects will become more pronounced.

Thus the objective of this work is to study the effect of implantation induced defects on transient enhanced




3


diffusion, and to provide an experimental basis for the improvements of process development, process simulators and ion implanters. These improvements would be of paramount importance to optimizing the device structure.


1.2 Backaround



1.2.1 Shallow Junctions


A cross section of a complementary metal oxide

semiconductor (CMOS) structure is shown in Fig. 1.1. The performance of this device is controlled by the gate electrode, the gate dielectric, the channel and well region and the source/drain junctions.


Gate

LDD LDD

Source Drain

poly-Si





--._ Channel
Interstitials



Substrate



Figure 1.1. Schematic of MOS transistor showing how interstitials are released from source/drain and LDD implants and affect channel voltage.




4


As mentioned in the previous section, the common trend in the development of solid state devices, either memory or microprocessor, is the decrease in device dimensions. MOS devices with sub-micron dimensions require junction depths of

0.1-~0.2pm. The most commonly used technique in forming shallow junctions is to implant dopant species at low energies into the silicon substrate.

To form shallow p+/n junctions, direct ion implantation typically requires a very low energy boron implant preceded by silicon or germanium amorphization to reduce channeling, or a low energy BF2+ implant which uses the F to amorphize the surface.22-24 The use of pre-amorphization by Si+ or Ge+ introduces another implant step. In addition, the amorphization of the surface will introduce end-of-range damage, which may not be annealed out even at high temperatures. This end-of-range damage is a source of interstitials and, if located across the junctions, will result in unacceptably high leakage currents. For the use of BF2+ implants, the fluorine ions can affect the permeability of the gate oxide to boron diffusion. This can cause a penetration problem of boron atoms from the gate region through the gate oxide to the channel region, therefore undesirably affecting the threshold voltage of the device. Thus one would like to avoid BF2+ implantation. One alternative is to simply use low energy B+ implants. However, there are several problems associated with B+ implants alone.




5


In addition to the channeling effect of B+ implantation without pre-amorphization which can significantly increase the junction depth, the post-implantation annealing results in deeper profiles greater than that allowed by intrinsic diffusion due to TED. TED is associated with the interaction of mobile silicon interstitials with boron atoms.18,25'26 Implants into a pre-amorphized substrate can eliminate channeling tails, but TED is still present. For B+ implant the amount of TED is determined by both the number of excess interstitials in the damage region and the partitioning of these excess interstitials into different forms such as mobile boron interstitial pairs, boron interstitial clusters, {311} defects and dislocation loops and the role of the surface.27-29 This implantation induced TED will be discussed in greater detail in section 1.2.4.

Shallow junctions are also formed in combination with

silicidation of source and drain. In general the implant and anneal are done before metal deposition an silicide formation. However, the implants could also be performed through a deposited metal layer (Co or Ti mainly) before silicide formation,30-34 or into an already formed ilicide layer that then acts as a diffusion source (SADS) upon subsequent annealing.35-37 In the first two cases, a silicide layer is formed after annealing with a steep implant profile and a shallow junction. During the silicide formation, a portion of the heavily doped silicon region is consumed and the silicide interface moves into silicon, while dopants





6


segregate into the silicide.30 This causes serious contact resistance problems.3134 Moreover, silicidation perturbs the point defect concentration in the substrate by injecting vacancies and depleting interstitials,38039 which should also be taken into consideration during the process design. In the case of SADS process, the dopant species is implanted entirely into the silicide and then driven into silicon at low temperature. This avoids both dopant channeling and damaging of the silicon. SADS is a very attractive method to form 300-400A junctions. However, extension implants are still necessary to define the channel length etc. The disadvantages of SADS include uneven doping due to rough interface and dopant precipitates in the silicide.

Thermal diffusion is also used for junction formation. While conventional diffusion does not create damage, it suffers from poor controllability and requires a high thermal budget. An alternative diffusion method is gas-immersion laser doping (GILD).40 An excimer laser is used to rapidly heat and melt the silicon surface. BF3 is then adsorbed, pyrolized and diffused into the melt. The junction depth is determined by the melt. A 100A junction depth has been achieved with this method, which would be attractive if technical, uniformity and cost issues can be resolved.

In summary, there are various ways to form a shallow junction, and each has its own advantages and drawbacks. This study is focused on the use of implantation of B+ into





7


crystalline silicon to investigate the effect of implant damage on dopant redistribution.


1.2.2 Implant Damaae and Types of Extended Defects


Ion implantation is a well-known technique for

controllably introducing dopants into the substrate during IC fabrication. The major disadvantage of ion implantation is the damage introduced by the energetic ions and the need for post-implantation annealing to activate the dopant. Different types of extended defects form after annealing, depending on both the implant and the annealing conditions. It is important to understand the defect formation kinetics in order to study the effect of defects on dopant diffusion.

As an incident ion penetrates into the substrate, it loses kinetic energy through elastic collisions with the nuclei and inelastic collisions with electrons of the target material until it stops. Elastic collisions are primarily responsible for the lattice displacements through the production of Frenkel pairs. If the kinetic energy transferred to the host atom is higher than the displacement threshold energy (-15eV for Si)41, the host atom leaves its lattice site (primary collision) and collides with other atoms (secondary collision), which in turn gives rise to more collisions. This sequence results in a collision cascade. For a sufficiently high concentration of damage, it is possible to amorphize the crystalline silicon near the surface. If the damage density is not high enough, some




8


isolated amorphous pockets may form without overlapping into a fully amorphized layer. Different damage levels will result in different types of extended defects after postimplantation annealing.

Residual implantation damage can be visible in the TEM or submicroscopic such as dopant clusters. Extended defects visible in a TEM are categorized according to two formation mechanisms: point defect condensation and solid phase regrowth.42"43 There are three types of extended defects that arise from the first mechanism and two types from the second mechanism, as will be discussed below.

I. Extended defects from point defect condensation

1. Sub-amorphization defects

When the implant damage is not sufficient to turn the silicon surface into a continuous amorphous layer, subamorphization defects form during subsequent annealing. These defects arise from a supersaturation of point defects due to the non-conservative nature of the implantation process. They form around the projected range of the implanted ions, where the supersaturation level is the highest. These defects have two common forms: rod-like (311} defects and dislocation loops. Sub-amorphization (311) defects are one of the main interests of the present study and they will be discussed in detail in the next section.

2. End-of-range defects

When the silicon surface has been amorphized during implantation, end-of-range defects form just below the




9


amorphous/crystalline interface upon annealing. End-of-range defects may consist of both dislocation loops and (311) defects if the annealing temperature is low. These defects are important in IC processing since amorphous layers are always present for P, As and BF2 implants under the implant conditions that are usually used. For more information about the formation and evolution of these defects, please refer to references 42-46.

3. Precipitation related defects

If the concentration of the implanted impurity is above its solid solubility in silicon, precipitation related defects may form around the projected range. The precipitation process may generate such a high concentration of intrinsic point defects that extended defects can form as a result. The most common example of this is after high dose arsenic implantation. 43,4748 Two layers of dislocation loops have been observed upon annealing, one at the projected range which is associated with As clustering, the other at the end of range which is associated with excess interstitials beyond the amorphous/crystalline interface.

II. Extended defects from solid phase regrowth

1. Regrowth defects

When the solid phase epitaxial regrowth process is imperfect, defects nucleate at the advancing amorphous/crystalline interface and propagate into the regrowing silicon. They do not appear to dramatically affect the diffusivity of dopants, but they can getter impurities




10


and provide leakage paths if they span across a junction.4249'50

2. Clamshell defects

If the implant energy is sufficient but the dose is not too high, it is possible to produce a buried amorphous layer. Upon annealing two amorphous/crystalline interfaces meet and extended defects may result. These defects consist of larger dislocation loops compared to end-of-range defects.4951

We have discussed briefly each type of extended defects due to ion implantation. Visible damage from boron implants employed in this study consisted predominantly of {311} defects in the annealing temperature range that we used. In the next section, we will review some previous studies on {311) defects in detail.


1.2.3 {3111 Rod-like Defects


As we discussed in the last section, (311) defects can form under both sub-amorphization and amorphization conditions, providing the annealing temperature is low and/or the annealing time is short. This is summarized in Fig. 1.2.

Because of the low mass of the boron ions, the silicon substrate surface is not amorphized if the dose is below 5x1015 cm-2. Under such sub-amorphization conditions, (311) defects form at lower doses than dislocation loops. They are relatively unstable. Upon annealing, they either evolve into dislocation dipoles through an unfaulting reaction or dissolve. The dissolution provides interstitials to the




11


Energy and/or Dose

sub-amorphized amorphized


RP- R,as-implant





low T and/or I X R
short t anneal'F "

type 1 {311) {311} loop typ II (E.O.R.)



high T and/orR 0 R"long t anneal o ype I
dipole loop dipole typ II (E.O.R.)


Figure 1.2. Schematic of the conditions under which {3111 defects can form and how they evolve with time. background environment (which can result in TED) and the growth of dislocation loops if they are present. Since (311) defects form after a short time and/or low temperature anneal with or without the presence of dislocation loops, and they dissolve fairly fast, they have been termed intermediate defect configurations (IDCs). As an example, Fig. 1.3 shows a plan-view transmission electron microscope picture of (311) defects in boron implanted silicon after an anneal at 7500C for 5 min. Under the imaging conditions used, there are





12


three directions of {311) defects. Two of them are orthogonal to each other while the third one forms a 450 angle to the first two sets. This morphology will be explained in the following paragraphs.



























Figure 1.3. Example of (311} defects in 30keV 1x1015cm-2 B+ implanted Si under weak-beam dark field imaging condition with g220 reflection, annealed at 7500C for 5min in nitrogen.



{311) defects are elongated in [110] directions and

consist of interstitials precipitating on {311) habit planes. A unit cell of silicon lattice with {311) defects lying in six [110] directions is schematically shown in Fig. 1.4. The reflection vector g was chosen to be 220 when plan-view transmission electron microscopy pictures were taken in this




13


study. Since {311} defects are out of contrast for g parallel to their length, those on one of the six directions are invisible. Thus viewing on a [100] plane, we observed the morphology shown in Fig.l.3.








I


17 010]


PO-.[100]







line direction u gxu visible(+) or invisible(-)
[110] #0 +
[-110] =0
[101] #0 + [10-1] #0 + [011] 0 + (01-1] 0 +

Figure 1.4. Schematic of {311} defect directions and defect visible conditions under reflection g=220.

Various atomic models incorporated by self-interstitials have been proposed for the (311) defects. Salisbury and Loretto52,53 suggested that self-interstitial atoms located on the tetrahedral sites are responsible for {311) defect




14


formation. Models of split self-interstitials with or without dangling bonds were proposed by Tsubokawa et al.54 and Pasemann et al.55 respectively. Tan56 developed a homogeneous nucleation model which shows that for extrinsic defects, a chain of interstitial atoms form (311) defects having nonsix-membered atomic rings with matrix atoms, while for intrinsic defects, a chain of matrix atoms is cut out and the remaining atoms surrounding the cut are used to form {311} defects having non-six-membered atomic rings. A schematic diagram for the different steps involved in the formation of extrinsic {311) defects is shown in Fig.l.5 (a)~(d). This non-six-membered atomic ring configuration minimizes the dangling bond that may present.














001
QA

.110

(a)

Figure 1.5. Schematic of four step {311) defect formation process: (a) The accepted split [100] interstitial chain, (b) The addition of one more split [100] chain, (c) Details of the bond rearrangement and(d) The resulting defect configuration.




15







4



y2







(b)







4
3 b Il



(c)

Figure 1.5. (Continued)





16




















(d)


Figure 1.5. (Continued)

The displacement vector of {311} defects is

perpendicular to the line and has a prominent [100] component. This vector has been determined to be a<100>/2,57 a<411>/6,58 a<311>/1152 or a<611>/k by various groups, where a is the lattice constant and k=2559 or 31.60 The discrepancy between these values reflects the inaccuracy of the measurement because of the narrowness of the defects. Different methods have been used to determine the displacement vector, including the traditional image contrast analysis and the energy minimization calculation based on the relaxation of the interstitial agglomerate model under the Stillinger-Weber potential.

{311} defects have been studied in various ion implanted or particle irradiated systems for more than two decades. They were found in silicon,61 boron,57,62-64 arsenic,65




17


phosphorous66-68 and BF269 implanted silicon, electron,52,70 neutron 71, proton72 and neon73 irradiated silicon as well as electron irradiated germanium.59 The implant temperature effect on {311} defect formation has also been studied. Tamura74 found that room temperature phosphorous implants in silicon at low energies resulted in the formation of dislocation loops, whereas elevated temperature implants (200-4000C) gave rise to the formation of rod-like {311} defects. This temperature effect has also been shown by Chadderton and Eisen.75 They found that dislocation loops formed in boron implanted silicon if the implantation was held at liquid nitrogen temperature, in contrast with the room-temperature implants which showed {311} defects. This behavior might arise from the self-annealing effect occurring at higher temperatures which reduces the number of free interstitials. Under this condition {311) defects may be easier to form than dislocation loops presumably because the energy to form a (311} defect is less than that of a loop.76

Back in the seventies, the formation of {311} defects was associated with the presence of boron by Seshan and Washburn.67'68 They observed {311} defects in phosphorous implanted boron-doped silicon. However, the {311} defects were absent from n-type silicon under the same implant and annealing conditions. Thus they concluded that boron is essential for the formation of {311} defects. Recently, a research group at Bell Laboratories (private communication) investigated the role of boron during {311} defect formation




18


by studying Si+ implanted silicon doped with various concentrations of boron. They measured the number of interstitials trapped in {311} defects after the same postimplantation anneal and found that the interstitial density actually decreased with the increase of boron concentration. We have also observed a higher formation threshold of {311} defects in boron implanted than that in silicon implanted silicon. It is believed that boron atoms pair with interstitials and form submicroscopic boron interstitial clusters and mobile boron interstitial pairs, which consume part of the interstitials and reduce the number available for {311) defect formation. This will be discussed further in the following chapters. It should also be mentioned that the new interest in {311} defects arises partly from the fact that they have been claimed to be the only source of interstitials for TED, which is still controversial at this time.


1.2.4 Dopant Diffusion in Silicon


The diffusion of dopant atoms can be described by
DA = Do exp(- QA) (1.1)
kT
where DA is dopant diffusivity, DO is the pre-exponential constant and QA is the activation energy for diffusion. It is generally agreed that on an atomic level, dopant atoms in silicon diffuse through interactions with silicon selfinterstitials and vacancies. The type of point defects which dominates the diffusion process determines the type of





19


diffusion mechanism, vacancy, interstitial or interstitialcy mechanism. In the vacancy mechanism, a substitutional dopant atom exchanges position with an empty lattice site. In the interstitial mechanism, the dopant atom is kicked out of the silicon lattice by a silicon self-interstitial and diffuses as a pure interstitial before returning to the lattice as a substitutional atom. In the interstitialcy mechanism, the silicon self-interstitial and the dopant atom form a diffusion pair. This process is illustrated in Fig. 1.6. Usually no distinction is made between the interstitial and the interstitialcy mechanisms.


Vacancy





Interstitial





Interstitialcy



Figure 1.6. Schematic diagram of different dopant diffusion mechanisms.

If an asterisk is used to represent atoms in their free states, then the interaction of dopant atoms and the point defects can be expressed as

A*+X* <-> AX (1.2) where X=I corresponds to interstitial/interstitialcy




20


mechanism while X=V refers to a vacancy mechanism. Mathematically the ratio of dopant diffusivity under nonequilibrium conditions over its intrinsic, equilibrium value can be written as
DA -f CI +AV CV.3)
DA* CI* CV*

where fAX=DAX*/DA* (X=I or V), fAI+fAV=l, CI and CV are the concentration of interstitials and vacancies under nonequilibrium conditions, CI* and CV* are the quantities under intrinsic, equilibrium conditions. Equation (1.3) shows that the diffusion of dopant atoms consists of two parts: diffusion by coupling with interstitials and with vacancies. The fraction of each is represented by fAI and fAV, respectively. While this treatment of diffusion allows us to mathematically explain the experimental results, it is possible that the actual mechanism is quite different.

Dopant diffusion in silicon has been under investigation for many years; however, the diffusion mechanisms responsible for some of the impurities are still under debate. It is generally accepted that antimony diffusion is dominated by a vacancy mechanism. Phosphorous and boron diffuse predominately by an interstitial/interstitialcy mechanism. Arsenic appears to diffuse by both vacancy and interstitial/interstitialcy mechanisms. The diffusion of Al, Ga and In is believed to have a strong interstitial component77. Elements such as Li, Na, He and H are believed to diffuse by an interstitial/interstitialcy mechanism as do





21


most of the transition metals8'9. The mechanism of Si selfdiffusion appears to be vacancy controlled at temperatures <10000C and interstitial controlled at higher temperatures.80

Boron is the diffusing species that will be discussed in this study. The set of nonequilibrium reactions involved in the kinetic model for boron diffusion is described as follows:

Bs + I <-> (BI) (1.4) (BI) + V <-> Bs (1.5) I + V <-> [0] (1.6) where Bs and (BI) are the substitutional boron and the boron interstitial pair, respectively, and [0] denotes the silicon lattice site.

The relationship between point defect concentration and dopant diffusion is given by Eq. (1.3). To understand dopant diffusion in silicon, it is necessary to understand how different processes affect point defect concentration. Because of its small value relative to the concentration of silicon atoms residing on the lattice sites, the concentration of point defects is not a quantity that can be measured directly. Diffusion studies are mostly used to gain information of the interstitials and vacancies. An enhanced diffusion of a species which diffuses by interstitial (or vacancy) mechanism indicates a supersaturation of interstitials (or vacancies) in the bulk. Ways of quantitatively monitoring these point defects will be discussed later.





22


Under equilibrium conditions at a particular

temperature, the possible sources of point defects are chemical reactions at the silicon surface, such as oxidation,81-86 nitridation87-90 and silicidation,91-96 the impurity clustering and precipitation,97-100 and the radiation damage,20',101-105 etc. Deviations from equilibrium will cause excess or depletion of either interstitials or vacancies, which will further affect dopant diffusion. For dopants that diffuse predominantly by an interstitial/interstitialcy mechanism, an excess of interstitials or depletion of vacancies will enhance the dopant diffusion, while for dopants that diffuse primarily by vacancy mechanism, the same conditions will retard dopant diffusion.

Two possible detectors for excess point defects in

silicon are dopant profiles and extended defects. Dopant profiles include both epitaxially grown doping superlattices or implant profiles. Boron and antimony superlattices can be used to monitor the in-diffusion of silicon interstitials or vacancies produced by processes such as surface oxidation, nitridation, silicidation and aluminization, or the implant damage. Since boron diffuses mostly via interstitial/interstitialcy mechanism, the boron epi layer spike will broaden after oxidation and implantation during which processes interstitials are injected into the bulk. On the contrary, silicidation, nitridation and aluminization will cause vacancy injection or interstitial depletion, therefore broadening the antimony epi layer spike. The





23


implant profiles behave in the same way as the epi layer spikes but are less sensitive because the epi layer profiles are thinner and possess a more ideal Gaussian shape. Simulation programs such as FLOOPS can then be used to calculate the diffusivity of the dopant atoms and quantify the enhancement or retardation effect.

Other than dopant profiles, extended defects can also be used to monitor the point defect perturbance. Early studies of the trapping of point defects by extended defects were conducted by using stacking faults,s2,106-108 which grow by absorbing interstitials and shrink by emitting interstitials, thereby giving a good indication of the change in point defect concentration. However, because of their large size (> 5 pm) and low density (< 106 cm-2), the stacking faults are not ideal for detecting small perturbations. Instead, post-implantation annealed dislocation loops (end-of-range loops) have the advantage of much smaller size (- 0.02 pm in diameter) and larger density ( 1010 cm-2), therefore being more sensitive detectors. In addition, the dislocation loops form at well-defined depth, which can not only be utilized to measure the net interstitial flux,'05,109 but have also been found to be an effective barrier to interstitials driving anomalous dopant diffusion.17,110,111

Extended defects themselves can act as a source of interstitials during annealing. Since dislocation loops dissolve at temperatures that are too high (>10000C) to explain TED which is completed within seconds at T>9500C, the




24


{311} rod-like defects have been examined as the possible sources of interstitials that cause TED. Cowern et al.03 claimed that the release of interstitials which drive TED of boron in silicon occurs at two different time scales. Interstitials emerging from the ion collision cascade give rise to a primary (ultra fast) pulse of diffusion, and at a sufficiently high displacement density, emission of interstitials from {311} defects causes a secondary (much slower) diffusion transient. Later, Eaglesham et al.112 and Stolk et al."3 reported that (311} defects are the only source of interstitials driving TED based on their study of Si+ implanted Si. They reasoned that interstitials are emitted from {311} defects in the damage region during annealing. This emission occurs at a rate which exhibits the same activation energy as that of the interstitial indiffusion, and the duration of {311} dissolution process is in close agreement with that of TED. Our group have been studying (311) defects and TED in boron implanted silicon. We found that TED occurs without the formation of {311} defects if the implant energy (4 keV) and dose (1x1014 cm-2) are low.114 The source of interstitials driving TED in this case was attributed to sub-microscopic mobile boron interstitial pairs. These B-I pairs play a key role in TED of boron in silicon, as will be discussed.

Therefore, understanding the correlation between

implantation induced defects (e.g., {311} defects and boron interstitial clusters) and TED is essential to the




25


understanding of dopant diffusion. This correlation will be further investigated in this study.


1.3 Ion Implantation. Characterization and Simulation Techniaues



1.3.1 Ion Imolantation


Ion implantation is the introduction of energetic charged particles into targets with enough energy to penetrate beyond the surface region. The ions enter the substrate, collide with the host atoms, gradually lose energy and finally come to rest at some depth within the substrate. Ion implantation is involved in several steps during the fabrication sequence of MOSFET and bipolar transistors when the controlled doping of selected areas is required.

An ion implanter consists of the following major

components: an ion source, an extracting and ion analyzing system, an accelerating column, a scanning system and an end station. The ions produced by the source are extracted by a small accelerating voltage and then injected into the analyzer magnet. A spatial separation of ions subjected to the Lorenz force occurs due to the differences in the mass and charge. The selected ions are injected into the accelerating column and the rest are screened out. In this case the system operates in the pre-analysis configuration. If the ions are accelerated to their full energy before the mass separation, then it works in the post-analysis





26


configuration. The extraction voltage ranges usually between 15 and 40kV. For low energy implants, a retarding voltage can be used after separation. The selected ions are accelerated by a static electric field and focused and shaped in the column and ready to be implanted.

Some important quantities during the ion implantation

process are listed below. The implant dose can be expressed as
It
Q, = -(1.7)
nqA

where I is the implant current, t is the time, n=l for singly ionized species and 2 for doubly ionized species, q is the charge and A is the area of the implanted surface. For an ion accelerated through a potential V and deflected through a magnetic field B, the radius of the ion's path is
1/2MV
r=-B q (1.8)
B q

where M is the mass of the dopant atom. In absence of crystal orientation effects, the range distribution is roughly Gaussian and to a first approximation the projected range distribution N(x) is described as a one-dimensional Gaussian profile characterized by the projected range Rp and the standard deviation ARp, as given by
Q 1x-R
N(x) =2 exp-] (1.9)
42irAR, 2 ARP

Schematic diagram of an implanted impurity profile is shown in Fig. 1.7. Due to the channeling effect, the use of a Gaussian profile is not accurate. The UT-Marlowe code (a




27


Monte Carlo simulator) yields a much more accurate picture of the ion distribution just after implantation.





/ I---Ix Rp)
N(x) N exP -2ALV






I I I

o.14 --- -I---



Distance into material. x

Figure 1.7. Schematic illustration of a Gaussian distribution resulting from ion implantation.


1.3.2 Transmission Electron Microscooy


The transmission electron microscope (TEM) is used to obtain information from samples which are thin enough to transmit electrons. A bright-field image is formed if the directly transmitted beam is selected and a dark-field image is formed if a diffracted beam is selected. Dark-field images normally have higher resolution than bright-field images and this is the mode we used in our study. The defect structure can be analyzed by both plan-view and crosssectional TEM. Plan-view TEM (PTEM) is used to study the defect evolution during annealing and cross-sectional TEM




28


(XTEM) is used to get the side view and the depth of the defect layer.

To make a PTEM sample, the sample is first cut into 3mm diameter pieces using an ultrasonic disk cutter. The small disk is then polished down to about half of its original thickness using a polishing jig with 15pm aluminum oxide powder. A coat of wax is then applied to the top surface (i.e. the implanted surface) of the sample to protect it during the etch. The sample is mounted with wax on a Teflon holder facing down and etched from the back side in a jet etcher using a HF:HNO3=25%:75% solution until a small hole with an electron transparent region around it is obtained. The wax is the removed from the sample using heptane and/or acetone and the sample is ready for analysis.

To make a XTEM specimen, a piece of silicon wafer is cut into 10mil thick strips using a dicing saw. The top surface of the two strips is coated with M600 Bond epoxy resin. The two top surfaces are then brought into close contact with each other. Two silicon dummy strips can also be glued on the side of each sample strip to improve the mechanical stability. The bonded strips are heated in a conventional oven at 1000C for lhr to harden the epoxy resin. The cured strips are then polished down to a thickness of about 40pm using 5pm aluminum oxide powder. A copper ring is mounted onto the thinned strips using the M600 Bond, leaving the interface of the sample strips exposed in the center of the ring. The specimen is then further thinned in a Gatan two-




29


stage ion mill by using two Ar+ ion guns at a gun voltage of

4 kV and a gun current of 0.5 mA until a small hole is obtained at the sample strips' interface. The region in the vicinity of the hole is sufficiently thin (<0.5 gm) to be transparent to the electron beam of the TEM.


1.3.3 Secondary Ion Mass SDectroscoDv


In secondary ion mass spectroscopy (SIMS), an energetic beam of focused ions is directed at the sample surface in a high or ultrahigh vacuum environment. The momentum transfer from the impinging primary ions to the sample surface causes sputtering of the surface atoms and molecules. The most commonly used sputtering ions are Cs+, 02+, 0+ and Ar+ in the energy range from 2 to 20 keV at angles of incidence between 450 and 90'. 02+ ions are used in this study to enhance secondary ion yield. Some of the sputtered species are ejected with positive or negative charges. These are termed secondary ions. The secondary ions are then mass analyzed using a mass spectrometer. SIMS can be used to obtain a depth concentration profile of the near-surface region to a resolution of 2-30 nm at a detection limit of 1012~1016 cm-3. For further information see ref.115


1.3.4 Florida Object Oriented Process Simulator


FLOOPS modeling program is a physically based twodimensional process simulator capable of modeling a complex device structure. It uses local point defect concentration




30


to determine dopant diffusivities. In order to keep track of the spatial distribution of point defects, FLOOPS solves the diffusion equations for the dopant atoms as well as the equivalent equations for the point defects. Since vacancies and interstitials can interact, the equations describing their motion are coupled. The equations used to determine dopant diffusion include
dC =-V* JA (1.10)


A= -K,,(C, A )
m
JA= -DAxCA Vln(CA n) (1.12)
x -x. Cx n,

where JA is the flux of the dopant, CA is the total concentration of the dopant species, the summation over X is for all possible diffusion mechanisms, Cx is the concentration of the X species, DAX is the diffusivity corresponding to the X mechanism, ni and n refer to the intrinsic and actual electron concentrations, CA1 and CA2 are the concentration of the dopant species on the two sides of a materials boundary, Ktr is the transport coefficient across the material boundary and m is the segregation coefficient of the dopant for the boundary materials. The equations solved for interstitials or vacancies are
Cx= -V* Jx Kb cICV -;C_ ) (1.13)

Gx = Jx + Kxs(Cx Cx*) (1.14) Jx=-DxCxV C (1.15)
cx

where x=I for interstitials and V for vacancies, JX is the flux of the X point defect species, CX* and CX are the





31


equilibrium and actual concentrations of X, GX is the surface generation rate for X, KXS is the surface recombination constant and Kbulk is the bulk recombination constant.

For the case of dopant diffusion following an

implantation, the initial distribution of point defects is determined by the implantation process. These point defects then diffuse according to the above equations. Anomalous diffusion can be observed until the point defect concentrations return to the equilibrium values, when only normal diffusion can be seen.


1.4 Thesis Statement


The contributions of this work are in the following areas:

1. Determination of the formation threshold of extended defects in low energy B+ implanted silicon.

2. Quantitative TEM studies of the annealing kinetics of {311} defects arising from B implantation.

3. Experimental investigation of the implant energy and dose effects on {311) defect behavior.

4. Experimental investigation of the implant energy and dose effects on TED.

5. Extraction of diffusivity enhancement for different implant energies, doses, annealing times and temperatures.

6. Experimental investigation of the sources of interstitials driving dopant diffusion.





32


7. Provision of experimental basis for the improvement of process development and process simulators.













CHAPTER 2
THE FORMATION THRESHOLD OF (311} DEFECTS AND SUBAMORPHIZATION DISLOCATION LOOPS


2.1 Overview


The formation of very shallow p-type region by ion

implantation requires the use of low energy boron implants. Although there have been many reports on the defect and diffusion behavior in boron implanted silicon, 63,116-118 none of them have systematically studied the defect formation threshold in the low energy implantation regime. For boron implantation, because of the low mass of the implanted ions, even if the ion fluences reach 2x1015 cm-2, the damaged surface layer still does not amorphize. In this case the extended defects that form after annealing can be classified as type I or sub-amorphization defects. The implant energy and dose combinations in this study are similar to those used in the semiconductor industry for B implants and fall in the sub-amorphization regime.

In order to understand how the sub-amorphization defects influence dopant diffusion, it is important to know the implant conditions under which these defects are actually formed. This is the purpose of this section of the current study.





33




34


2.2 Experimental Procedure


Czochralski-grown (100) n-type 8-20 fcm 150 mm wafers were implanted with boron ions at energies of 5 keV, 10 keV, 20 keV, 30 keV and 40 keV to doses of 5x1013 cm-2, 1x1014 cm-2, 2x1014 cm-2, 5x1014 cm-2 and 1x1015 cm-2. The tilt/rotation angles were 50/00. The implant current was

3 mA. During ion implantation, the samples were kept at room temperature using water cooling. Furnace anneals were then performed at 7500C for 5 min or 9000C for 15 min in a nitrogen ambient to study the formation threshold of {3111 defects and sub-amorphization dislocation loops respectively. Plan-view transmission electron microscopy (PTEM) samples were prepared using standard jet-etching procedures and Cross-sectional TEM (XTEM) samples were prepared using ionmilling. Micrographs were taken from each sample to examine the presence or absence of secondary defects. The g220 reflection was used to acquire all the micrographs under weak-beam dark field imaging conditions. A defect density of 1.2x107 cm-2 was used to distinguish between samples with and without extended defects, i.e., if no defect is observed in three randomly-picked areas each with a size of 7x10 cm2 under a magnification of 50000X, then it is stated that there are "no" defects in that sample.





35


2.3 Defect Microstructure


Figure 2.1 shows PTEM micrographs of some of the samples in the implant matrix after an anneal at 7500C for 5 min. Micrographs of 5 keV, 20 keV and 40 keV implants with doses of 1x1014 cm-2, 2x1014 cm-2 and 5x1014 cm-2 are shown in the figure. At a dose of 1xl014 cm-2, there are no {311) defects in the 5 keV sample and the defect density increases with increasing energy. When the dose is doubled to 2x1014 cm-2, there are still no {3111 defects in the 5 keV sample, but more defects are observed in the other two implants. Increasing the dose further to 5x1014 cm-2 results in {311} defect formation at 5 keV. Figure 2.1 clearly presents a transition from samples without to those with (311} defects. It is obvious that as the implant energy increases, the critical dose for forming {311} defects decreases. From this figure it is also obvious that the interstitial supersaturation necessary to nucleate (311} defects is far less than that for dislocation loops. This is consistent with Eaglesham 's observation (private communication) of {311} defects at 7x1012 cm-2 for 40 keV Si implants and the observation of Jones et al.42 that stable dislocation loops (approximately the same implant energy) do not form until a dose of 2x1014 cm-2. In addition, the threshold dose for {311} defects in a 20 keV B implant is around 2x1014 cm-2, which is much greater than the threshold dose of 7x1012 cm-2 for a comparable depth of 40 keV Si implant. This is





36


presumably due to the formation of immobile boron interstitial clusters and mobile boron interstitial pairs that reduces the concentration of free interstitials available to form {311} defects.

PTEM micrographs of selective samples annealed at 9000C for 15min are shown in Fig. 2.2. This temperature/time condition was chosen to reveal stable sub-amorphization dislocation loops and dipoles that evolve from unstable (311} defects. The {311} defects have completely dissolved after the anneal. A submatrix of 5 keV, 20 keV and 40 keV implants with doses of 2x1014 cm-2, 5x1014 cm-2 and 1x1015 cm-2 are shown in Fig. 2.2. The 40 keV sample has dislocation loops and dipoles at a dose of 2x1014 cm-2. The 20 keV and 5 key samples do not show loops or dipoles until the dose is increased to 5x1014 cm-2. The critical dose for type I loop formation appears to decrease with increasing implant energy. The threshold dose at 40 keV (2x1014 cm-2) matches the previous results of Jones et al.'s for 30 key boron. Table 2.1 lists the formation threshold for both {311} defects and [(110] loops for the whole implant matrix.

XTEM micrographs of 10 keV, 20 keV, 30 keV and 40 keV implants to a dose of 1x1015 cm-2 after an anneal at 7500C for 5 min are shown in Fig. 2.3. The reason for choosing the highest dose in our implant matrix is that both {311) defects and dislocation loops are present in these implants under this annealing condition. A layer of sub-amorphization extended defects is expected to form at the projected range





37


Rp. Compared with the SIMS profiles which will be shown later in Chapter 4, the depth of the defect layer was found to correspond to Rp. Both the distance between the defect layer and the surface and the width of the defect layer decrease as implant energy decreases. This decrease of distance between the damage region and the surface may induce a more prominent recombination effect at the surface, as will be discussed later.


2.4 Criteria for the formation of {311} Defects and Subamorphization LooPs


Various suggestions have been made for the criterion for secondary defect formation. Tamura et al.119 have stated that implant dose can be used as such a criterion. They studied defect formation in 1-2 MeV B, P and As implanted silicon and found that secondary defects form if the dose range of 2x1013-1x1014 cm-2 is exceeded, independent of the implant species. Other experiments, however, show that a higher dose is required for defect formation if the ion mass is smaller. Gibbons120 studied B, Al, N, P, Sb, Ne and Si implanted silicon at energies from 40 keV to 100 keV. The threshold for small loop formation for Sb implant is about 5x1013 cm-2 while for B implant it is above 1x1014 cm-2. This implies that the implant dose cannot be the sole criterion for loop formation. Furthermore, the results of studies on channeling versus random implants show that the defect formation is not dependent only on impurity dose but on the damage introduced





38


during implantation. Raineri et al.121 studied random and [100] channeling implants with B and P ions. 100 keV P implants with doses from 5x1013 cm-2 to 2x1014 cm-2 does not show defects under channeling conditions after an anneal at 5000C for 1 hr. The same implants in a random direction give rise to a network of dislocation loops. A systematic study on implant damage using TEM, RBS and TRIM simulations was reported by Schreutelkamp et al.122 They claimed that if the total number of silicon atoms displaced by the implant ions exceeds a critical value, stable sub-amorphization dislocation loops are observed. They implanted Si with B, Si, P, Ga, As, In and Sb ions at keV and MeV energies. For B implants with energies ranging from 50 keV to 190 keV, the critical value of displaced Si atoms was 4.7x1016 cm-2 based upon TRIM calculations and 1.4x1016 cm-2 based upon Rutherford Backscattering (RBS) analysis. This value increases with the mass of the implant ions. To the author's best knowledge, there has been no report on the formation threshold of {311) defects or stable loops in the low energy (<20 keV) regime of B implanted silicon.

We have calculated the number of displaced silicon atoms using TRIM for our implant matrix and the results are listed in Table 2.2. As can be seen, when the displaced atom density reaches about 1.5x1016 cm-2, (311} defects form, and when it reaches about 2.6x1016 cm-2, sub-amorphization loops form. The latter value is less than the result of Schreutelkamp et al., but considering the fact that we used a




39


new version of TRIM, the results are in reasonable agreement. It thus appears that the number of displaced silicon atoms can be used effectively as the criterion for both (311} defect and sub-amorphization dislocation loop formation.

If the above argument is true, we expect to see more

extended defects after annealing in samples containing more displaced silicon atoms after implantation. This indeed is the case for both anneal conditions with only two exceptions. The first is the 5 keV 1x1015 cm2 implant after an anneal at 7500C for 5 min (Fig. 2.1). Keeping the dose at 1x1015 cm-2, we see less (311} defects but more dislocation loops in the 5 key implant than in the higher energies. The reason might be that the 5 keV 1x1015 cm-2 implant has the highest interstitial concentration in the whole implant matrix. When the number of free interstitials reaches a critical value, the system may form loop nuclei. If the energetics are such that interstitials would rather be in a loop, then the loop/{311} ratio increases, i.e., loop formation might be more favorable than {311} defect formation. The second exception to the damage alone being the reason for threshold changes arises when one compares the 5 keV and the 20 keV implant. The displaced atom density for 5 keV 5x1014 cm-2 implant (2.56x1016 cm-2) is slightly less than that of the 20 keV 2x1014 cm-2 implant (2.94x1016 cm-2) (Table 2.2), but after an anneal at 9000C for 15 min, the 5 keV sample shows a low density of half loops while the 20 keV implant shows no defects. The reason might be that these displaced atom




40


densities are near the threshold for defect formation so the behavior is due to the other fluctuations. However, the reason may also be due to the above argument for why loops are seen most prominently for 5 keV implant.

Besides the effect of implant energy on damage

production, interstitial recombination at the sample surface may also affect the formation of secondary defects. Figure

2.4 shows the projected range Rp as a function of implant energy obtained from the XTEM micrographs and the SIMS profiles which will be shown in Chapter 4. Rp increases with increasing energy. It is about 230A for the 5keV implant and about 1750A for the 40keV implant according to SIMS. As the implant energy decreases, the damage region gets closer to the sample surface. Surface recombination could then become more efficient and thus reduce the number of free interstitials available to form secondary defects. It is difficult to determine which effect, decreased damage, increased boron interstitial cluster formation due to higher interstitial supersaturation in the dopant peak region or increased surface recombination, is primarily responsible for the increased threshold dose with decreasing implant energy. Additional experiments to sort out these effects will be necessary.

In summary, the formation threshold for both (3111 rodlike defects and [110] dislocation loops in low energy boron implanted silicon have been investigated. A matrix was chosen with implant energies ranging from 5keV to 40keV and




41


doses from 5x1013cm2 to 1x1015cm-2. TEM was used to evaluate the presence of both types of defects. It was observed that the threshold dose for both {311} defect and subamorphization dislocation loop formation increases significantly with decreasing implant energy. This threshold is higher for boron implants than for silicon implants presumably because of the formation of boron interstitial clusters and mobile boron interstitial pairs. As discussed earlier there are several possible explanations for the higher threshold dose with decreasing implant energy. These include a decrease in the displaced atom density, increased formation of boron interstitial clusters and/or increased surface recombination. Experimentally a displaced atom density of about 1.5x1016 cm-2 for {311) defects and

2.6x1016 cm-2 for sub-amorphization loops from TRIM calculations reasonably predicts defect formation for these implant conditions. However, the results of Zhang et al.114 and Eaglesham et al.(private communication) clearly indicate that boron is a very effective interstitial trap and that increasing the boron concentration clearly results in an increase in the concentration of interstitial trapping.






S.2jim
lxl014cm-2 2xl014cm-2 5x1014cm-2




5keV








20keV








40keV





Figure 2.1. Weak beam dark field (g220) PTEM images of 5keV, 20keV and 40keV B+ implanted Si to doses of 1x1014cm-2, 2x1014cm-2 and 5x1014cm-2, after an anneal at 7500C for 5min in N2.







2xl014cm-2 5xl014cm-2 lxl015cm-2




5keV








20keV








40keV





Figure 2.2. Weak beam dark field (g220) PTEM images of 5keV, 20keV and 40keV B+ implanted Si to doses of 2x1014cm-2, 5x1014cm-2 and lxl015cm-2, after an anneal at 9000C for 15min in N2.








Table 2.1 Types of extended defects formed in B+ implanted silicon 5keV 10keV 20keV 30keV 40keV 5x1013cm-2 none none none none none lxl014cm-2 none none {311 })s { 311 }s 311 }s 2x1014cm-2 none {311)s {311 }s {311}s {311}s Loops Loops

5x1014cm-2 1311)s (311)}s 1311)s 1311)s (311)s
Loops Loops Loops Loops Loops lxl015cm-2 {3111}s {311}s {311)s (311)s {311}s
Loops Loops Loops Loops Loops




45




Sm

lOkeV







S

20keV








30keV










*-..40keV







Figure 2.3. Weak beam dark field (g220) XTEM images of 10keV, 20keV, 30keV and 40keV B+ implanted Si to a dose of lxl015cm-2,
after an anneal at 7500C for 5min in N2.












Table 2.2 Displaced atom density (#/cm2) from TRIM calculations (B+ implanted silicon)

5keV 10keV 20keV 30keV 40keV 5x1013cm-2 2.56e15 4.51e15 7.35e15 9.60e15 1.16e16 lx1014cm-2 5.12e15 9.02e15 1.47e16 1.92e16 2.31e16 2x1014cm-2 1.02e16 1.80e16 2.94e16 3.84e16 4.62e16 5x1014cm-2 2.56e16 4.5 e16 7.35e16 9.60e 16 1.16e17 lx1015cm-2 5.12e16 9.02e 16 1.47e 17 1.92e 17 2.31e17




47






1800
----from XTEM
1600 -- from SIMS

< 1400
CL
m 1200
CD
S1000

800
0
a 600

400

200
0 10 20 30 40 50 Implant Energy (keV)










Figure2.4. Projected range Rp of 5keV to 40keV B+ implanted Si, determined from XTEM micrographs and SIMS profiles.














CHAPTER 3
EFFECT OF ANNEALING ON DEFECT EVOLUTION AND TRANSIENT ENHANCED DIFFUSION


3.1 Overview


Due to the limitation that TED imposes on the minimum device dimensions, it has been the subject of considerable studies for many years, yet its mechanism is still under debate. TED has been attributed to the excess point defects introduced by the implantation process. The magnitude of TED is related to the implantation conditions, the annealing conditions and the surface and bulk recombination effects. Recently, Stolk et al.11 have suggested that TED occurs by the emission of silicon self-interstitials from rod-like [311} defects during annealing. The activation energy (3.60.1 eV) for the {311} defect dissolution process relates closely with that of TED (3.5 eV). Later, Zhang et al.114 reported that they observed TED in B+ implanted samples without (311} defects and they attributed the interstitial source of TED to submicroscopic clusters. More recently, Cowern et al.23 studied the role of carbon and boron clusters on the diffusion process. The number of self-interstitials trapped per clustered impurity atom is about 1.15 for carbon and about 1 for boron. This provided further evidence to the




48





49


possibility that there was more than one storage mechanism for excess interstitials that contribute to TED.

The purpose of this section is to discuss the effect of annealing on defect behavior and TED in order to investigate the correlation between the two.


3.2 Effect of Annealina on Defect Behavior



3.2.1 Experimental Procedure


In order to study the dissolution kinetics of {311} defects from B+ implantation without the complication of stable loop formation, we picked one sample (20 keV 2x1014 cm-2) from the implant matrix discussed in Chapter 2 (refer to section 2.2). Subsequent furnace anneals were performed in a nitrogen ambient using fast pushes and pulls. The annealing temperatures and times were chosen as 6500C for

2 hr, 6 hr, 14 hr, 24 hr, 36 hr, 48 hr; 7000C for 30 min, 1 hr, 2 hr, 4 hr, 8 hr; 7500C for 5 min, 20 min, 30 min, 1 hr,

1.5 hr and 8000C for 3 min, 5 min, 8 min, 10 min, 13 min. PTEM micrographs were taken under weak beam dark field imaging condition using a g220 reflection. The density and length of {311} rod-like defects were measured directly from the TEM micrographs. The areal interstitial density bound by {311} defects after each anneal was calculated by multiplying the defect length by the density of interstitials per unit length. The time constant for the interstitial decay process at each temperature was plotted in an Arrhenius graph as a





50


function of the reciprocal of temperature to obtain the activation energy for the {311) defect dissolution process.


3.2.2 Defect Evolution: Microstructure. Interstitial Density
and the Activation Energy of Interstitial Decay


Figure 3.1 (a)~(d) shows the PTEM micrographs of the sample after anneals at 6500C, 700*C, 750"C and 800'C for various times. The formation and coarsening/dissolution processes of {311) defects are faster at higher temperatures. After 2 hr at 6500C, these defects are not fully formed yet. The defects in their initial stages exhibit a "black dot" contrast. After 14 hr, (311) defects start to form. However, Ostwald ripening along with the dissolution process is so slow at this low temperature that the number of the defects is relatively large and their size is relatively small even after 48 hr of annealing. If the annealing temperature is increased to 7500C, {311) defects are observed after only 5 min and the dissolution process becomes more rapid. At 8000C, {311) defects almost dissolve completely after 13 min of annealing.

We have measured the net interstitials trapped in {311) defects from the PTEM micrographs after each anneal. Figure

3.2 shows the {311} defect dissolution process at each temperature. As we can see, the dissolution process shows an exponential dependence on annealing time. The dots in Fig.

3.2 are the actual data points and the lines are exponential fitting curves of the form





51


Si(t) = Si,(O)exp(--) (3.1)

where T is defined as the characteristic time constant of the decay. Although the decay process is strongly dependent on annealing temperature, the initial density of interstitials at all four temperatures is very close to the same value of 3~4x1013 cm-2, which is only 15-20% of the implant dose of 2x1014 cm-2. According to the "plus one" model,124 the initial density of interstitials should closely mimic the implant dose, because each implanted ion is supposed to give rise to one excess interstitial during annealing. This model assumes that the implant ion induces a series of Frenkel pair formation and then comes to rest in the lattice at an interstitial site. Subsequent annealing results in annihilation of all vacancy-interstitial pairs with just the one interstitial remaining. This theory is proved to be reasonable by the study of Si implanted silicon of Listebarger et al.'05 and valid to a remarkable degree by the study of Eaglesham et al.125. Listebarger et al. used dislocation loops as point defect detectors to measure the flux of interstitials introduced by a 30keV B+ implant to doses of 7x1013 cm-2, 1x1014 cm-2 and 2x1014 cm-2. The loops layer was formed by 100 keV Gel implant to a dose of 1x1015 cm-2 followed by an anneal at 5500C for 16 hr to regrow the amorphous surface and another anneal at 8000C for 30 min to coarsen the loops. They reported that the net flux of interstitials measured using the loop detectors is very




52


similar to the dose for each B+ implant. Eaglesham et al. measured the density of interstitials trapped in {311} defects in 40 keV Si implanted silicon to a dose of 5x1013 cm-2. The interstitial density versus dose plot shows a slope of 1.5, indicating slightly more than one interstitial per implanted ion. However, in our B+ implanted silicon study, the observed interstitials only account for approximately one tenth of the dose. We therefore believe that there is another type of defects, i.e., boron interstitial complexes, which trap the missing interstitials. These complexes are sub-microscopic and are not resolvable in the TEM. There appear to be more than one form of boron interstitial complexes. From our low energy low dose implants (as will be discussed later), we observed as did Zhang et al.114 that there is a significant fraction of the implant profile peak that is immobile. This could be explained by the formation of B-B-interstitial clusters. A second portion of the profile is mobile and may simply be Binterstitial pairs as suggested by Cowern.123 Both of these non-visible defects could help account for the "missing" interstitials in the (311} defects.

An Arrhenius plot of the characteristic rate constant
k=l/T of {311) defect dissolution process as a function of 1/T is shown in Fig. 3.3. This temperature dependence shows an activation energy of 3.8 eV. Along with our results, Fig.

3.3 also presents the temperature dependence of {311) defect dissolution from the study of Stolk et al.13 of Si+ implanted





53


silicon. It is interesting to notice that our activation energy is in close agreement with theirs of 3.60.1 eV, yet there is a y-axis shift of the two curves, i.e., the {311} defects in our sample are more stable. The similar activation energies implies the process of dissociation of an interstitial from {311) defects is the same, independent of the implant species. The reasons for the change in stability could be: (1) The dose for our implant (2x1014 cm-2) is much higher than what they used (5x1013 cm-3) and thus there are more interstitials in our sample to stabilize {311) defects.

(2) The dissolution of boron interstitial clusters in our sample during the anneal induces a higher concentration of interstitials in the background environment and again delays the dissolution of {311) defects.


3.2.2 Defect Evolution: Density. Size and Distribution


Quantitative results that one can obtain from the PTEM micrographs include the {311} defect density, the average defect length and the defect distribution after annealing.

Figure 3.4 shows the {311) defect density as a function of annealing time for each temperature. The defect density decreases dramatically with increasing annealing time. Although the rate is just slightly higher for higher temperature, it takes much longer for a lower temperature to reach a particular density. For example, the density drops to 109 cm-2 after about 10 min at 8000C, but it takes more than 10 hr at 7000C. This rapid decrease in the number of





54


defects explains the decay of interstitials trapped in {311}s even when the average defect size increases.

The average (311} defect size is shown in Fig. 3.5 as a function of annealing time for each temperature. The width change of these defects have been found to be very small during the anneal by high resolution TEM studies conducted by Eaglesham et al.125, so that the size change is represented by the change of the length. As can be seen from Fig. 3.5, the average defect size increases initially with annealing time and then starts to decreases at some point during the dissolution process. This breaking point can be seen for anneals at 7000C and 7500C but not for anneals at 6500C or 8000C. We believe this final drop of defect size indicates a pure dissolution stage which should be observed as well at 6500C and 8000C if the anneals are long enough. Again, to reach a particular defect size, it take longer for lower temperatures. For example, for 8000C anneal, the defect size reaches 400 A after about 3min, while at 7000C, it takes about 1.5 hr.

To further illustrate the {311} defect evolution

process, the defect distributions after each anneal at each temperature are shown in Fig. 3.6 (a)~(d). The distribution moves to a larger size with increasing time at each temperature and the defect density N(1) at every size 1 shifts to a lower value. This movement is faster for a higher temperature. During the anneal, the interstitial concentration in the vicinity of large defects is lower than





55


that in the vicinity of small defects. As a result, interstitials are emitted from the small defects and move toward the large ones. The large defects then absorb these interstitials and grow at the expense of the shrinkage of the small defects. This is the so-called Ostwald ripening process. Meanwhile, since the interstitial concentration in the area surrounding a defect is higher than that in the background environment, interstitials tend to leave the defect and move into the bulk. As a result, defect shrink and system approaches equilibrium condition. This is the defect dissolution process. During the Ostwald ripening process, defect density is expected to decrease, defect size is expected to increase and the total number of interstitials trapped in the defects is expected to remain as a constant. On the other hand, during the dissolution process, all three quantities are expected to decrease. In our experiment, we observed a decreasing interstitial concentration, an increasing defect size and a decreasing defect density with increasing annealing time. This indicates an Ostwald ripening process along with dissolution process. The distribution plots for longer time anneals (e.g., 8000C 13 min) show considerable scatter rather than a regular distribution because the total number of defects is small and the counting results are not as representative as those of the short time anneals.

The driving force for the Ostwald ripening process is

the self-energy difference between the defects with different





56


sizes. Self-energy of the {311} defect was calculated by Parisini and Bourret76 by approximating the rod-like shape with a dipole of edge dislocations of the same sense and opposite Burgers vector. It is given by
W = 2 [In( ) + cos20]OL (3.2)
27r(-v) 2p
where g is the shear modulus, b the Burgers vector, V the Possions ratio, q the distance between the two edge dislocations (i.e., the width of the {311} defect), p=b/8 the cut-off distance, 0 the angle between the Burgers vector and the normal to the plane containing the {311} defect and L the length of the {311} defect. To obtain the energy per interstitial, one can divide the previous expression by the number of interstitials N trapped in a {311) defect with a length of L. N is given by

N= p,Ly (3.3) where pl is the atomic density on the {311} plane. The energy of one interstitial is thus given by

W, = [In( ) + cos20]---4
2(1- v) 2p p

As we can see from eq.(3.4), WI is independent of the length L but is dependent on the width 1. Since ln(1/p) decays slower than n11, WI decreases when I increases. We thus speculate that the width of a shorter {311} defect is smaller than that of a longer one. As a result, the self-energy of an interstitial trapped in a shorter {311) defect is higher than that in a longer one. Interstitials would thus flow from shorter rods to longer ones. This explains the Ostwald ripening process.




57


In fact we have observed the {311) defect width change under the TEM. PTEM pictures were taken on the <311> zone with g=220. The following trend was found: the width is about 20 A when the length L is 200-400 A. It increases to about 40 A when L is around 1000 A, to about 80 A when the L is around 1200-1400 A, then to about 100 A for L=1600~1800 A and about 120 A when L~3000 A. These measurements were performed on samples annealed at 750'C for 5 min, 15 min and 1 hr. The error range might be large because of the narrow width of these defects. However, the width change with length can be used to account for the interstitial flow during the coarsening process. It should be mentioned that in this study this width change is ignored when the areal density of interstitials trapped in (311} defects was measured from the length of {311) defects and the interstitial density per unit length. In fact it is almost impossible to measure both the width and the length of every defect to obtain a more accurate result.

In summary, in this section of Chapter 3 we have

discussed the {311} defect evolution process during anneals at 6500C to 8000C for various times. The areal density of interstitials bound by (311} defects decays exponentially with annealing time with a characteristic time constant which is larger at lower temperatures. This dependence has an activation energy of 3.8 eV. The number of interstitial trapped in (311} defects is only 15-20% of the implant dose, implying the presence of boron interstitial complexes,




58


including both immobile clusters and mobile boron interstitial pairs. The defect density decreases with increasing annealing time and the average defect size increases initially and then decreases with time. The distribution of {311) defects shows a shift of the average defect size to a larger value and a decrease of density at each size. The defect evolution shows the characteristics of Ostwald ripening process accompanied by a defect dissolution process.


3.3 Effect of Annealing on Transient Enhanced Diffusion



3.3.1 Experimental Procedure


The same wafer (20 keV 2x1014 cm-2 B+ implant) which we used to study the effect of annealing on defect evolution was also used for diffusion study. Subsequent anneals were again performed in a nitrogen ambient at 6500C, 7000C, 7500C and 8000C for various times. SIMS analysis was performed on the CAMECA IMS4f system. B+ ions were monitored under 02+ bombardment at an impact angle of about 420. Secondary ions were collected from the center 12% of a 150 pm x 150 pm rastered area to avoid edge effects. Atomic concentrations were calculated using a relative sensitivity factor (RSF) for boron determined from an implanted standard. Stylus profilometry was used to calibrate the depth scale for the profiles. The dopant profiles were then imported into




59


Florida Object Oriented Process Simulator (FLOOPS) and the diffusivity enhancement was extracted for each anneal.


3.3.2 Dopant Diffusion during Annealing


3.3.2.1 Static Peak at High Concentration

As an example of the dopant diffusion behavior during

annealing, Fig.3.7 (a)~(d) shows the boron profiles obtained through SIMS after anneals at 6500C for 2 hr, 14 hr, 24 hr, at 7000C for 30 min, 2 hr and 4 hr, at 7500C for 20 min, 1 hr, 2 hr and at 8000C for 3 min, 30 min and 1 hr. We can see that a critical concentration exists under every annealing condition, separating the low concentration tail region where transient diffusion occurs, from the high concentration peak region where the profile is relatively immobile. The breaking point between the mobile and immobile region (Cenh) moves to higher concentration when annealing temperature increases. It is about 2.5x1018 cm-3 at 7000C and moves up to about 7x1018 cm-3 at 8000C. Figure 3.8 shows Cenh as a function of inverse temperature for our study and some of the previous studies. Also plotted in Fig. 3.8 are the solid solubility for boron in silicon Cs and the intrinsic carrier concentration ni.

There are several groups which have reported the peak clustering behavior of an implant profile. Michel et al.126 studied 60 keV 2x1014 cm-2 B+ implanted silicon. Over an annealing temperature range of 8000C-950'C, their measurements show that both the maximum concentration of the





60


displaced boron and the electrical activity are related to the intrinsic carrier concentration. Later, Fair127 has extended the correlation between Cenh and ni to the temperature range of 6000C-8000C. It was observed that when T>8500C, Cenh approaches B solid solubility Cs. A model was proposed by Fair, suggesting that the existence of Cenh is due to the Fermi level and the charge state of the defect affecting diffusion. For boron diffusion the location of the donor level of self-interstitials I+ in the energy gap causes Cenh to equal ni for T<8000C. Under this condition, a significant fraction of the excess interstitials are in the neutral charge state Ix. For doping under Cenh-ni, diffusion is dominated by Ix. Since DIX/DI+ is 10-100, more enhanced diffusion occurs at C850aC, the EI+ level is sufficiently above EF and the majority of injected interstitial exist in I+ state. Then the thermally assisted, concentration dependent diffusion via I+ becomes dominant and Cenh is limited by B solid solubility Cs. The study of Cowern et al.128 on 25 keV 2x1014 cm-2 B+ implants agreed with Fair's results. Cenh lies within a factor of 2 of ni in a temperature range of 5500C-9000C. A relative insensitivity of Cenh to the implant energy and dose was suggested by all three groups. As shown in Fig. 3.8, our data of 20 keV 2x1014 cm-2 B+ implants also agree with those of the above groups to a reasonable degree within 6500C-800aC. However, the data of Zhang et al. for 4 keV 1x1014 cm-2 B+ implants"114 do not match the intrinsic carrier concentration but are




61


rather higher than that, especially at higher temperatures (T>7000C). This mismatch can not be explained by Fair's model. We therefore speculate that the correlation between Cenh and ni might just be a coincidence. A better explanation is that clustering effect induces an immobile fraction of the profile. As for the kinetics of peak clustering, Cowern et al.128 suggested the following reactions:

Bs + I <-> BI (3.5) mBs + nI <-> IDC (3.6) where the intermediate defect configurations (IDCs), which was originally proposed by Tan,56 might include boron pairs or clusters, rodlike (311} defects, certain stacking faults, etc. Silicon interstitials (I) pair or kick out substitutional boron atoms (Bs) to form migrating interstitial boron atoms (BI). During the transient phase, the concentration of BI is high enough to allow homogeneous nucleation of boron containing defects. The defect formation is most favorable near the peak region where the B concentration is highest. Stolk et al.113 proposed another clustering reaction given by

Bs + mBI <-> Bc + nI (3.7) where Bc refers to immobile clusters containing m+l boron atoms, and n is the number of interstitials injected upon clustering to allow stress relief (n



62


available for diffusion at high B and I concentrations. Comparing eq.(3.6) and (3.7) will raise the question of whether the immobile B-I complexes contain Bs-BI-I clusters or Bs-Bs-I clusters. Diaz de la Rubia et al. (private communication) have shown through ab initio calculation that the BI-Bs pair has a positive binding energy of about 1.8 eV. This BI-Bs pair also traps silicon interstitial. It is immobile and electrically inactive. The Bs-Bs pair, on the other hand, is found to be energetically unfavorable with a binding energy of -1.7 eV and electrically active. According to this calculation, the immobile, electrically inactive implant profile peak might be consisted of the BI-Bs-I clusters.

It should be mentioned that rod-like {311} defects also locate in a depth corresponding to the profile peak region, so do the sub-amorphization dislocation loops. However, the {311} defects113 dissolve on a time scale comparable to the transient diffusion (e.g. <30 min at 8000C), while the static B-I clusters dissolve much slower (e.g. > 4 hr at 8000C128). {311} defects and unstable loops contribute to transient enhanced diffusion before it saturates. When the clusters and stable loops finally break up, there will be a second burst of enhanced diffusion. The slight increase of Cenh with increasing temperature is due to a decreasing interstitial supersaturation level which results in less clustering or even breaks up some existing clusters. There is no experimental evidence on how many interstitials are





63


trapped in one Bs-Bx cluster. In our 20 keV 2x1014 cm-2 B+ implant, after an anneal at 7500C for 2 hr, {311} defects almost dissolved completely. The density of activated boron atoms is estimated to be < 7x1013 cm-2 by integrating the non-clustered region of the implant profile. Assuming a valid "plus 1" model and assuming that one activated B atom pairs with one interstitial, the number of interstitials trapped in the B-I clusters must exceed 1.3x1014 cm-2. Thus we speculate that one Bs-B, cluster might trap more than two interstitials.

The annealing time scale we used in this study is much shorter than that is needed to break up the clusters and stable loops, so our focus is on TED at low concentration after relatively short anneals.

3.3.2.2 Broadened Tail at Low Concentration

From Fig. 3.7 we notice that significant diffusion

occurred in the tail region (C




64


2 hr at 7000C. Lower temperature annealing causes more diffusion before the saturation point is reached.

The SIMS profiles of boron in the as-implanted and annealed samples were imported into FLOOPS and the enhancement of diffusivity /DB* after each anneal was obtained after the simulations. Here is the timeaveraged boron diffusivity and DB* is the intrinsic equilibrium value. DB* has the form of

DB*=0.757exp(-3.46eV/kT) (3.8)

as a default value in FLOOPS, which was originally taken from Fair's review article29. This value is about

1.03x10-19 cm2/sec, 9.64x10-19 cm2/sec, 7.22x10-18 cm2/sec and 4.48x10-17 cm2/sec at 6500C, 700'C, 7500C and 8000C, respectively. Since the dopant profiles are not perfectly Gaussian, instead of choosing the whole profile, we chose a portion of the tail region (approximately 1600 A-2600 A) as the target profile during simulation. Figure 3.9 shows these simulation results of /DB* as a function of annealing time for each temperature. As we can see, the time-averaged diffusivity enhancement decreases with increasing annealing time. Since /DB* is proportional to the interstitial supersaturation level, /CI*, in the case of B diffusion, the decrease of /DB* represents a process during which the interstitial concentration returns to its equilibrium value. Here is the time-averaged interstitial concentration in the bulk, and CI* is the equilibrium value. CI* is smaller and /CI* is higher at lower temperatures





65


so that /DB* is higher. It requires longer time for the system to reach equilibrium at a lower temperature. However, the time point at which the equilibrium is reached cannot be read directly from the /DB* graph mainly because they are the time-averaged values. This time will be determined from the diffusion length vs. time plot as will be presented in the following section, from which we will derive the activation energy for the diffusion process.


3.3.3 The Activation Energyav for the Diffusion Saturation
Process


The diffusion length can be calculated as (t) from the diffusivity enhancement, the intrinsic diffusivity DB* and the annealing time at each temperature. Figure 3.10 summarizes the results. The dots in Fig. 3.10 are the actual data points and the lines are the fitting lines in the form of aq(l-exp(-t/T)), where a and T are constants and t is the annealing time. This is obtained by assuming the instantaneous value of DB is proportional to exp(-t/T), and the diffusion length L of boron is the square root of the time integration of DB, i.e.,
I
D,(t) ce r (3.9)

L= t= D((t)dt,
Jed, c __(3.10) = oe 'dt (1-e )

The diffusion length increases initially with time and then plateaus out. The time till the "knee" region, i.e., just before it plateaus out is 2T. If we choose 2T as the time





66


constant for the TED saturation process, from Fig. 3.10 we can see that this constant increases dramatically as annealing temperature decreases. It is about 4.5 hr at 6500C and decreases more than one order of magnitude to about 15min at 8000C.

The temperature dependence of the time constant for the TED saturation process is shown in Fig. 3.11. Such a dependence exhibits an activation energy of about 1.6eV. This value is smaller than the 3.1-3.7eV reported by Packan,130 3.5 eV reported by Stolk et al.113 yet it is greater than the < 1 eV we calculated from the data of Zhang et al.114 as shown together in Fig. 3.10. The difference between these experiments is that both Packan and Stolk et al. studied silicon self-implanted silicon. In these samples the interstitial source for TED is believed to be only {311} defects. Zhang et al. studied low energy (4 keV) boron implanted silicon where the energy and dose combination resulted in no extended defects observed by TEM. The source of interstitials driving TED in this case was claimed to be sub-microscopic clusters, which may be as simple as just mobile B-I pairs that diffuse until the pair breaks up. In our case, the samples were implanted with boron, but at a higher energy and dose so that not only {311} defects are present, B-I pairs might also exist. The activation energy of the TED saturation process reflects a complex process, including the release of interstitials from B-I pairs and {311} defects, the migration of interstitials through the





67


bulk, the surface recombination effect, impurity trapping, etc. Variation in the activation energy is thus not unexpected in these different systems. However, we still believe that in our case, not only {311} defects play a role in terms of releasing interstitials during annealing and thereby driving TED, boron interstitial pairs also play an important role.

In summary, 20 keV 2x1014 cm-2 boron implanted silicon was studied for the purpose of identifying the possible sources of interstitials driving TED. The static peak region was presumably consisted of immobile BI-Bs-I clusters which breaks up in times much longer than normal TED. Boron TED in the tail region was measured through SIMS analysis and FLOOPS simulation. The duration of TED increases dramatically when the annealing temperature decreases. The activation energy for the TED saturation process is about 1.6 eV, higher than that in the sample presumably with only B-I pairs contributing to TED and lower than that in samples with only {311} defects. The interstitials driving TED in our sample are believed to come from both mobile boron interstitial pairs and {311) defects. The dissolution of {311} defects and B-I pairs may interact with each other and result in a single activation energy lying in between.








Q-2=






2hr 6hr








00


14hr 24hr






(a)
Figure 3.1. Weak beam dark field (g220) PTEM micrographs of {311} defects in 20keV
2x1014cm-2 B+ implanted Si after anneals at (a) 650*C, (b) 7000C, (c) 7500C and (d) 8000C for various times.













lhr 2hr











4hr 8hr






(b)
Figure 3.1. (Continued)













5min
5min 20min











30min lhr






(c)
Figure 3.1. (Continued)














3min 5min












10min 13min






(d)

Figure 3.1. (Continued)





72










I I I I II Iill I I I 5I l I 1 5 I ilit

C,,
E "+1"
14 10




too
0U
c,

.00

0 012 ---800C S---o--- 7500C
i a -*- 700'C
6500C
4'
-- 1011 ... I. I. ....II I I I. ... ll I . ... l
1046
100 1000 104 10s 106
Annealing Time (sec)









Figure 3.2. The exponential decay of interstitials trapped in (311} defects in 20keV B+ implanted Si after anneals at 6500C to 8000C for various times.




73









Temperature (oC) 10838 780 727 679 636
10' II 1 I I
---o--- 20 keV B 2e14
1--o---40 keV Si 5e13* 02 .


S10-3 3.6 eV


13.8 eV


10-5



106 I I "
9 9.5 10 10.5 11 104'/T (K-')








Figure 3.3. Rate of {311} dissolution process and the activation energy.




74









1011


N1
'E

S1010
(- )
2 \I


10'
i.. ----.. 800oc
8001
S- -v- 7500C
-- 7000C
-6-- 6500C
o
1 0 8 . .. .... . ..... . . .... .,... .
10 103 104 10s 106 Annealing Time (sec)








Figure 3.4. {311} defect density as a function of annealing time in 20keV B+ implanted Si for different annealing conditions.




75










103
? ------ 8000C
-- -7500C -- 7000C
( -- 6500C
N





Cl) V
10' ** **'V * * ** **
C.)
a) I
Q ,



I1

0)2



102 103 104 10s 106

Annealing Time (sec)








Figure 3.5. Average {311) defect size as a function of annealing time in 20keV B+ implanted Si for different annealing conditions.




76



12
11 ....___ _CE 10 6500C6hr
,E 1011
1010
a ) 109
O 10 60 260 460 660 860 106012601460


C 10'
E 101 6500C14hr
a 10

a 10
0 1 IAEI .a ... I ... I *l... l *..IM122
S 60 260 460 660 860 106012601460



E 10 6500C36hr
t 0 10



% & 1010
0




109L ...... ..L...
10 60 260 460 660 860 1060 1260 1460
10o12
-.1E 011 650C48hr
2: 101
D -- 0
(D108LR .
0 60 260 460 660 860 1060 1260 1460

Average {311} Size (A)
(a)
Figure 3.6. {311} defect size distribution during annealing at (a) 6500C, (b) 7000C, (c) 7500C and (d) 8000C.




77



111
101 7000C30min E 10 ,- 010 c : 109

O 10 60 260 460 660 860 1060 1260 1460



S10 7000C60min
E 10 10




10IAIIIE// 44l EE,, 41 IO
,loI ,10 60 260 460 660 860 1060 1260 1460


1011
S 010 7000C4hr E.. 10 2) 109



.55 8c 10
0)
Q 10 60 260 460 660 860 1060 1260 1460


E 10 7000C8hr E a 1010

(D 7I... Aa I.
0 10 60 260 460 660 860 1060 1260 1460
Average {311} Size (A)
(b)
Figure 3.6. (Continued)




78



1011
cE 101 7500C5min


Q 107 .r. I S10 60 260 460 660 860 1060 1260 1460



t 1011
E 10 7500C20min
C- 0 10 r
" E 10
a) 7 _ _ _ _ _ _ _ _ _
a 10 60 260 460 660 860 1060 1260 1460



101
E 10i 7500C30min
>, 1o0 m m=9.
6 108
0101 60 260 460 660 860 1060 1260 1460



c10 7500C60min
E 1010 .
o
c 108 TR ME 1 08
O 10 60 260 460 660 860 106012601460
Average {31 1} Size (A)
(c)
Figure 3.6. (Continued)




79



101
E 10 8000C3min
5 0 10 E 10- 101 h

S107
S10 60 260 460 660 860 1060 1260 1460

---101
E 101 8000C5min .-E 109


10
160 260 460 660 860 106012601460

^ 101
'E 0o 8000C10min .0 109
10 -- 0 8 . I . . 1 .




60 260 460 660 860 1060 1260 1460



E 101 8000C13min D 100
c: 10 r

7 I . I . I . I .. I. .
60 260 460 660 860 1060 1260 1460
Average {311} Size (A)

(d)
Figure 3.6. (Continued)




80



1020 l Ila
102 .......... _________________--1-- as-implanted E 1- 2hr
Q --.--14hr
1018 24hr
.o

10 .
ca~

C 16
10

0 101s
0
c

10 14 I I* i i a s i a s i s e i a s i .
0 1000200030004000 5000 6000 7000 8000 Depth (A)
(a)


1 0 2 0 I l l 1tl l
-0 9 --- as-implanted
E ------ 30 min
-10 --*--2hr
c 101
0 O
.o - -A 4 h r

17
S10

0 ol
0 1016

c0 1 0 s
oO
0
1014

0 1000200030004000 5000 6000 7000 8000 Depth (A)
(b)
Figure 3.7. Dopant SIMS profiles of 20keV 2x1014cm-2 B+ implanted Si after anneals at (a) 6500C, (b) 7000C, (c) 7500C and (d) 8000C for various times.




81



1020
--- as-implanted 1019 E 1 -v--- 20min ..W --*--lhr
C 1018 2hr

1017
0
C 16 t-0 1


2 1 615
0 10
0
1014

0 1000 2000 3000 4000 5000 6000 7000 8000 Depth (A)

(c)

1020
--.- as-implanted
19
E 10 --V--- 3min S -- 30min c' 018 h
0 1 1hr
o
17 01 1017 o 1016 mm
2 15

1014 "** i .. I s l. . .
0 10002000300040005000 6000 7000 8000 Depth (A)

(d)
Figure 3.7. (Continued)




82









a0 m this study
102 v Cowern
Michel
--A--Zhang
a Fair
ni
"Cs
E
1019





1018


8 8.5 9 9.5 10 10.5 11
10000/T (K)







Figure 3.8. The breaking point between the mobile profile tail and the immobile profile peak Cenh as a function of inverse temperature.





83












410 4 .. .. .I ..
--- 800oC
S-~V 750*C
0 -*- 700C
V A
S ___ 6500C

C (1)
E
103





o
c 4


C
0
M 100 .. * "
100 1000 10000 100000
Annealing Time (sec)








Figure 3.9. Time-averaged Boron diffusivity enhancement /DB* as a function of annealing time for different annealing temperatures.




84










10 . .... . . . .
--m-- 8000C
750*C
-- -7001C
--A--650oC

0A



0
Co -1 0 0 . , ] fi*i a ,
100 1000 10000 100000
Annealing Time (sec)






Figure 3.10. Boron diffusion length vs. annealing time for different annealing temperatures.





85









10
-- this study
UJ
-- ---Packan lel4 3.7eV
o -- -Packan 5e13
2 - Zhang w/TED1.6eV
0- 1.6eV
.- g - Zhang w/o TED
C 04___o 10











E 100 . , 1.. .,
Ca) W









9 8.5 9 9.5 10 10.5 11
1010000/T(K)

















Figure 3.11. Time constant and activation energy of TED saturation process, comparison with previous studies.
C
C
o A

E 100 1
I- 8.5 9 9.5 10 10.5 11

1 0000/T(K)












Figure 3.11. Time constant and activation energy of TED saturation process, comparison with previous studies.













CHAPTER 4
THE INFLUENCE OF IMPLANT ENERGY ON DEFECT BEHAVIOR AND
TRANSIENT ENHANCED DIFFUSION


4.1 Overview


It is believed that transient enhanced diffusion is a direct result of the damage created by the implantation process. Characteristics of the implanted ions (such as the mass), the substrate conditions (such as crystalline versus amorphous), and variations in implantation parameters (such as energy, dose, dose rate and temperature) would all cause variations in the damage and therefore affect dopant diffusion. This section discusses the effect of implant energy on defect behavior and TED.

It has been claimed that the amount of damage created by the implantation process is strongly dependent on the dose and the mass of the implant species but only weakly dependent on the implant energy. 42 Although higher energy does mean more kinetic energy per implanted ion, the resulting implant profile is more spread out and thus the "energy density" deposited per volume is lessened by this dilution effect. The effect of increasing energy is thus weakened and the number of interstitial-vacancy pairs would not be increased much provided the dose is the same. However, increasing energy can indeed change the location of the damage and the


86





87


depth of the interstitial and vacancy profiles and the separation between these two profiles. These changes can affect the defect behavior and TED, as will be discussed later.


4.2 Implant Energy Effect on Defect Behavior



4.2.1 Experimental Procedure


A single row of the implant matrix discussed in Chapter

2 was used to study the implant energy on {311} defect evolution. The implant dose was kept at 2x1014 cm-2 and the energy was varied from 5 key to 40 keV. Subsequent furnace anneals were conducted in a nitrogen ambient for times between 5 min and 1 hr at 7500C. The {311) defect density, average defect size and, areal density of interstitials trapped by {311) defects and the defect distribution during annealing were measured from the PTEM micrographs.


4.2.2 Defect Microstructure and Dissolution Process


PTEM micrographs of the defects after implantation at

10 keV to 40 keV and anneals at 750'C from 5 min to 1 hr are shown in Fig. 4.1(a)-(d). No defects were observed for the

5 keV implant and therefore the pictures for those samples are not included. {311} defects are the only type of extended defects observed under the TEM in the 10 keV and 20 keV implants. A majority of {311} defects together with just a few dislocation loops are observed in the 30 keV and 40 key





88


implants. For each of those implants showing {311} defects, the number of the defects decreases with increasing annealing time. This dissolution process is more rapid for a lower implant energy.

The areal density of interstitials bound by (311}

defects during the annealing process was measured from the PTEM micrographs and the results are presented in Fig. 4.2. The dots in the plot are the actual data points and the lines are the exponential fitting curves for the data with the form of eq. (3.1). The emission of interstitials from (311} defects clearly shows an exponential time-dependence, which is stronger for lower implant energies. After a 30 min anneal at 7500C, {311) defects in the 10 keV implant almost dissolved completely. The density of the remaining trapped interstitials is about 1.8x1011 cm-2. For the 40 keV implant, after an anneal of 1 hr, there is still an areal interstitial density of about 1.2x1012 cm-2 contained in {311} defects. The maximum density of trapped interstitials observed was about 2x1013 cm-2 for the 10 keV implant and about 3.5x1013 cm-2 for the 40 keV implant after an anneal for 5 min and these values are much lesser than the implant dose of 2x1014 cm-2. This implies the existence of immobile boron interstitial clusters and/or mobile boron interstitial pairs which might have trapped the missing interstitials assuming that the "plus one" model holds. There is, however, another possible configuration that could consume interstitials, i.e., silicon self-interstitial complex--two





89


interstitials split along <110> direction occupying one lattice site. According the ab initio calculations conducted by Diaz de la Rubia et al. (private communication) this configuration is energetically unfavorable with a formation energy of about 3.7 eV. Interstitials are more likely to pair with boron atoms because the binding energy of a Bs-I pair is about 1.1 eV.

The lack of {311) defects in the 5 keV sample and the

lower density of interstitials trapped by {311) defects in a lower energy implant could be attributed to at least the following four factors: (1) Higher energy might introduce more as-implanted damage which may further induce more trapped interstitials in extended defects (if formed) after annealing. (2) Higher energy causes larger separation between the as-implant vacancy profile (closer to the surface) and the interstitial profile (further away from the surface) because of the larger forward momentum carried by the implanted ions. The result of this might be that the I-V recombination rate is lower and the number of interstitials available to form extended defects is higher. (3) According to a series of proposed defect reactions113 which describe the generation and clustering of mobile boron atoms given by

eq.(3.5) and (3.7),

Bs+I <-> BI (3.5) Bs+mBI <-> Bc+nI (3.7) at a lower implant energy, the concentrations of Bs or BI and silicon interstitials in the implant peak region are higher,




90


which favors the formation of boron clusters which bind excess interstitials. This will reduce the total number of interstitials available for forming (311} defects. (4) The surface recombination effect is more efficient for a lower energy when the damage region is closer to the surface. More interstitials are drawn to the surface and less are available for defect formation during annealing.

If the time constant of the {311) defect dissolution process is chosen as the time required for the number of interstitials to decrease to 1/e of its maximum value, then it would be about 5 min for the 10 keV implant and 16 min for the 40 keV implant. This time constant is plotted in Fig.

4.3 as a function of energy. It is interesting to note that, if we derive from Fig. 4.3 the time constant of the 5 keV implant, it would be about 2.6 min. If {311} defects existed in the 5 keV implant from the beginning, we should be able to see them after a 5 min anneal, even though the defect density would be low, yet we do not see any defects. This could be attributed to one or a combination of the factors discussed above.


4.2.3 Defect Density, Size and Distribution during Annealing


The {311} defect density for different implants as a

function of annealing time measured from PTEM images is shown in Fig. 4.4. If the annealing time is short (e.g., less than 10 min), the lower energy implant shows a slightly higher density of {311) defects. This is probably due to earlier




91


formation and coarsening of {311} defects in the higher energy implants. However, the defects dissolve faster for a lower energy. As annealing time increases further, defect density is higher for a higher energy implant. The difference in defect density becomes more significant when the annealing time increases. After a 30 min anneal, the 10 keV implant shows a defect density of only 3.2x108 cm-2, while the 40 keV implant shows a density of 2.7x109 cm-2.

The average {311) defect size as a function of annealing time for different implant energies is shown in Fig. 4.5. The average defect size increases with annealing time initially, and then decreases when the defects enter a purely dissolution regime. This final dissolution regime is only observed for the 10keV implant for the current annealing times. As the implant energy increases, the average defect size increases. After an anneal of 15 min, the 20 key implant shows an average defect size of about 518 A, while for the 40 keV implant, it is about 952 A.

Figure 4.6 (a)-(d) show the size distribution of {311) defects after anneals at 7500C for various times for 10 keV to 40 keV implants. Higher energy implants have a broader distribution than lower energy implants. The figure also shows a larger average defect size and a slower dissolution rate for a higher energy. For each energy, the average defect size increases while the defect density at each size decreases during annealing. This behavior together with the fact that the total number of interstitials trapped in (3111





92


defect decreases with increasing time indicates that the Ostwald ripening process occurs along with the defect dissolution process. Large (311} defects grow by absorbing the interstitials released from small defects as a result of higher local interstitial concentration around small defects. Meanwhile, large defects themselves release interstitials into the bulk. This resulted in the final defect distribution we have observed.

As a summary, the {311) defect microstructure and

annealing behavior as a function of implant energy (5 key ~ 40 keV) have been studied using TEM. No defects are observed in the 5 keV implant. There are more interstitials bound by {311} defects in higher energy implants although the maximum density observed (about 2-3.5x1013 cm-2 after an anneal of

5 min at 7500C) are all far less than the implant dose (2x1014 cm-2). The defect density is higher for a lower implant energy after a short anneal time (10 min) and then becomes lower as time increases. The average defect size is larger for a higher implant energy. The size distribution of defects shows a co-occurrence of the Ostwald ripening process and the dissolution process. The lack of {311) defects in the 5 keV implant and the decrease in density of trapped interstitials by (311) defects with decreasing implant energy might be the result of one or a combination of the following factors: decreased damage product with decreasing energy, increased I-V recombinations, increased boron interstitial cluster formation and an increased surface recombination.




Full Text
133
Bt = Bt KeqI/(1+KeqI) (5.3)
The above equation shows that for large damage dose, Br
approaches a maximum value of BT. Further increase in the
dose will not produce more B!. Effectively all the B is
paired and further increase in I do not enhance Db.
At this time we do not know which mechanism, the
loop/cluster trapping of interstitials or dopant interstitial
pair saturation, is valid in our case. Quantification of the
interstitials in the clustered region will shed light on this
issue. If the extra interstitials due to the dose increase
from 2x1()14 cm2 to lxlO^^ cm2 are all trapped in the
loops/clusters, then the first mechanism might dominate.
Otherwise, we should also consider the second mechanism. For
our APCVD grown boron spike, the area integration under the
spike shows a dose of lxlC)13 cm_2. Because this is only 5%
of the 2x10^4 cm"2 dose and 1% of the lxlO^^ cm"2 dose, it is
quite possible that dopant interstitial pair saturation is
reached. But we also observed more clustering for the higher
dose in Fig. 5.7 and the trap of interstitials by loops in
Fig. 5.4. Both mechanisms might function together and either
one of them could give rise to the current diffusion
behavior.
The diffusion length of the buried boron spike V(t)
as a function of annealing time for each implant is shown in
Fig. 5.10. The dots are the data points and the lines are
the fits in the form of aV (1-exp (-t/T)) where a and x are
constants and t is the annealing time (please refer to


58
including both immobile clusters and mobile boron
interstitial pairs. The defect density decreases with
increasing annealing time and the average defect size
increases initially and then decreases with time. The
distribution of {311} defects shows a shift of the average
defect size to a larger value and a decrease of density at
each size. The defect evolution shows the characteristics of
Ostwald ripening process accompanied by a defect dissolution
process.
3.3 Effect of Annealing on Transient Enhanced Diffusion
3.3.1 Experimental Procedure
The same wafer (20 keV 2x10^-^ cm~2 b+ implant) which we
used to study the effect of annealing on defect evolution was
also used for diffusion study. Subsequent anneals were again
performed in a nitrogen ambient at 650C, 700C, 750C and
800C for various times. SIMS analysis was performed on the
CAMECA IMS4f system. B+ ions were monitored under C>2 +
bombardment at an impact angle of about 42. Secondary ions
were collected from the center 12% of a 150 |im x 150 |im
rastered area to avoid edge effects. Atomic concentrations
were calculated using a relative sensitivity factor (RSF) for
boron determined from an implanted standard. Stylus
profilometry was used to calibrate the depth scale for the
profiles. The dopant profiles were then imported into


3.3.2 Dopant Diffusion during Annealing 59
3.3.2.1 Static Peak at High Concentration 59
3.3.2.2 Broadened Tail at Low Concentration 63
3.3.3 The Activation Energy for the Diffusion
Saturation Process 65
4 THE INFLUENCE OF IMPLANT ENERGY ON DEFECT BEHAVIOR
AND TRANSIENT ENHANCED DIFFUSION 86
4.1 Overview 86
4.2 Implant Energy Effect on Defect Behavior 87
4.2.1 Experimental Procedures 87
4.2.2 Defect Microstructure and Dissolution
Process 87
4.2.3 Defect Density, Size and Size Distribution
during Annealing 90
4.3 Implant Energy Effect on Transient Enhanced
Diffusion 93
4.3.1 Experimental Procedures 93
4.3.2 Dopant Diffusion Behavior 94
4.3.3 Correlation Between Defects and TED 96
5 THE INFLUENCE OF IMPLANT DOSE ON DEFECT BEHAVIOR AND
TRANSIENT ENHANCED DIFFUSION 120
5.1 Overview 120
5.2 Implant Dose Effect on Defect Behavior 122
5.2.1 Experimental Procedures 122
5.2.2 Defect Microstructure and Dissolution
Process 122
5.2.3 Defect Density, Size and Size Distribution
during Annealing 126
5.3 Implant Dose Effect on Transient Enhanced
Diffusion 129
5.3.1 Experimental Procedures 129
5.3.2 Dopant Diffusion Behavior 130
5.3.3 TED Saturation at Higher Dose 132
6 SUMMARY AND FUTURE WORK 151
6.1 Summary 151
6.2 Future Work 158
LIST OF REFERENCES 163
BIOGRAPHICAL SKETCH 171
vi


Defect Defect Defect Defect
Density (cm'2) Density (cm'2) Density (cm'2) Density (cm'2)
112
triiiiii i |i-1 i| i ir-i
40keV 5min =
a ""'"'"""'Tniingiir 11 i i i i i i i .
10 60 260 460 660 860 1060 1260 1460
1011
1010
109
108
107
60 260 460 660 860 1060 1260 1460
1011
101
109
108
107
T 1I 1II 1| 1I 1
40keV 30min
60 260 460 660 860 1060 1260 1460
j oil
IU r 1 TTrrTTl 1"1 I 1 1 1T '1 I 111 I 1 1 1 I 1 1
p 40keV 1 hrj
_ili i.,!)i..i.illi .run rali i
60 260 460 660 860 1060 1260 1460
{311} Defect Size ()
(d)
Figure 4.6. (Continued)


127
decreasing interstitial density (Fig. 5.2) and decreasing
defect density are the characteristics of defect dissolution
process. The two coupled processes give rise to the defect
distribution shown in Fig. 5.7. If we compare Fig. 5.7(a)
and (b) with Fig. 4.6(b) (the size distribution of
2xl014 cm-2 implant), we see that the standard deviation of
the size distribution increases as dose increases. This
again might be due to the larger number of defect nuclei in
the sample implanted with higher dose. Some nuclei might
grow faster than the others which results in more possible
defect sizes than the sample with fewer nuclei.
As a summary, the {311} defect microstructure and
annealing behavior as a function of implant dose (5xl013 cm-2
~ lxlC)15 cm-2) have been studied using TEM. After anneals at
750C for times ranging from 5 min to 2 hr, no {311} defects
are observed in the 5x10-*--^ cm-2 implant, a few in the
lxlO-*-4 cm-2 implant and more in the 2xl0-*-4 cm-2 implant. A
majority of {311} defects and a few dislocation loops are
present at the higher doses. Assuming a valid "plus 1"
model, the number of interstitials bound by boron
interstitial complexes is estimated to be about 88% of the
implant dose for 2xl044 cm-2, 5xl0^-4 cm-2 and lxlO-4^ cm-2
implants after an anneal for 5 min. The {311} defect
dissolution process shows a smaller time constant for a lower
dose. Dislocation loops expedite {311} defect dissolution at
doses > 5xl014citT2 by acting as an effective interstitial
sink and therefore reducing the time constant increase rate


26
configuration. The extraction voltage ranges usually between
15 and 40kV. For low energy implants, a retarding voltage
can be used after separation. The selected ions are
accelerated by a static electric field and focused and shaped
in the column and ready to be implanted.
Some important quantities during the ion implantation
process are listed below. The implant dose can be expressed
as
where I is the implant current, t is the time, n=l for singly
ionized species and 2 for doubly ionized species, q is the
charge and A is the area of the implanted surface. For an
ion accelerated through a potential V and deflected through a
magnetic field B, the radius of the ion's path is
1 2MV
(1.8)
q
where M is the mass of the dopant atom. In absence of
crystal orientation effects, the range distribution is
roughly Gaussian and to a first approximation the projected
range distribution N(x) is described as a one-dimensional
Gaussian profile characterized by the projected range Rp and
the standard deviation ARp, as given by
(1.9)
Schematic diagram of an implanted impurity profile is shown
in Fig. 1.7. Due to the channeling effect, the use of a
Gaussian profile is not accurate. The UT-Marlowe code (a


16
(d)
Figure 1.5. (Continued)
The displacement vector of {311} defects is
perpendicular to the line and has a prominent [100]
component. This vector has been determined to be a<100>/2,57
a<411>/6,58 a<311>/ll52 or a<611>/k by various groups, where a
is the lattice constant and k=2559 or 31.60 The discrepancy
between these values reflects the inaccuracy of the
measurement because of the narrowness of the defects.
Different methods have been used to determine the
displacement vector, including the traditional image contrast
analysis and the energy minimization calculation based on the
relaxation of the interstitial agglomerate model under the
Stillinger-Weber potential.
{311} defects have been studied in various ion implanted
or particle irradiated systems for more than two decades.
They were found in silicon,61 boron, 57,62'64 arsenic,65


20
mechanism while X=V refers to a vacancy mechanism.
Mathematically the ratio of dopant diffusivity under
nonequilibrium conditions over its intrinsic, equilibrium
value can be written as
Cv
_ T1A\/
Da* Cj*
av
Cy*
(1.3)
where fAX=DAX*/A* (X=I or V) fAI+fAV=l/ CI and Cy are the
concentration of interstitials and vacancies under
nonequilibrium conditions, Ci* and Cy* are the quantities
under intrinsic, equilibrium conditions. Equation (1.3)
shows that the diffusion of dopant atoms consists of two
parts: diffusion by coupling with interstitials and with
vacancies. The fraction of each is represented by fAl and
fAV, respectively. While this treatment of diffusion allows
us to mathematically explain the experimental results, it is
possible that the actual mechanism is quite different.
Dopant diffusion in silicon has been under investigation
for many years; however, the diffusion mechanisms responsible
for some of the impurities are still under debate. It is
generally accepted that antimony diffusion is dominated by a
vacancy mechanism. Phosphorous and boron diffuse
predominately by an interstitial/interstitialcy mechanism.
Arsenic appears to diffuse by both vacancy and
interstitial/interstitialcy mechanisms. The diffusion of Al,
Ga and In is believed to have a strong interstitial
component77. Elements such as Li, Na, He and H are believed
to diffuse by an interstitial/interstitialcy mechanism as do


8
isolated amorphous pockets may form without overlapping into
a fully amorphized layer. Different damage levels will
result in different types of extended defects after post
implantation annealing.
Residual implantation damage can be visible in the TEM
or submicroscopic such as dopant clusters. Extended defects
visible in a TEM are categorized according to two formation
mechanisms: point defect condensation and solid phase
regrowth. 42,43 There are three types of extended defects that
arise from the first mechanism and two types from the second
mechanism, as will be discussed below.
I. Extended defects from point defect condensation
1. Sub-amorphization defects
When the implant damage is not sufficient to turn the
silicon surface into a continuous amorphous layer, sub-
amorphization defects form during subsequent annealing.
These defects arise from a supersaturation of point defects
due to the non-conservative nature of the implantation
process. They form around the projected range of the
implanted ions, where the supersaturation level is the
highest. These defects have two common forms: rod-like {311}
defects and dislocation loops. Sub-amorphization {311}
defects are one of the main interests of the present study
and they will be discussed in detail in the next section.
2. End-of-range defects
When the silicon surface has been amorphized during
implantation, end-of-range defects form just below the


9
amorphous/crystalline interface upon annealing. End-of-range
defects may consist of both dislocation loops and {311}
defects if the annealing temperature is low. These defects
are important in IC processing since amorphous layers are
always present for P, As and BF2 implants under the implant
conditions that are usually used. For more information about
the formation and evolution of these defects, please refer to
references 42-46.
3. Precipitation related defects
If the concentration of the implanted impurity is above
its solid solubility in silicon, precipitation related
defects may form around the projected range. The
precipitation process may generate such a high concentration
of intrinsic point defects that extended defects can form as
a result. The most common example of this is after high dose
arsenic implantation. 43,47,48 Two layers of dislocation loops
have been observed upon annealing, one at the projected range
which is associated with As clustering, the other at the end
of range which is associated with excess interstitials beyond
the amorphous/crystalline interface.
II. Extended defects from solid phase regrowth
1. Regrowth defects
When the solid phase epitaxial regrowth process is
imperfect, defects nucleate at the advancing
amorphous/crystalline interface and propagate into the
regrowing silicon. They do not appear to dramatically affect
the diffusivity of dopants, but they can getter impurities


15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
164
F. F. Morehead and R. T. Hodgson, Mat. Res. Soc. Svxno.
Proc. 35. 341 (1985).
J, F. Marchiando and J. Albers, J. Appl. Phvs. 61. 1380
(1987) .
T. 0. Sedgwick, A. E. Michel, V. R. Deline, S. A. Cohen,
and J. B. Lasky, J. AppI. Phvs. 63. 1452 (1988).
S. Solmi, R. Angelucci, F. Cembali, M. Servidori, and M.
Anderle, AppI. Phvs. Lett. 51. 331 (1987).
S. Nishikawa, A. Tanaka, and T. Yamaji, AppI. Phvs. Lett.
£0# 2270 (1992).
Q. Guo, X. Bao, J. Hong, Y. Yan, and D. Feng, AppI. Phvs.
Lett. 54. 1433 (1989).
C. S. Rafferty, H.-H. Vuong, S. A. Eshragi, M. D. Giles,
M. R. Pinto, and S. J. Hillenius, IEDM Technical Digest.
311 (1993).
M. C. Ozturk and J. J. Wortman, AppI. Phvs. Lett. 52. 963
(1988) .
M. Kase, M. Kimura, H. Mori, and T. Ogawa, AppI. Phvs.
Lett. 56. 1231 (1990).
A. Bousetta, J. A. van den Berg, D. G. Amour, and P. C.
Zalm, AppI. Phvs. Lett. 58. 1626 (1991).
0. M. Holland and J. Narayan, Nucl. Inst. Meth, B 40/41,
537 (1989).
R. G. Wilson, J. AppI. Phvs. 54. 6879 (1983).
E. Ganin and A. Marwick, Mat. Res. Soc. 147. 13 (1989) .
A. C. Ajmera and G. A. Rozgonyi, AppI. Phvs. Lett. 49.
1269 (1986).
L. Laanab, C. Bergaud, C. Bonafos, A. Martinez, and A.
Claverie, Nucl. Inst. Meth. B 96. 236 (1995) .
R. G. Taylor, C. Andre, T. Salama, and P. Ratnam, IEDM
Tech. Dig., p.644 (1988).
D. C. Chen, T. R. Cass, J. E. Turner, P. P. Merchant, and
K. Y. Chiu, IEEE Trans. Electron Devices ED-33. 1463
(1986) .


135
Later on these trapped interstitials are released from the
loops and make contribution to boron diffusion.
Interstitials are easier to escape the trap of dislocation
loops at higher temperatures and therefore it takes shorter
time to observe a larger diffusion at a higher dose.
Our results are also comparable with those of Solmi et
al.132. They used 20 keV to 30 keV B+ implantations to doses
from 2xl044 cm-2 to 5x10-*-^ cm-2 to study the diffusion of the
B implant itself. They observed a less displacement in the
profile tail region for the higher doses (> 5xl014 cm-2
compared with < 5xl044 cm-2) within a time range of 5min to
2hrs at 800C. The displacements of the 2xl015 cm-2 and the
5x10-*-5 cm-2 implants are very close. They attributed the
reduction of diffusion to the extended defects which might
act as a sink of the free interstitials.
In summary, 8-doping boron marker layer was used to
study dopant diffusion behavior after B+ implantation at 20
keV to various doses. This APCVD grown layer acted as a
sensitive diffusion monitor for diffusion because of the
steep dopant gradient. The doses chosen cover the transition
regions from sample with no {311} defects (5x10-*--^ cm-2), to
that with {311} defects (2xl014 cm-2), then to that with both
dislocation loops and {311} defects (1x10^-^ cm-2). The
lowest dose shows the lowest diffusion enhancement and the
smallest diffusion length for all the anneals. The diffusion
enhancement and the diffusion length saturate for doses above
2xl014 cm-2. There are two mechanisms which can both explain


53
silicon. It is interesting to notice that our activation
energy is in close agreement with theirs of 3.60.1 eV, yet
there is a y-axis shift of the two curves, i.e., the {311}
defects in our sample are more stable. The similar
activation energies implies the process of dissociation of an
interstitial from {311} defects is the same, independent of
the implant species. The reasons for the change in stability
could be: (1) The dose for our implant (2x1014 cm-2) is much
higher than what they used (5xl0-*-3 cm-2) and thus there are
more interstitials in our sample to stabilize {311} defects.
(2) The dissolution of boron interstitial clusters in our
sample during the anneal induces a higher concentration of
interstitials in the background environment and again delays
the dissolution of {311} defects.
3,2.2 Defect Evolution: Density, Size and Distribution
Quantitative results that one can obtain from the PTEM
micrographs include the {311} defect density, the average
defect length and the defect distribution after annealing.
Figure 3.4 shows the {311} defect density as a function
of annealing time for each temperature. The defect density
decreases dramatically with increasing annealing time.
Although the rate is just slightly higher for higher
temperature, it takes much longer for a lower temperature to
reach a particular density. For example, the density drops
to 109 cm-2 after about 10 min at 800C, but it takes more
than 10 hr at 700C. This rapid decrease in the number of


Diffusion Length (Dt)0'5 ()
84
Annealing Time (sec)
Figure 3.10. Boron diffusion length vs. annealing time for
different annealing temperatures.


132
either {311} defects or invisible mobile boron interstitial
pairs. Besides this trapping effect of interstitials by-
loops, more interstitials might be trapped in the boron
interstitial clusters in the implant damage peak region
because more clusters are formed in the higher dose implant.
As a result of both trapping effects, the number of
interstitials available for TED is reduced and so is the
enhancement of diffusion. This can explain why we do not
observe a decrease in /Db* for times < 1 hr even though
the dose is increased four times.
Another explanation of the TED saturation behavior with
increasing dose was proposed by Griffin et al133. They
claimed that this saturation can occur if most of the dopant
atoms are already pushed into a mobile state so that more
damage produces no increase in diffusivity. They used a
simple chemical reaction model to explain the phenomena. The
interaction of a substitutional dopant atom Bs with a free
interstitial I to produce a mobile Bx can be described by
Bs + I <-> Bj (3.5)
If this reaction is in equilibrium, then we have
Bi KeqBsI (5.1)
where Keq is the equilibrium reaction constant. The
substitutional Bs and mobile BI are related to the total
dopant by
BT = Bs + Bj (5.2)
Using eq.(5.1) and (5.2) we have


11
Energy and/or Dose
sub-amorphized amorphized
as-implant
Figure 1.2. Schematic of the conditions under which {311}
defects can form and how they evolve with time.
background environment (which can result in TED) and the
growth of dislocation loops if they are present. Since {311}
defects form after a short time and/or low temperature anneal
with or without the presence of dislocation loops, and they
dissolve fairly fast, they have been termed intermediate
defect configurations (IDCs). As an example, Fig. 1.3 shows a
plan-view transmission electron microscope picture of {311}
defects in boron implanted silicon after an anneal at 750C
for 5 min. Under the imaging conditions used, there are


0.2|im
(b)
Figure 5.1 (Continued)


122
anneals and when annealing time is increased to 15 min, these
defects completely dissolve. Defect behavior of the
2xl014 cm 2 implant was shown in the last chapter (Fig. 4.1
(b)) and is left out here. A majority of {311} defects
together with just a few dislocation loops are observed in
the 5xl0^-4 cm-2 sample and the loop density increases for the
lxiol- cm-2 implant. For each of the implants showing {311}
defects, the number of the defects decreases with increasing
annealing time. This dissolution process is much more rapid
for a lower implant dose.
The areal density of interstitials trapped in {311}
defects during the annealing process was quantified from the
defect measurements and the results are shown in Fig. 5.2.
Results for the lxlO14 cm-2 implant are not included because
the defect density is too low even after a 5min anneal. The
emission of interstitials from {311} defects again shows an
exponential time-dependence, which is much stronger for lower
implant dose. After a 30 min anneal, {311} defects in the
2x101* cm-2 implant dissolved to a very low density. The
number of remaining trapped interstitials is about
7.7xl0H cm-2. If the dose is increased to 5xl0l4 cm-2,
after an anneal of lhr, there is still an areal interstitial
density of about 1.7xlC)l3 cm-2. If we compare Fig. 5.2 with
Fig. 4.2, we will find that doubling the dose decreased the
dissolution rate much more than doubling the energy. Since
the number of displaced atoms per implanted ion is
proportional to the implant energy, such an observation might


80
O
i
E
c
o
(
1
c
O
c
o
O
c
o
o
CQ
0 1000 2000 3000 4000 5000 6000 7000 8000
Depth ()
(a)
c
CD
O
c
o
O
c
o
o
CQ
10
20
1 1 1 I 1 1 1 1 I
E
c
o
co
10
19
10
18
£ 1017 L-
106 1
2 10'5 1
lili l'l I I I | I I I I 1 T' I II.
as-implanted
30 min
----2hr
*4hr
1 Q14 I I I I I I I
0 1000 2000 3000 4000 5000 6000 7000 8000
Depth ()
(b)
Figure 3.7. Dopant SIMS profiles of 20keV 2xlC)14cin-2 B+
implanted Si after anneals at (a) 650C, (b) 700C, (c) 750C
and (d) 800C for various times.


25
understanding of dopant diffusion. This correlation will be
further investigated in this study.
1.3 Ion Implantation. Characterization and Simulation
Techniques
1.3.1 Ion Implantation
Ion implantation is the introduction of energetic
charged particles into targets with enough energy to
penetrate beyond the surface region. The ions enter the
substrate, collide with the host atoms, gradually lose energy
and finally come to rest at some depth within the substrate.
Ion implantation is involved in several steps during the
fabrication sequence of MOSFET and bipolar transistors when
the controlled doping of selected areas is required.
An ion implanter consists of the following major
components: an ion source, an extracting and ion analyzing
system, an accelerating column, a scanning system and an end
station. The ions produced by the source are extracted by a
small accelerating voltage and then injected into the
analyzer magnet. A spatial separation of ions subjected to
the Lorenz force occurs due to the differences in the mass
and charge. The selected ions are injected into the
accelerating column and the rest are screened out. In this
case the system operates in the pre-analysis configuration.
If the ions are accelerated to their full energy before the
mass separation, then it works in the post-analysis


115
m
m
Q
'a

Q
V
c
Q)
E
CD
O
c
CO
.c
c:
LU
>
'in
3
5=
Depth ()
(a)
*
CD
Q
'a
m
Q
V
-*<
c
CD
E
CD
O
c
CO
sz
c
LU
>
w
3
it
b
Depth ()
(b)
Figure 4.8. Enhancement of B diffusivity /Db* in B+
implanted Si superlattices at an energy of (a) 5keV, (b)
lOkeV, (c) 20keV and (d) 40keV, annealed at 750C.


169
104 G. Hobler and S. Selberherr, IEEE Trans, on Comp. Aid.
Des. 7. 174 (1988).
105 J. K. Listebarger, K. S. Jones, and J. A. Slinkman J.
AppI. Phvs. 73. 4815 (1993).
106 S. Prussin, J. AppI. Phvs. 43. 2850 (1972).
107 C. M. Drum and W. van Gelder, J. AppI. Phvs. 43. 4465
(1972) .
108 O. L. Krivanek and D. M. Maher, AppI. Phvs. Lett. 32. 451
(1978) .
109 H. L. Meng, S. Prussin, M. E. Law, and K. S. Jones, J.
AppI. Phvs. 73. 955 (1993).
110 M. Servidori, Z. Sourek, and S. Solmi, J. AppI. Phvs. 62.
1723 (1987).
111 H. Chao, P. Griffin, and J. Plummer, AppI. Phvs. Lett.
M# 3570 (1996).
112 D. J. Eaglesham, P. A. Stolk, H.-J. Gossman, and J. M.
Poate, AppI. Phvs. Lett. 65. 2305 (1994).
113 P. A. Stolk, H.-J. Gossmann, D. J. Eaglesham, D. C.
Jacobson, H. S. Luftman, and J. M. Poate, Mat. Res. Soc.
Sypip. Prog. 354 (1995) .
114 L. H. Zhang, K. S. Jones, P. H. Chi, and D. S. Simons,
AppI. Phvs. Lett. 67. 2025 (1995).
115 Encyclopedia of Materials Characterization (Butterworth-
Heinemann, Stoneham, MA 02180, 1992).
116 Y. M. Kim, G. Q. Lo, H. Kinoshita, D. L. Kwong, H. H.
Tseng, and R. Hance, J. Electrochem. Soc. 138. 1122
(1991).
117 H. Kinoshita, G. Q. Lo, and D. L. Kwong, J. Electrochem.
Soc. 140(1). 248 (1993).
118 H. U. Jager, J. AppI. Phvs. 78. 176 (1995).
119 M. Tamura, N. Natsuaki, Y. Wada, and E. Mitani, Nucl.
Inst. Meth. B 21. 438 (1987).
120J. F. Gibbons, Proc. IEEE 60. 1062 (1972).


Defect Defect Defect Defect
Density (cm'2) Density (cm'2) Density (cm'2) Density (cm'2)
ill
* i i i i
60 260 460 660 860 1060 1260 1460
1i i ir-r tiIiiiiirr-ii i ir-r i iiiir-|
30keV 15min !
lULi uJ
L
260 460 660 860 1060 1260 1460
7-111-71r111i1-7
30keV 30min
i
111
SSL
L
260 460 660 860 1060 1260 1460
JO11 1 1 1 1-1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 t 1 1 1 1
30keV 1 hri
1010
r

109
r
108
r
107
i
60
{311} Defect Size ()
(c)
Figure 4.6. (Continued)


1
2
3
4
5
6
7
8
9
10
11
12
13
14
LIST OF REFERENCES
C. M. Osburn, J. Electron. Mater. 19. 67 (1980) .
H. Jiang, C. M. Osburn, P. Smith, Z.-G. Xiao, D. Griffis,
G. McGuire, and G. A. Rozgonyi, J. Electrochem. Soc. 139.
196 (1992).
S. N. Hong, G. A. Ruggles, J. J. Wortman, and M. C.
Ozturk, IEEE Trans. Electron Devices ED-38. 476 (1991) .
R. H. Dennard, F. H. Gaensslen, N. H. Yu, V. L. Rideout,
E. Bassous, and A. R. LeBlanc, IEEE J. Solid-State
Circuits SC-9. 256 (1974) .
J. R. Brews, W. Fichtner, E. H. Nicollian, and S. M. Sze,
IEEE Electron Device Lett, EDL-lf 2 (1980) .
S. N. Hong, G. A. Ruggles, J. J. Paulos, J. J. Wortman,
and M. C. Ozturk, Appl. Phvs, Lett. 53. 1741 (1988).
E. Ling, P. D. Maguire, H. S. Gamble, and B. M.
Armstrong, IEEE Electron Device Lett. EDL-8. 96 (1987) .
K. T. Kim and C. K. Kim, IEEE Electron Device Lett. EDL-
£, 569 (1987) .
H. J. Bohm, H. Wendt, H. Oppolzer, K. Masseli, and R.
Kassing, J. Appl. Phvs. 62. 2784 (1987) .
M. Horiuchi and K. Yamaguchi, IEEE Trans. Electron
Devices ED-33. 260 (1986).
P. G. Carey, T. W. Sigmon, R. L. Press, and T. S. Fahlen,
IEEE Electron Device Lett. EDL-6. 291 (1985) .
Y. El-Mansy, IEEE.Trans. Electron Devices ED-29. 567
(1982) .
Y. Kim, H. Z. Massoud, and R. B. Fair, AppI. Phvs. Lett.
3, 2197 (1988).
S. Guimaraes, E. Landi, and S. Solmi, Phvs. Stat. Sol
a) 95. 589 (1986).
163


ACKNOWLEDGMENTS
This dissertation would not have been possible without
the guidance and support of my advisor, Dr. Kevin Jones, who
not only provides us illuminating advice but also creates a
stress-free environment for us to work in and do our best. I
would like to thank Drs. Steve Pearton, Cammy Albernathy,
Rajiv Singh and Mark Law for serving on my committee, and Dr.
Robert Park for sitting in my defense. Special gratitude is
due to Dr. Jian Chen, who patiently helped me get into the
door when I first joined the group four years ago and gave me
many valuable suggestions in the years following. Special
gratitude is also due to Dr. Wish Krishnamoorthy for many
stimulating discussions and a thorough correction on my
dissertation draft. I would like to thank all my colleagues,
especially members of our silicon group, Brad Herner and
Sushil Bharatan, for many helpful discussions, and my
officemate as well as good friend, Ranju Datta, for keeping
me company and making my time at school enjoyable.
It would have been impossible to finish this work
without the assistance of many other people. Dr. Lenny Rubin
at Eaton Corporation and Craig Jasper at Motorola helped me
with the implantations. Dr. Jinghong Shi and Joe Bennett
from SEMATECH helped me with the SIMS analysis. Dr. Hans
in


87
depth of the interstitial and vacancy profiles and the
separation between these two profiles. These changes can
affect the defect behavior and TED, as will be discussed
later.
4.2 Implant Energy Effect on Defect Behavior
4.2.1 Experimental Procedure
A single row of the implant matrix discussed in Chapter
2 was used to study the implant energy on {311} defect
evolution. The implant dose was kept at 2x10-*-^ cm-2 and the
energy was varied from 5 keV to 40 keV. Subsequent furnace
anneals were conducted in a nitrogen ambient for times
between 5 min and 1 hr at 750C. The {311} defect density,
average defect size and, areal density of interstitials
trapped by {311} defects and the defect distribution during
annealing were measured from the PTEM micrographs.
4.2.2 Defect Microstructure and Dissolution Process
PTEM micrographs of the defects after implantation at
10 keV to 40 keV and anneals at 750C from 5 min to 1 hr are
shown in Fig. 4.1(a)~(d). No defects were observed for the
5 keV implant and therefore the pictures for those samples
are not included. {311} defects are the only type of
extended defects observed under the TEM in the 10 keV and 20
keV implants. A majority of {311} defects together with just
a few dislocation loops are observed in the 30 keV and 40 keV


Average {311} Defect Size ()
75
Annealing Time (sec)
Figure 3.5. Average {311}
annealing time in 20keV B+
annealing conditions.
defect size as a function of
implanted Si for different


Defect Defect Defect Defect
Density (cm ) Density (cm'2) Density (cm'2) Density (cm2)
76
650C6hr
1 1 1 1 1 1
60 260 460 660 860 1060 1260 1460
650C14hr
1 1 1
i
60 260 460 660 860 1060 1260 1460
650C36hr
10 60 260 460 660 860 1060 1260 1460
12
10
1011 r
650C48hr
10
10
10
10e
9
Mi. M
j-i- i 111
60 260 460 660 860 1060 1260 1460
Average {311} Size ()
(a)
Figure 3.6. {311} defect size distribution during annealing
at (a) 650C, (b) 700C, (c) 750C and (d) 800C.


93
Each of these factors could decrease the number of
interstitials available for the formation of {311} defects.
4.3 Implant Eneren/ Effect on Transient Enhanced Diffusion
4.3.1 Experimental Procedure
In order to study the effect of interstitial release
during TED independent of the diffusing profile, doping
superlattices were used. Boron 8-doping superlattices were
grown by low temperature molecular beam epitaxy (LTMBE) on
float zone Si (100) substrate with boron-doped to a
resistivity of 1000 ilem. The custom-made MBE system had a
base pressure of 4xl0-11 Torr. The temperature during growth
was set through the choice of the heater power, which in turn
had been calibrated in terms of temperature by laser
interferometry. The samples contained six 100 wide box
shaped boron spikes with a separation of about 1000 between
each spike. The peak boron concentration of these spikes was
about 1.4x10^ cm3. This low concentration kept boron
diffusion intrinsic for the as-grown control samples during
the following anneals and avoided the complication of
extrinsic effects. Boron ions were implanted into the
superlattices at energies of 5 keV, 10 keV, 20 keV and 40 keV
to a dose of 2x10^4 cm-2 at room temperature. The samples
were then annealed at 750C for times of 3 min, 15 min and
2 hr in a nitrogen ambient. The temperature error range is
estimated to be 10C. Due to the finite rise time of the


95
furnace temperature and SIMS analysis accuracy and the
ability to predict Db*. The diffusion distance for the
control sample is indeed too small to be extracted reliably
since that the SIMS depth has an estimated error of 5%.
Figure 4.8 (a)~(d) are plots of /Db* as a function
of depth for different implant energies after each anneal.
The program assumes that the total sheet concentration is
preserved for each spike during diffusion, i.e., the depth
integration of boron concentration under each spike remains a
constant after implantation and annealing. However, the
influx from the surface implant into the spike region
actually violates the conservation. The error bars of
/Db* to be shown do not include this effect. The dots in
Fig. 4.8 are the data points and the lines are exponential
fitting curves. We can see that the value of /Db* decays
exponentially with increasing depth, and the decay length is
smaller at a shorter annealing time (3 min) than longer times
(15 min and 2 hr), i.e., the /Db* versus depth curve is
steeper. The cross of /Db* curves for the 3 min and
15 min anneals implies that interstitials have not diffused
deep enough to cause a large diffusion enhancement after a
shorter time. The slightly positive slopes of the /Db*
curves for the 15 min and 2 hr anneals might be due to the
SIMS analysis and the simulation error. After a short time
(3 min) anneal, for each particular spike the value of
/Db* increases with increase in implant energy. For
example, for a boron spike at a depth of around 3800 A, after


40
densities are near the threshold for defect formation so the
behavior is due to the other fluctuations. However, the
reason may also be due to the above argument for why loops
are seen most prominently for 5 keV implant.
Besides the effect of implant energy on damage
production, interstitial recombination at the sample surface
may also affect the formation of secondary defects. Figure
2.4 shows the projected range Rp as a function of implant
energy obtained from the XTEM micrographs and the SIMS
profiles which will be shown in Chapter 4. Rp increases
with increasing energy. It is about 230 for the 5keV
implant and about 1750 for the 40keV implant according to
SIMS. As the implant energy decreases, the damage region
gets closer to the sample surface. Surface recombination
could then become more efficient and thus reduce the number
of free interstitials available to form secondary defects.
It is difficult to determine which effect, decreased damage,
increased boron interstitial cluster formation due to higher
interstitial supersaturation in the dopant peak region or
increased surface recombination, is primarily responsible for
the increased threshold dose with decreasing implant energy.
Additional experiments to sort out these effects will be
necessary.
In summary, the formation threshold for both {311} rod
like defects and [110] dislocation loops in low energy boron
implanted silicon have been investigated. A matrix was
chosen with implant energies ranging from 5keV to 40keV and


57
In fact we have observed the {311} defect width change
under the TEM. PTEM pictures were taken on the <311> zone
with g=220. The following trend was found: the width is
about 20 A when the length L is 200-400 A. It increases to
about 40 A when L is around 1000 A, to about 80 A when the L
is around 1200-1400 A, then to about 100 A for L=1600~1800 A
and about 120 A when L-3000 A. These measurements were
performed on samples annealed at 750C for 5 min, 15 min and
1 hr. The error range might be large because of the narrow
width of these defects. However, the width change with
length can be used to account for the interstitial flow
during the coarsening process. It should be mentioned that
in this study this width change is ignored when the areal
density of interstitials trapped in {311} defects was
measured from the length of {311} defects and the
interstitial density per unit length. In fact it is almost
impossible to measure both the width and the length of every
defect to obtain a more accurate result.
In summary, in this section of Chapter 3 we have
discussed the {311} defect evolution process during anneals
at 650C to 800C for various times. The areal density of
interstitials bound by {311} defects decays exponentially
with annealing time with a characteristic time constant which
is larger at lower temperatures. This dependence has an
activation energy of 3.8 eV. The number of interstitial
trapped in {311} defects is only 15-20% of the implant dose,
implying the presence of boron interstitial complexes,


Density of Interstitials Trapped in {311} Defects (cm'2)
140
Annealing time (sec)
Figure 5.2. The exponential decay of interstitials trapped
in {311} defects in 20keV B+ implanted Si to various doses.


152
The interstitials trapped in {311} defects decay
exponentially with annealing time with a characteristic time
constant smaller at higher temperature. The activation
energy of {311} defect dissolution process was determined to
be 3.8eV. The defect density and the average defect size
were also quantified. The size distribution of {311} defects
demonstrates an Ostwald ripening process along with a defect
dissolution process. Boron diffusion was measured through
SIMS and the diffusivity enhancement /Db* was extracted
from FLOOPS simulations. /Db* decreases with increasing
annealing time. The diffusion length increases initially
during the annealing and then reaches a saturation point when
it starts to plateau out. The time constant decreases
dramatically when annealing temperature increases. The
activation energy for the TED saturation process was
determined to be 1.6eV. Both mobile boron interstitial pair
complexes and {311} defects are believed to be the source of
interstitials driving TED. The static implant profile peak
is believed to be due to the Bs-Bx-I clustering effect. The
breaking point between the mobile tail region and the
immobile peak region Cenh moves to higher concentration when
temperature increases due to a decreasing interstitial
supersaturation level and thus less cluster formation.
The influence of implant energy on {311} defect
evolution was investigated by studying samples implanted to a
single dose (2x10-*-^ cm-^) at various energies (5~40 keV) .
{311} defects were observed at energies greater than 5 keV.


59
Florida Object Oriented Process Simulator (FLOOPS) and the
diffusivity enhancement was extracted for each anneal.
3.3.2 Dopant Diffusion during Annealing
3.3.2.1 Static Peak at High Concentration
As an example of the dopant diffusion behavior during
annealing, Fig.3.7 (a)~(d) shows the boron profiles obtained
through SIMS after anneals at 650C for 2 hr, 14 hr, 24 hr,
at 700C for 30 min, 2 hr and 4 hr, at 750C for 20 min, 1
hr, 2 hr and at 800C for 3 min, 3 0 min and 1 hr. We can see
that a critical concentration exists under every annealing
condition, separating the low concentration tail region where
transient diffusion occurs, from the high concentration peak
region where the profile is relatively immobile. The
breaking point between the mobile and immobile region (Cenh)
moves to higher concentration when annealing temperature
increases. It is about 2.5x10-*-^ cm-^ at 700C and moves up
to about 7xlC)18 cm~3 at 800C. Figure 3.8 shows Cenh as a
function of inverse temperature for our study and some of the
previous studies. Also plotted in Fig. 3.8 are the solid
solubility for boron in silicon Cs and the intrinsic carrier
concentration ni.
There are several groups which have reported the peak
clustering behavior of an implant profile. Michel et al.126
studied 60 keV 2x10^4 cirT^ b+ implanted silicon. Over an
annealing temperature range of 800C~950C, their
measurements show that both the maximum concentration of the


14
formation. Models of split self-interstitials with or
without dangling bonds were proposed by Tsubokawa et al.54 and
Pasemann et al.55 respectively. Tan56 developed a homogeneous
nucleation model which shows that for extrinsic defects, a
chain of interstitial atoms form {311} defects having non-
six-membered atomic rings with matrix atoms, while for
intrinsic defects, a chain of matrix atoms is cut out and the
remaining atoms surrounding the cut are used to form {311}
defects having non-six-membered atomic rings. A schematic
diagram for the different steps involved in the formation of
extrinsic {311} defects is shown in Fig.1.5 (a)~(d). This
non-six-membered atomic ring configuration minimizes the
dangling bond that may present.
(a)
Figure 1.5. Schematic of four step {311} defect formation
process: (a) The accepted split [100] interstitial chain, (b)
The addition of one more split [100] chain, (c) Details of
the bond rearrangement and(d) The resulting defect
configuration.


lxlO^cirr2
5keV
20keV
\ '
40kcV
. C -
\ : >
A % \ ^
*
v.
0.2^im
2xl014cm*2 5xl014cm'2
Figure 2.1. Weak beam dark field (g220> PTEM images of 5keV, 20keV and 40keV B+ implanted
Si to doses of lxl0^-4cm-2, 2xl0^-4cm-2 and 5xl0^-4cm-2, after an anneal at 750 C for 5min in
N2.
N>


0-2|im
(a)
Figure 3.1. Weak beam dark field (g220) PTEM micrographs of {311} defects in 20keV
2xl014cirT2 B+ implanted Si after anneals at (a) 650C, (b) 700C, (c) 750C and (d) 800C
for various times.


LD
1780
199/
.L7S35
UNIVERSITY OF FLORIDA
3 1262 08554 9607


51
Si,(t) = 5//(0)exp(-) (3.1)
T
where X is defined as the characteristic time constant of the
decay. Although the decay process is strongly dependent on
annealing temperature, the initial density of interstitials
at all four temperatures is very close to the same value of
3~4xl0^3 cm-2, which is only 15-20% of the implant dose of
2xl014 cm-2. According to the "plus one" model,124 the
initial density of interstitials should closely mimic the
implant dose, because each implanted ion is supposed to give
rise to one excess interstitial during annealing. This model
assumes that the implant ion induces a series of Frenkel pair
formation and then comes to rest in the lattice at an
interstitial site. Subsequent annealing results in
annihilation of all vacancy-interstitial pairs with just the
one interstitial remaining. This theory is proved to be
reasonable by the study of Si implanted silicon of
Listebarger et al.105 and valid to a remarkable degree by the
study of Eaglesham et al.125. Listebarger et al. used
dislocation loops as point defect detectors to measure the
flux of interstitials introduced by a 30keV B+ implant to
doses of 7x10-*-3 cm-2, lxlO-*-4 cm-2 and 2xl0^-4 cm-2. The loops
layer was formed by 100 keV Ge+ implant to a dose of
lxlO-*-^ cm-2 followed by an anneal at 550C for 16 hr to
regrow the amorphous surface and another anneal at 800C for
30 min to coarsen the loops. They reported that the net flux
of interstitials measured using the loop detectors is very


Q)
ON
to
Figure 6.2. A schematic of different sources of interstitials for TED that presumably
exist in B+ implanted Si.


56
sizes. Self-energy of the {311} defect was calculated by
Parisini and Bourret76 by approximating the rod-like shape
with a dipole of edge dislocations of the same sense and
opposite Burgers vector. It is given by
W = [ln() + cos20]L (3.2)
2n{\-v) 2 p
where (l is the shear modulus, b the Burgers vector, v the
Possions ratio, r| the distance between the two edge
dislocations (i.e., the width of the {311} defect), p=b/8 the
cut-off distance, 9 the angle between the Burgers vector and
the normal to the plane containing the {311} defect and L the
length of the {311} defect. To obtain the energy per
interstitial, one can divide the previous expression by the
number of interstitials N trapped in a {311} defect with a
length of L. N is given by
N = p,Lri (3.3)
where p¡ is the atomic density on the {311} plane. The energy
of one interstitial is thus given by
W, = ^ [In A + cos2 0]
2/r(l-v) 2 p p,r¡
(3.4)
As we can see from eq. (3.4), Wx is independent of the length L
but is dependent on the width T). Since ln(T|/p) decays slower
than T|, Wi decreases when r| increases. We thus speculate that
the width of a shorter {311} defect is smaller than that of a
longer one. As a result, the self-energy of an interstitial
trapped in a shorter {311} defect is higher than that in a
longer one. Interstitials would thus flow from shorter rods
to longer ones. This explains the Ostwald ripening process.


60
displaced boron and the electrical activity are related to
the intrinsic carrier concentration. Later, Fair127 has
extended the correlation between Cenh and ni to the
temperature range of 600C~800C. It was observed that when
T>850C, Cenh approaches B solid solubility Cs. A model was
proposed by Fair, suggesting that the existence of Cenh is
due to the Fermi level and the charge state of the defect
affecting diffusion. For boron diffusion the location of the
donor level of self-interstitials I+ in the energy gap causes
Cenh to equal ni for T<800C. Under this condition, a
significant fraction of the excess interstitials are in the
neutral charge state Ix. For doping under Cenh~ni> diffusion
is dominated by Ix. Since Dix/Di+ is 10-100, more enhanced
diffusion occurs at C850C, the Ei+ level is
sufficiently above Ep and the majority of injected
interstitial exist in I+ state. Then the thermally assisted,
concentration dependent diffusion via I+ becomes dominant and
Cenh is limited by B solid solubility Cs The study of
Cowern et al.128 on 25 keV 2xl014 cm-2 B+ implants agreed with
Fair's results. Cenh lies within a factor of 2 of ni in a
temperature range of 550C~900C. A relative insensitivity
of Cenh to the implant energy and dose was suggested by all
three groups. As shown in Fig. 3.8, our data of 20 keV
2xl014 cm 2 B+ implants also agree with those of the above
groups to a reasonable degree within 650C~800C. However,
the data of Zhang et al. for 4 keV lxlO14 cm-2 B+ implants114
do not match the intrinsic carrier concentration but are


28
(XTEM) is used to get the side view and the depth of the
defect layer.
To make a PTEM sample, the sample is first cut into 3mm
diameter pieces using an ultrasonic disk cutter. The small
disk is then polished down to about half of its original
thickness using a polishing jig with 15|im aluminum oxide
powder. A coat of wax is then applied to the top surface
(i.e. the implanted surface) of the sample to protect it
during the etch. The sample is mounted with wax on a Teflon
holder facing down and etched from the back side in a jet
etcher using a HF: HNC>3 = 25%: 75% solution until a small hole
with an electron transparent region around it is obtained.
The wax is the removed from the sample using heptane and/or
acetone and the sample is ready for analysis.
To make a XTEM specimen, a piece of silicon wafer is cut
into lCmil thick strips using a dicing saw. The top surface
of the two strips is coated with M600 Bond epoxy resin. The
two top surfaces are then brought into close contact with
each other. Two silicon dummy strips can also be glued on
the side of each sample strip to improve the mechanical
stability. The bonded strips are heated in a conventional
oven at 100C for lhr to harden the epoxy resin. The cured
strips are then polished down to a thickness of about 40|lm
using 5|im aluminum oxide powder. A copper ring is mounted
onto the thinned strips using the M600 Bond, leaving the
interface of the sample strips exposed in the center of the
ring. The specimen is then further thinned in a Gatan two-


153
The lack of {311} defects in the 5 keV implant might be due
to a combination of factors including surface recombination
effect, large density of boron interstitial clusters and less
implant damage and high I-V recombination. The time constant
for the {311} dissolution process is larger for higher
energy. Boron 8-doping superlattices were used to monitor
the diffusion behavior for each implant energy. /Db*
decays exponentially with depth with a decay length larger at
shorter annealing time before equilibrium is reached. TED
lasts longer and the amount of TED is larger for E>5 keV than
the 5keV implant.
The influence of implant dose on {311} defect evolution
was studied by using samples implanted at the same energy
(20keV) to various doses (5xlC)13~lxlO-*-5 cm2). {311} defects
dissolve much slower at higher dose. The ratio of the
interstitials trapped in {311} defects and loops to the
implant dose stays about the same (12%) after a 5min anneal
at 750C regardless of the dose change. Both defect density
and average defect size increase with increasing dose.
Wafers with a buried boron 8-doping layer were implanted to
the same doses at the same energy to study the diffusion
behavior. The amount of diffusion increases with increasing
dose before dislocation loops are formed. Further increase
in the dose showed no additional increase in TED presumably
due to the trap of interstitials by the loops or the
saturation of dopant atoms paired with interstitials.


134
Chapter 3 for the derivation). Another data point (t=0,
V(t)=0) is added on each set of data for the convenience
of fitting. It is obvious that TED saturates after a shorter
time (around 16 min) at the low dose of 5x10^-^ cm" 2 than at
higher doses (37 min at 2x10^-^ cm"2 and 4 min at
lxlC)15 cm-2). The diffusion length at 2x10-*-^ cm"2 dose is
very close to that at 1x10-*-^ cm"2 dose due to the factors we
discussed above.
A similar dose dependent diffusion behavior was observed
by Packan130. As was mentioned in the overview section in
this chapter, the diffusion of a buried implanted boron
profile was studied by using Si+ implanted silicon with doses
ranging from 1x10-*-^ cm"2 to 2x10-*-^ cm"^ at an energy of 200
keV. He presented the diffusion length versus annealing time
plots for doses of 5x10-^^ cm"2, 1x10-*-4 cm"2 and 2x10-*-^ cm" 2
after anneals at 750C, 800C and 850C for various times.
The diffusion length for the 5x10^-^ cm"2 implant is the
lowest for all the anneals. The diffusion lengths for the
1x1014 cm-2 and the 2x10^4 cm"2 implants stay about the same
for t < 330 min at 750C, t < 60 min at 800C and t < lOmin
at 850C. After these times higher dose gives rise to larger
diffusion length. Comparing our results with Packan*s, it is
reasonable to believe that for 200 keV Si+ implantation, the
threshold dose for the formation of sub-amorphization
dislocation loops is between lxlO^ Cm"2 and 2x10^^ cm"2.
During the early stage of annealing these loops absorb
interstitials and reduce the enhancement of boron diffusion.


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E3BS0EYSX_PG7YYG INGEST_TIME 2015-04-06T18:41:04Z PACKAGE AA00029750_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES


170
121 V. Raineri, R. J. Schreutelkamp, F. W. Saris, R. E. Kaim,
and K. T. F. Janssen, NIMPR B59/60. 1056 (1991).
122 R. J. Schreutelkamp, J. S. Custer, J. R. Liefting, W. X.
Lu, and F. W. Saris, Mat. Sci. Reports 6. 275 (1991).
123 N. E. B. Cowern, A. Cacciato, J. S. Custer, F. W. Saris,
and W. Vandervorst, AdpI. Phvs. Lett. 68. 1150 (1996).
124 M. D. Giles, J. Electrochem. Soc. 138. 1160-1165 (1991).
125 D. J. Eaglesham, P. A. Stolk, H. J. Gossmann, T. E.
Haynes, and J. M. Poate, Nucl. Inst. Meth. B 106. 191
(1995) .
126 A. E. Michel, W. Rausch, P. A. Ronsheim, and R. H. Kastl,
AppI. Phvs. Lett. 50. 416 (1987).
127 R. B. Fair, J. of Electrochem. Soc. 137. 667 (1990) .
128 N. E. B. Cowern, K. T. F. Janssen, and H. F. F. Jos, J.
Appl. Phvs. 68, 6191 (1990).
129 R. B. Fair, in Impurity Doping Processes in Silicon,
edited by F. F. Y. Wang (North Holland, New York, 1981),
p.968, 315.
130 P. Packan, Ph.D. Dissertation. Stanford University
(1989).
131 K. S. Jones, V. Krishnamoorthy, L. H. Zhang, M. Law, D.
S. Simons, P. H. Chi, L. Rubin, and R. G. Elliman, AppI.
Phvs. Lett. 68. 2672 (1996).
132 S. Solmi, F. Baruffaldi, and R. Canteri, J. AppI. Phvs.
£2, 2135 (1991).
133 P. B. Griffin, R. F. Lever, P. A. Packan, and J. D.
Plummer, Appl. Phvs. Lett. 64. 1242 (1994).
134 C. S. Rafferty, G. H. Gilmer, M. Jaraiz, D. Eaglesham,
and H.-J. Gossmann, AppI. Phvs. Lett. 68. 2395 (1996).
135H. S. Chao, C. S. Rafferty, P. B. Griffin, and J. D.
Plummer, unpublished paper.


{311} Defect Density (cm'2)
107
Annealing Time (sec)
Figure 4.4. Density of {311} defects as a function of
annealing time for different implant energies.


131
of annealing time for different implant doses after each
anneal. Before the simulation, dose normalization was
conducted on the spike for the as-implanted and annealed
samples regardless of the small influx of boron from the
surface implant into the spike region which actually violates
the dose conservation. From Fig.5.9 we can see that the
value of /Db* decreases with increasing annealing time
for each implant. At all annealing times, /Db* is lower
for the 5x10^3 cm-2 implant than the two higher doses. After
an anneal at 750C for 4 hr, /Db* increases from 409 for
the 5xl0l3c m2 implant to 854 for the 2x10^4 Cm"2 implant.
For B diffusion, /Db* is proportional to the interstitial
supersaturation level /Ci*. If the "plus 1" model is
valid, the difference in /Db* simply should be four
times. Since only a two times increase is seen yet all the
{311} defects have dissolved, stable B-I cluster formation in
the 2xl0l4 cm2 implant could account for the small increase
in /Db*.
5.3.3 TED Saturation at Higher Dose
It is interesting to notice that /Db* is close for
the two higher doses, 2x10-'-^ cm-2 and lxlO-^5 cm-^. As we
discussed before, the formation threshold for the sub-
amorphization dislocation loops lies between doses of
2x10-1-4 cm2 and 5x10-*-^ cm-^. When the dose reaches
lxlO15 cm 2f quite a few loops form after annealing. These
loops can act as a sink for interstitials released from


19
diffusion mechanism, vacancy, interstitial or interstitialcy
mechanism. In the vacancy mechanism, a substitutional dopant
atom exchanges position with an empty lattice site. In the
interstitial mechanism, the dopant atom is kicked out of the
silicon lattice by a silicon self-interstitial and diffuses
as a pure interstitial before returning to the lattice as a
substitutional atom. In the interstitialcy mechanism, the
silicon self-interstitial and the dopant atom form a
diffusion pair. This process is illustrated in Fig. 1.6.
Usually no distinction is made between the interstitial and
the interstitialcy mechanisms.
Vacancy
Interstitial
Interstitialcy
Figure 1.6. Schematic diagram of different dopant diffusion
mechanisms.
If an asterisk is used to represent atoms in their free
states, then the interaction of dopant atoms and the point
defects can be expressed as
A*+X* <-> AX (1.2)
where X=I corresponds to interstitial/interstitialcy


Diffusivity Enhancement /D
116
Depth ()
(c)
Depth ()
(d)
Figure 4.8. (Continued)


0-2um
Figure 3.1. (Continued)


Table 2.1 Types of extended defects formed in B+ implanted silicon
5keV
lOkeV
20keV
30keV
40keV
5xl0l3cm'2
none
none
none
none
none
Ixl0l4cm"2
none
none
{311 }s
{311 }s
{311 }s
2xl0l4cnr2
none
{311 }s
{311 }s
{311 }s
{311 }s
Loops
Loops
5xl0^cm_2
{311}s
{311}s
{311}s
{311}s
{311}s
Loops
Loops
Loops
Loops
Loops
lxl()15cm"2
{311}s
{311} s
{311 }s
{311 }s
{311 }s
Loops
Loops
Loops
Loops
Loops


0.2(im
Figure 5.1 (Continued)


Boron Concentration (cm'3) Boron Concentration (cm'3)
148
0 2000 4000 6000 8000 1 104 1.2 104
Depth ()
(c)
0 2000 4000 6000 8000 1 104 1.2 104
Depth ()
(d)
Figure 5.8. (Continued)


52
similar to the dose for each B+ implant. Eaglesham et al.
measured the density of interstitials trapped in {311}
defects in 40 keV Si implanted silicon to a dose of
5xlOi;^ cm2. The interstitial density versus dose plot shows
a slope of 1.5, indicating slightly more than one
interstitial per implanted ion. However, in our B+ implanted
silicon study, the observed interstitials only account for
approximately one tenth of the dose. We therefore believe
that there is another type of defects, i.e., boron
interstitial complexes, which trap the missing interstitials.
These complexes are sub-microscopic and are not resolvable in
the TEM. There appear to be more than one form of boron
interstitial complexes. From our low energy low dose
implants (as will be discussed later), we observed as did
Zhang et al.114 that there is a significant fraction of the
implant profile peak that is immobile. This could be
explained by the formation of B-B-interstitial clusters. A
second portion of the profile is mobile and may simply be B-
interstitial pairs as suggested by Cowern.123 Both of these
non-visible defects could help account for the "missing"
interstitials in the {311} defects.
An Arrhenius plot of the characteristic rate constant
k=l/T of {311} defect dissolution process as a function of 1/T
is shown in Fig. 3.3. This temperature dependence shows an
activation energy of 3.8 eV. Along with our results, Fig.
3.3 also presents the temperature dependence of {311} defect
dissolution from the study of Stolk et al.113
of Si+ implanted


5min
30min
0.2[im
15min
lhr
o
(d)
Figure 4.1 (Continued)


30
to determine dopant diffusivities. In order to keep track of
the spatial distribution of point defects, FLOOPS solves the
diffusion equations for the dopant atoms as well as the
equivalent equations for the point defects. Since vacancies
and interstitials can interact, the equations describing
their motion are coupled. The equations used to
determine dopant diffusion include
(1.10)
A
dt
(1.11)
m
(1.12)
where Ja is the flux of the dopant, Ca is the total
concentration of the dopant species, the summation over X is
for all possible diffusion mechanisms, Cx is the
concentration of the X species, Dax is the diffusivity
corresponding to the X mechanism, ni and n refer to the
intrinsic and actual electron concentrations, Cai and Ca2 are
the concentration of the dopant species on the two sides of a
materials boundary, Ktr is the transport coefficient across
the material boundary and m is the segregation coefficient of
the dopant for the boundary materials. The equations solved
for interstitials or vacancies are
(1.14)
(1.13)
(1.15)
where x=I for interstitials and V for vacancies, Jx is the
flux of the X point defect species, Cx* and Cx are the


DEFECT AND DIFFUSION STUDY IN BORON IMPLANTED SILICON
By
JINNING LIU
A THESIS 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
1996


0-2um
(a)
Figure 4.1 Weak beam dark field PTEM (g^) micrographs of 2xlOl4cm'2 B+ implanted Si at an
energy of (a) lOkeV, (b) 20kkeV, (c) 30keV and (d) 40keV.


22
Under equilibrium conditions at a particular
temperature, the possible sources of point defects are
chemical reactions at the silicon surface, such as
oxidation,81-86 nitridation87-90 and silicidation,91-96 the
impurity clustering and precipitation, 97-100 and the radiation
damage, 20,101-105 etc. Deviations from equilibrium will cause
excess or depletion of either interstitials or vacancies,
which will further affect dopant diffusion. For dopants that
diffuse predominantly by an interstitial/interstitialcy
mechanism, an excess of interstitials or depletion of
vacancies will enhance the dopant diffusion, while for
dopants that diffuse primarily by vacancy mechanism, the same
conditions will retard dopant diffusion.
Two possible detectors for excess point defects in
silicon are dopant profiles and extended defects. Dopant
profiles include both epitaxially grown doping superlattices
or implant profiles. Boron and antimony superlattices can be
used to monitor the in-diffusion of silicon interstitials or
vacancies produced by processes such as surface oxidation,
nitridation, silicidation and aluminization, or the implant
damage. Since boron diffuses mostly via
interstitial/interstitialcy mechanism, the boron epi layer
spike will broaden after oxidation and implantation during
which processes interstitials are injected into the bulk. On
the contrary, silicidation, nitridation and aluminization
will cause vacancy injection or interstitial depletion,
therefore broadening the antimony epi layer spike. The


24
{311} rod-like defects have been examined as the possible
sources of interstitials that cause TED. Cowern et al.103
claimed that the release of interstitials which drive TED of
boron in silicon occurs at two different time scales.
Interstitials emerging from the ion collision cascade give
rise to a primary (ultra fast) pulse of diffusion, and at a
sufficiently high displacement density, emission of
interstitials from {311} defects causes a secondary (much
slower) diffusion transient. Later, Eaglesham et al.112 and
Stolk et al.113 reported that {311} defects are the only
source of interstitials driving TED based on their study of
Si+ implanted Si. They reasoned that interstitials are
emitted from {311} defects in the damage region during
annealing. This emission occurs at a rate which exhibits the
same activation energy as that of the interstitial in
diffusion, and the duration of {311} dissolution process is
in close agreement with that of TED. Our group have been
studying {311} defects and TED in boron implanted silicon.
We found that TED occurs without the formation of {311}
defects if the implant energy (4 keV) and dose (lxlO14 cm"2)
are low.114 The source of interstitials driving TED in this
case was attributed to sub-microscopic mobile boron
interstitial pairs. These B-I pairs play a key role in TED
of boron in silicon, as will be discussed.
Therefore, understanding the correlation between
implantation induced defects (e.g., {311} defects and boron
interstitial clusters) and TED is essential to the


88
implants. For each of those implants showing {311} defects,
the number of the defects decreases with increasing annealing
time. This dissolution process is more rapid for a lower
implant energy.
The areal density of interstitials bound by {311}
defects during the annealing process was measured from the
PTEM micrographs and the results are presented in Fig. 4.2.
The dots in the plot are the actual data points and the lines
are the exponential fitting curves for the data with the form
of eq. (3.1). The emission of interstitials from {311}
defects clearly shows an exponential time-dependence, which
is stronger for lower implant energies. After a 30 min
anneal at 750C, {311} defects in the 10 keV implant almost
dissolved completely. The density of the remaining trapped
interstitials is about 1.8xlO cm-^. For the 40 keV
implant, after an anneal of 1 hr, there is still an areal
interstitial density of about 1.2xl0-*-2 cm~2 contained in
{311} defects. The maximum density of trapped interstitials
observed was about 2xl0l3 cm2 for the 10 keV implant and
about 3.5xl0l3 cm-^ for the 40 keV implant after an anneal
for 5 min and these values are much lesser than the implant
dose of 2x10-*-^ cm2. This implies the existence of immobile
boron interstitial clusters and/or mobile boron interstitial
pairs which might have trapped the missing interstitials
assuming that the "plus one" model holds. There is, however,
another possible configuration that could consume
interstitials, i.e.,
silicon self-interstitial complex--two


(d)
Figure 3.1. (Continued)


130
therefore stays immobile for the two higher doses. The
breaking point between the mobile and immobile part is at a
concentration of about 3x10-*-^ cirT^, the same for both
implants and for every anneal.
The SIMS data of the control and implanted samples were
imported into FLOOPS. Diffusivity enhancement /Db* was
obtained by matching the initial profile with the target
profile assuming an intrinsic diffusivity with the form
Db*=0.757exp(-3.46eV/kT) (eq.(3.8)). We extracted the
enhancement factor for control sample 1 (as-grown plus 800C
1 hr anneal), control sample 2 (pre-anneal plus 750C 1 hr
anneal) and control sample 3 (pre-anneal plus 750C 4 hr
anneal). The purpose of control sample 1 was to examine how
much enhancement would come from the excess interstitials
induced by the APCVD growth process. The purpose of control
samples 2 and 3 was to test whether the nitrogen ambient
indeed functions as an inert ambient. The enhancement factor
for these three samples was 4.1, 4.9 and 4.7 respectively.
This small enhancement might be due to the residue oxygen in
the nitrogen ambient. However, it might also just result
from the experimental errors, including the SIMS depth error
range of 5% and the annealing temperature accuracy of 10C.
The number of excess interstitials introduced by the APCVD
process is obviously not high enough to cause much enhanced
diffusion.
Fig.5.9 is a plot of the time-averaged diffusivity
enhancement /Db* of the buried boron spike as a function


Projected Range R ()
47
Implant Energy (keV)
Figure 2.4. Projected range Rp of 5keV to 40keV B+ implanted Si,
determined from XTEM micrographs and SIMS profiles.


123
support the idea of "plus 1" model over the idea of collision
cascade as the source of the interstitials. However, this
time constant increase with increasing dose seems to become
less pronounced when the dose is increased further. From
2xl014 cm 2 to 5xl0^-4 cm-2, the time constant changes from
424 sec to 3029 sec. But from 5x10-*-4 cm-2 to 1x10^-^ cm-2,
the time constant only changes from 3029 sec to 4604 sec. As
we discussed in Chapter 2, the threshold dose for the
formation of stable sub-amorphization (type I) dislocation
loops lies somewhere between 2xl0^-4 cm-2 and 5xl0-*-4 cm-2 at
20 keV (see Table 2.2). The presence of loops seems to
expedite the dissolution process of {311} defects.
Interstitials emitted from {311} defects will feed the
dislocation loops and the sink effect of loops makes the
{311} dissolution faster.
By measuring the loop size on the PTEM micrographs, the
concentration of interstitials trapped by dislocation loops
was obtained. By assuming a circular loop, the radius of
each loop or partial loop was measured along its longest axis
and the corresponding loop area was calculated. The
concentration of atoms bound by loops could be estimated by
multiplying the fraction of loop area by the atomic density
of atoms on the {111} plane, which is 1.6xl015 cm-2.
The density of interstitials trapped in {311} defects
and dislocation loops after anneals for 5 min and 2 hr at
750C as a function of implant dose is shown in Fig. 5.4. We
can see that the density of interstitial in {311} defects or


15
(c)
Figure 1.5. (Continued)


121
I) dislocation loops. Our aim is to investigate the effect
of varying dose on the defect behavior as well as on
transient enhanced diffusion.
5.2 Implant Dose Effect on Defect Behavior
5.2.1 Experimental Procedure
To study the dose effect on defect evolution, we
selected a single row from the implant matrix (see section
2.2). The implant energy was kept as 20 keV and the dose was
varied from 5x10-'-^ cm" 2 to 1x10^-^ cm" 2. Subsequent furnace
anneals were carried out in a nitrogen ambient for times
between 5 min and 2 hr at 750C. PTEM micrographs were taken
under weak beam dark field imaging condition using g220 beam.
The {311} defect density, average defect size, areal density
of interstitials trapped by {311} defects and the defect
distribution after annealing were measured from the TEM
micrographs.
5.2.2 Defect Microstructure and Dissolution Process
PTEM micrographs of the defects after boron implantation
at 20 keV to doses of 1x10-*-^ cm"2, 5x10^-^ cm"2, and
lxlO15 cm2 and anneals at 750C from 5 min to 2 hr are shown
in Fig. 5.1 (a)-(c). No defects were observed for the lowest
dose, 5x10^3 cm2, and therefore the pictures of these
samples are not included. The 1x1014 cm-2 implant contains
only a small number of {311} defects after 5 min and 10 min


CHAPTER 6
SUMMARY AND FUTURE WORK
6.1 Summary
Transient enhanced diffusion presents a major obstacle
for the optimization of shallow junction devices. The excess
interstitials introduced by an ion implantation process are
the driving force for TED. In order to gain better control
of dopant diffusion, it is crucial to understand the behavior
of ion implantation induced defects and their correlation
with dopant diffusion. This work represents an effort
towards meeting this goal. A brief summary of this work is
given below.
A low energy B implant matrix has been employed in this
study. The formation threshold doses of both {311} rod-like
defects and [110] sub-amorphization loops were examined using
TEM. Both threshold doses increase with decreasing implant
energy and the one for {311} defect formation is lower than
that for loop formation. Displaced atom density calculated
from TRIM seems to be able to predict the formation threshold
very well.
One particular sample in the matrix was used to study
the effect of annealing on {311} defect evolution and TED
behavior. {311} defects dissolve rapidly during annealing.
151


154
This work supports the following models for {311} defect
evolution and for interstitial storage and release during
annealing in B+ implanted silicon.
Rafferty et al.134 proposed the following theory to
explain the {311} defect formation and dissolution. The
model can be represented by
n r
^ = 4jcaaD,I(x,t)C(x,t) Cf exp(h~)
at a kT
(6.1)
where a is the capture radius of an interstitial by a {311}
defect expressed in units of the average interatomic spacing
a. D^Doexp (-Em/kT) is the interstitial diffusivity. I(x,t)
is the concentration of free interstitials and C(x,t) is the
concentration of interstitials trapped in {311} defects. The
first term on the right side of eq.(6.1) is the {311} defect
formation term. A defect grow whenever diffusing
interstitials are trapped by it. The second term is the
dissolution term. An interstitial in a {311} defect may
escape at a rate given by the interstitial hopping frequency
and the binding energy to the defect. At steady state, the
first term is balanced by the second. The interstitial
supersaturation Ie/I* is given by
/ 1 Eh Ef
~ ;exP(
I 4nar
kT
(6.2)
where T is a dimensionless prefactor that arises from the
entropy of formation of an interstitial and Ef is the
formation energy of an interstitial. The time to dissolve
the {311} defects is given by


Defect Defect Defect Defect
Density (cm ) Density (cm ) Density (cm'2) Density (cm'2)
110
| I I1 I IIi[niIiir
20keV 10min
60 260 460 660 860 1060 1260 1460
1011
1010
109
108
107
60 260 460 660 860 1060 1260 1460
11
10
101 j.
109 j.
108 j.
^07
tIiIIIr
J' l l I '| l ll | l l I
20keV 30min .
60
260 460 660 860 1060 1260 1460
{311} Defect Size (A)
(b)
Figure 4.6. (Continued)


Boron Concentration (cm'3) Boron concentration (cm'3)
114
I 1I 1 I II I I 1 I I I II 1IIU I I lifMi 1 i I I i i
0 1000 2000 3000 4000 5000 6000
Depth ()
(c)
Depth ()
(d)
Figure 4.7. (continued)


Defect Defect Defect Defect
Density (cm 2) Density (cm'2) Density (cm'2) Density (cm'2)
109
1011
1010
109
108
107
60 260 460 660 860 1060 1260 1460
1011 |1 1 ~r i
io10 r
10a
10a
I I 1 I 1 '
10keV 30min
Mi
MKMMl I I I I I
10 60 260 460 660 860 1060 1260 1460
{311} Defect Size ()
(a)
Figure 4.6. {311} defect size distribution for (a)10keV
(b)20keV (c)30keV and (d)40keV implants, annealed at 750C.


149
Annealing Time (min)
Figure 5.9. Time-averaged boron diffusivity enhancement
/Db* as a function of annealing time for different
implant doses.


Time Constant of {311} Defect Dissolution (sec)
106
Implant Energy (keV)
Figure 4.3. Time constant of the dissolution process of the
interstitials trapped in {311} defects as a function of
implant energy, annealed at 750C.


125
interstitials either diffuse to the surface or into the bulk.
For the 5xl0l4 cm2 implant, the number of interstitials
released from {311} defects between 5 min to 2 hr is higher
than that trapped in the loops. Because of the small density
of the loops they do not act as a very strong interstitial
sink. While for the 1x10-*-^ cm~2 implant, loops trap more
interstitials than those released from {311} defects.
Although the number of interstitials introduced into the bulk
is the highest for this dose, there is no dramatic increase
in transient diffusion compared with a lower dose (e.g.
2x10-1-4 cm-2) lue to the presence of larger density of
interstitial traps--dislocation loops. This will be
discussed in detail later.
5.2.3 Defect Density. Size and Distribution during Annealing
The {311} defect density of different implants as a
function of annealing time measured from PTEM images is shown
in Fig. 5.5. For all annealing times, the defect density
increases with increasing dose. After a 5 min anneal, the
{311} defect density is about 3.3x10-1-0 cm-2 for the
2x10-1-4 cm 2
implant,
and about
1.1x10-1-1 cm 2
for
the
1x10-1-5 cm-2
implant.
For each
implant dose,
the
defect
density decreases with increasing annealing time, with the
decrease occurring faster for a lower dose. The reason for
the strong dose dependence of defect density might be that in
a sample implanted with higher dose, the interstitial
supersaturation level is higher. As a result, the number of


29
stage ion mill by using two Ar+ ion guns at a gun voltage of
4 kV and a gun current of 0.5 mA until a small hole is
obtained at the sample strips' interface. The region in the
vicinity of the hole is sufficiently thin (<0.5 |im) to be
transparent to the electron beam of the TEM.
1.3.3 Secondary Ion Mass Spectroscopy
In secondary ion mass spectroscopy (SIMS), an energetic
beam of focused ions is directed at the sample surface in a
high or ultrahigh vacuum environment. The momentum transfer
from the impinging primary ions to the sample surface causes
sputtering of the surface atoms and molecules. The most
commonly used sputtering ions are Cs+( 02+, 0+ and Ar+ in the
energy range from 2 to 20 keV at angles of incidence between
45 and 90. 02+ ions are used in this study to enhance
secondary ion yield. Some of the sputtered species are
ejected with positive or negative charges. These are termed
secondary ions. The secondary ions are then mass analyzed
using a mass spectrometer. SIMS can be used to obtain a
depth concentration profile of the near-surface region to a
resolution of 2~30 nm at a detection limit of 10^-10^ cm-^ .
For further information see ref.115
1.3.4 Florida Object Oriented Process Simulator
FLOOPS modeling program is a physically based two-
dimensional process simulator capable of modeling a complex
device structure. It uses local point defect concentration


144
Annealing Time (sec)
Figure 5.6. Average {311} defect size as a function of
annealing time in 20keV B+ implanted Si for different doses.


160
Boron doped silicon has been found to show less {311}
defects when the doping level is increased. The doping of
other species, such as P, may also influence the defect
behavior.
5. Surface recombination effect
We have discussed that the effect of surface
recombination coupled with other factors which altogether
result in the lack of {311} defects in the 5keV implant and a
smaller enhanced diffusion than higher energy implants. It
would be interesting if the surface recombination effect can
be quantified and sorted out from the others.


91
formation and coarsening of {311} defects in the higher
energy implants. However, the defects dissolve faster for a
lower energy. As annealing time increases further, defect
density is higher for a higher energy implant. The
difference in defect density becomes more significant when
the annealing time increases. After a 30 min anneal, the
10 keV implant shows a defect density of only 3.2x10 cm-2,
while the 40 keV implant shows a density of 2.7x10 cm2.
The average {311} defect size as a function of annealing
time for different implant energies is shown in Fig. 4.5.
The average defect size increases with annealing time
initially, and then decreases when the defects enter a purely
dissolution regime. This final dissolution regime is only
observed for the lOkeV implant for the current annealing
times. As the implant energy increases, the average defect
size increases. After an anneal of 15 min, the 20 keV
implant shows an average defect size of about 518 while
for the 40 keV implant, it is about 952 .
Figure 4.6 (a)~(d) show the size distribution of {311}
defects after anneals at 750C for various times for 10 keV
to 40 keV implants. Higher energy implants have a broader
distribution than lower energy implants. The figure also
shows a larger average defect size and a slower dissolution
rate for a higher energy. For each energy, the average
defect size increases while the defect density at each size
decreases during annealing. This behavior together with the
fact that the total number of interstitials trapped in {311}


96
an anneal of 3 min, the 5 keV implant shows an enhancement of
about 260, whereas the 40 keV implant shows a much greater
enhancement of about 2800. As the anneal time increases to
longer than 3 min, this difference is less distinguishable
for energies above 5 keV.
4,3.3 Correlation Between Defects and TED
The average boron diffusivity at a particular depth of
the sample can be expressed as
(4.1)
where Ci is the instantaneous value of interstitial
concentration at time t and t is the diffusion time. If the
instantaneous interstitial supersaturation level Cj/Ci* is
greater than 1, we will see the enhancement of diffusivity.
If we assume that the excess interstitials introduced by an
implant are trapped in interstitial-contained defects (in our
case, {311} defects and mobile B-I pairs) in the initial
stage of the annealing, then we can consider these defects as
an interstitial reservoir that keeps Cj/Cj* constant for a
short period of time (tto all the {311}
defects have dissolved and all B-I pairs have broken up, then
the reservoir is empty and Cj/Cj* drops abruptly to 1.
Mathematically this can be expressed as
Cj_ 1 + E...(t < t0)
c; l...(i>r0)
(4.2)
where E is the supersaturation of interstitials. The
integral of eg. (4.1) would then become


50
function of the reciprocal of temperature to obtain the
activation energy for the {311} defect dissolution process.
3.2.2 Defect Evolution: Microstructure. Interstitial Density
and the Activation Energy of Interstitial Decay
Figure 3.1 (a)~(d) shows the PTEM micrographs of the
sample after anneals at 650C, 700C, 750C and 800C for
various times. The formation and coarsening/dissolution
processes of {311} defects are faster at higher temperatures.
After 2 hr at 650C, these defects are not fully formed yet.
The defects in their initial stages exhibit a "black dot"
contrast. After 14 hr, {311} defects start to form.
However, Ostwald ripening along with the dissolution process
is so slow at this low temperature that the number of the
defects is relatively large and their size is relatively
small even after 48 hr of annealing. If the annealing
temperature is increased to 750C, {311} defects are observed
after only 5 min and the dissolution process becomes more
rapid. At 800C, {311} defects almost dissolve completely
after 13 min of annealing.
We have measured the net interstitials trapped in {311}
defects from the PTEM micrographs after each anneal. Figure
3.2 shows the {311} defect dissolution process at each
temperature. As we can see, the dissolution process shows an
exponential dependence on annealing time. The dots in Fig.
3.2are the actual data points and the lines are exponential
fitting curves of the form


10
and provide leakage paths if they span across a
junction. 42,49,50
2. Clamshell defects
If the implant energy is sufficient but the dose is not
too high, it is possible to produce a buried amorphous layer.
Upon annealing two amorphous/crystalline interfaces meet and
extended defects may result. These defects consist of larger
dislocation loops compared to end-of-range defects.49,51
We have discussed briefly each type of extended defects
due to ion implantation. Visible damage from boron implants
employed in this study consisted predominantly of {311}
defects in the annealing temperature range that we used. In
the next section, we will review some previous studies on
{311} defects in detail.
1.2.3 (311) Rod-like Defects
As we discussed in the last section, {311} defects can
form under both sub-amorphization and amorphization
conditions, providing the annealing temperature is low and/or
the annealing time is short. This is summarized in Fig. 1.2.
Because of the low mass of the boron ions, the silicon
substrate surface is not amorphized if the dose is below
5xl015 cm 2. Under such sub-amorphization conditions, {311}
defects form at lower doses than dislocation loops. They are
relatively unstable. Upon annealing, they either evolve into
dislocation dipoles through an unfaulting reaction or
dissolve. The dissolution provides interstitials to the


147
i
E
o
c
o
cC
c
0
o
c
o
O
c
o
o
CQ
0 2000 4000 6000 8000 1 104
Depth ()
(a)
Depth ()
(b)
Figure 5.8. SIMS profiles of boron in 20keV B+ implanted Si
superlattice at a dose of (b)5xl013cm-2, (c) 2xl014cm-2, (d)
lxl0-*-^cm-2. (a) shows the control sample profiles.


161
Surface Recombination, B-B-I Clustering,
I-V Recombination

Dose
Implant Damage, {311}/Loop Formation
Energy-
Figure 6.1. A summary of different types of defects that
presumably exist in B+ implanted Si, effects of implant
energies and doses.
Implant Damage, B-B-I Clustering,
{311}/Loop Formation


21
most of the transition metals78,79 The mechanism of Si self
diffusion appears to be vacancy controlled at temperatures
<1000C and interstitial controlled at higher temperatures.80
Boron is the diffusing species that will be discussed in
this study. The set of nonequilibrium reactions involved in
the kinetic model for boron diffusion is described as
follows:
Bs + I <-> (BI)
(1.4)
(BI) + V <-> Bs
(1.5)
I + V <-> [0]
(1.6)
where Bs and (BI) are the substitutional boron and the boron
interstitial pair, respectively, and [0] denotes the silicon
lattice site.
The relationship between point defect concentration and
dopant diffusion is given by Eq. (1.3). To understand dopant
diffusion in silicon, it is necessary to understand how
different processes affect point defect concentration.
Because of its small value relative to the concentration of
silicon atoms residing on the lattice sites, the
concentration of point defects is not a quantity that can be
measured directly. Diffusion studies are mostly used to gain
information of the interstitials and vacancies. An enhanced
diffusion of a species which diffuses by interstitial (or
vacancy) mechanism indicates a supersaturation of
interstitials (or vacancies) in the bulk. Ways of
quantitatively monitoring these point defects will be
discussed later.


155
T =
4p*,6 +
exp(-
)
(6.3)
where Rp is the projected range, Q is the implant dose,
Ns=a3 is the number of lattice sites and Em is the migration
energy of an interstitial. This model predicts that the
dissolution time of {311} defects is proportional to the
projected range, and therefore to the implant energy (see
Fig.2.4), and is proportional to the same degree to the
implant dose. A strong temperature dependence is shown by
the exponential term. This agrees with our results, which
show an increase in T with increasing energy and dose and with
decreasing temperature. However, we observed that doubling
the dose increases T more than doubling the energy. This can
not be explained by the current model.
In Chapter 3, we used the variation of the energy of an
interstitial trapped in {311} defects to explain the Ostwald
ripening process. In fact the experimentally observed
Ostwald ripening process is also in agreement with the model
discussed above. Since larger {311} defects are more stable
than smaller ones, Eb must be larger for a larger defect.
From eq.(6.2) we can see that Ie/I* is lower for larger Eb-
This lower interstitial supersaturation is the driving force
for the interstitial flow from smaller defects to larger
ones, i.e., the Ostwald ripening process.
Recently a homogeneous nucleation mechanism for the
{311} defects was proposed by Chao et al.135. They claimed
that the diffusion enhancement, which is proportional to the


158
annealing temperature is high enough or the time is long
enough (e.g. 800C > 1 hr), the loops start to dissolve,
release interstitials and give rise to more diffusion. The
dissolution of loops occurs earlier than the de-clustering of
the immobile profile peak, but later than the first
saturation of TED.
Assuming the presence of immobile B-I clusters and
mobile B-I pairs, a summary of different types of defects
formed in samples implanted with various energies and doses
based on our implant matrix is shown in Fig.6.1. and a
schematic of different sources of TED during an annealing
sequence is shown in Fig.6.2. Quantification of how the
interstitials partition themselves into B-I clusters or B-I
pairs will allow us to gain better understanding on dopant
diffusion.
6.2 Future Work
There is still a significant amount of research to be
done in the area of defect and diffusion behavior after ion
implantation. The following are some suggestions.
1. High annealing temperature regime
An activation energy of 1.6eV for the TED saturation
process was obtained through the diffusion study of 20keV
2xl014cm-2 B+ implanted silicon after furnace anneals at
650~800C for various times (see Chapter 3). The sample
contains {311} defects only. A much higher activation energy
of 4.3eV has been reported by Michel for 60keV 2xl0l4cm-2 B+


Table 2.2 Displaced atom density (#/cm2) from TRIM calculations
(B+ implanted silicon)
5keV
lOkeV
20keV
30keV
40keV
5xl0l3cm'2
2.56el5
4.5 lel5
7.35el5
9.60el5
1.16e16
lxl0l4Cm-2
5.12el5
9.02el5
1.47el6
1.92el6
2.3 lel6
2xl0^cm'2
1.02el6
1.80el6
2.94el6
3.84el6
4.62el6
5xl0l4crrr2
2.56el6
4.51el6
7.35el6
9.60el6
1.16e17
lxl0^cm_2
5.12el6
9.02el6
1.47el7
1.92el7
2.31el7


124
loops is much less than the implant dose. Assuming that the
net number of excess interstitials is roughly equal to the
dose (plus 1 model) and the missing interstitials are all
trapped in sub-microscopic boron interstitial complexes,
including mobile B-I pairs and immobile B-I clusters, then
the percentage of the interstitials trapped in these boron
interstitial complexes after a 5 min anneal at 750C is
87.5%, 87.8% and 88.6% for the 2xl014 cm"2, 5xl014 cm-2 and
lxlO45 cm-2 implant respectively. Within the experimental
error range we can say that this percentage does not change
with implant dose. This is different from the energy
dependence which we discussed in Chapter 4. The percentage
of interstitials trapped in boron interstitial complexes
decreases with increasing energy because of the lower boron
peak concentration due to the spreading out of the dopant
profile. Here, when implant dose is increased, both the
boron peak concentration and the implant damage increase
proportionally. As a result more boron interstitial
complexes and more {311} defects/loops are formed and the
percentage of interstitials trapped in B-I complexes versus
{311}s/loops stays constant. This argument, however, cannot
be used to predict the phenomenon in a sample without {311}
defects or loops.
It is interesting to notice from Fig. 5.4 how
interstitials re-distribute themselves during the annealing.
The {311} defects for the 2xl014 cm-2 implant dissolve almost
completely after 2 hr anneal at 750C. The trapped


12
three directions of {311} defects. Two of them are
orthogonal to each other while the third one forms a 45
angle to the first two sets. This morphology will be
explained in the following paragraphs.
Figure 1.3. Example of {311} defects in 30keV lxlO-'-^cm-^ b+
implanted Si under weak-beam dark field imaging condition
with g220 reflection, annealed at 750C for 5min in nitrogen.
{311} defects are elongated in [110] directions and
consist of interstitials precipitating on {311} habit planes.
A unit cell of silicon lattice with {311} defects lying in
six [110] directions is schematically shown in Fig. 1.4. The
reflection vector g was chosen to be 220 when plan-view
transmission electron microscopy pictures were taken in this


94
sample temperature, the error range for the 3 min sample
could be 30C. The initial boron profile as well as the
profiles after implantation and annealing were measured by
SIMS with 3.0 keV C>2+ at a current of 50 nA. The raster size
was 150x150 |lm2 and the analyzed area was 60x60 Jim2. The
broadening of the B profiles was simulated using the process
simulator PROPHET. The spikes that merged into the B implant
profile were excluded from the simulation.
4.3.2 Dopant Diffusion Behavior
SIMS profiles of boron in the superlattices implanted
with 5 keV to 40 keV boron ions are shown in Fig. 4.7(a)~(d).
As can be seen, a higher energy results in a deeper implant
profile and a greater enhanced diffusion of the buried boron
spikes. The time-averaged boron diffusivities for a
particular spike, , were extracted by finding the value
of Db in the simulation that resulted in the best match
between diffused and target profiles. Using the literature
value of Db*=0.757exp(-3.46/kT) (eq. (3.8)), diffusivity
enhancement /Db* is thus obtained. For the inert ambient
annealed control samples, it is expected that no diffusion
enhancement occurs, i.e., /Db*=1 or =Db*. We have
extracted via simulation for the 2hr annealed control
sample. Taking an average over different spikes since
is depth-independent, we found that is about
1.6x1017 cm2/s, about 2 times greater than the predicted
value (7.2xl0-18 cm2/s), which is within the error range of


/D
117
1 10 100 1000
Annealing Time (min)
Figure 4.9. Enhancement of B diffusivity /Db* at a depth
of 3600 away from Rp as a function of annealing time for
different implant energies.


Defect Defect Defect Defect
Density (cm'2) Density (cm'2) Density (cm'2) Density (cm'2)
79
1060 1260 1460
800C5min
i i i i.i.i.i
Imlii
260 460 660 860 1060 1260 1460
800C1 Omin
iM MB
260 460 660 860 1060 1260 1460
11
10
10
10£
1081
1071
10
800C13min
j. 1 ii ill ii Ii.llI i i
60 260 460 660 860 1060 1260 1460
Average {311} Size ()
Figure 3.6.
(d)
(Continued)


4
As mentioned in the previous section, the common trend
in the development of solid state devices, either memory or
microprocessor, is the decrease in device dimensions. MOS
devices with sub-micron dimensions require junction depths of
0.1~0.2)lm. The most commonly used technique in forming
shallow junctions is to implant dopant species at low
energies into the silicon substrate.
To form shallow p+/n junctions, direct ion implantation
typically requires a very low energy boron implant preceded
by silicon or germanium amorphization to reduce channeling,
or a low energy BF2+ implant which uses the F to amorphize
the surface. 22-24 The use of pre-amorphization by Si+ or Ge+
introduces another implant step. In addition, the
amorphization of the surface will introduce end-of-range
damage, which may not be annealed out even at high
temperatures. This end-of-range damage is a source of
interstitials and, if located across the junctions, will
result in unacceptably high leakage currents. For the use of
BF2+ implants, the fluorine ions can affect the permeability
of the gate oxide to boron diffusion. This can cause a
penetration problem of boron atoms from the gate region
through the gate oxide to the channel region, therefore
undesirably affecting the threshold voltage of the device.
Thus one would like to avoid BF2+ implantation. One
alternative is to simply use low energy B+ implants.
However, there are several problems associated with B+
implants alone.


CHAPTER 3
EFFECT OF ANNEALING ON DEFECT EVOLUTION AND TRANSIENT
ENHANCED DIFFUSION
3.1 Overview
Due to the limitation that TED imposes on the minimum
device dimensions, it has been the subject of considerable
studies for many years, yet its mechanism is still under
debate. TED has been attributed to the excess point defects
introduced by the implantation process. The magnitude of TED
is related to the implantation conditions, the annealing
conditions and the surface and bulk recombination effects.
Recently, Stolk et al.113 have suggested that TED occurs by
the emission of silicon self-interstitials from rod-like
[311} defects during annealing. The activation energy
(3.60.1 eV) for the {311} defect dissolution process relates
closely with that of TED (3.5 eV) Later, Zhang et al.114
reported that they observed TED in B+ implanted samples
without {311} defects and they attributed the interstitial
source of TED to submicroscopic clusters. More recently,
Cowern et al.123 studied the role of carbon and boron clusters
on the diffusion process. The number of self-interstitials
trapped per clustered impurity atom is about 1.15 for carbon
and about 1 for boron. This provided further evidence to the
48


27
Monte Carlo simulator) yields a much more accurate picture of
the ion distribution just after implantation.
Figure 1.7. Schematic illustration of a Gaussian
distribution resulting from ion implantation.
1.3.2 Transmission Electron Microscopy
The transmission electron microscope (TEM) is used to
obtain information from samples which are thin enough to
transmit electrons. A bright-field image is formed if the
directly transmitted beam is selected and a dark-field image
is formed if a diffracted beam is selected. Dark-field
images normally have higher resolution than bright-field
images and this is the mode we used in our study. The defect
structure can be analyzed by both plan-view and cross-
sectional TEM. Plan-view TEM (PTEM) is used to study the
defect evolution during annealing and cross-sectional TEM


128
of defect dissolution. The interstitial sink effect in the
lxlO-*-^ cm2 implant is more prominent than that in the
5xl014 cm"2 implant due to the higher loop density resulted
from the higher dose. Both defect density and average defect
size increases when implant dose increases. The evolution of
{311} defects indicates an Ostwald ripening process and a
defect dissolution process. The standard deviation of defect
size distribution is larger for higher doses.
5.3 Implant Dose Effect on Transient Enhanced Diffusion
5.3.1 Experimental Procedure
It is critical to understand how all the defect
evolutionary changes affect the transient enhanced diffusion
process. In order to study this again we separated the
implant profile from the diffusion marker layer. A thin 8-
doping boron marker layer was used to monitor boron diffusion
behavior after ion implantation and annealing. The layer was
grown via atmospheric pressure chemical vapor deposition
(APCVD) on p-type (100) Czochralski-grown silicon substrate,
followed by growth of an overlayer of approximately 7000 A of
undoped silicon. The full width at half maximum (FWHM) of
the boron spike is about 500 A and peak concentration is
about 2xl0l8 cm"3. This low concentration allows dopant
diffusion to remain in the intrinsic regime at our processing
temperature and also reduces boron interstitial cluster
formation in the doping spike. The as-grown wafers were pre-


120
energy of 200 keV. Since the annealing temperatures were
above 750C and the annealing times were long enough to
remove the {311} defects that would possibly be seen after
low budget anneals, it was claimed that few extended defects
were present in the sample. He studied the effect of varying
Si implant dose on the diffusion behavior of a separately
implanted deeper boron profile and found that the amount of
diffusion increased with increasing dose but doubling the
dose did not double the diffusion coefficient. This sub-
linear dependence suggests a saturation of enhanced diffusion
with increasing implant dose.
Solmi et al.132 studied boron diffusion in 20 keV and
30 keV B+ implanted silicon to doses from 2x10^4 Cm2 to
5x10-L5 cm-2. Thermal treatments were carried out at 800C to
1000C. They observed a reduced displacement of the profile
for the higher dose associated with extended defect
formation. The transient time for the enhanced diffusion was
reduced by a factor of 2 for doses higher than 5x10^-^ cm2.
They attributed this reduction to the formation of extended
defects at the projected range at these higher doses. The
extended defects act as a sink for the interstitials, thus
reducing the interstitial supersaturation level.
In this section we will discuss our study on boron
diffusion after boron implantation at low and medium doses
and anneals at a relatively low temperature (750C).
Different doses give rise to different types of extended
defects, including {311} defects and sub-amorphization (type


CHAPTER 1
INTRODUCTION
1.1 Motivation and Objective
Since the development of the first integrated circuits,
the minimum feature sizes of devices have been reduced
dramatically. This miniaturization is expected to continue
and by the year 2000, the number of transistors per chip is
expected to reach one billion. The scaling of devices to
smaller dimensions both vertically and horizontally implies
lower voltages, thinner dielectrics and more abrupt and
shallower dopant profiles.1-3 The new devices will possess
the properties of higher packing density, higher speed and
lower power consumption.4,5 However, before the realization
of this goal, many challenges remain in processing and
packaging. One of them, for example, is to scale the
source/drain junction depth and the lateral penetration under
the gate in order to maintain a threshold voltage for the
MOSFET independent of the channel length.
Among the various methods available for forming shallow
junctions,6-11 ion implantation is currently the most widely
used one, due to its precise controllability, uniformity and
reproducibility. However, future devices require shallow
p+/n junctions with depths less than 1000 to minimize short
1


41
doses from SxlO-^cm2 to lxiol^cm-2. TEM was use to evaluate
the presence of both types of defects. It was observed that
the threshold dose for both {311} defect and sub-
amorphization dislocation loop formation increases
significantly with decreasing implant energy. This threshold
is higher for boron implants than for silicon implants
presumably because of the formation of boron interstitial
clusters and mobile boron interstitial pairs. As discussed
earlier there are several possible explanations for the
higher threshold dose with decreasing implant energy. These
include a decrease in the displaced atom density, increased
formation of boron interstitial clusters and/or increased
surface recombination. Experimentally a displaced atom
density of about 1.5xl0l6 cm-2 for {311} defects and
2.6xl01^ cm-2 for sub-amorphization loops from TRIM
calculations reasonably predicts defect formation for these
implant conditions. However, the results of Zhang et al.114
and Eaglesham et al.(private communication) clearly indicate
that boron is a very effective interstitial trap and that
increasing the boron concentration clearly results in an
increase in the concentration of interstitial trapping.


5
In addition to the channeling effect of B+ implantation
without pre-amorphization which can significantly increase
the junction depth, the post-implantation annealing results
in deeper profiles greater than that allowed by intrinsic
diffusion due to TED. TED is associated with the interaction
of mobile silicon interstitials with boron atoms.18'25,26
Implants into a pre-amorphized substrate can eliminate
channeling tails, but TED is still present. For B+ implant
the amount of TED is determined by both the number of excess
interstitials in the damage region and the partitioning of
these excess interstitials into different forms such as
mobile boron interstitial pairs, boron interstitial clusters,
{311} defects and dislocation loops and the role of the
surface. 27-29 This implantation induced TED will be discussed
in greater detail in section 1.2.4.
Shallow junctions are also formed in combination with
silicidation of source and drain. In general the implant and
anneal are done before metal deposition an silicide
formation. However, the implants could also be performed
through a deposited metal layer (Co or Ti mainly) before
silicide formation, 30-34 or into an already formed silicide
layer that then acts as a diffusion source (SADS) upon
subsequent annealing. 35-37 In the first two cases, a silicide
layer is formed after annealing with a steep implant profile
and a shallow junction. During the silicide formation, a
portion of the heavily doped silicon region is consumed and
the silicide interface moves into silicon, while dopants


64
2 hr at 700C. Lower temperature annealing causes more
diffusion before the saturation point is reached.
The SIMS profiles of boron in the as-implanted and
annealed samples were imported into FLOOPS and the
enhancement of diffusivity /Db* after each anneal was
obtained after the simulations. Here is the time-
averaged boron diffusivity and Db* is the intrinsic
equilibrium value. Db* has the form of
Db*=0.757exp(-3.46eV/kT) (3.8)
as a default value in FLOOPS, which was originally taken from
Fair's review article129. This value is about
1.03x10-*-9 cm^/sec, 9.64xl0--*-^ cm^/sec, 7.22x10^-^ cm^/sec
and 4.48xl0"l7 cm^/sec at 650C, 700C, 750C and 800C,
respectively. Since the dopant profiles are not perfectly
Gaussian, instead of choosing the whole profile, we chose a
portion of the tail region (approximately 1600 A-2600 ) as
the target profile during simulation. Figure 3.9 shows these
simulation results of /Db* as a function of annealing
time for each temperature. As we can see, the time-averaged
diffusivity enhancement decreases with increasing annealing
time. Since /Db* is proportional to the interstitial
supersaturation level, /Ci*, in the case of B diffusion,
the decrease of /Db* represents a process during which
the interstitial concentration returns to its equilibrium
value. Here is the time-averaged interstitial
concentration in the bulk, and Ci* is the equilibrium value.
Ci* is smaller and /Ci* is higher at lower temperatures


34
2.2 Experimental Procedure
Czochralski-grown (100) n-type 8-20 cm 150 mm wafers
were implanted with boron ions at energies of 5 keV, 10 keV,
20 keV, 30 keV and 40 keV to doses of 5x10^--^ cm-2,
lxl0-*-4 cm-2, 2x10-'-^ cm-2, 5x10^4 cm-2 and lxlO^S cm-2. The
tilt/rotation angles were 5/0. The implant current was
3 mA. During ion implantation, the samples were kept at room
temperature using water cooling. Furnace anneals were then
performed at 750C for 5 min or 900C for 15 min in a
nitrogen ambient to study the formation threshold of {311}
defects and sub-amorphization dislocation loops respectively.
Plan-view transmission electron microscopy (PTEM) samples
were prepared using standard jet-etching procedures and
Cross-sectional TEM (XTEM) samples were prepared using ion
milling. Micrographs were taken from each sample to examine
the presence or absence of secondary defects. The g220
reflection was used to acquire all the micrographs under
weak-beam dark field imaging conditions. A defect density of
1.2xl07 cm-2 was used to distinguish between samples with and
without extended defects, i.e., if no defect is observed in
three randomly-picked areas each with a size of 7x10 cm2
under a magnification of 50000X, then it is stated that there
are "no" defects in that sample.


0.2|im


92
defect decreases with increasing time indicates that the
Ostwald ripening process occurs along with the defect
dissolution process. Large {311} defects grow by absorbing
the interstitials released from small defects as a result of
higher local interstitial concentration around small defects.
Meanwhile, large defects themselves release interstitials
into the bulk. This resulted in the final defect
distribution we have observed.
As a summary, the {311} defect microstructure and
annealing behavior as a function of implant energy (5 keV ~
40 keV) have been studied using TEM. No defects are observed
in the 5 keV implant. There are more interstitials bound by
{311} defects in higher energy implants although the maximum
density observed (about 2~3.5xl0-*-3 cm-2 after an anneal of
5 min at 750C) are all far less than the implant dose
(2x10-*-^ cm2). The defect density is higher for a lower
implant energy after a short anneal time (10 min) and then
becomes lower as time increases. The average defect size is
larger for a higher implant energy. The size distribution of
defects shows a co-occurrence of the Ostwald ripening process
and the dissolution process. The lack of {311} defects in
the 5 keV implant and the decrease in density of trapped
interstitials by (311} defects with decreasing implant energy
might be the result of one or a combination of the following
factors: decreased damage product with decreasing energy,
increased I-V recombinations, increased boron interstitial
cluster formation and an increased surface recombination.


Defect Defect Defect Defect
Density (cm'2) Density (cm'2) Density (cm'2) Density (cm2)
145
1011
1010
109
108
107
60 260 460 660 860 1060 1260 1460
11
10
10
10
10s
10£
a rJ I 111,, 1^11 i 1 1 1 L1-X..-1Li-Jj I 1 I I L
10 60 260 460 660 860 1060 1260 1460
T|III | III|IIII'" I I1|
5e14 1hr:
60 260 460 660 860 1060 1260 1460
1011
101
109
108
107
i tt I i i i | n | i i i | i i i | n i i iiiinr"|
5e14 2hrs -
60 260 460 660 860 1060 1260 1460
{311} Defect Size ()
(a)
Figure 5.7. {311} defect size distribution for (a)5xl014cm-2
and (b)lxl015cm-2 implants, annealed at 750C.


159
implant after furnace and rapid thermal anneals at
800~1000C. His sample, according to our defect formation
threshold study, should contain both {311} defects and
dislocation loops. It would be interesting to anneal our
samples in that high temperature regime to investigate
possible difference in diffusion mechanism and the effect of
loops on TED.
2. Electrical activity measurements
Through spreading resistance profiling (SRP)
measurements we can determine the electrically active portion
of the implanted boron atoms. This can help us gain
information regarding the clustering/de-clustering motion of
the boron peak region and its effect on TED in the tail
region.
3. Effect of implant species, dose rate and implant
temperature
Our group's preliminary results show that P implants
have higher formation threshold than Si and B implant.
Implant dose rate, if changing within one order of magnitude,
does not affect TED behavior in crystalline silicon, however,
higher dose rate does result in faster {311} defect
dissolution under the same annealing conditions.
Implantation temperature, if changing between 5~40C, does
not affect TED behavior either. Further studies would be
necessary to investigate the effects of these variables on
defect evolution and dopant diffusion.
4. Effect of background doping


Areal Density of Interstitials
Trapped in {311} Defects or Loops (cm'2)
142
0 10 2 1014 4 1014 6 1014 8 1014 1 1015 1.2 1015
Implant Dose (cm'2)
Figure 5.4. Areal density of interstitials trapped in {311}
defects and dislocation loops as a function of dose.


{311} Defect Density (cm')
74
Annealing Time (sec)
Figure 3.4. {311} defect density as a function of annealing
time in 20keV B+ implanted Si for different annealing
conditions.


Boron Diffusivity Enhancement /D
83
Annealing Time (sec)
Figure 3.9. Time-averaged Boron diffusivity enhancement
/Db* as a function of annealing time for different
annealing temperatures.


66
constant for the TED saturation process, from Fig. 3.10 we
can see that this constant increases dramatically as
annealing temperature decreases. It is about 4.5 hr at 650C
and decreases more than one order of magnitude to about 15min
at 800C.
The temperature dependence of the time constant for the
TED saturation process is shown in Fig. 3.11. Such a
dependence exhibits an activation energy of about 1.6eV.
This value is smaller than the 3.1-3.7eV reported by
Packan, 130 3.5 eV reported by Stolk et al.113 yet it is greater
than the < 1 eV we calculated from the data of Zhang et al.114
as shown together in Fig. 3.10. The difference between these
experiments is that both Packan and Stolk et al. studied
silicon self-implanted silicon. In these samples the
interstitial source for TED is believed to be only {311}
defects. Zhang et al. studied low energy (4 keV) boron
implanted silicon where the energy and dose combination
resulted in no extended defects observed by TEM. The source
of interstitials driving TED in this case was claimed to be
sub-microscopic clusters, which may be as simple as just
mobile B-I pairs that diffuse until the pair breaks up. In
our case, the samples were implanted with boron, but at a
higher energy and dose so that not only {311} defects are
present, B-I pairs might also exist. The activation energy
of the TED saturation process reflects a complex process,
including the release of interstitials from B-I pairs and
{311} defects, the migration of interstitials through the


38
during implantation. Raineri et al.121 studied random and
[100] channeling implants with B and P ions. 100 keV P
implants with doses from 5xl012 cm-2 to 2x10^4 cm-2 does not
show defects under channeling conditions after an anneal at
500C for 1 hr. The same implants in a random direction give
rise to a network of dislocation loops. A systematic study
on implant damage using TEM, RBS and TRIM simulations was
reported by Schreutelkamp et al.122 They claimed that if the
total number of silicon atoms displaced by the implant ions
exceeds a critical value, stable sub-amorphization
dislocation loops are observed. They implanted Si with B,
Si, P, Ga, As, In and Sb ions at keV and MeV energies. For B
implants with energies ranging from 50 keV to 190 keV, the
critical value of displaced Si atoms was 4.7x10-*-^ cm-2 based
upon TRIM calculations and 1.4x10-1-6 cm-2 based upon
Rutherford Backscattering (RBS) analysis. This value
increases with the mass of the implant ions. To the author's
best knowledge, there has been no report on the formation
threshold of {311} defects or stable loops in the low energy
(<20 keV) regime of B implanted silicon.
We have calculated the number of displaced silicon atoms
using TRIM for our implant matrix and the results are listed
in Table 2.2. As can be seen, when the displaced atom
density reaches about 1.5x10-1-6 cm-2, {311} defects form, and
when it reaches about 2.6xl016 cm-2, sub-amorphization loops
form. The latter value is less than the result of
Schreutelkamp et al., but considering the fact that we used a


62
available for diffusion at high B and I concentrations.
Comparing eq.(3.6) and (3.7) will raise the question of
whether the immobile B-I complexes contain Bs-B];-! clusters or
Bs-Bs-I clusters. Diaz de la Rubia et al. (private
communication) have shown through ab initio calculation that
the Bj-Bs pair has a positive binding energy of about 1.8 eV.
This Bj-Bs pair also traps silicon interstitial. It is
immobile and electrically inactive. The Bs-Bs pair, on the
other hand, is found to be energetically unfavorable with a
binding energy of -1.7 eV and electrically active. According
to this calculation, the immobile, electrically inactive
implant profile peak might be consisted of the Bj-Bs-I
clusters.
It should be mentioned that rod-like {311} defects also
locate in a depth corresponding to the profile peak region,
so do the sub-amorphization dislocation loops. However, the
{311} defects113 dissolve on a time scale comparable to the
transient diffusion (e.g. <30 min at 800C) while the static
B-I clusters dissolve much slower (e.g. > 4 hr at 800C128).
{311} defects and unstable loops contribute to transient
enhanced diffusion before it saturates. When the clusters
and stable loops finally break up, there will be a second
burst of enhanced diffusion. The slight increase of Cenh
with increasing temperature is due to a decreasing
interstitial supersaturation level which results in less
clustering or even breaks up some existing clusters. There
is no experimental evidence on how many interstitials are


7
crystalline silicon to investigate the effect of implant
damage on dopant redistribution.
1.2.2 Implant Damage and Types of Extended Defects
Ion implantation is a well-known technique for
controllably introducing dopants into the substrate during IC
fabrication. The major disadvantage of ion implantation is
the damage introduced by the energetic ions and the need for
post-implantation annealing to activate the dopant.
Different types of extended defects form after annealing,
depending on both the implant and the annealing conditions.
It is important to understand the defect formation kinetics
in order to study the effect of defects on dopant diffusion.
As an incident ion penetrates into the substrate, it
loses kinetic energy through elastic collisions with the
nuclei and inelastic collisions with electrons of the target
material until it stops. Elastic collisions are primarily
responsible for the lattice displacements through the
production of Frenkel pairs. If the kinetic energy
transferred to the host atom is higher than the displacement
threshold energy (~15eV for Si)41, the host atom leaves its
lattice site (primary collision) and collides with other
atoms (secondary collision), which in turn gives rise to more
collisions. This sequence results in a collision cascade.
For a sufficiently high concentration of damage, it is
possible to amorphize the crystalline silicon near the
surface. If the damage density is not high enough, some


13
study. Since {311} defects are out of contrast for g
parallel to their length, those on one of the six directions
are invisible. Thus viewing on a [100] plane, we observed
the morphology shown in Fig.1.3.
line direction u
gxu
[110]
*0
[-110]
=0
[101]
*0
[10-1]
*0
[011]
*0
[01-1]
*0
visible(+) or invisible(-)
+
+
+
+
+
Figure 1.4. Schematic of {311} defect directions and defect
visible conditions under reflection g=220.
Various atomic models incorporated by self-interstitials
have been proposed for the {311} defects. Salisbury and
Loretto52,53 suggested that self-interstitial atoms located on
the tetrahedral sites are responsible for {311} defect


5keV
20keV
40keV
0.2nm
2xl0^cm2 5xl0^cm'2 lxlO^cm'^
1>J
Figure 2.2. Weak beam dark field (g220) PTEM images of 5keV, 20keV and 40keV B+ implanted
Si to doses of 2xl0^^cm~2> SxlO^cm-^ and lxlO^^cm-^, after an anneal at 900C for 15min in
N2.


Intersititals Bound by {311} Defects (cm'2)
72
Annealing Time (sec)
Figure 3.2. The exponential decay of interstitials trapped
in {311} defects in 20keV B+ implanted Si after anneals at
650C to 800C for various times.


32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
165
E. K. Broadbent, R. F. Inrani, A. E. Morgan, and P.
Maillot, IEEE Trans. Electron Devices ED-36, 2440 (1989).
J. Lin, S. Barnerjee, J. Lee, and C. Teng, IEEE Electron
Device EDL-11. 191 (1990).
V. Ilderem and R. Reif, J. Electrochem. Soc. 136, 2989
(1989).
R. Liu, D. S. Williams, and W. T. Lynch, J. Aool. Phvs.
£2, 1990 (1988).
V. Probst, H. Schaber, A. Mitwalski, H. Kabza, L. Van den
hove, and K. Maex, J. Appl. Phvs. 70. 708 (1991) .
M. H. Juang and H. C. Cheng, J. Aool. Phvs. 71. 1271
(1992).
S. B. Herner, K. S. Jones, H.-J. Gossman, R. T. Tung, J.
M. Poate, and H. S. Luftman, in The effect of TiSi2 film
thickness and growth on the point defect perturbance in
Si, 1996 (Electrochemical Society), p.337-347.
D.-S. Wen, P. Smith, C. M. Osburn, and G. A. Rozgonyi, J.
Electrochem. Soc. 136, 466 (1989) .
K. H. Weiner, P. G. Garey, A. M. McCarthy, and T. W.
Sigmon, IEEE Electron Device Lett. EDL-13. 369 (1991).
J. F. Gibbons, Proc. of the IEEE 60. 1062-1096 (1972).
K. S. Jones, S. Prussin, and E. R. Weber, AppI. Phvs. A
45, 1 (1988) .
K. S. Jones and J. Gyulai, Ion Implantation Science and
Technology (Ion Implantation Technology Co., Yorktown,
NY, 1996).
J. Liu, M. E. Law, and K. S. Jones, Solid-State
Electronics 38. 1305 (1995).
D. Venables and K. S. Jones, Nucl. Inst. Meth. Phv, Res.
B59/60. 1019 (1991).
L. Laanab, C. Bergaud, M. M. Faye, J. Faure, A. Martinez,
and A. Claverie, Mat. Res. Soc. Svmp. Proc. 279. 381
(1993) .
O. Dokumaci, P. Rousseau, S. Luning, V. Krishnamoorthy,
K. S. Jones, and M. E. Law, J. AppI. Phvs. 78. 828
(1995) .


CHAPTER 5
THE INFLUENCE OF IMPLANT DOSE ON DEFECT BEHAVIOR AND
TRANSIENT ENHANCED DIFFUSION
5,1 Overview
Implant dose is one of the most important parameters
that strongly affect the form of implantation damage. Since
transient enhanced diffusion is directly related to the
damage created by the implantation, variation of implant dose
is also expected to influence diffusion behavior.
For elements heavier than carbon, at low doses, the
damage is in the form of point defects or small point defect
clusters. At higher doses, extended defects begin to form.
At very high doses, the surface of the substrate can be
amorphized. The amorphization threshold depends on the
implant condition and ion mass. Most of the experiments
investigating TED due to implantation damage have used high
dose amorphizing implants. In this case, the end-of-range
damage can act as both a source and a sink of excess
interstitials. For more information regarding the influence
of implant damage on TED in amorphized sample, please refer
to ref.17,111,131.
In the case of low dose non-amorphizing implants,
Packan130 studied boron diffusion in Si+ implanted silicon
with doses ranging from 1x10^2 cm-2 (-0 2x10^4 cm2 at an
119


54
defects explains the decay of interstitials trapped in {311}s
even when the average defect size increases.
The average {311} defect size is shown in Fig. 3.5 as a
function of annealing time for each temperature. The width
change of these defects have been found to be very small
during the anneal by high resolution TEM studies conducted by
Eaglesham et al.125, so that the size change is represented by
the change of the length. As can be seen from Fig. 3.5, the
average defect size increases initially with annealing time
and then starts to decreases at some point during the
dissolution process. This breaking point can be seen for
anneals at 700C and 750C but not for anneals at 650C or
800C. We believe this final drop of defect size indicates a
pure dissolution stage which should be observed as well at
650C and 800C if the anneals are long enough. Again, to
reach a particular defect size, it take longer for lower
temperatures. For example, for 800C anneal, the defect size
reaches 400 after about 3min, while at 700C, it takes
about 1.5 hr.
To further illustrate the {311} defect evolution
process, the defect distributions after each anneal at each
temperature are shown in Fig. 3.6 (a)~(d). The distribution
moves to a larger size with increasing time at each
temperature and the defect density N(l) at every size 1
shifts to a lower value. This movement is faster for a
higher temperature. During the anneal, the interstitial
concentration in the vicinity of large defects is lower than


90
which favors the formation of boron clusters which bind
excess interstitials. This will reduce the total number of
interstitials available for forming {311} defects. (4) The
surface recombination effect is more efficient for a lower
energy when the damage region is closer to the surface. More
interstitials are drawn to the surface and less are available
for defect formation during annealing.
If the time constant of the {311} defect dissolution
process is chosen as the time required for the number of
interstitials to decrease to 1/e of its maximum value, then
it would be about 5 min for the 10 keV implant and 16 min for
the 40 keV implant. This time constant is plotted in Fig.
4.3 as a function of energy. It is interesting to note that,
if we derive from Fig. 4.3 the time constant of the 5 keV
implant, it would be about 2.6 min. If {311} defects existed
in the 5 keV implant from the beginning, we should be able to
see them after a 5 min anneal, even though the defect density
would be low, yet we do not see any defects. This could be
attributed to one or a combination of the factors discussed
above.
4^2^3..Defect Density, Size and Distribution during Annealing
The {311} defect density for different implants as a
function of annealing time measured from PTEM images is shown
in Fig. 4.4. If the annealing time is short (e.g., less than
10 min), the lower energy implant shows a slightly higher
density of {311} defects. This is probably due to earlier


31
equilibrium and actual concentrations of X, Gx is the surface
generation rate for X, Kxs is the surface recombination
constant and Kfc>ulk is the bulk recombination constant.
For the case of dopant diffusion following an
implantation, the initial distribution of point defects is
determined by the implantation process. These point defects
then diffuse according to the above equations. Anomalous
diffusion can be observed until the point defect
concentrations return to the equilibrium values, when only
normal diffusion can be seen.
1.4 Thesis Statement
The contributions of this work are in the following
areas:
1. Determination of the formation threshold of extended
defects in low energy B+ implanted silicon.
2. Quantitative TEM studies of the annealing kinetics of
{311} defects arising from B implantation.
3. Experimental investigation of the implant energy and
dose effects on {311} defect behavior.
4. Experimental investigation of the implant energy and
dose effects on TED.
5. Extraction of diffusivity enhancement for different
implant energies, doses, annealing times and temperatures.
6. Experimental investigation of the sources of
interstitials driving dopant diffusion.


Defect Defect Defect Defect
Density (cm'2) Density (cm'2) Density (cm'2) Density (cm'2)
78
750C5min
i L
60 260 460 660 860 1060 1260 1460
260 460 660 860 1060'126b'1460'
1011
1010
109
108
107
L
750C30min
i
LLJ
111
60 260 460 660 860 1060 1260 1460
260 460 660 860 1060 1260 1460
Average {311} Size ()
(c)
Figure 3.6. (Continued)


CHAPTER 2
THE FORMATION THRESHOLD OF {311} DEFECTS AND SUB-
AMORPHIZATION DISLOCATION LOOPS
2,1 Overview
The formation of very shallow p-type region by ion
implantation requires the use of low energy boron implants.
Although there have been many reports on the defect and
diffusion behavior in boron implanted silicon,63,116-118 none of
them have systematically studied the defect formation
threshold in the low energy implantation regime. For boron
implantation, because of the low mass of the implanted ions,
even if the ion fluences reach 2xl015 cm~2( the damaged
surface layer still does not amorphize. In this case the
extended defects that form after annealing can be classified
as type I or sub-amorphization defects. The implant energy
and dose combinations in this study are similar to those used
in the semiconductor industry for B implants and fall in the
sub-amorphization regime.
In order to understand how the sub-amorphization defects
influence dopant diffusion, it is important to know the
implant conditions under which these defects are actually
formed. This is the purpose of this section of the current
study.
33


Gossmann kindly provided me with the boron doping
superlattices as well as his PROPHET simulation results. Dr.
Heemyong Park, Dr. Samir Chaudhry and Omer Dokumaci showed me
how to run the simulation program FLOOPS. I thank all of
them.
I am especially grateful to Joe and Alisa Zhu, John and
Sherry Ding for their long-lasting friendship, which has
driven away much of my lonely time and made my stay in
Gainesville full of joy.
I am unable to express my gratitude to my parents Zezhan
and Yafan and my sister Jinqi for the sacrifice they have
made for me to make it so far.
Finally, I would like to thank my husband Jun Zhou for
his love and understanding.
IV


136
this saturation effect. One is that dislocation loops and
more B-I clusters form at the highest dose, which act as a
sink of interstitials for a limited period of time, during
which dopant diffusion is reduced, and after which
interstitials are emitted again and give rise to more
diffusion. The other is that all the boron atoms are paired
with interstitials and more interstitials would not cause any
further diffusion. Our results are comparable to those
reported by Packan, where lower dose Si+ implantation was
used to examine the dose dependence of dopant diffusion, and
to those of Solmi et al., where higher dose B+ implantation
was used and the formation of dislocation loops was claimed
to result in a reduction in TED.


100
2 hr, i.e. after equilibrium is reached, the interstitial
supersaturation level /Ci* varies little when implant
energy is above 5 keV. The factors that we discussed in the
previous section on why {311} defects do not form in the 5
keV implant can also be used to explain the TED behavior.
These factors include the formation of boron interstitial
clusters, the implant damage, the I-V recombination rate and
the surface recombination effect variations with implant
energy. Since the first three factors are expected to change
gradually with the changing of energy, surface recombination
effect is believed to be the most significant factor that
reduces the amount of diffusion in the 5keV implant. The
energy threshold below which the surface starts to play an
important role lies between 5keV and lOkeV.
In summary, boron 5-doping superlattices have been used
to study the effect of implant energy (5 keV to 40 keV) on
TED of boron implanted silicon. Higher implant energies
( > 5 keV) cause more total diffusion enhancement and longer
TED duration. It is speculated that a threshold exists at an
implant energy between 5keV and lOkeV, below which the
surface recombination effect strongly affects diffusion. It
is also speculated that both mobile boron interstitial pairs
and {311} defects existed in higher energy ( > 5 keV)
implants and they function together to provide interstitials
for TED. Hence the formation of {311} defects appears to
result in a dramatic increase in the duration of TED.


35
2.3 Defect Microstructure
Figure 2.1 shows PTEM micrographs of some of the samples
in the implant matrix after an anneal at 750C for 5 min.
Micrographs of 5 keV, 20 keV and 40 keV implants with doses
of lxlO-*-^ cm-2, 2x10^4 cm-2 and 5x10-*-^ cm-2 are shown in the
figure. At a dose of 1x10-*-4 cm'^, there are no {311} defects
in the 5 keV sample and the defect density increases with
increasing energy. When the dose is doubled to 2x10^4 cm-2,
there are still no {311} defects in the 5 keV sample, but
more defects are observed in the other two implants.
Increasing the dose further to 5xl014 cm-2 results in {311}
defect formation at 5 keV. Figure 2.1 clearly presents a
transition from samples without to those with {311} defects.
It is obvious that as the implant energy increases, the
critical dose for forming {311} defects decreases. From this
figure it is also obvious that the interstitial
supersaturation necessary to nucleate {311} defects is far
less than that for dislocation loops. This is consistent
with Eaglesham 's observation (private communication) of
{311} defects at 7xl012 cm-2 for 40 keV Si implants and the
observation of Jones et al.42 that stable dislocation loops
(approximately the same implant energy) do not form until a
dose of 2xl0l4 cm-2. In addition, the threshold dose for
{311} defects in a 20 keV B implant is around 2x10^^ cm-2,
which is much greater than the threshold dose of 7xl012 cm-2
for a comparable depth of 40 keV Si implant. This is


32
7. Provision of experimental basis for the improvement
of process development and process simulators.


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
DEFECT AND DIFFUSION STUDY IN BORON IMPLANTED SILICON
By
Jinning Liu
December, 1996
Chairman: Dr. Kevin S. Jones
Major Department: Materials Science and Engineering
The aim of this work is to better understand the nature
of transient enhanced diffusion (TED) of boron in silicon.
A matrix of sub-amorphization boron implants was used in
this work with energies ranged from 5 to 40 keV and the doses
ranged from 5x10-*--^ to 1x10-*-^ cm-^. The formation threshold
of {311} rod-like defects and sub-amorphization dislocation
loops are higher for lower implant energies. For the same
energy, the threshold dose for {311} defect formation is
lower than that for loop formation.
The study of annealing condition effect on {311} defect
behavior and boron TED shows that {311} defects dissolve more
rapidly at higher temperatures. The concentration of
interstitials trapped in {311} defects was below the value
predicted from previous studies of Si+ implants, indicating
Vll


118
*
O
v
in
05
CO
3
c
Implant Energy (keV)
(a)
Figure 4.10. Supersaturation of interstitials as a function
of implant energy after an anneal at 750C for (a) 3min and
(b) 2hrs.


17
phosphorous66'68 and BF269 implanted silicon, electron,52,70
neutron71, proton72 and neon73 irradiated silicon as well as
electron irradiated germanium.59 The implant temperature
effect on {311} defect formation has also been studied.
Tamura74 found that room temperature phosphorous implants in
silicon at low energies resulted in the formation of
dislocation loops, whereas elevated temperature implants
(200~400C) gave rise to the formation of rod-like {311}
defects. This temperature effect has also been shown by
Chadderton and Eisen.75 They found that dislocation loops
formed in boron implanted silicon if the implantation was
held at liquid nitrogen temperature, in contrast with the
room-temperature implants which showed {311} defects. This
behavior might arise from the self-annealing effect occurring
at higher temperatures which reduces the number of free
interstitials. Under this condition {311} defects may be
easier to form than dislocation loops presumably because the
energy to form a {311} defect is less than that of a loop.76
Back in the seventies, the formation of {311} defects
was associated with the presence of boron by Seshan and
Washburn.67,68 They observed {311} defects in phosphorous
implanted boron-doped silicon. However, the {311} defects
were absent from n-type silicon under the same implant and
annealing conditions. Thus they concluded that boron is
essential for the formation of {311} defects. Recently, a
research group at Bell Laboratories (private communication)
investigated the role of boron during {311} defect formation


6
segregate into the silicide.30 This causes serious contact
resistance problems.31-34 Moreover, silicidation perturbs the
point defect concentration in the substrate by injecting
vacancies and depleting interstitials,38,39 which should also
be taken into consideration during the process design. In
the case of SADS process, the dopant species is implanted
entirely into the silicide and then driven into silicon at
low temperature. This avoids both dopant channeling and
damaging of the silicon. SADS is a very attractive method to
form 300-400 junctions. However, extension implants are
still necessary to define the channel length etc. The
disadvantages of SADS include uneven doping due to rough
interface and dopant precipitates in the silicide.
Thermal diffusion is also used for junction formation.
While conventional diffusion does not create damage, it
suffers from poor controllability and requires a high thermal
budget. An alternative diffusion method is gas-immersion
laser doping (GILD).40 An excimer laser is used to rapidly
heat and melt the silicon surface. BF3 is then adsorbed,
pyrolized and diffused into the melt. The junction depth is
determined by the melt. A 100 junction depth has been
achieved with this method, which would be attractive if
technical, uniformity and cost issues can be resolved.
In summary, there are various ways to form a shallow
junction, and each has its own advantages and drawbacks.
This study is focused on the use of implantation of B+ into


Boron Concentration (cm'3) Boron Concentration (cm'3)
81
Depth ()
(c)
0 1000 2000 3000 4000 5000 6000 7000 8000
Depth ()
(d)
Figure 3.7. (Continued)


Defect Defect Defect Defect
Density (cm'2) Density (cm'2) Density (cm'2) Density (cm2)
146
tii|ii1\ iin
1e15 bmin -
J I I I L
60 260 460 660 860 1060 1260 1460
1011
1010
109
108
107
60 260 460 660 860 1060 1260 1460
11
10
10
10
10s
10
10
r
8 '
7
60
i | 1 1 1 I 1 1 1 l 1 1 1 I 1 1 1 I 1 1 1 I 1 1 1 !
1e15 1hr
260 460 660 860 1060 1260 1460
1011 rr"r 1 i 1 1 1 i 1 1 1 i l"1 1 i 1 1 1 i 1 1 i 1 1 i 1 1 i
", q1o i- 1e15 *hrs -I
60 260 460 660 860 1060 1260 1460
{311} Defect Size ()
(b)
Figure 5.7. (Continued)


167
66 F. Cembali, L. Dori, R. Galloni, M. Servidori, and F.
Zignani, Radiation Effects 36. Ill (1978).
67 K. Seshan and J. Washburn, Rad. Eff. 26. 31 (1975) .
68 K. Seshan and J. Washburn, Radiation Effects 14. 267
(1972) .
69 I. W. Wu and L. J. Chen, J. Appl. Phvs. 58, 3032 (1985).
70 E. Nes and J. Washburn, J, Appl. Phvs. 42. 3559 (1971).
71 J. Narayan and J. Fletcher, MRS Svmo. Proc. 2. 191
(1981) .
72 M. D. Matthews and S. J. Ashby, Phil. Mag. 27. 1313
(1973).
73 S. M. Davidson and G. R. Booker, Rad. Effects 6. 33
(1970) .
74 M. Tamura, Appl, Phvs. Lett. 23# 651 (1973).
75 L. T. Chadderton and F. H. Eisen, Radiation Effects 7.
129 (1971).
76 A. Parisini and A. Bourret, Phil. Mag. A 67. 605 (1992).
77 T. Y. Tan and U. Gosele, J. AppI. Phvs. 37. 1 (1985).
78 S. Wolf and R. N. Tauber, Silicon Processing for the VLSI
Era. Vol. 1 (Proc. Tech. Latt. Press, Sunset Beach, FL,
1983).
79 W. Frank, U. Gosele, H. Maher, and A. Seeger, Diffusion
in Si and Ge (Academic Press, Orlando, FL, 1983).
80 S. K. Ghandhi, VLSI Fabrication Principles. John Wiley &
Sons, Inc. (1982).
81 C. M. Hsieh and D. M. Maher, J. AppI. Phvs. 44. 1302
(1973) .
82 S. M. Hu, J. AppI. Phvs. 45. 1567 (1974).
83 T. Y. Tan and U. Gosele, Appl. Phvs. Lett. 39. 86 (1981).
84 D. A. Antoniadis and I. Moskowitz, J. AppI. Phvs. 53.
6788 (1982).
85 P. A. Packan and J. D. Plummer, J. AppI. Phvs. 68. 4327
(1990) .


97

D'b iff'0
-J\\ + E)dt]=l + E...(t -[ f'(1 + E)dtx +\'dt] = 1 +^E...(t > t0)
t '*0 t
(4.3)
From the above equation we can see that if /Db*-1 is
plotted as a function of time in a log-log scale, it should
be a curve that is flat for t 1/t for t>to and the breaking point to is the time when the
interstitial concentration reaches equilibrium value and TED
is completed. This of course is a simplified picture since
TED sources do not turn off abruptly, but rather the
interstitial supersaturation decays smoothly over the time
period studied. In other words we may not see only one
breaking point to but more than one, e.g., to, ti, t2, etc..
Figure 4.9 shows a plot of /Db*-1 versus annealing time
for different implant energies. Because of the finite speed
of interstitial diffusion, we have to pick a fixed distance
from the interstitial source to make the comparison. A
distance of 3600 A from the projected range of each implant
was thus chosen in this plot. Because of the long interval
between the annealing times, we did see just one breaking
point to for E > 5 keV. According to the above discussion,
TED stops before 15 min for the 5 keV implant and within
15 min to 2 hr for higher energy implants.
The energy dependence of the time-averaged interstitial
supersaturation /Ci* after 3min and 2hr anneals is shown
in Fig. 4.10 (a) and (b). For the 3 min anneal the
interstitial supersaturation near the implant damage region


113
r
E
c
o
ro
c
CD
O
c
o
O
c
o
l
o
en
0 1000 2000 3000 4000 5000 6000
Depth ()
(a)
0 1000 2000 3000 4000 5000 6000
Depth ()
(b)
Figure 4.7. SIMS profiles of boron in 2xl014cm-2 B+
Implanted Si superlattices at an energy of (a)5keV, (b)
lOkeV, (c) 20keV and (d) 40keV, annealed at 750C.


63
trapped in one Bs-Bj cluster. In our 20 keV 2x10-*-^ cm 2 b+
implant, after an anneal at 750C for 2 hr, {311} defects
almost dissolved completely. The density of activated boron
atoms is estimated to be < 7x10^3 cm-2 by integrating the
non-clustered region of the implant profile. Assuming a
valid "plus 1" model and assuming that one activated B atom
pairs with one interstitial, the number of interstitials
trapped in the B-I clusters must exceed 1.3x10-*-^ cm-2. Thus
we speculate that one Bs-Bj cluster might trap more than two
interstitials.
The annealing time scale we used in this study is much
shorter than that is needed to break up the clusters and
stable loops, so our focus is on TED at low concentration
after relatively short anneals.
3 .3.2.2 Broadened Tail at Low Concentration
From Fig. 3.7 we notice that significant diffusion
occurred in the tail region (C four temperatures. This is caused by the pairing of excess
interstitials induced by ion implantation with dopant atoms.
The boron profile displacement at this region changes with
the temperature. A 4 hr anneal at 700C resulted in a
displacement of about 2200 at a concentration of lxlO1^
atoms/cm^ relative to the as-implanted profile, whereas a lhr
anneal at 800C resulted in a displacement of about 1750
for the same concentration. As will be shown later, TED is
complete after 1 hr at 800C but is still continuing after


129
annealed in a nitrogen ambient at 800C for lhr to remove the
excess point defects introduced by the APCVD process. The
wafers were then implanted with 20 keV boron ions to doses of
5xl0l3 cm-2, 2xl0-'-4 cm-2 and 1x10^-^ cm-2. Subsequent anneals
were performed in nitrogen at 750C for 15 min, 30 min, 1 hr,
2 hr and 4 hr. SIMS analysis was performed on the CAMECA
IMS4f system. The parameters used during the SIMS analysis
were the same as discussed in section 3.3.1. The dopant
profiles were then imported into FLOOPS and the diffusivity
enhancement was extracted for each anneal.
5.3.2 Dopant Diffusion Behavior
SIMS profiles of boron in the as-grown and pre-annealed
samples as well as in the implanted and annealed samples are
shown in Fig. 5.8(a)~(d). The small displacement of the
profile of the pre-annealed sample with respect to that of
the as-grown sample is believed to be due to boron intrinsic
diffusion. The buried boron spike in the 5xlC)l2 cm-2
implanted sample shows an enhanced diffusion but the profile
displacement is smaller compared with those at higher doses.
The 30min annealed profile of the 2x10^^ cm-2 implanted
sample shows an enhanced diffusion to a similar degree as
that of the lxlO^-5 cm-2 implant. But the profile after a 4hr
anneal shows a smaller enhancement than that at lxlO15 cm-2.
Besides the buried boron spike, the tail of implanted boron
profile also shows an enhanced diffusion in each sample. The
peak of this implanted profile, however, is clustered and


TABLE OF CONTENTS
ACKNOWLEDGMENTS iii
ABSTRACT vii
CHAPTERS
1 INTRODUCTION 1
1.1 Motivation and Objective 1
1.2 Background 3
1.2.1 Shallow Junctions 3
1.2.2 Implant Damage and Types of Extended Defects 7
1.2.3 {311} Rod-like Defects 10
1.2.4 Dopant Diffusion in Silicon 18
1.3 Implantation, Characterization and Simulation
Techniques 25
1.3.1 Ion Implantation 25
1.3.2 Transmission Electron Microscopy 27
1.3.3 Secondary Ion Spectroscopy 29
1.3.4 Florida Object Oriented Process Simulator 3 0
1.4 Thesis Statement 31
2 THE FORMATION THRESHOLD OF {311} DEFECTS AND
SUB AMOR PHI Z AT I ON DISLOCATION LOOPS 33
2.1 Overview 33
2.2 Experimental Procedures 34
2.3 Defect Microstructure 3 5
2.4 Criteria for the Formation of {311} Defects and
Sub-amorphization Loops 37
3 EFFECT OF ANNEALING ON DEFECT EVOLUTION AND TRANSIENT
ENHANCED DIFFUSION 48
3.1 Overview 48
3.2 Effect of Annealing on Defect Behavior 49
3.2.1 Experimental Procedures 49
3.2.2 Defect Evolution: Microstructure,
Interstitial Density and the Activation Energy
of Interstitial Decay 50
3.2.3 Defect Evolution: Density, Size and Size
Distribution 53
3.3 Effect of Annealing on Transient Enhanced
Diffusion 58
3.3.1Experimental Procedures 58
v


Time Constant of TED Saturation Process (sec)
85
10000/T(K)
Figure 3.11. Time constant and activation energy of TED
saturation process, comparison with previous studies.


157
equilibrium. When the annealing temperature is high enough
and the time is long enough to break up the immobile B
clusters (e.g. 800C >4hrs), more interstitials are released
from these clusters and a second burst of TED will be
observed. Only this time the breaking point between the
mobile and immobile regions moves up from Cenh to Cs, the
solid solubility limit of B in silicon.
When the implant dose and/or energy increase so that
{311} defects form during annealing, interstitials are not
only trapped in the B-I clusters and mobile B-I pairs as
discussed above, but also trapped in {311} defects. {311}
defects dissolve rapidly during annealing and emit
interstitials. These interstitials together with those
emitted by the B-I pairs are available to drive dopant
diffusion. Because of the co-existence of the two sources of
interstitials, the activation energy of TED saturation
process is lower than that in Si+ implanted samples with only
{311} defects and higher than that in low energy low dose B+
implanted sample without {311} defects but presumably with B-
I pairs.
When the implant dose and/or energy increase further,
stable dislocation loops form in addition to {311} defects.
Unlike the {311} defects, the loops dissolve relatively slow.
They may trap a small or a large amount of interstitials,
depending on the density of the loops. As a result of the
interstitial sink effect the loops present, the total number
of interstitials available to drive TED is reduced. When the


CHAPTER 4
THE INFLUENCE OF IMPLANT ENERGY ON DEFECT BEHAVIOR AND
TRANSIENT ENHANCED DIFFUSION
4.1 Overview
It is believed that transient enhanced diffusion is a
direct result of the damage created by the implantation
process. Characteristics of the implanted ions (such as the
mass), the substrate conditions (such as crystalline versus
amorphous), and variations in implantation parameters (such
as energy, dose, dose rate and temperature) would all cause
variations in the damage and therefore affect dopant
diffusion. This section discusses the effect of implant
energy on defect behavior and TED.
It has been claimed that the amount of damage created by
the implantation process is strongly dependent on the dose
and the mass of the implant species but only weakly dependent
on the implant energy.42 Although higher energy does mean
more kinetic energy per implanted ion, the resulting implant
profile is more spread out and thus the "energy density"
deposited per volume is lessened by this dilution effect.
The effect of increasing energy is thus weakened and the
number of interstitial-vacancy pairs would not be increased
much provided the dose is the same. However, increasing
energy can indeed change the location of the damage and the
86


126
small defect nuclei is much larger during the initial defect
formation process. Subsequently more defects are grown from
these nuclei and therefore more defects are observed in a
higher dose implant.
The average {311} defect size as a function of annealing
time for different implant doses is shown in Fig. 5.6. The
average defect size increases with annealing time. For a
particular time, the defect size is larger for a higher dose.
After an anneal of 15 min, the 2x10-*-^ cm2 implant shows an
average defect size of about 517 A, while for the 1x10^5 cm-2
implant, it is about 613 A. Unlike the defect density
dependence, the size dependence on dose is not very strong.
This is not surprising because the total number of
interstitials introduced into the substrate represents the
product of the defect density and the defect size. Given a
fixed amount of total interstitials, a strong dose dependence
of the defect density must be accompanied by a weak dose
dependence of the defect size.
The size distribution of {311} defects after anneals at
750C for various times for 5xlC)14 cm2 and 1x10-*-^ cm"2
implants is shown in Fig. 5.7 (a) and (b). For each implant
we can see the same trend as we observed in the previous
chapters: a shift of the average {311} defect size to a
larger value and a decrease of the defect density at a
particular size with increasing annealing time. The
decreasing defect density and the increasing defect size are
the characteristics of the Ostwald ripening process. The


36
presumably due to the formation of immobile boron
interstitial clusters and mobile boron interstitial pairs
that reduces the concentration of free interstitials
available to form {311} defects.
PTEM micrographs of selective samples annealed at 900C
for 15min are shown in Fig. 2.2. This temperature/time
condition was chosen to reveal stable sub-amorphization
dislocation loops and dipoles that evolve from unstable {311}
defects. The {311} defects have completely dissolved after
the anneal. A submatrix of 5 keV, 20 keV and 40 keV implants
with doses of 2x10^4 cm-^, 5x10-'-^ cm-2 and 1x10^^ cm-2 are
shown in Fig. 2.2. The 40 keV sample has dislocation loops
and dipoles at a dose of 2x10^4 cm2. The 20 keV and 5 keV
samples do not show loops or dipoles until the dose is
increased to 5x10^4 Cm2. The critical dose for type I loop
formation appears to decrease with increasing implant energy.
The threshold dose at 40 keV (2x10^4 cm-^) matches the
previous results of Jones et al.'s for 30 keV boron. Table
2.1 lists the formation threshold for both {311} defects and
[110] loops for the whole implant matrix.
XTEM micrographs of 10 keV, 20 keV, 30 keV and 40 keV
implants to a dose of lxlO^^ cm~2 after an anneal at 750C
for 5 min are shown in Fig. 2.3. The reason for choosing the
highest dose in our implant matrix is that both {311} defects
and dislocation loops are present in these implants under
this annealing condition. A layer of sub-amorphization
extended defects is expected to form at the projected range


2
channel effects and to maintain low contact resistance.12
These shallow junctions will be difficult to form by ion
implantation. One of the reasons is that the implantation
process induces a supersaturation of silicon interstitials,
which will eventually lead to anomalous dopant diffusion,
known as transient enhanced diffusion (TED).13-20 TED is
caused by the pairing of excess silicon interstitials with
dopant atoms which reduces the energy barrier for the
migration of dopant atoms through the bulk. As a result, the
dopant penetration depth from diffusion is greatly increased.
TED lasts for only a limited time at typical annealing
temperatures so it is transient in nature.
TED will affect the source/drain dopant profiles both
vertically and laterally. The vertical diffusion results in
deeper profiles, which will weaken the control of the gate
voltage. The lateral diffusion reduces the effective channel
length which may result in punch-through phenomenon. Besides
these two effects, the enhanced diffusion of channel dopant
near the gate edge from excess interstitials created during
the lightly doped drain (LDD) or source/drain implant causes
an increase in the threshold voltage for shorter gate length
devices.21 This is the so-called reverse short channel
effect, i.e., an anomalous increase in the threshold voltage
with a decrease in device dimension. As device dimension
shrinks further, these effects will become more pronounced.
Thus the objective of this work is to study the effect
of implantation induced defects on transient enhanced


BIOGRAPHICAL SKETCH
The author was born on September 30, 1968, in Shenyang,
China. She entered Wuhan University in China at the age of
17 and obtained a BS in Physics degree with honors in July
1989. She came to the United States in August 1991 as a
graduate student in the Physics Department at Tulane
University, New Orleans. In July 1992 she transferred to the
Department of Materials Science and Engineering at the
University of Florida, specializing in electronic materials
under the supervision of Dr. Kevin Jones. She received her
MS degree in August 1994 and will complete her Ph.D. study in
August 1996.
171


18
by studying Si+ implanted silicon doped with various
concentrations of boron. They measured the number of
interstitials trapped in {311} defects after the same post
implantation anneal and found that the interstitial density
actually decreased with the increase of boron concentration.
We have also observed a higher formation threshold of {311}
defects in boron implanted than that in silicon implanted
silicon. It is believed that boron atoms pair with
interstitials and form submicroscopic boron interstitial
clusters and mobile boron interstitial pairs, which consume
part of the interstitials and reduce the number available for
{311} defect formation. This will be discussed further in
the following chapters. It should also be mentioned that the
new interest in {311} defects arises partly from the fact
that they have been claimed to be the only source of
interstitials for TED, which is still controversial at this
time.
1.2.4 Dopant Diffusion in Silicon
The diffusion of dopant atoms can be described by
DA = D0exp(--^-) (1.1)
where Da is dopant diffusivity, Do is the pre-exponential
constant and Qa is the activation energy for diffusion. It
is generally agreed that on an atomic level, dopant atoms in
silicon diffuse through interactions with silicon self
interstitials and vacancies. The type of point defects which
dominates the diffusion process determines the type of


45
30keV
4keV
Figure 2.3. Weak beam dark field (g220) XTEM images of lOkeV,
s
lOkeV
20keV
20keV, 30keV and 40keV B+ implanted Si to a dose of lxlC)15crn-2f
after an anneal at 750C for 5min in N2.


67
bulk, the surface recombination effect, impurity trapping,
etc. Variation in the activation energy is thus not
unexpected in these different systems. However, we still
believe that in our case, not only {311} defects play a role
in terms of releasing interstitials during annealing and
thereby driving TED, boron interstitial pairs also play an
important role.
In summary, 20 keV 2x10-*-^ cm-^ boron implanted silicon
was studied for the purpose of identifying the possible
sources of interstitials driving TED. The static peak region
was presumably consisted of immobile B^Bs-I clusters which
breaks up in times much longer than normal TED. Boron TED in
the tail region was measured through SIMS analysis and FLOOPS
simulation. The duration of TED increases dramatically when
the annealing temperature decreases. The activation energy
for the TED saturation process is about 1.6 eV, higher than
that in the sample presumably with only B-I pairs
contributing to TED and lower than that in samples with only
{311} defects. The interstitials driving TED in our sample
are believed to come from both mobile boron interstitial
pairs and {311} defects. The dissolution of {311} defects
and B-I pairs may interact with each other and result in a
single activation energy lying in between.


Figure 5.1. Weak beam dark field PTEM (g^) micrographs of 20keVcni2 B+ implanted Si to a dose
of (a) lxl014cm'2, (b) 5xl014cm'2 and (c) lxl015cm'2.


the presence of submicroscopic boron interstitial complexes.
The interstitial emission from {311} defects during anneals
shows an exponential time dependence with an activation
energy of 3.8 eV. The time constant for TED saturation
process is smaller at higher temperatures and the dependence
shows an activation energy of 1.6 eV. Both {311} defects and
boron interstitial pairs are believed to be the source of
interstitials driving TED.
The implant energy effect on {311} defects and TED was
also studied. The dissolution of {311} defects is slower for
higher energy implants. This may be due to either an
increase in the surface recombination or an increase in the
ratio of boron interstitial clusters to {311} defects with
decreasing energy. Lower diffusion enhancement and shorter
duration of TED were observed for the 5 keV implant than for
all the higher energies at a dose of 2xl01^ cm-2. The {311}
defects might anneal slower than mobile boron interstitial
pairs so that the increase in {311} defect concentration may
account for the increase in TED duration. Surface
recombination is believed to play an important role for the 5
keV implant so that the amount of TED is reduced.
The study of dose effect on defect and TED shows that
{311} defects dissolve slower at higher doses. Diffusion
inceases with increasing dose initially. Boron interstitial
clusters and dislocation loops formed at high doses suppress
any further increase in TED with increasing dose above
2xl014 cm2 at 20 keV by acting as interstitial sink.
viii


Average {311} Defect Size ()
108
Annealing Time (sec)
Figure 4.5. Average size of {311} defects as a function of
annealing time for different implant energies.


89
interstitials split along <110> direction occupying one
lattice site. According the ab initio calculations conducted
by Diaz de la Rubia et al. (private communication) this
configuration is energetically unfavorable with a formation
energy of about 3.7 eV. Interstitials are more likely to
pair with boron atoms because the binding energy of a Bs-I
pair is about 1.1 eV.
The lack of {311} defects in the 5 keV sample and the
lower density of interstitials trapped by {311} defects in a
lower energy implant could be attributed to at least the
following four factors: (1) Higher energy might introduce
more as-implanted damage which may further induce more
trapped interstitials in extended defects (if formed) after
annealing. (2) Higher energy causes larger separation
between the as-implant vacancy profile (closer to the
surface) and the interstitial profile (further away from the
surface) because of the larger forward momentum carried by
the implanted ions. The result of this might be that the I-V
recombination rate is lower and the number of interstitials
available to form extended defects is higher. (3) According
to a series of proposed defect reactions113 which describe the
generation and clustering of mobile boron atoms given by
eq. (3.5) and (3.7),
Bs+I <-> Bj (3.5)
Bs+mBj <-> Bc+nl (3.7)
at a lower implant energy, the concentrations of Bs or Bz and
silicon interstitials in the implant peak region are higher,


(C)
Figure 4.1 (Continued)


Time Constant of {311} Dissolution Process (sec)
141
Implant Dose (cm'2)
Figure 5.3. Time constant of {311} dissolution process as a
function of implant dose, annealed at 750C.


23
implant profiles behave in the same way as the epi layer
spikes but are less sensitive because the epi layer profiles
are thinner and possess a more ideal Gaussian shape.
Simulation programs such as FLOOPS can then be used to
calculate the diffusivity of the dopant atoms and quantify
the enhancement or retardation effect.
Other than dopant profiles, extended defects can also be
used to monitor the point defect perturbance. Early studies
of the trapping of point defects by extended defects were
conducted by using stacking faults, 82,106_108 which grow by
absorbing interstitials and shrink by emitting interstitials,
thereby giving a good indication of the change in point
defect concentration. However, because of their large size
(> 5 |im) and low density (< 10^ cm-2), the stacking faults
are not ideal for detecting small perturbations. Instead,
post-implantation annealed dislocation loops (end-of-range
loops) have the advantage of much smaller size (~ 0.02 |lm in
diameter) and larger density (> 10^-0 cm-^) t therefore being
more sensitive detectors. In addition, the dislocation loops
form at well-defined depth, which can not only be utilized to
measure the net interstitial flux, 105,109 but have also been
found to be an effective barrier to interstitials driving
anomalous dopant diffusion.17'110,111
Extended defects themselves can act as a source of
interstitials during annealing. Since dislocation loops
dissolve at temperatures that are too high (>1000C) to
explain TED which is completed within seconds at T>950C, the


Density of Interstitials Trapped in {311} Defects (cm'2)
105
Annealing Time (sec)
Figure 4.2. Density of interstitials trapped in {311}
defects in 10keV~40keV B+ implanted Si, annealed at 750C.


48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
166
K. Krishnamoorthy, D. Venables, and K. S. Jones, in Proc.
Ion Implantation Technology, Austin, TX, 1996.
R. G. Elliman, J. S. Williams, D. M. Maher, and W. L.
Brown, Mat. Res. Soc. Svmp. Proc. 51 (1986).
F. F. Komarov, A. P. Novikov, T. T. Samoilyuk, V. S.
Solov'yev, and S. U. Shiryaev, Rad. Eff. 90. 307 (1985).
J. Linnros and G. Holmes, Mat. Res. Soc. Svmp. Proc. 51
(1986).
I. G. Salisbury and M. H. Loretto, Phil. Maa. 39. 317
(1979) .
I. G. Salisbury and M. H. Loretto, Rad. Eff. 59. 59
(1981).
Y. Tsubokawa, M. Kuwabara, H. Endoh, and H. Hashimoto, J.
Electron Microscopy 35 Suool, 953 (1986).
M. Pasemann, D. Hoehl, A. L. Aseev, and 0. P. Pchelyakov,
Phvs. Status Solidi a80. 135 (1983).
T. Y. Tan, Phil. Maa. 44. 101 (1981).
P. K. Madden and S. M. Davidson, Radiation Effects 14.
271 (1972).
K. Seshan and J. Washburn, Phvs. Stat. Sol. 26. 345-352
(1974).
C. A. Ferreira Lima and A. Howie, Phil. Mag. 34. 1057
(1976).
S. Takeda, Jpn, J, AppI. Phvs. 30. L639 (1991) .
K. S. Jones, J. Liu, L. Zhang, V. Krishnamoorthy, and R.
T. DeHoff, Nucl. Inst. Meth. B 106. 227 (1995) .
J. A. Lambert and P. S. Dodson, Phil. Maa. 44. 1043
(1981) .
W. Wu and J. Washburn, J. Appl. Phvs. 48. 3742 (1977) .
J. J. Comer, Radiation Effects 36. 57 (1978).
H. Foell, T. Y. Tan, and W. Krakow, Mat. Res. Soc. Svmp.
Proc. 2. 173 (1981).


61
rather higher than that, especially at higher temperatures
(T>700C) This mismatch can not be explained by Fair's
model. We therefore speculate that the correlation between
Cenh and ni might just be a coincidence. A better
explanation is that clustering effect induces an immobile
fraction of the profile. As for the kinetics of peak
clustering, Cowern et al.128 suggested the following
reactions:
Bs + I <-> Bx (3.5)
mBs + ni <-> IDC (3.6)
where the intermediate defect configurations (IDCs), which
was originally proposed by Tan,56 might include boron pairs or
clusters, rodlike {311} defects, certain stacking faults,
etc. Silicon interstitials (I) pair or kick out
substitutional boron atoms (Bs) to form migrating
interstitial boron atoms (B!) During the transient phase,
the concentration of Bj is high enough to allow homogeneous
nucleation of boron containing defects. The defect formation
is most favorable near the peak region where the B
concentration is highest. Stolk et al.113 proposed another
clustering reaction given by
Bg + mBj <-> Be + ni (3.7)
where Bc refers to immobile clusters containing m+1 boron
atoms, and n is the number of interstitials injected upon
clustering to allow stress relief (n sustains the enhanced diffusion of BSf eqs.(3.6) and (3.7)
lead to cluster formation and reduce the amount of Bs


39
new version of TRIM, the results are in reasonable agreement.
It thus appears that the number of displaced silicon atoms
can be used effectively as the criterion for both {311}
defect and sub-amorphization dislocation loop formation.
If the above argument is true, we expect to see more
extended defects after annealing in samples containing more
displaced silicon atoms after implantation. This indeed is
the case for both anneal conditions with only two exceptions.
The first is the 5 keV 1x10^^ cm2 implant after an anneal at
750C for 5 min (Fig. 2.1). Keeping the dose at 1x10^-^ cm-2,
we see less {311} defects but more dislocation loops in the
5 keV implant than in the higher energies. The reason might
be that the 5 keV lxlO^-S cm-2 implant has the highest
interstitial concentration in the whole implant matrix. When
the number of free interstitials reaches a critical value,
the system may form loop nuclei. If the energetics are such
that interstitials would rather be in a loop, then the
loop/{311} ratio increases, i.e., loop formation might be
more favorable than {311} defect formation. The second
exception to the damage alone being the reason for threshold
changes arises when one compares the 5 keV and the 20 keV
implant. The displaced atom density for 5 keV 5x10^-^ cm-2
implant (2.56xl01^ cm-2) is slightly less than that of the
20 keV 2xl014 cm-2 implant (2.94xl01^ cm-2) (Table 2.2), but
after an anneal at 900C for 15 min, the 5 keV sample shows a
low density of half loops while the 20 keV implant shows no
defects. The reason might be that these displaced atom


I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Mark Law
Associate Professor of
Electrical Engineering
This dissertation was submitted to the Graduate Faculty
of the College of Engineering and to the Graduate School and
was accepted as partial fulfillment of the requirements for
the degree of Doctor of Philosophy.
December, 1996
Winfred M. Phillips
Dean, College of Engineering
Karen A. Holbrook
Dean, Graduate School


DEFECT AND DIFFUSION STUDY IN BORON IMPLANTED SILICON
By
JINNING LIU
A THESIS 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
1996

dedicated to my parents

ACKNOWLEDGMENTS
This dissertation would not have been possible without
the guidance and support of my advisor, Dr. Kevin Jones, who
not only provides us illuminating advice but also creates a
stress-free environment for us to work in and do our best. I
would like to thank Drs. Steve Pearton, Cammy Albernathy,
Rajiv Singh and Mark Law for serving on my committee, and Dr.
Robert Park for sitting in my defense. Special gratitude is
due to Dr. Jian Chen, who patiently helped me get into the
door when I first joined the group four years ago and gave me
many valuable suggestions in the years following. Special
gratitude is also due to Dr. Wish Krishnamoorthy for many
stimulating discussions and a thorough correction on my
dissertation draft. I would like to thank all my colleagues,
especially members of our silicon group, Brad Herner and
Sushil Bharatan, for many helpful discussions, and my
officemate as well as good friend, Ranju Datta, for keeping
me company and making my time at school enjoyable.
It would have been impossible to finish this work
without the assistance of many other people. Dr. Lenny Rubin
at Eaton Corporation and Craig Jasper at Motorola helped me
with the implantations. Dr. Jinghong Shi and Joe Bennett
from SEMATECH helped me with the SIMS analysis. Dr. Hans
in

Gossmann kindly provided me with the boron doping
superlattices as well as his PROPHET simulation results. Dr.
Heemyong Park, Dr. Samir Chaudhry and Omer Dokumaci showed me
how to run the simulation program FLOOPS. I thank all of
them.
I am especially grateful to Joe and Alisa Zhu, John and
Sherry Ding for their long-lasting friendship, which has
driven away much of my lonely time and made my stay in
Gainesville full of joy.
I am unable to express my gratitude to my parents Zezhan
and Yafan and my sister Jinqi for the sacrifice they have
made for me to make it so far.
Finally, I would like to thank my husband Jun Zhou for
his love and understanding.
IV

TABLE OF CONTENTS
ACKNOWLEDGMENTS iii
ABSTRACT vii
CHAPTERS
1 INTRODUCTION 1
1.1 Motivation and Objective 1
1.2 Background 3
1.2.1 Shallow Junctions 3
1.2.2 Implant Damage and Types of Extended Defects 7
1.2.3 {311} Rod-like Defects 10
1.2.4 Dopant Diffusion in Silicon 18
1.3 Implantation, Characterization and Simulation
Techniques 25
1.3.1 Ion Implantation 25
1.3.2 Transmission Electron Microscopy 27
1.3.3 Secondary Ion Spectroscopy 29
1.3.4 Florida Object Oriented Process Simulator 3 0
1.4 Thesis Statement 31
2 THE FORMATION THRESHOLD OF {311} DEFECTS AND
SUB AMOR PHI Z AT I ON DISLOCATION LOOPS 33
2.1 Overview 33
2.2 Experimental Procedures 34
2.3 Defect Microstructure 3 5
2.4 Criteria for the Formation of {311} Defects and
Sub-amorphization Loops 37
3 EFFECT OF ANNEALING ON DEFECT EVOLUTION AND TRANSIENT
ENHANCED DIFFUSION 48
3.1 Overview 48
3.2 Effect of Annealing on Defect Behavior 49
3.2.1 Experimental Procedures 49
3.2.2 Defect Evolution: Microstructure,
Interstitial Density and the Activation Energy
of Interstitial Decay 50
3.2.3 Defect Evolution: Density, Size and Size
Distribution 53
3.3 Effect of Annealing on Transient Enhanced
Diffusion 58
3.3.1Experimental Procedures 58
v

3.3.2 Dopant Diffusion during Annealing 59
3.3.2.1 Static Peak at High Concentration 59
3.3.2.2 Broadened Tail at Low Concentration 63
3.3.3 The Activation Energy for the Diffusion
Saturation Process 65
4 THE INFLUENCE OF IMPLANT ENERGY ON DEFECT BEHAVIOR
AND TRANSIENT ENHANCED DIFFUSION 86
4.1 Overview 86
4.2 Implant Energy Effect on Defect Behavior 87
4.2.1 Experimental Procedures 87
4.2.2 Defect Microstructure and Dissolution
Process 87
4.2.3 Defect Density, Size and Size Distribution
during Annealing 90
4.3 Implant Energy Effect on Transient Enhanced
Diffusion 93
4.3.1 Experimental Procedures 93
4.3.2 Dopant Diffusion Behavior 94
4.3.3 Correlation Between Defects and TED 96
5 THE INFLUENCE OF IMPLANT DOSE ON DEFECT BEHAVIOR AND
TRANSIENT ENHANCED DIFFUSION 120
5.1 Overview 120
5.2 Implant Dose Effect on Defect Behavior 122
5.2.1 Experimental Procedures 122
5.2.2 Defect Microstructure and Dissolution
Process 122
5.2.3 Defect Density, Size and Size Distribution
during Annealing 126
5.3 Implant Dose Effect on Transient Enhanced
Diffusion 129
5.3.1 Experimental Procedures 129
5.3.2 Dopant Diffusion Behavior 130
5.3.3 TED Saturation at Higher Dose 132
6 SUMMARY AND FUTURE WORK 151
6.1 Summary 151
6.2 Future Work 158
LIST OF REFERENCES 163
BIOGRAPHICAL SKETCH 171
vi

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
DEFECT AND DIFFUSION STUDY IN BORON IMPLANTED SILICON
By
Jinning Liu
December, 1996
Chairman: Dr. Kevin S. Jones
Major Department: Materials Science and Engineering
The aim of this work is to better understand the nature
of transient enhanced diffusion (TED) of boron in silicon.
A matrix of sub-amorphization boron implants was used in
this work with energies ranged from 5 to 40 keV and the doses
ranged from 5x10-*--^ to 1x10-*-^ cm-^. The formation threshold
of {311} rod-like defects and sub-amorphization dislocation
loops are higher for lower implant energies. For the same
energy, the threshold dose for {311} defect formation is
lower than that for loop formation.
The study of annealing condition effect on {311} defect
behavior and boron TED shows that {311} defects dissolve more
rapidly at higher temperatures. The concentration of
interstitials trapped in {311} defects was below the value
predicted from previous studies of Si+ implants, indicating
Vll

the presence of submicroscopic boron interstitial complexes.
The interstitial emission from {311} defects during anneals
shows an exponential time dependence with an activation
energy of 3.8 eV. The time constant for TED saturation
process is smaller at higher temperatures and the dependence
shows an activation energy of 1.6 eV. Both {311} defects and
boron interstitial pairs are believed to be the source of
interstitials driving TED.
The implant energy effect on {311} defects and TED was
also studied. The dissolution of {311} defects is slower for
higher energy implants. This may be due to either an
increase in the surface recombination or an increase in the
ratio of boron interstitial clusters to {311} defects with
decreasing energy. Lower diffusion enhancement and shorter
duration of TED were observed for the 5 keV implant than for
all the higher energies at a dose of 2xl01^ cm-2. The {311}
defects might anneal slower than mobile boron interstitial
pairs so that the increase in {311} defect concentration may
account for the increase in TED duration. Surface
recombination is believed to play an important role for the 5
keV implant so that the amount of TED is reduced.
The study of dose effect on defect and TED shows that
{311} defects dissolve slower at higher doses. Diffusion
inceases with increasing dose initially. Boron interstitial
clusters and dislocation loops formed at high doses suppress
any further increase in TED with increasing dose above
2xl014 cm2 at 20 keV by acting as interstitial sink.
viii

CHAPTER 1
INTRODUCTION
1.1 Motivation and Objective
Since the development of the first integrated circuits,
the minimum feature sizes of devices have been reduced
dramatically. This miniaturization is expected to continue
and by the year 2000, the number of transistors per chip is
expected to reach one billion. The scaling of devices to
smaller dimensions both vertically and horizontally implies
lower voltages, thinner dielectrics and more abrupt and
shallower dopant profiles.1-3 The new devices will possess
the properties of higher packing density, higher speed and
lower power consumption.4,5 However, before the realization
of this goal, many challenges remain in processing and
packaging. One of them, for example, is to scale the
source/drain junction depth and the lateral penetration under
the gate in order to maintain a threshold voltage for the
MOSFET independent of the channel length.
Among the various methods available for forming shallow
junctions,6-11 ion implantation is currently the most widely
used one, due to its precise controllability, uniformity and
reproducibility. However, future devices require shallow
p+/n junctions with depths less than 1000 to minimize short
1

2
channel effects and to maintain low contact resistance.12
These shallow junctions will be difficult to form by ion
implantation. One of the reasons is that the implantation
process induces a supersaturation of silicon interstitials,
which will eventually lead to anomalous dopant diffusion,
known as transient enhanced diffusion (TED).13-20 TED is
caused by the pairing of excess silicon interstitials with
dopant atoms which reduces the energy barrier for the
migration of dopant atoms through the bulk. As a result, the
dopant penetration depth from diffusion is greatly increased.
TED lasts for only a limited time at typical annealing
temperatures so it is transient in nature.
TED will affect the source/drain dopant profiles both
vertically and laterally. The vertical diffusion results in
deeper profiles, which will weaken the control of the gate
voltage. The lateral diffusion reduces the effective channel
length which may result in punch-through phenomenon. Besides
these two effects, the enhanced diffusion of channel dopant
near the gate edge from excess interstitials created during
the lightly doped drain (LDD) or source/drain implant causes
an increase in the threshold voltage for shorter gate length
devices.21 This is the so-called reverse short channel
effect, i.e., an anomalous increase in the threshold voltage
with a decrease in device dimension. As device dimension
shrinks further, these effects will become more pronounced.
Thus the objective of this work is to study the effect
of implantation induced defects on transient enhanced

3
diffusion, and to provide an experimental basis for the
improvements of process development, process simulators and
ion implanters. These improvements would be of paramount
importance to optimizing the device structure.
1.2 Background
1.2.1 Shallow Junctions
A cross section of a complementary metal oxide
semiconductor (CMOS) structure is shown in Fig. 1.1. The
performance of this device is controlled by the gate
electrode, the gate dielectric, the channel and well region
and the source/drain junctions.
Gate
Figure 1.1. Schematic of MOS transistor showing how
interstitials are released from source/drain and LDD implants
and affect channel voltage.

4
As mentioned in the previous section, the common trend
in the development of solid state devices, either memory or
microprocessor, is the decrease in device dimensions. MOS
devices with sub-micron dimensions require junction depths of
0.1~0.2)lm. The most commonly used technique in forming
shallow junctions is to implant dopant species at low
energies into the silicon substrate.
To form shallow p+/n junctions, direct ion implantation
typically requires a very low energy boron implant preceded
by silicon or germanium amorphization to reduce channeling,
or a low energy BF2+ implant which uses the F to amorphize
the surface. 22-24 The use of pre-amorphization by Si+ or Ge+
introduces another implant step. In addition, the
amorphization of the surface will introduce end-of-range
damage, which may not be annealed out even at high
temperatures. This end-of-range damage is a source of
interstitials and, if located across the junctions, will
result in unacceptably high leakage currents. For the use of
BF2+ implants, the fluorine ions can affect the permeability
of the gate oxide to boron diffusion. This can cause a
penetration problem of boron atoms from the gate region
through the gate oxide to the channel region, therefore
undesirably affecting the threshold voltage of the device.
Thus one would like to avoid BF2+ implantation. One
alternative is to simply use low energy B+ implants.
However, there are several problems associated with B+
implants alone.

5
In addition to the channeling effect of B+ implantation
without pre-amorphization which can significantly increase
the junction depth, the post-implantation annealing results
in deeper profiles greater than that allowed by intrinsic
diffusion due to TED. TED is associated with the interaction
of mobile silicon interstitials with boron atoms.18'25,26
Implants into a pre-amorphized substrate can eliminate
channeling tails, but TED is still present. For B+ implant
the amount of TED is determined by both the number of excess
interstitials in the damage region and the partitioning of
these excess interstitials into different forms such as
mobile boron interstitial pairs, boron interstitial clusters,
{311} defects and dislocation loops and the role of the
surface. 27-29 This implantation induced TED will be discussed
in greater detail in section 1.2.4.
Shallow junctions are also formed in combination with
silicidation of source and drain. In general the implant and
anneal are done before metal deposition an silicide
formation. However, the implants could also be performed
through a deposited metal layer (Co or Ti mainly) before
silicide formation, 30-34 or into an already formed silicide
layer that then acts as a diffusion source (SADS) upon
subsequent annealing. 35-37 In the first two cases, a silicide
layer is formed after annealing with a steep implant profile
and a shallow junction. During the silicide formation, a
portion of the heavily doped silicon region is consumed and
the silicide interface moves into silicon, while dopants

6
segregate into the silicide.30 This causes serious contact
resistance problems.31-34 Moreover, silicidation perturbs the
point defect concentration in the substrate by injecting
vacancies and depleting interstitials,38,39 which should also
be taken into consideration during the process design. In
the case of SADS process, the dopant species is implanted
entirely into the silicide and then driven into silicon at
low temperature. This avoids both dopant channeling and
damaging of the silicon. SADS is a very attractive method to
form 300-400 junctions. However, extension implants are
still necessary to define the channel length etc. The
disadvantages of SADS include uneven doping due to rough
interface and dopant precipitates in the silicide.
Thermal diffusion is also used for junction formation.
While conventional diffusion does not create damage, it
suffers from poor controllability and requires a high thermal
budget. An alternative diffusion method is gas-immersion
laser doping (GILD).40 An excimer laser is used to rapidly
heat and melt the silicon surface. BF3 is then adsorbed,
pyrolized and diffused into the melt. The junction depth is
determined by the melt. A 100 junction depth has been
achieved with this method, which would be attractive if
technical, uniformity and cost issues can be resolved.
In summary, there are various ways to form a shallow
junction, and each has its own advantages and drawbacks.
This study is focused on the use of implantation of B+ into

7
crystalline silicon to investigate the effect of implant
damage on dopant redistribution.
1.2.2 Implant Damage and Types of Extended Defects
Ion implantation is a well-known technique for
controllably introducing dopants into the substrate during IC
fabrication. The major disadvantage of ion implantation is
the damage introduced by the energetic ions and the need for
post-implantation annealing to activate the dopant.
Different types of extended defects form after annealing,
depending on both the implant and the annealing conditions.
It is important to understand the defect formation kinetics
in order to study the effect of defects on dopant diffusion.
As an incident ion penetrates into the substrate, it
loses kinetic energy through elastic collisions with the
nuclei and inelastic collisions with electrons of the target
material until it stops. Elastic collisions are primarily
responsible for the lattice displacements through the
production of Frenkel pairs. If the kinetic energy
transferred to the host atom is higher than the displacement
threshold energy (~15eV for Si)41, the host atom leaves its
lattice site (primary collision) and collides with other
atoms (secondary collision), which in turn gives rise to more
collisions. This sequence results in a collision cascade.
For a sufficiently high concentration of damage, it is
possible to amorphize the crystalline silicon near the
surface. If the damage density is not high enough, some

8
isolated amorphous pockets may form without overlapping into
a fully amorphized layer. Different damage levels will
result in different types of extended defects after post
implantation annealing.
Residual implantation damage can be visible in the TEM
or submicroscopic such as dopant clusters. Extended defects
visible in a TEM are categorized according to two formation
mechanisms: point defect condensation and solid phase
regrowth. 42,43 There are three types of extended defects that
arise from the first mechanism and two types from the second
mechanism, as will be discussed below.
I. Extended defects from point defect condensation
1. Sub-amorphization defects
When the implant damage is not sufficient to turn the
silicon surface into a continuous amorphous layer, sub-
amorphization defects form during subsequent annealing.
These defects arise from a supersaturation of point defects
due to the non-conservative nature of the implantation
process. They form around the projected range of the
implanted ions, where the supersaturation level is the
highest. These defects have two common forms: rod-like {311}
defects and dislocation loops. Sub-amorphization {311}
defects are one of the main interests of the present study
and they will be discussed in detail in the next section.
2. End-of-range defects
When the silicon surface has been amorphized during
implantation, end-of-range defects form just below the

9
amorphous/crystalline interface upon annealing. End-of-range
defects may consist of both dislocation loops and {311}
defects if the annealing temperature is low. These defects
are important in IC processing since amorphous layers are
always present for P, As and BF2 implants under the implant
conditions that are usually used. For more information about
the formation and evolution of these defects, please refer to
references 42-46.
3. Precipitation related defects
If the concentration of the implanted impurity is above
its solid solubility in silicon, precipitation related
defects may form around the projected range. The
precipitation process may generate such a high concentration
of intrinsic point defects that extended defects can form as
a result. The most common example of this is after high dose
arsenic implantation. 43,47,48 Two layers of dislocation loops
have been observed upon annealing, one at the projected range
which is associated with As clustering, the other at the end
of range which is associated with excess interstitials beyond
the amorphous/crystalline interface.
II. Extended defects from solid phase regrowth
1. Regrowth defects
When the solid phase epitaxial regrowth process is
imperfect, defects nucleate at the advancing
amorphous/crystalline interface and propagate into the
regrowing silicon. They do not appear to dramatically affect
the diffusivity of dopants, but they can getter impurities

10
and provide leakage paths if they span across a
junction. 42,49,50
2. Clamshell defects
If the implant energy is sufficient but the dose is not
too high, it is possible to produce a buried amorphous layer.
Upon annealing two amorphous/crystalline interfaces meet and
extended defects may result. These defects consist of larger
dislocation loops compared to end-of-range defects.49,51
We have discussed briefly each type of extended defects
due to ion implantation. Visible damage from boron implants
employed in this study consisted predominantly of {311}
defects in the annealing temperature range that we used. In
the next section, we will review some previous studies on
{311} defects in detail.
1.2.3 (311) Rod-like Defects
As we discussed in the last section, {311} defects can
form under both sub-amorphization and amorphization
conditions, providing the annealing temperature is low and/or
the annealing time is short. This is summarized in Fig. 1.2.
Because of the low mass of the boron ions, the silicon
substrate surface is not amorphized if the dose is below
5xl015 cm 2. Under such sub-amorphization conditions, {311}
defects form at lower doses than dislocation loops. They are
relatively unstable. Upon annealing, they either evolve into
dislocation dipoles through an unfaulting reaction or
dissolve. The dissolution provides interstitials to the

11
Energy and/or Dose
sub-amorphized amorphized
as-implant
Figure 1.2. Schematic of the conditions under which {311}
defects can form and how they evolve with time.
background environment (which can result in TED) and the
growth of dislocation loops if they are present. Since {311}
defects form after a short time and/or low temperature anneal
with or without the presence of dislocation loops, and they
dissolve fairly fast, they have been termed intermediate
defect configurations (IDCs). As an example, Fig. 1.3 shows a
plan-view transmission electron microscope picture of {311}
defects in boron implanted silicon after an anneal at 750C
for 5 min. Under the imaging conditions used, there are

12
three directions of {311} defects. Two of them are
orthogonal to each other while the third one forms a 45
angle to the first two sets. This morphology will be
explained in the following paragraphs.
Figure 1.3. Example of {311} defects in 30keV lxlO-'-^cm-^ b+
implanted Si under weak-beam dark field imaging condition
with g220 reflection, annealed at 750C for 5min in nitrogen.
{311} defects are elongated in [110] directions and
consist of interstitials precipitating on {311} habit planes.
A unit cell of silicon lattice with {311} defects lying in
six [110] directions is schematically shown in Fig. 1.4. The
reflection vector g was chosen to be 220 when plan-view
transmission electron microscopy pictures were taken in this

13
study. Since {311} defects are out of contrast for g
parallel to their length, those on one of the six directions
are invisible. Thus viewing on a [100] plane, we observed
the morphology shown in Fig.1.3.
line direction u
gxu
[110]
*0
[-110]
=0
[101]
*0
[10-1]
*0
[011]
*0
[01-1]
*0
visible(+) or invisible(-)
+
+
+
+
+
Figure 1.4. Schematic of {311} defect directions and defect
visible conditions under reflection g=220.
Various atomic models incorporated by self-interstitials
have been proposed for the {311} defects. Salisbury and
Loretto52,53 suggested that self-interstitial atoms located on
the tetrahedral sites are responsible for {311} defect

14
formation. Models of split self-interstitials with or
without dangling bonds were proposed by Tsubokawa et al.54 and
Pasemann et al.55 respectively. Tan56 developed a homogeneous
nucleation model which shows that for extrinsic defects, a
chain of interstitial atoms form {311} defects having non-
six-membered atomic rings with matrix atoms, while for
intrinsic defects, a chain of matrix atoms is cut out and the
remaining atoms surrounding the cut are used to form {311}
defects having non-six-membered atomic rings. A schematic
diagram for the different steps involved in the formation of
extrinsic {311} defects is shown in Fig.1.5 (a)~(d). This
non-six-membered atomic ring configuration minimizes the
dangling bond that may present.
(a)
Figure 1.5. Schematic of four step {311} defect formation
process: (a) The accepted split [100] interstitial chain, (b)
The addition of one more split [100] chain, (c) Details of
the bond rearrangement and(d) The resulting defect
configuration.

15
(c)
Figure 1.5. (Continued)

16
(d)
Figure 1.5. (Continued)
The displacement vector of {311} defects is
perpendicular to the line and has a prominent [100]
component. This vector has been determined to be a<100>/2,57
a<411>/6,58 a<311>/ll52 or a<611>/k by various groups, where a
is the lattice constant and k=2559 or 31.60 The discrepancy
between these values reflects the inaccuracy of the
measurement because of the narrowness of the defects.
Different methods have been used to determine the
displacement vector, including the traditional image contrast
analysis and the energy minimization calculation based on the
relaxation of the interstitial agglomerate model under the
Stillinger-Weber potential.
{311} defects have been studied in various ion implanted
or particle irradiated systems for more than two decades.
They were found in silicon,61 boron, 57,62'64 arsenic,65

17
phosphorous66'68 and BF269 implanted silicon, electron,52,70
neutron71, proton72 and neon73 irradiated silicon as well as
electron irradiated germanium.59 The implant temperature
effect on {311} defect formation has also been studied.
Tamura74 found that room temperature phosphorous implants in
silicon at low energies resulted in the formation of
dislocation loops, whereas elevated temperature implants
(200~400C) gave rise to the formation of rod-like {311}
defects. This temperature effect has also been shown by
Chadderton and Eisen.75 They found that dislocation loops
formed in boron implanted silicon if the implantation was
held at liquid nitrogen temperature, in contrast with the
room-temperature implants which showed {311} defects. This
behavior might arise from the self-annealing effect occurring
at higher temperatures which reduces the number of free
interstitials. Under this condition {311} defects may be
easier to form than dislocation loops presumably because the
energy to form a {311} defect is less than that of a loop.76
Back in the seventies, the formation of {311} defects
was associated with the presence of boron by Seshan and
Washburn.67,68 They observed {311} defects in phosphorous
implanted boron-doped silicon. However, the {311} defects
were absent from n-type silicon under the same implant and
annealing conditions. Thus they concluded that boron is
essential for the formation of {311} defects. Recently, a
research group at Bell Laboratories (private communication)
investigated the role of boron during {311} defect formation

18
by studying Si+ implanted silicon doped with various
concentrations of boron. They measured the number of
interstitials trapped in {311} defects after the same post
implantation anneal and found that the interstitial density
actually decreased with the increase of boron concentration.
We have also observed a higher formation threshold of {311}
defects in boron implanted than that in silicon implanted
silicon. It is believed that boron atoms pair with
interstitials and form submicroscopic boron interstitial
clusters and mobile boron interstitial pairs, which consume
part of the interstitials and reduce the number available for
{311} defect formation. This will be discussed further in
the following chapters. It should also be mentioned that the
new interest in {311} defects arises partly from the fact
that they have been claimed to be the only source of
interstitials for TED, which is still controversial at this
time.
1.2.4 Dopant Diffusion in Silicon
The diffusion of dopant atoms can be described by
DA = D0exp(--^-) (1.1)
where Da is dopant diffusivity, Do is the pre-exponential
constant and Qa is the activation energy for diffusion. It
is generally agreed that on an atomic level, dopant atoms in
silicon diffuse through interactions with silicon self
interstitials and vacancies. The type of point defects which
dominates the diffusion process determines the type of

19
diffusion mechanism, vacancy, interstitial or interstitialcy
mechanism. In the vacancy mechanism, a substitutional dopant
atom exchanges position with an empty lattice site. In the
interstitial mechanism, the dopant atom is kicked out of the
silicon lattice by a silicon self-interstitial and diffuses
as a pure interstitial before returning to the lattice as a
substitutional atom. In the interstitialcy mechanism, the
silicon self-interstitial and the dopant atom form a
diffusion pair. This process is illustrated in Fig. 1.6.
Usually no distinction is made between the interstitial and
the interstitialcy mechanisms.
Vacancy
Interstitial
Interstitialcy
Figure 1.6. Schematic diagram of different dopant diffusion
mechanisms.
If an asterisk is used to represent atoms in their free
states, then the interaction of dopant atoms and the point
defects can be expressed as
A*+X* <-> AX (1.2)
where X=I corresponds to interstitial/interstitialcy

20
mechanism while X=V refers to a vacancy mechanism.
Mathematically the ratio of dopant diffusivity under
nonequilibrium conditions over its intrinsic, equilibrium
value can be written as
Cv
_ T1A\/
Da* Cj*
av
Cy*
(1.3)
where fAX=DAX*/A* (X=I or V) fAI+fAV=l/ CI and Cy are the
concentration of interstitials and vacancies under
nonequilibrium conditions, Ci* and Cy* are the quantities
under intrinsic, equilibrium conditions. Equation (1.3)
shows that the diffusion of dopant atoms consists of two
parts: diffusion by coupling with interstitials and with
vacancies. The fraction of each is represented by fAl and
fAV, respectively. While this treatment of diffusion allows
us to mathematically explain the experimental results, it is
possible that the actual mechanism is quite different.
Dopant diffusion in silicon has been under investigation
for many years; however, the diffusion mechanisms responsible
for some of the impurities are still under debate. It is
generally accepted that antimony diffusion is dominated by a
vacancy mechanism. Phosphorous and boron diffuse
predominately by an interstitial/interstitialcy mechanism.
Arsenic appears to diffuse by both vacancy and
interstitial/interstitialcy mechanisms. The diffusion of Al,
Ga and In is believed to have a strong interstitial
component77. Elements such as Li, Na, He and H are believed
to diffuse by an interstitial/interstitialcy mechanism as do

21
most of the transition metals78,79 The mechanism of Si self
diffusion appears to be vacancy controlled at temperatures
<1000C and interstitial controlled at higher temperatures.80
Boron is the diffusing species that will be discussed in
this study. The set of nonequilibrium reactions involved in
the kinetic model for boron diffusion is described as
follows:
Bs + I <-> (BI)
(1.4)
(BI) + V <-> Bs
(1.5)
I + V <-> [0]
(1.6)
where Bs and (BI) are the substitutional boron and the boron
interstitial pair, respectively, and [0] denotes the silicon
lattice site.
The relationship between point defect concentration and
dopant diffusion is given by Eq. (1.3). To understand dopant
diffusion in silicon, it is necessary to understand how
different processes affect point defect concentration.
Because of its small value relative to the concentration of
silicon atoms residing on the lattice sites, the
concentration of point defects is not a quantity that can be
measured directly. Diffusion studies are mostly used to gain
information of the interstitials and vacancies. An enhanced
diffusion of a species which diffuses by interstitial (or
vacancy) mechanism indicates a supersaturation of
interstitials (or vacancies) in the bulk. Ways of
quantitatively monitoring these point defects will be
discussed later.

22
Under equilibrium conditions at a particular
temperature, the possible sources of point defects are
chemical reactions at the silicon surface, such as
oxidation,81-86 nitridation87-90 and silicidation,91-96 the
impurity clustering and precipitation, 97-100 and the radiation
damage, 20,101-105 etc. Deviations from equilibrium will cause
excess or depletion of either interstitials or vacancies,
which will further affect dopant diffusion. For dopants that
diffuse predominantly by an interstitial/interstitialcy
mechanism, an excess of interstitials or depletion of
vacancies will enhance the dopant diffusion, while for
dopants that diffuse primarily by vacancy mechanism, the same
conditions will retard dopant diffusion.
Two possible detectors for excess point defects in
silicon are dopant profiles and extended defects. Dopant
profiles include both epitaxially grown doping superlattices
or implant profiles. Boron and antimony superlattices can be
used to monitor the in-diffusion of silicon interstitials or
vacancies produced by processes such as surface oxidation,
nitridation, silicidation and aluminization, or the implant
damage. Since boron diffuses mostly via
interstitial/interstitialcy mechanism, the boron epi layer
spike will broaden after oxidation and implantation during
which processes interstitials are injected into the bulk. On
the contrary, silicidation, nitridation and aluminization
will cause vacancy injection or interstitial depletion,
therefore broadening the antimony epi layer spike. The

23
implant profiles behave in the same way as the epi layer
spikes but are less sensitive because the epi layer profiles
are thinner and possess a more ideal Gaussian shape.
Simulation programs such as FLOOPS can then be used to
calculate the diffusivity of the dopant atoms and quantify
the enhancement or retardation effect.
Other than dopant profiles, extended defects can also be
used to monitor the point defect perturbance. Early studies
of the trapping of point defects by extended defects were
conducted by using stacking faults, 82,106_108 which grow by
absorbing interstitials and shrink by emitting interstitials,
thereby giving a good indication of the change in point
defect concentration. However, because of their large size
(> 5 |im) and low density (< 10^ cm-2), the stacking faults
are not ideal for detecting small perturbations. Instead,
post-implantation annealed dislocation loops (end-of-range
loops) have the advantage of much smaller size (~ 0.02 |lm in
diameter) and larger density (> 10^-0 cm-^) t therefore being
more sensitive detectors. In addition, the dislocation loops
form at well-defined depth, which can not only be utilized to
measure the net interstitial flux, 105,109 but have also been
found to be an effective barrier to interstitials driving
anomalous dopant diffusion.17'110,111
Extended defects themselves can act as a source of
interstitials during annealing. Since dislocation loops
dissolve at temperatures that are too high (>1000C) to
explain TED which is completed within seconds at T>950C, the

24
{311} rod-like defects have been examined as the possible
sources of interstitials that cause TED. Cowern et al.103
claimed that the release of interstitials which drive TED of
boron in silicon occurs at two different time scales.
Interstitials emerging from the ion collision cascade give
rise to a primary (ultra fast) pulse of diffusion, and at a
sufficiently high displacement density, emission of
interstitials from {311} defects causes a secondary (much
slower) diffusion transient. Later, Eaglesham et al.112 and
Stolk et al.113 reported that {311} defects are the only
source of interstitials driving TED based on their study of
Si+ implanted Si. They reasoned that interstitials are
emitted from {311} defects in the damage region during
annealing. This emission occurs at a rate which exhibits the
same activation energy as that of the interstitial in
diffusion, and the duration of {311} dissolution process is
in close agreement with that of TED. Our group have been
studying {311} defects and TED in boron implanted silicon.
We found that TED occurs without the formation of {311}
defects if the implant energy (4 keV) and dose (lxlO14 cm"2)
are low.114 The source of interstitials driving TED in this
case was attributed to sub-microscopic mobile boron
interstitial pairs. These B-I pairs play a key role in TED
of boron in silicon, as will be discussed.
Therefore, understanding the correlation between
implantation induced defects (e.g., {311} defects and boron
interstitial clusters) and TED is essential to the

25
understanding of dopant diffusion. This correlation will be
further investigated in this study.
1.3 Ion Implantation. Characterization and Simulation
Techniques
1.3.1 Ion Implantation
Ion implantation is the introduction of energetic
charged particles into targets with enough energy to
penetrate beyond the surface region. The ions enter the
substrate, collide with the host atoms, gradually lose energy
and finally come to rest at some depth within the substrate.
Ion implantation is involved in several steps during the
fabrication sequence of MOSFET and bipolar transistors when
the controlled doping of selected areas is required.
An ion implanter consists of the following major
components: an ion source, an extracting and ion analyzing
system, an accelerating column, a scanning system and an end
station. The ions produced by the source are extracted by a
small accelerating voltage and then injected into the
analyzer magnet. A spatial separation of ions subjected to
the Lorenz force occurs due to the differences in the mass
and charge. The selected ions are injected into the
accelerating column and the rest are screened out. In this
case the system operates in the pre-analysis configuration.
If the ions are accelerated to their full energy before the
mass separation, then it works in the post-analysis

26
configuration. The extraction voltage ranges usually between
15 and 40kV. For low energy implants, a retarding voltage
can be used after separation. The selected ions are
accelerated by a static electric field and focused and shaped
in the column and ready to be implanted.
Some important quantities during the ion implantation
process are listed below. The implant dose can be expressed
as
where I is the implant current, t is the time, n=l for singly
ionized species and 2 for doubly ionized species, q is the
charge and A is the area of the implanted surface. For an
ion accelerated through a potential V and deflected through a
magnetic field B, the radius of the ion's path is
1 2MV
(1.8)
q
where M is the mass of the dopant atom. In absence of
crystal orientation effects, the range distribution is
roughly Gaussian and to a first approximation the projected
range distribution N(x) is described as a one-dimensional
Gaussian profile characterized by the projected range Rp and
the standard deviation ARp, as given by
(1.9)
Schematic diagram of an implanted impurity profile is shown
in Fig. 1.7. Due to the channeling effect, the use of a
Gaussian profile is not accurate. The UT-Marlowe code (a

27
Monte Carlo simulator) yields a much more accurate picture of
the ion distribution just after implantation.
Figure 1.7. Schematic illustration of a Gaussian
distribution resulting from ion implantation.
1.3.2 Transmission Electron Microscopy
The transmission electron microscope (TEM) is used to
obtain information from samples which are thin enough to
transmit electrons. A bright-field image is formed if the
directly transmitted beam is selected and a dark-field image
is formed if a diffracted beam is selected. Dark-field
images normally have higher resolution than bright-field
images and this is the mode we used in our study. The defect
structure can be analyzed by both plan-view and cross-
sectional TEM. Plan-view TEM (PTEM) is used to study the
defect evolution during annealing and cross-sectional TEM

28
(XTEM) is used to get the side view and the depth of the
defect layer.
To make a PTEM sample, the sample is first cut into 3mm
diameter pieces using an ultrasonic disk cutter. The small
disk is then polished down to about half of its original
thickness using a polishing jig with 15|im aluminum oxide
powder. A coat of wax is then applied to the top surface
(i.e. the implanted surface) of the sample to protect it
during the etch. The sample is mounted with wax on a Teflon
holder facing down and etched from the back side in a jet
etcher using a HF: HNC>3 = 25%: 75% solution until a small hole
with an electron transparent region around it is obtained.
The wax is the removed from the sample using heptane and/or
acetone and the sample is ready for analysis.
To make a XTEM specimen, a piece of silicon wafer is cut
into lCmil thick strips using a dicing saw. The top surface
of the two strips is coated with M600 Bond epoxy resin. The
two top surfaces are then brought into close contact with
each other. Two silicon dummy strips can also be glued on
the side of each sample strip to improve the mechanical
stability. The bonded strips are heated in a conventional
oven at 100C for lhr to harden the epoxy resin. The cured
strips are then polished down to a thickness of about 40|lm
using 5|im aluminum oxide powder. A copper ring is mounted
onto the thinned strips using the M600 Bond, leaving the
interface of the sample strips exposed in the center of the
ring. The specimen is then further thinned in a Gatan two-

29
stage ion mill by using two Ar+ ion guns at a gun voltage of
4 kV and a gun current of 0.5 mA until a small hole is
obtained at the sample strips' interface. The region in the
vicinity of the hole is sufficiently thin (<0.5 |im) to be
transparent to the electron beam of the TEM.
1.3.3 Secondary Ion Mass Spectroscopy
In secondary ion mass spectroscopy (SIMS), an energetic
beam of focused ions is directed at the sample surface in a
high or ultrahigh vacuum environment. The momentum transfer
from the impinging primary ions to the sample surface causes
sputtering of the surface atoms and molecules. The most
commonly used sputtering ions are Cs+( 02+, 0+ and Ar+ in the
energy range from 2 to 20 keV at angles of incidence between
45 and 90. 02+ ions are used in this study to enhance
secondary ion yield. Some of the sputtered species are
ejected with positive or negative charges. These are termed
secondary ions. The secondary ions are then mass analyzed
using a mass spectrometer. SIMS can be used to obtain a
depth concentration profile of the near-surface region to a
resolution of 2~30 nm at a detection limit of 10^-10^ cm-^ .
For further information see ref.115
1.3.4 Florida Object Oriented Process Simulator
FLOOPS modeling program is a physically based two-
dimensional process simulator capable of modeling a complex
device structure. It uses local point defect concentration

30
to determine dopant diffusivities. In order to keep track of
the spatial distribution of point defects, FLOOPS solves the
diffusion equations for the dopant atoms as well as the
equivalent equations for the point defects. Since vacancies
and interstitials can interact, the equations describing
their motion are coupled. The equations used to
determine dopant diffusion include
(1.10)
A
dt
(1.11)
m
(1.12)
where Ja is the flux of the dopant, Ca is the total
concentration of the dopant species, the summation over X is
for all possible diffusion mechanisms, Cx is the
concentration of the X species, Dax is the diffusivity
corresponding to the X mechanism, ni and n refer to the
intrinsic and actual electron concentrations, Cai and Ca2 are
the concentration of the dopant species on the two sides of a
materials boundary, Ktr is the transport coefficient across
the material boundary and m is the segregation coefficient of
the dopant for the boundary materials. The equations solved
for interstitials or vacancies are
(1.14)
(1.13)
(1.15)
where x=I for interstitials and V for vacancies, Jx is the
flux of the X point defect species, Cx* and Cx are the

31
equilibrium and actual concentrations of X, Gx is the surface
generation rate for X, Kxs is the surface recombination
constant and Kfc>ulk is the bulk recombination constant.
For the case of dopant diffusion following an
implantation, the initial distribution of point defects is
determined by the implantation process. These point defects
then diffuse according to the above equations. Anomalous
diffusion can be observed until the point defect
concentrations return to the equilibrium values, when only
normal diffusion can be seen.
1.4 Thesis Statement
The contributions of this work are in the following
areas:
1. Determination of the formation threshold of extended
defects in low energy B+ implanted silicon.
2. Quantitative TEM studies of the annealing kinetics of
{311} defects arising from B implantation.
3. Experimental investigation of the implant energy and
dose effects on {311} defect behavior.
4. Experimental investigation of the implant energy and
dose effects on TED.
5. Extraction of diffusivity enhancement for different
implant energies, doses, annealing times and temperatures.
6. Experimental investigation of the sources of
interstitials driving dopant diffusion.

32
7. Provision of experimental basis for the improvement
of process development and process simulators.

CHAPTER 2
THE FORMATION THRESHOLD OF {311} DEFECTS AND SUB-
AMORPHIZATION DISLOCATION LOOPS
2,1 Overview
The formation of very shallow p-type region by ion
implantation requires the use of low energy boron implants.
Although there have been many reports on the defect and
diffusion behavior in boron implanted silicon,63,116-118 none of
them have systematically studied the defect formation
threshold in the low energy implantation regime. For boron
implantation, because of the low mass of the implanted ions,
even if the ion fluences reach 2xl015 cm~2( the damaged
surface layer still does not amorphize. In this case the
extended defects that form after annealing can be classified
as type I or sub-amorphization defects. The implant energy
and dose combinations in this study are similar to those used
in the semiconductor industry for B implants and fall in the
sub-amorphization regime.
In order to understand how the sub-amorphization defects
influence dopant diffusion, it is important to know the
implant conditions under which these defects are actually
formed. This is the purpose of this section of the current
study.
33

34
2.2 Experimental Procedure
Czochralski-grown (100) n-type 8-20 cm 150 mm wafers
were implanted with boron ions at energies of 5 keV, 10 keV,
20 keV, 30 keV and 40 keV to doses of 5x10^--^ cm-2,
lxl0-*-4 cm-2, 2x10-'-^ cm-2, 5x10^4 cm-2 and lxlO^S cm-2. The
tilt/rotation angles were 5/0. The implant current was
3 mA. During ion implantation, the samples were kept at room
temperature using water cooling. Furnace anneals were then
performed at 750C for 5 min or 900C for 15 min in a
nitrogen ambient to study the formation threshold of {311}
defects and sub-amorphization dislocation loops respectively.
Plan-view transmission electron microscopy (PTEM) samples
were prepared using standard jet-etching procedures and
Cross-sectional TEM (XTEM) samples were prepared using ion
milling. Micrographs were taken from each sample to examine
the presence or absence of secondary defects. The g220
reflection was used to acquire all the micrographs under
weak-beam dark field imaging conditions. A defect density of
1.2xl07 cm-2 was used to distinguish between samples with and
without extended defects, i.e., if no defect is observed in
three randomly-picked areas each with a size of 7x10 cm2
under a magnification of 50000X, then it is stated that there
are "no" defects in that sample.

35
2.3 Defect Microstructure
Figure 2.1 shows PTEM micrographs of some of the samples
in the implant matrix after an anneal at 750C for 5 min.
Micrographs of 5 keV, 20 keV and 40 keV implants with doses
of lxlO-*-^ cm-2, 2x10^4 cm-2 and 5x10-*-^ cm-2 are shown in the
figure. At a dose of 1x10-*-4 cm'^, there are no {311} defects
in the 5 keV sample and the defect density increases with
increasing energy. When the dose is doubled to 2x10^4 cm-2,
there are still no {311} defects in the 5 keV sample, but
more defects are observed in the other two implants.
Increasing the dose further to 5xl014 cm-2 results in {311}
defect formation at 5 keV. Figure 2.1 clearly presents a
transition from samples without to those with {311} defects.
It is obvious that as the implant energy increases, the
critical dose for forming {311} defects decreases. From this
figure it is also obvious that the interstitial
supersaturation necessary to nucleate {311} defects is far
less than that for dislocation loops. This is consistent
with Eaglesham 's observation (private communication) of
{311} defects at 7xl012 cm-2 for 40 keV Si implants and the
observation of Jones et al.42 that stable dislocation loops
(approximately the same implant energy) do not form until a
dose of 2xl0l4 cm-2. In addition, the threshold dose for
{311} defects in a 20 keV B implant is around 2x10^^ cm-2,
which is much greater than the threshold dose of 7xl012 cm-2
for a comparable depth of 40 keV Si implant. This is

36
presumably due to the formation of immobile boron
interstitial clusters and mobile boron interstitial pairs
that reduces the concentration of free interstitials
available to form {311} defects.
PTEM micrographs of selective samples annealed at 900C
for 15min are shown in Fig. 2.2. This temperature/time
condition was chosen to reveal stable sub-amorphization
dislocation loops and dipoles that evolve from unstable {311}
defects. The {311} defects have completely dissolved after
the anneal. A submatrix of 5 keV, 20 keV and 40 keV implants
with doses of 2x10^4 cm-^, 5x10-'-^ cm-2 and 1x10^^ cm-2 are
shown in Fig. 2.2. The 40 keV sample has dislocation loops
and dipoles at a dose of 2x10^4 cm2. The 20 keV and 5 keV
samples do not show loops or dipoles until the dose is
increased to 5x10^4 Cm2. The critical dose for type I loop
formation appears to decrease with increasing implant energy.
The threshold dose at 40 keV (2x10^4 cm-^) matches the
previous results of Jones et al.'s for 30 keV boron. Table
2.1 lists the formation threshold for both {311} defects and
[110] loops for the whole implant matrix.
XTEM micrographs of 10 keV, 20 keV, 30 keV and 40 keV
implants to a dose of lxlO^^ cm~2 after an anneal at 750C
for 5 min are shown in Fig. 2.3. The reason for choosing the
highest dose in our implant matrix is that both {311} defects
and dislocation loops are present in these implants under
this annealing condition. A layer of sub-amorphization
extended defects is expected to form at the projected range

37
Rp. Compared with the SIMS profiles which will be shown
later in Chapter 4, the depth of the defect layer was found
to correspond to Rp. Both the distance between the defect
layer and the surface and the width of the defect layer
decrease as implant energy decreases. This decrease of
distance between the damage region and the surface may induce
a more prominent recombination effect at the surface, as will
be discussed later.
2.4 Criteria for the formation of (311) Defects and Sub-
amorohization Loops
Various suggestions have been made for the criterion for
secondary defect formation. Tamura et al.119 have stated that
implant dose can be used as such a criterion. They studied
defect formation in 1~2 MeV B, P and As implanted silicon and
found that secondary defects form if the dose range of
2xlO-*-3~lxlC)14 cm2 is exceeded, independent of the implant
species. Other experiments, however, show that a higher dose
is required for defect formation if the ion mass is smaller.
Gibbons120 studied B, Al, N, P, Sb, Ne and Si implanted
silicon at energies from 40 keV to 100 keV. The threshold
for small loop formation for Sb implant is about 5x10-1-3 cm~2
while for B implant it is above lxlO14 cm2. This implies
that the implant dose cannot be the sole criterion for loop
formation. Furthermore, the results of studies on channeling
versus random implants show that the defect formation is not
dependent only on impurity dose but on the damage introduced

38
during implantation. Raineri et al.121 studied random and
[100] channeling implants with B and P ions. 100 keV P
implants with doses from 5xl012 cm-2 to 2x10^4 cm-2 does not
show defects under channeling conditions after an anneal at
500C for 1 hr. The same implants in a random direction give
rise to a network of dislocation loops. A systematic study
on implant damage using TEM, RBS and TRIM simulations was
reported by Schreutelkamp et al.122 They claimed that if the
total number of silicon atoms displaced by the implant ions
exceeds a critical value, stable sub-amorphization
dislocation loops are observed. They implanted Si with B,
Si, P, Ga, As, In and Sb ions at keV and MeV energies. For B
implants with energies ranging from 50 keV to 190 keV, the
critical value of displaced Si atoms was 4.7x10-*-^ cm-2 based
upon TRIM calculations and 1.4x10-1-6 cm-2 based upon
Rutherford Backscattering (RBS) analysis. This value
increases with the mass of the implant ions. To the author's
best knowledge, there has been no report on the formation
threshold of {311} defects or stable loops in the low energy
(<20 keV) regime of B implanted silicon.
We have calculated the number of displaced silicon atoms
using TRIM for our implant matrix and the results are listed
in Table 2.2. As can be seen, when the displaced atom
density reaches about 1.5x10-1-6 cm-2, {311} defects form, and
when it reaches about 2.6xl016 cm-2, sub-amorphization loops
form. The latter value is less than the result of
Schreutelkamp et al., but considering the fact that we used a

39
new version of TRIM, the results are in reasonable agreement.
It thus appears that the number of displaced silicon atoms
can be used effectively as the criterion for both {311}
defect and sub-amorphization dislocation loop formation.
If the above argument is true, we expect to see more
extended defects after annealing in samples containing more
displaced silicon atoms after implantation. This indeed is
the case for both anneal conditions with only two exceptions.
The first is the 5 keV 1x10^^ cm2 implant after an anneal at
750C for 5 min (Fig. 2.1). Keeping the dose at 1x10^-^ cm-2,
we see less {311} defects but more dislocation loops in the
5 keV implant than in the higher energies. The reason might
be that the 5 keV lxlO^-S cm-2 implant has the highest
interstitial concentration in the whole implant matrix. When
the number of free interstitials reaches a critical value,
the system may form loop nuclei. If the energetics are such
that interstitials would rather be in a loop, then the
loop/{311} ratio increases, i.e., loop formation might be
more favorable than {311} defect formation. The second
exception to the damage alone being the reason for threshold
changes arises when one compares the 5 keV and the 20 keV
implant. The displaced atom density for 5 keV 5x10^-^ cm-2
implant (2.56xl01^ cm-2) is slightly less than that of the
20 keV 2xl014 cm-2 implant (2.94xl01^ cm-2) (Table 2.2), but
after an anneal at 900C for 15 min, the 5 keV sample shows a
low density of half loops while the 20 keV implant shows no
defects. The reason might be that these displaced atom

40
densities are near the threshold for defect formation so the
behavior is due to the other fluctuations. However, the
reason may also be due to the above argument for why loops
are seen most prominently for 5 keV implant.
Besides the effect of implant energy on damage
production, interstitial recombination at the sample surface
may also affect the formation of secondary defects. Figure
2.4 shows the projected range Rp as a function of implant
energy obtained from the XTEM micrographs and the SIMS
profiles which will be shown in Chapter 4. Rp increases
with increasing energy. It is about 230 for the 5keV
implant and about 1750 for the 40keV implant according to
SIMS. As the implant energy decreases, the damage region
gets closer to the sample surface. Surface recombination
could then become more efficient and thus reduce the number
of free interstitials available to form secondary defects.
It is difficult to determine which effect, decreased damage,
increased boron interstitial cluster formation due to higher
interstitial supersaturation in the dopant peak region or
increased surface recombination, is primarily responsible for
the increased threshold dose with decreasing implant energy.
Additional experiments to sort out these effects will be
necessary.
In summary, the formation threshold for both {311} rod
like defects and [110] dislocation loops in low energy boron
implanted silicon have been investigated. A matrix was
chosen with implant energies ranging from 5keV to 40keV and

41
doses from SxlO-^cm2 to lxiol^cm-2. TEM was use to evaluate
the presence of both types of defects. It was observed that
the threshold dose for both {311} defect and sub-
amorphization dislocation loop formation increases
significantly with decreasing implant energy. This threshold
is higher for boron implants than for silicon implants
presumably because of the formation of boron interstitial
clusters and mobile boron interstitial pairs. As discussed
earlier there are several possible explanations for the
higher threshold dose with decreasing implant energy. These
include a decrease in the displaced atom density, increased
formation of boron interstitial clusters and/or increased
surface recombination. Experimentally a displaced atom
density of about 1.5xl0l6 cm-2 for {311} defects and
2.6xl01^ cm-2 for sub-amorphization loops from TRIM
calculations reasonably predicts defect formation for these
implant conditions. However, the results of Zhang et al.114
and Eaglesham et al.(private communication) clearly indicate
that boron is a very effective interstitial trap and that
increasing the boron concentration clearly results in an
increase in the concentration of interstitial trapping.

lxlO^cirr2
5keV
20keV
\ '
40kcV
. C -
\ : >
A % \ ^
*
v.
0.2^im
2xl014cm*2 5xl014cm'2
Figure 2.1. Weak beam dark field (g220> PTEM images of 5keV, 20keV and 40keV B+ implanted
Si to doses of lxl0^-4cm-2, 2xl0^-4cm-2 and 5xl0^-4cm-2, after an anneal at 750 C for 5min in
N2.
N>

5keV
20keV
40keV
0.2nm
2xl0^cm2 5xl0^cm'2 lxlO^cm'^
1>J
Figure 2.2. Weak beam dark field (g220) PTEM images of 5keV, 20keV and 40keV B+ implanted
Si to doses of 2xl0^^cm~2> SxlO^cm-^ and lxlO^^cm-^, after an anneal at 900C for 15min in
N2.

Table 2.1 Types of extended defects formed in B+ implanted silicon
5keV
lOkeV
20keV
30keV
40keV
5xl0l3cm'2
none
none
none
none
none
Ixl0l4cm"2
none
none
{311 }s
{311 }s
{311 }s
2xl0l4cnr2
none
{311 }s
{311 }s
{311 }s
{311 }s
Loops
Loops
5xl0^cm_2
{311}s
{311}s
{311}s
{311}s
{311}s
Loops
Loops
Loops
Loops
Loops
lxl()15cm"2
{311}s
{311} s
{311 }s
{311 }s
{311 }s
Loops
Loops
Loops
Loops
Loops

45
30keV
4keV
Figure 2.3. Weak beam dark field (g220) XTEM images of lOkeV,
s
lOkeV
20keV
20keV, 30keV and 40keV B+ implanted Si to a dose of lxlC)15crn-2f
after an anneal at 750C for 5min in N2.

Table 2.2 Displaced atom density (#/cm2) from TRIM calculations
(B+ implanted silicon)
5keV
lOkeV
20keV
30keV
40keV
5xl0l3cm'2
2.56el5
4.5 lel5
7.35el5
9.60el5
1.16e16
lxl0l4Cm-2
5.12el5
9.02el5
1.47el6
1.92el6
2.3 lel6
2xl0^cm'2
1.02el6
1.80el6
2.94el6
3.84el6
4.62el6
5xl0l4crrr2
2.56el6
4.51el6
7.35el6
9.60el6
1.16e17
lxl0^cm_2
5.12el6
9.02el6
1.47el7
1.92el7
2.31el7

Projected Range R ()
47
Implant Energy (keV)
Figure 2.4. Projected range Rp of 5keV to 40keV B+ implanted Si,
determined from XTEM micrographs and SIMS profiles.

CHAPTER 3
EFFECT OF ANNEALING ON DEFECT EVOLUTION AND TRANSIENT
ENHANCED DIFFUSION
3.1 Overview
Due to the limitation that TED imposes on the minimum
device dimensions, it has been the subject of considerable
studies for many years, yet its mechanism is still under
debate. TED has been attributed to the excess point defects
introduced by the implantation process. The magnitude of TED
is related to the implantation conditions, the annealing
conditions and the surface and bulk recombination effects.
Recently, Stolk et al.113 have suggested that TED occurs by
the emission of silicon self-interstitials from rod-like
[311} defects during annealing. The activation energy
(3.60.1 eV) for the {311} defect dissolution process relates
closely with that of TED (3.5 eV) Later, Zhang et al.114
reported that they observed TED in B+ implanted samples
without {311} defects and they attributed the interstitial
source of TED to submicroscopic clusters. More recently,
Cowern et al.123 studied the role of carbon and boron clusters
on the diffusion process. The number of self-interstitials
trapped per clustered impurity atom is about 1.15 for carbon
and about 1 for boron. This provided further evidence to the
48

49
possibility that there was more than one storage mechanism
for excess interstitials that contribute to TED.
The purpose of this section is to discuss the effect of
annealing on defect behavior and TED in order to investigate
the correlation between the two.
3.2 Effect of Annealing on Defect Behavior
3.2.1 Experimental Procedure
In order to study the dissolution kinetics of {311}
defects from B+ implantation without the complication of
stable loop formation, we picked one sample (20 keV
2x10-*-^ cm2) from the implant matrix discussed in Chapter 2
(refer to section 2.2). Subsequent furnace anneals were
performed in a nitrogen ambient using fast pushes and pulls.
The annealing temperatures and times were chosen as 650C for
2 hr, 6 hr, 14 hr, 24 hr, 36 hr, 48 hr; 700C for 30 min, 1
hr, 2 hr, 4 hr, 8 hr; 750C for 5 min, 20 min, 30 min, 1 hr,
1.5 hr and 800C for 3 min, 5 min, 8 min, 10 min, 13 min.
PTEM micrographs were taken under weak beam dark field
imaging condition using a g220 reflection. The density and
length of {311} rod-like defects were measured directly from
the TEM micrographs. The areal interstitial density bound by
{311} defects after each anneal was calculated by multiplying
the defect length by the density of interstitials per unit
length. The time constant for the interstitial decay process
at each temperature was plotted in an Arrhenius graph as a

50
function of the reciprocal of temperature to obtain the
activation energy for the {311} defect dissolution process.
3.2.2 Defect Evolution: Microstructure. Interstitial Density
and the Activation Energy of Interstitial Decay
Figure 3.1 (a)~(d) shows the PTEM micrographs of the
sample after anneals at 650C, 700C, 750C and 800C for
various times. The formation and coarsening/dissolution
processes of {311} defects are faster at higher temperatures.
After 2 hr at 650C, these defects are not fully formed yet.
The defects in their initial stages exhibit a "black dot"
contrast. After 14 hr, {311} defects start to form.
However, Ostwald ripening along with the dissolution process
is so slow at this low temperature that the number of the
defects is relatively large and their size is relatively
small even after 48 hr of annealing. If the annealing
temperature is increased to 750C, {311} defects are observed
after only 5 min and the dissolution process becomes more
rapid. At 800C, {311} defects almost dissolve completely
after 13 min of annealing.
We have measured the net interstitials trapped in {311}
defects from the PTEM micrographs after each anneal. Figure
3.2 shows the {311} defect dissolution process at each
temperature. As we can see, the dissolution process shows an
exponential dependence on annealing time. The dots in Fig.
3.2are the actual data points and the lines are exponential
fitting curves of the form

51
Si,(t) = 5//(0)exp(-) (3.1)
T
where X is defined as the characteristic time constant of the
decay. Although the decay process is strongly dependent on
annealing temperature, the initial density of interstitials
at all four temperatures is very close to the same value of
3~4xl0^3 cm-2, which is only 15-20% of the implant dose of
2xl014 cm-2. According to the "plus one" model,124 the
initial density of interstitials should closely mimic the
implant dose, because each implanted ion is supposed to give
rise to one excess interstitial during annealing. This model
assumes that the implant ion induces a series of Frenkel pair
formation and then comes to rest in the lattice at an
interstitial site. Subsequent annealing results in
annihilation of all vacancy-interstitial pairs with just the
one interstitial remaining. This theory is proved to be
reasonable by the study of Si implanted silicon of
Listebarger et al.105 and valid to a remarkable degree by the
study of Eaglesham et al.125. Listebarger et al. used
dislocation loops as point defect detectors to measure the
flux of interstitials introduced by a 30keV B+ implant to
doses of 7x10-*-3 cm-2, lxlO-*-4 cm-2 and 2xl0^-4 cm-2. The loops
layer was formed by 100 keV Ge+ implant to a dose of
lxlO-*-^ cm-2 followed by an anneal at 550C for 16 hr to
regrow the amorphous surface and another anneal at 800C for
30 min to coarsen the loops. They reported that the net flux
of interstitials measured using the loop detectors is very

52
similar to the dose for each B+ implant. Eaglesham et al.
measured the density of interstitials trapped in {311}
defects in 40 keV Si implanted silicon to a dose of
5xlOi;^ cm2. The interstitial density versus dose plot shows
a slope of 1.5, indicating slightly more than one
interstitial per implanted ion. However, in our B+ implanted
silicon study, the observed interstitials only account for
approximately one tenth of the dose. We therefore believe
that there is another type of defects, i.e., boron
interstitial complexes, which trap the missing interstitials.
These complexes are sub-microscopic and are not resolvable in
the TEM. There appear to be more than one form of boron
interstitial complexes. From our low energy low dose
implants (as will be discussed later), we observed as did
Zhang et al.114 that there is a significant fraction of the
implant profile peak that is immobile. This could be
explained by the formation of B-B-interstitial clusters. A
second portion of the profile is mobile and may simply be B-
interstitial pairs as suggested by Cowern.123 Both of these
non-visible defects could help account for the "missing"
interstitials in the {311} defects.
An Arrhenius plot of the characteristic rate constant
k=l/T of {311} defect dissolution process as a function of 1/T
is shown in Fig. 3.3. This temperature dependence shows an
activation energy of 3.8 eV. Along with our results, Fig.
3.3 also presents the temperature dependence of {311} defect
dissolution from the study of Stolk et al.113
of Si+ implanted

53
silicon. It is interesting to notice that our activation
energy is in close agreement with theirs of 3.60.1 eV, yet
there is a y-axis shift of the two curves, i.e., the {311}
defects in our sample are more stable. The similar
activation energies implies the process of dissociation of an
interstitial from {311} defects is the same, independent of
the implant species. The reasons for the change in stability
could be: (1) The dose for our implant (2x1014 cm-2) is much
higher than what they used (5xl0-*-3 cm-2) and thus there are
more interstitials in our sample to stabilize {311} defects.
(2) The dissolution of boron interstitial clusters in our
sample during the anneal induces a higher concentration of
interstitials in the background environment and again delays
the dissolution of {311} defects.
3,2.2 Defect Evolution: Density, Size and Distribution
Quantitative results that one can obtain from the PTEM
micrographs include the {311} defect density, the average
defect length and the defect distribution after annealing.
Figure 3.4 shows the {311} defect density as a function
of annealing time for each temperature. The defect density
decreases dramatically with increasing annealing time.
Although the rate is just slightly higher for higher
temperature, it takes much longer for a lower temperature to
reach a particular density. For example, the density drops
to 109 cm-2 after about 10 min at 800C, but it takes more
than 10 hr at 700C. This rapid decrease in the number of

54
defects explains the decay of interstitials trapped in {311}s
even when the average defect size increases.
The average {311} defect size is shown in Fig. 3.5 as a
function of annealing time for each temperature. The width
change of these defects have been found to be very small
during the anneal by high resolution TEM studies conducted by
Eaglesham et al.125, so that the size change is represented by
the change of the length. As can be seen from Fig. 3.5, the
average defect size increases initially with annealing time
and then starts to decreases at some point during the
dissolution process. This breaking point can be seen for
anneals at 700C and 750C but not for anneals at 650C or
800C. We believe this final drop of defect size indicates a
pure dissolution stage which should be observed as well at
650C and 800C if the anneals are long enough. Again, to
reach a particular defect size, it take longer for lower
temperatures. For example, for 800C anneal, the defect size
reaches 400 after about 3min, while at 700C, it takes
about 1.5 hr.
To further illustrate the {311} defect evolution
process, the defect distributions after each anneal at each
temperature are shown in Fig. 3.6 (a)~(d). The distribution
moves to a larger size with increasing time at each
temperature and the defect density N(l) at every size 1
shifts to a lower value. This movement is faster for a
higher temperature. During the anneal, the interstitial
concentration in the vicinity of large defects is lower than

55
that in the vicinity of small defects. As a result,
interstitials are emitted from the small defects and move
toward the large ones. The large defects then absorb these
interstitials and grow at the expense of the shrinkage of the
small defects. This is the so-called Ostwald ripening
process. Meanwhile, since the interstitial concentration in
the area surrounding a defect is higher than that in the
background environment, interstitials tend to leave the
defect and move into the bulk. As a result, defect shrink
and system approaches equilibrium condition. This is the
defect dissolution process. During the Ostwald ripening
process, defect density is expected to decrease, defect size
is expected to increase and the total number of interstitials
trapped in the defects is expected to remain as a constant.
On the other hand, during the dissolution process, all three
quantities are expected to decrease. In our experiment, we
observed a decreasing interstitial concentration, an
increasing defect size and a decreasing defect density with
increasing annealing time. This indicates an Ostwald
ripening process along with dissolution process. The
distribution plots for longer time anneals (e.g., 800C 13
min) show considerable scatter rather than a regular
distribution because the total number of defects is small and
the counting results are not as representative as those of
the short time anneals.
The driving force for the Ostwald ripening process is
the self-energy difference between the defects with different

56
sizes. Self-energy of the {311} defect was calculated by
Parisini and Bourret76 by approximating the rod-like shape
with a dipole of edge dislocations of the same sense and
opposite Burgers vector. It is given by
W = [ln() + cos20]L (3.2)
2n{\-v) 2 p
where (l is the shear modulus, b the Burgers vector, v the
Possions ratio, r| the distance between the two edge
dislocations (i.e., the width of the {311} defect), p=b/8 the
cut-off distance, 9 the angle between the Burgers vector and
the normal to the plane containing the {311} defect and L the
length of the {311} defect. To obtain the energy per
interstitial, one can divide the previous expression by the
number of interstitials N trapped in a {311} defect with a
length of L. N is given by
N = p,Lri (3.3)
where p¡ is the atomic density on the {311} plane. The energy
of one interstitial is thus given by
W, = ^ [In A + cos2 0]
2/r(l-v) 2 p p,r¡
(3.4)
As we can see from eq. (3.4), Wx is independent of the length L
but is dependent on the width T). Since ln(T|/p) decays slower
than T|, Wi decreases when r| increases. We thus speculate that
the width of a shorter {311} defect is smaller than that of a
longer one. As a result, the self-energy of an interstitial
trapped in a shorter {311} defect is higher than that in a
longer one. Interstitials would thus flow from shorter rods
to longer ones. This explains the Ostwald ripening process.

57
In fact we have observed the {311} defect width change
under the TEM. PTEM pictures were taken on the <311> zone
with g=220. The following trend was found: the width is
about 20 A when the length L is 200-400 A. It increases to
about 40 A when L is around 1000 A, to about 80 A when the L
is around 1200-1400 A, then to about 100 A for L=1600~1800 A
and about 120 A when L-3000 A. These measurements were
performed on samples annealed at 750C for 5 min, 15 min and
1 hr. The error range might be large because of the narrow
width of these defects. However, the width change with
length can be used to account for the interstitial flow
during the coarsening process. It should be mentioned that
in this study this width change is ignored when the areal
density of interstitials trapped in {311} defects was
measured from the length of {311} defects and the
interstitial density per unit length. In fact it is almost
impossible to measure both the width and the length of every
defect to obtain a more accurate result.
In summary, in this section of Chapter 3 we have
discussed the {311} defect evolution process during anneals
at 650C to 800C for various times. The areal density of
interstitials bound by {311} defects decays exponentially
with annealing time with a characteristic time constant which
is larger at lower temperatures. This dependence has an
activation energy of 3.8 eV. The number of interstitial
trapped in {311} defects is only 15-20% of the implant dose,
implying the presence of boron interstitial complexes,

58
including both immobile clusters and mobile boron
interstitial pairs. The defect density decreases with
increasing annealing time and the average defect size
increases initially and then decreases with time. The
distribution of {311} defects shows a shift of the average
defect size to a larger value and a decrease of density at
each size. The defect evolution shows the characteristics of
Ostwald ripening process accompanied by a defect dissolution
process.
3.3 Effect of Annealing on Transient Enhanced Diffusion
3.3.1 Experimental Procedure
The same wafer (20 keV 2x10^-^ cm~2 b+ implant) which we
used to study the effect of annealing on defect evolution was
also used for diffusion study. Subsequent anneals were again
performed in a nitrogen ambient at 650C, 700C, 750C and
800C for various times. SIMS analysis was performed on the
CAMECA IMS4f system. B+ ions were monitored under C>2 +
bombardment at an impact angle of about 42. Secondary ions
were collected from the center 12% of a 150 |im x 150 |im
rastered area to avoid edge effects. Atomic concentrations
were calculated using a relative sensitivity factor (RSF) for
boron determined from an implanted standard. Stylus
profilometry was used to calibrate the depth scale for the
profiles. The dopant profiles were then imported into

59
Florida Object Oriented Process Simulator (FLOOPS) and the
diffusivity enhancement was extracted for each anneal.
3.3.2 Dopant Diffusion during Annealing
3.3.2.1 Static Peak at High Concentration
As an example of the dopant diffusion behavior during
annealing, Fig.3.7 (a)~(d) shows the boron profiles obtained
through SIMS after anneals at 650C for 2 hr, 14 hr, 24 hr,
at 700C for 30 min, 2 hr and 4 hr, at 750C for 20 min, 1
hr, 2 hr and at 800C for 3 min, 3 0 min and 1 hr. We can see
that a critical concentration exists under every annealing
condition, separating the low concentration tail region where
transient diffusion occurs, from the high concentration peak
region where the profile is relatively immobile. The
breaking point between the mobile and immobile region (Cenh)
moves to higher concentration when annealing temperature
increases. It is about 2.5x10-*-^ cm-^ at 700C and moves up
to about 7xlC)18 cm~3 at 800C. Figure 3.8 shows Cenh as a
function of inverse temperature for our study and some of the
previous studies. Also plotted in Fig. 3.8 are the solid
solubility for boron in silicon Cs and the intrinsic carrier
concentration ni.
There are several groups which have reported the peak
clustering behavior of an implant profile. Michel et al.126
studied 60 keV 2x10^4 cirT^ b+ implanted silicon. Over an
annealing temperature range of 800C~950C, their
measurements show that both the maximum concentration of the

60
displaced boron and the electrical activity are related to
the intrinsic carrier concentration. Later, Fair127 has
extended the correlation between Cenh and ni to the
temperature range of 600C~800C. It was observed that when
T>850C, Cenh approaches B solid solubility Cs. A model was
proposed by Fair, suggesting that the existence of Cenh is
due to the Fermi level and the charge state of the defect
affecting diffusion. For boron diffusion the location of the
donor level of self-interstitials I+ in the energy gap causes
Cenh to equal ni for T<800C. Under this condition, a
significant fraction of the excess interstitials are in the
neutral charge state Ix. For doping under Cenh~ni> diffusion
is dominated by Ix. Since Dix/Di+ is 10-100, more enhanced
diffusion occurs at C850C, the Ei+ level is
sufficiently above Ep and the majority of injected
interstitial exist in I+ state. Then the thermally assisted,
concentration dependent diffusion via I+ becomes dominant and
Cenh is limited by B solid solubility Cs The study of
Cowern et al.128 on 25 keV 2xl014 cm-2 B+ implants agreed with
Fair's results. Cenh lies within a factor of 2 of ni in a
temperature range of 550C~900C. A relative insensitivity
of Cenh to the implant energy and dose was suggested by all
three groups. As shown in Fig. 3.8, our data of 20 keV
2xl014 cm 2 B+ implants also agree with those of the above
groups to a reasonable degree within 650C~800C. However,
the data of Zhang et al. for 4 keV lxlO14 cm-2 B+ implants114
do not match the intrinsic carrier concentration but are

61
rather higher than that, especially at higher temperatures
(T>700C) This mismatch can not be explained by Fair's
model. We therefore speculate that the correlation between
Cenh and ni might just be a coincidence. A better
explanation is that clustering effect induces an immobile
fraction of the profile. As for the kinetics of peak
clustering, Cowern et al.128 suggested the following
reactions:
Bs + I <-> Bx (3.5)
mBs + ni <-> IDC (3.6)
where the intermediate defect configurations (IDCs), which
was originally proposed by Tan,56 might include boron pairs or
clusters, rodlike {311} defects, certain stacking faults,
etc. Silicon interstitials (I) pair or kick out
substitutional boron atoms (Bs) to form migrating
interstitial boron atoms (B!) During the transient phase,
the concentration of Bj is high enough to allow homogeneous
nucleation of boron containing defects. The defect formation
is most favorable near the peak region where the B
concentration is highest. Stolk et al.113 proposed another
clustering reaction given by
Bg + mBj <-> Be + ni (3.7)
where Bc refers to immobile clusters containing m+1 boron
atoms, and n is the number of interstitials injected upon
clustering to allow stress relief (n sustains the enhanced diffusion of BSf eqs.(3.6) and (3.7)
lead to cluster formation and reduce the amount of Bs

62
available for diffusion at high B and I concentrations.
Comparing eq.(3.6) and (3.7) will raise the question of
whether the immobile B-I complexes contain Bs-B];-! clusters or
Bs-Bs-I clusters. Diaz de la Rubia et al. (private
communication) have shown through ab initio calculation that
the Bj-Bs pair has a positive binding energy of about 1.8 eV.
This Bj-Bs pair also traps silicon interstitial. It is
immobile and electrically inactive. The Bs-Bs pair, on the
other hand, is found to be energetically unfavorable with a
binding energy of -1.7 eV and electrically active. According
to this calculation, the immobile, electrically inactive
implant profile peak might be consisted of the Bj-Bs-I
clusters.
It should be mentioned that rod-like {311} defects also
locate in a depth corresponding to the profile peak region,
so do the sub-amorphization dislocation loops. However, the
{311} defects113 dissolve on a time scale comparable to the
transient diffusion (e.g. <30 min at 800C) while the static
B-I clusters dissolve much slower (e.g. > 4 hr at 800C128).
{311} defects and unstable loops contribute to transient
enhanced diffusion before it saturates. When the clusters
and stable loops finally break up, there will be a second
burst of enhanced diffusion. The slight increase of Cenh
with increasing temperature is due to a decreasing
interstitial supersaturation level which results in less
clustering or even breaks up some existing clusters. There
is no experimental evidence on how many interstitials are

63
trapped in one Bs-Bj cluster. In our 20 keV 2x10-*-^ cm 2 b+
implant, after an anneal at 750C for 2 hr, {311} defects
almost dissolved completely. The density of activated boron
atoms is estimated to be < 7x10^3 cm-2 by integrating the
non-clustered region of the implant profile. Assuming a
valid "plus 1" model and assuming that one activated B atom
pairs with one interstitial, the number of interstitials
trapped in the B-I clusters must exceed 1.3x10-*-^ cm-2. Thus
we speculate that one Bs-Bj cluster might trap more than two
interstitials.
The annealing time scale we used in this study is much
shorter than that is needed to break up the clusters and
stable loops, so our focus is on TED at low concentration
after relatively short anneals.
3 .3.2.2 Broadened Tail at Low Concentration
From Fig. 3.7 we notice that significant diffusion
occurred in the tail region (C four temperatures. This is caused by the pairing of excess
interstitials induced by ion implantation with dopant atoms.
The boron profile displacement at this region changes with
the temperature. A 4 hr anneal at 700C resulted in a
displacement of about 2200 at a concentration of lxlO1^
atoms/cm^ relative to the as-implanted profile, whereas a lhr
anneal at 800C resulted in a displacement of about 1750
for the same concentration. As will be shown later, TED is
complete after 1 hr at 800C but is still continuing after

64
2 hr at 700C. Lower temperature annealing causes more
diffusion before the saturation point is reached.
The SIMS profiles of boron in the as-implanted and
annealed samples were imported into FLOOPS and the
enhancement of diffusivity /Db* after each anneal was
obtained after the simulations. Here is the time-
averaged boron diffusivity and Db* is the intrinsic
equilibrium value. Db* has the form of
Db*=0.757exp(-3.46eV/kT) (3.8)
as a default value in FLOOPS, which was originally taken from
Fair's review article129. This value is about
1.03x10-*-9 cm^/sec, 9.64xl0--*-^ cm^/sec, 7.22x10^-^ cm^/sec
and 4.48xl0"l7 cm^/sec at 650C, 700C, 750C and 800C,
respectively. Since the dopant profiles are not perfectly
Gaussian, instead of choosing the whole profile, we chose a
portion of the tail region (approximately 1600 A-2600 ) as
the target profile during simulation. Figure 3.9 shows these
simulation results of /Db* as a function of annealing
time for each temperature. As we can see, the time-averaged
diffusivity enhancement decreases with increasing annealing
time. Since /Db* is proportional to the interstitial
supersaturation level, /Ci*, in the case of B diffusion,
the decrease of /Db* represents a process during which
the interstitial concentration returns to its equilibrium
value. Here is the time-averaged interstitial
concentration in the bulk, and Ci* is the equilibrium value.
Ci* is smaller and /Ci* is higher at lower temperatures

65
so that /Db* is higher. It requires longer time for the
system to reach equilibrium at a lower temperature. However,
the time point at which the equilibrium is reached cannot be
read directly from the /Db* graph mainly because they are
the time-averaged values. This time will be determined from
the diffusion length vs. time plot as will be presented in
the following section, from which we will derive the
activation energy for the diffusion process.
3.3.3 The Activation Eneren/ for the Diffusion Saturation
Process
The diffusion length can be calculated as \(t) from
the diffusivity enhancement, the intrinsic diffusivity Db*
and the annealing time at each temperature. Figure 3.10
summarizes the results. The dots in Fig. 3.10 are the actual
data points and the lines are the fitting lines in the form
of aV ( 1-exp (-t/T)) where a and T are constants and t is the
annealing time. This is obtained by assuming the
instantaneous value of Db is proportional to exp(-t/T), and
the diffusion length L of boron is the square root of the
time integration of Db, i.e.,
I
DB{t)~ex (3.9)
L = Jt=i]j'DB(t,)dtl
r-r (3-10)
=^[0 'i, -Vd-i >
The diffusion length increases initially with time and then
plateaus out. The time till the "knee" region, i.e., just
before it plateaus out is 2T. If we choose 2T as the time

66
constant for the TED saturation process, from Fig. 3.10 we
can see that this constant increases dramatically as
annealing temperature decreases. It is about 4.5 hr at 650C
and decreases more than one order of magnitude to about 15min
at 800C.
The temperature dependence of the time constant for the
TED saturation process is shown in Fig. 3.11. Such a
dependence exhibits an activation energy of about 1.6eV.
This value is smaller than the 3.1-3.7eV reported by
Packan, 130 3.5 eV reported by Stolk et al.113 yet it is greater
than the < 1 eV we calculated from the data of Zhang et al.114
as shown together in Fig. 3.10. The difference between these
experiments is that both Packan and Stolk et al. studied
silicon self-implanted silicon. In these samples the
interstitial source for TED is believed to be only {311}
defects. Zhang et al. studied low energy (4 keV) boron
implanted silicon where the energy and dose combination
resulted in no extended defects observed by TEM. The source
of interstitials driving TED in this case was claimed to be
sub-microscopic clusters, which may be as simple as just
mobile B-I pairs that diffuse until the pair breaks up. In
our case, the samples were implanted with boron, but at a
higher energy and dose so that not only {311} defects are
present, B-I pairs might also exist. The activation energy
of the TED saturation process reflects a complex process,
including the release of interstitials from B-I pairs and
{311} defects, the migration of interstitials through the

67
bulk, the surface recombination effect, impurity trapping,
etc. Variation in the activation energy is thus not
unexpected in these different systems. However, we still
believe that in our case, not only {311} defects play a role
in terms of releasing interstitials during annealing and
thereby driving TED, boron interstitial pairs also play an
important role.
In summary, 20 keV 2x10-*-^ cm-^ boron implanted silicon
was studied for the purpose of identifying the possible
sources of interstitials driving TED. The static peak region
was presumably consisted of immobile B^Bs-I clusters which
breaks up in times much longer than normal TED. Boron TED in
the tail region was measured through SIMS analysis and FLOOPS
simulation. The duration of TED increases dramatically when
the annealing temperature decreases. The activation energy
for the TED saturation process is about 1.6 eV, higher than
that in the sample presumably with only B-I pairs
contributing to TED and lower than that in samples with only
{311} defects. The interstitials driving TED in our sample
are believed to come from both mobile boron interstitial
pairs and {311} defects. The dissolution of {311} defects
and B-I pairs may interact with each other and result in a
single activation energy lying in between.

0-2|im
(a)
Figure 3.1. Weak beam dark field (g220) PTEM micrographs of {311} defects in 20keV
2xl014cirT2 B+ implanted Si after anneals at (a) 650C, (b) 700C, (c) 750C and (d) 800C
for various times.

0-2um
Figure 3.1. (Continued)

QJum
-j
o
Figure 3.1.
(Continued)

(d)
Figure 3.1. (Continued)

Intersititals Bound by {311} Defects (cm'2)
72
Annealing Time (sec)
Figure 3.2. The exponential decay of interstitials trapped
in {311} defects in 20keV B+ implanted Si after anneals at
650C to 800C for various times.

73
Temperature (C)
Figure 3.3. Rate of {311} dissolution process and the
activation energy.

{311} Defect Density (cm')
74
Annealing Time (sec)
Figure 3.4. {311} defect density as a function of annealing
time in 20keV B+ implanted Si for different annealing
conditions.

Average {311} Defect Size ()
75
Annealing Time (sec)
Figure 3.5. Average {311}
annealing time in 20keV B+
annealing conditions.
defect size as a function of
implanted Si for different

Defect Defect Defect Defect
Density (cm ) Density (cm'2) Density (cm'2) Density (cm2)
76
650C6hr
1 1 1 1 1 1
60 260 460 660 860 1060 1260 1460
650C14hr
1 1 1
i
60 260 460 660 860 1060 1260 1460
650C36hr
10 60 260 460 660 860 1060 1260 1460
12
10
1011 r
650C48hr
10
10
10
10e
9
Mi. M
j-i- i 111
60 260 460 660 860 1060 1260 1460
Average {311} Size ()
(a)
Figure 3.6. {311} defect size distribution during annealing
at (a) 650C, (b) 700C, (c) 750C and (d) 800C.

Defect Defect Defect Defect
Density (cm'2) Density (cm'2) Density (cm'2) Density (cm'2)
77
10
11
10
10
109
108
107
700C30min
1 1 i 1
i i i
60 260 460 660 860 1060 1260 1460
Min
1.,
700C60min
60 260 46(
) 660
860 1060 1260 1460
tail
All
LJ
700C4hr
60 260 46(
3 660
860 1
060 1260 1460
r!!!
700C8hr
Average {311} Size ()
(b)
Figure 3.6. (Continued)

Defect Defect Defect Defect
Density (cm'2) Density (cm'2) Density (cm'2) Density (cm'2)
78
750C5min
i L
60 260 460 660 860 1060 1260 1460
260 460 660 860 1060'126b'1460'
1011
1010
109
108
107
L
750C30min
i
LLJ
111
60 260 460 660 860 1060 1260 1460
260 460 660 860 1060 1260 1460
Average {311} Size ()
(c)
Figure 3.6. (Continued)

Defect Defect Defect Defect
Density (cm'2) Density (cm'2) Density (cm'2) Density (cm'2)
79
1060 1260 1460
800C5min
i i i i.i.i.i
Imlii
260 460 660 860 1060 1260 1460
800C1 Omin
iM MB
260 460 660 860 1060 1260 1460
11
10
10
10£
1081
1071
10
800C13min
j. 1 ii ill ii Ii.llI i i
60 260 460 660 860 1060 1260 1460
Average {311} Size ()
Figure 3.6.
(d)
(Continued)

80
O
i
E
c
o
(
1
c
O
c
o
O
c
o
o
CQ
0 1000 2000 3000 4000 5000 6000 7000 8000
Depth ()
(a)
c
CD
O
c
o
O
c
o
o
CQ
10
20
1 1 1 I 1 1 1 1 I
E
c
o
co
10
19
10
18
£ 1017 L-
106 1
2 10'5 1
lili l'l I I I | I I I I 1 T' I II.
as-implanted
30 min
----2hr
*4hr
1 Q14 I I I I I I I
0 1000 2000 3000 4000 5000 6000 7000 8000
Depth ()
(b)
Figure 3.7. Dopant SIMS profiles of 20keV 2xlC)14cin-2 B+
implanted Si after anneals at (a) 650C, (b) 700C, (c) 750C
and (d) 800C for various times.

Boron Concentration (cm'3) Boron Concentration (cm'3)
81
Depth ()
(c)
0 1000 2000 3000 4000 5000 6000 7000 8000
Depth ()
(d)
Figure 3.7. (Continued)

82
10000/T (K)
Figure 3.8. The breaking point between the mobile profile
tail and the immobile profile peak Cenh as a function of
inverse temperature.

Boron Diffusivity Enhancement /D
83
Annealing Time (sec)
Figure 3.9. Time-averaged Boron diffusivity enhancement
/Db* as a function of annealing time for different
annealing temperatures.

Diffusion Length (Dt)0'5 ()
84
Annealing Time (sec)
Figure 3.10. Boron diffusion length vs. annealing time for
different annealing temperatures.

Time Constant of TED Saturation Process (sec)
85
10000/T(K)
Figure 3.11. Time constant and activation energy of TED
saturation process, comparison with previous studies.

CHAPTER 4
THE INFLUENCE OF IMPLANT ENERGY ON DEFECT BEHAVIOR AND
TRANSIENT ENHANCED DIFFUSION
4.1 Overview
It is believed that transient enhanced diffusion is a
direct result of the damage created by the implantation
process. Characteristics of the implanted ions (such as the
mass), the substrate conditions (such as crystalline versus
amorphous), and variations in implantation parameters (such
as energy, dose, dose rate and temperature) would all cause
variations in the damage and therefore affect dopant
diffusion. This section discusses the effect of implant
energy on defect behavior and TED.
It has been claimed that the amount of damage created by
the implantation process is strongly dependent on the dose
and the mass of the implant species but only weakly dependent
on the implant energy.42 Although higher energy does mean
more kinetic energy per implanted ion, the resulting implant
profile is more spread out and thus the "energy density"
deposited per volume is lessened by this dilution effect.
The effect of increasing energy is thus weakened and the
number of interstitial-vacancy pairs would not be increased
much provided the dose is the same. However, increasing
energy can indeed change the location of the damage and the
86

87
depth of the interstitial and vacancy profiles and the
separation between these two profiles. These changes can
affect the defect behavior and TED, as will be discussed
later.
4.2 Implant Energy Effect on Defect Behavior
4.2.1 Experimental Procedure
A single row of the implant matrix discussed in Chapter
2 was used to study the implant energy on {311} defect
evolution. The implant dose was kept at 2x10-*-^ cm-2 and the
energy was varied from 5 keV to 40 keV. Subsequent furnace
anneals were conducted in a nitrogen ambient for times
between 5 min and 1 hr at 750C. The {311} defect density,
average defect size and, areal density of interstitials
trapped by {311} defects and the defect distribution during
annealing were measured from the PTEM micrographs.
4.2.2 Defect Microstructure and Dissolution Process
PTEM micrographs of the defects after implantation at
10 keV to 40 keV and anneals at 750C from 5 min to 1 hr are
shown in Fig. 4.1(a)~(d). No defects were observed for the
5 keV implant and therefore the pictures for those samples
are not included. {311} defects are the only type of
extended defects observed under the TEM in the 10 keV and 20
keV implants. A majority of {311} defects together with just
a few dislocation loops are observed in the 30 keV and 40 keV

88
implants. For each of those implants showing {311} defects,
the number of the defects decreases with increasing annealing
time. This dissolution process is more rapid for a lower
implant energy.
The areal density of interstitials bound by {311}
defects during the annealing process was measured from the
PTEM micrographs and the results are presented in Fig. 4.2.
The dots in the plot are the actual data points and the lines
are the exponential fitting curves for the data with the form
of eq. (3.1). The emission of interstitials from {311}
defects clearly shows an exponential time-dependence, which
is stronger for lower implant energies. After a 30 min
anneal at 750C, {311} defects in the 10 keV implant almost
dissolved completely. The density of the remaining trapped
interstitials is about 1.8xlO cm-^. For the 40 keV
implant, after an anneal of 1 hr, there is still an areal
interstitial density of about 1.2xl0-*-2 cm~2 contained in
{311} defects. The maximum density of trapped interstitials
observed was about 2xl0l3 cm2 for the 10 keV implant and
about 3.5xl0l3 cm-^ for the 40 keV implant after an anneal
for 5 min and these values are much lesser than the implant
dose of 2x10-*-^ cm2. This implies the existence of immobile
boron interstitial clusters and/or mobile boron interstitial
pairs which might have trapped the missing interstitials
assuming that the "plus one" model holds. There is, however,
another possible configuration that could consume
interstitials, i.e.,
silicon self-interstitial complex--two

89
interstitials split along <110> direction occupying one
lattice site. According the ab initio calculations conducted
by Diaz de la Rubia et al. (private communication) this
configuration is energetically unfavorable with a formation
energy of about 3.7 eV. Interstitials are more likely to
pair with boron atoms because the binding energy of a Bs-I
pair is about 1.1 eV.
The lack of {311} defects in the 5 keV sample and the
lower density of interstitials trapped by {311} defects in a
lower energy implant could be attributed to at least the
following four factors: (1) Higher energy might introduce
more as-implanted damage which may further induce more
trapped interstitials in extended defects (if formed) after
annealing. (2) Higher energy causes larger separation
between the as-implant vacancy profile (closer to the
surface) and the interstitial profile (further away from the
surface) because of the larger forward momentum carried by
the implanted ions. The result of this might be that the I-V
recombination rate is lower and the number of interstitials
available to form extended defects is higher. (3) According
to a series of proposed defect reactions113 which describe the
generation and clustering of mobile boron atoms given by
eq. (3.5) and (3.7),
Bs+I <-> Bj (3.5)
Bs+mBj <-> Bc+nl (3.7)
at a lower implant energy, the concentrations of Bs or Bz and
silicon interstitials in the implant peak region are higher,

90
which favors the formation of boron clusters which bind
excess interstitials. This will reduce the total number of
interstitials available for forming {311} defects. (4) The
surface recombination effect is more efficient for a lower
energy when the damage region is closer to the surface. More
interstitials are drawn to the surface and less are available
for defect formation during annealing.
If the time constant of the {311} defect dissolution
process is chosen as the time required for the number of
interstitials to decrease to 1/e of its maximum value, then
it would be about 5 min for the 10 keV implant and 16 min for
the 40 keV implant. This time constant is plotted in Fig.
4.3 as a function of energy. It is interesting to note that,
if we derive from Fig. 4.3 the time constant of the 5 keV
implant, it would be about 2.6 min. If {311} defects existed
in the 5 keV implant from the beginning, we should be able to
see them after a 5 min anneal, even though the defect density
would be low, yet we do not see any defects. This could be
attributed to one or a combination of the factors discussed
above.
4^2^3..Defect Density, Size and Distribution during Annealing
The {311} defect density for different implants as a
function of annealing time measured from PTEM images is shown
in Fig. 4.4. If the annealing time is short (e.g., less than
10 min), the lower energy implant shows a slightly higher
density of {311} defects. This is probably due to earlier

91
formation and coarsening of {311} defects in the higher
energy implants. However, the defects dissolve faster for a
lower energy. As annealing time increases further, defect
density is higher for a higher energy implant. The
difference in defect density becomes more significant when
the annealing time increases. After a 30 min anneal, the
10 keV implant shows a defect density of only 3.2x10 cm-2,
while the 40 keV implant shows a density of 2.7x10 cm2.
The average {311} defect size as a function of annealing
time for different implant energies is shown in Fig. 4.5.
The average defect size increases with annealing time
initially, and then decreases when the defects enter a purely
dissolution regime. This final dissolution regime is only
observed for the lOkeV implant for the current annealing
times. As the implant energy increases, the average defect
size increases. After an anneal of 15 min, the 20 keV
implant shows an average defect size of about 518 while
for the 40 keV implant, it is about 952 .
Figure 4.6 (a)~(d) show the size distribution of {311}
defects after anneals at 750C for various times for 10 keV
to 40 keV implants. Higher energy implants have a broader
distribution than lower energy implants. The figure also
shows a larger average defect size and a slower dissolution
rate for a higher energy. For each energy, the average
defect size increases while the defect density at each size
decreases during annealing. This behavior together with the
fact that the total number of interstitials trapped in {311}

92
defect decreases with increasing time indicates that the
Ostwald ripening process occurs along with the defect
dissolution process. Large {311} defects grow by absorbing
the interstitials released from small defects as a result of
higher local interstitial concentration around small defects.
Meanwhile, large defects themselves release interstitials
into the bulk. This resulted in the final defect
distribution we have observed.
As a summary, the {311} defect microstructure and
annealing behavior as a function of implant energy (5 keV ~
40 keV) have been studied using TEM. No defects are observed
in the 5 keV implant. There are more interstitials bound by
{311} defects in higher energy implants although the maximum
density observed (about 2~3.5xl0-*-3 cm-2 after an anneal of
5 min at 750C) are all far less than the implant dose
(2x10-*-^ cm2). The defect density is higher for a lower
implant energy after a short anneal time (10 min) and then
becomes lower as time increases. The average defect size is
larger for a higher implant energy. The size distribution of
defects shows a co-occurrence of the Ostwald ripening process
and the dissolution process. The lack of {311} defects in
the 5 keV implant and the decrease in density of trapped
interstitials by (311} defects with decreasing implant energy
might be the result of one or a combination of the following
factors: decreased damage product with decreasing energy,
increased I-V recombinations, increased boron interstitial
cluster formation and an increased surface recombination.

93
Each of these factors could decrease the number of
interstitials available for the formation of {311} defects.
4.3 Implant Eneren/ Effect on Transient Enhanced Diffusion
4.3.1 Experimental Procedure
In order to study the effect of interstitial release
during TED independent of the diffusing profile, doping
superlattices were used. Boron 8-doping superlattices were
grown by low temperature molecular beam epitaxy (LTMBE) on
float zone Si (100) substrate with boron-doped to a
resistivity of 1000 ilem. The custom-made MBE system had a
base pressure of 4xl0-11 Torr. The temperature during growth
was set through the choice of the heater power, which in turn
had been calibrated in terms of temperature by laser
interferometry. The samples contained six 100 wide box
shaped boron spikes with a separation of about 1000 between
each spike. The peak boron concentration of these spikes was
about 1.4x10^ cm3. This low concentration kept boron
diffusion intrinsic for the as-grown control samples during
the following anneals and avoided the complication of
extrinsic effects. Boron ions were implanted into the
superlattices at energies of 5 keV, 10 keV, 20 keV and 40 keV
to a dose of 2x10^4 cm-2 at room temperature. The samples
were then annealed at 750C for times of 3 min, 15 min and
2 hr in a nitrogen ambient. The temperature error range is
estimated to be 10C. Due to the finite rise time of the

94
sample temperature, the error range for the 3 min sample
could be 30C. The initial boron profile as well as the
profiles after implantation and annealing were measured by
SIMS with 3.0 keV C>2+ at a current of 50 nA. The raster size
was 150x150 |lm2 and the analyzed area was 60x60 Jim2. The
broadening of the B profiles was simulated using the process
simulator PROPHET. The spikes that merged into the B implant
profile were excluded from the simulation.
4.3.2 Dopant Diffusion Behavior
SIMS profiles of boron in the superlattices implanted
with 5 keV to 40 keV boron ions are shown in Fig. 4.7(a)~(d).
As can be seen, a higher energy results in a deeper implant
profile and a greater enhanced diffusion of the buried boron
spikes. The time-averaged boron diffusivities for a
particular spike, , were extracted by finding the value
of Db in the simulation that resulted in the best match
between diffused and target profiles. Using the literature
value of Db*=0.757exp(-3.46/kT) (eq. (3.8)), diffusivity
enhancement /Db* is thus obtained. For the inert ambient
annealed control samples, it is expected that no diffusion
enhancement occurs, i.e., /Db*=1 or =Db*. We have
extracted via simulation for the 2hr annealed control
sample. Taking an average over different spikes since
is depth-independent, we found that is about
1.6x1017 cm2/s, about 2 times greater than the predicted
value (7.2xl0-18 cm2/s), which is within the error range of

95
furnace temperature and SIMS analysis accuracy and the
ability to predict Db*. The diffusion distance for the
control sample is indeed too small to be extracted reliably
since that the SIMS depth has an estimated error of 5%.
Figure 4.8 (a)~(d) are plots of /Db* as a function
of depth for different implant energies after each anneal.
The program assumes that the total sheet concentration is
preserved for each spike during diffusion, i.e., the depth
integration of boron concentration under each spike remains a
constant after implantation and annealing. However, the
influx from the surface implant into the spike region
actually violates the conservation. The error bars of
/Db* to be shown do not include this effect. The dots in
Fig. 4.8 are the data points and the lines are exponential
fitting curves. We can see that the value of /Db* decays
exponentially with increasing depth, and the decay length is
smaller at a shorter annealing time (3 min) than longer times
(15 min and 2 hr), i.e., the /Db* versus depth curve is
steeper. The cross of /Db* curves for the 3 min and
15 min anneals implies that interstitials have not diffused
deep enough to cause a large diffusion enhancement after a
shorter time. The slightly positive slopes of the /Db*
curves for the 15 min and 2 hr anneals might be due to the
SIMS analysis and the simulation error. After a short time
(3 min) anneal, for each particular spike the value of
/Db* increases with increase in implant energy. For
example, for a boron spike at a depth of around 3800 A, after

96
an anneal of 3 min, the 5 keV implant shows an enhancement of
about 260, whereas the 40 keV implant shows a much greater
enhancement of about 2800. As the anneal time increases to
longer than 3 min, this difference is less distinguishable
for energies above 5 keV.
4,3.3 Correlation Between Defects and TED
The average boron diffusivity at a particular depth of
the sample can be expressed as
(4.1)
where Ci is the instantaneous value of interstitial
concentration at time t and t is the diffusion time. If the
instantaneous interstitial supersaturation level Cj/Ci* is
greater than 1, we will see the enhancement of diffusivity.
If we assume that the excess interstitials introduced by an
implant are trapped in interstitial-contained defects (in our
case, {311} defects and mobile B-I pairs) in the initial
stage of the annealing, then we can consider these defects as
an interstitial reservoir that keeps Cj/Cj* constant for a
short period of time (tto all the {311}
defects have dissolved and all B-I pairs have broken up, then
the reservoir is empty and Cj/Cj* drops abruptly to 1.
Mathematically this can be expressed as
Cj_ 1 + E...(t < t0)
c; l...(i>r0)
(4.2)
where E is the supersaturation of interstitials. The
integral of eg. (4.1) would then become

97

D'b iff'0
-J\\ + E)dt]=l + E...(t -[ f'(1 + E)dtx +\'dt] = 1 +^E...(t > t0)
t '*0 t
(4.3)
From the above equation we can see that if /Db*-1 is
plotted as a function of time in a log-log scale, it should
be a curve that is flat for t 1/t for t>to and the breaking point to is the time when the
interstitial concentration reaches equilibrium value and TED
is completed. This of course is a simplified picture since
TED sources do not turn off abruptly, but rather the
interstitial supersaturation decays smoothly over the time
period studied. In other words we may not see only one
breaking point to but more than one, e.g., to, ti, t2, etc..
Figure 4.9 shows a plot of /Db*-1 versus annealing time
for different implant energies. Because of the finite speed
of interstitial diffusion, we have to pick a fixed distance
from the interstitial source to make the comparison. A
distance of 3600 A from the projected range of each implant
was thus chosen in this plot. Because of the long interval
between the annealing times, we did see just one breaking
point to for E > 5 keV. According to the above discussion,
TED stops before 15 min for the 5 keV implant and within
15 min to 2 hr for higher energy implants.
The energy dependence of the time-averaged interstitial
supersaturation /Ci* after 3min and 2hr anneals is shown
in Fig. 4.10 (a) and (b). For the 3 min anneal the
interstitial supersaturation near the implant damage region

98
is much higher than that for larger depth and as a result the
boron peaks near the implant damage region broaden much more
than the deeper boron peaks. As time increases, this
difference becomes less marked since the longer time allows
for the excess interstitials to diffuse to deeper regions.
For the 3min anneal, the overall trend shows /Ci*
decreasing with increasing depth and increasing with
increasing implant energy. For the 2 hr anneal, the depth
and energy dependencies of /Cj* become much less
prominent except for the lowest energy, 5 keV. The reason
for the much shorter duration in TED and a less amount of
diffusion enhancement at 5 keV may be related to several
factors, including the source of the interstitials, as will
be discussed later.
As we discussed in the first section of this chapter,
there are no {311} defects in the 5 keV implant but they are
the only type of visible defects in the 10 and 20 keV
implants and comprise the majority of the defects in 30 and
40 keV implants. It is important to note that although there
are no {311} defects in the 5 keV implant, TED of boron does
occur. For this 5 keV sample annealed at 750C, for times of
3 min to 2 hr, the diffusivity enhancement reduces from 980
to 17 for the second shallowest spike and from 100 to 11 for
the deepest one. We believe TED occurs because interstitials
are released from mobile boron interstitial pairs during
annealing. Because of the relative proximity of the implant
damage to the surface at this low energy, a large part of the

99
interstitials could be attracted to the surface so that the
supersaturation of remaining interstitials is relatively low,
however these remaining interstitials can still diffuse into
the bulk and enhance dopant diffusion. Zhang et al.114 have
previously reported that TED was observed in 4keV lxlo!4cm-2
B+ implanted Si without the existence of {311} defects. They
attribute the source of interstitials driving TED to sub-
microscopic clusters. They determined the duration of TED by
noting when TED stops within the bounds of the SIMS
resolution, and their TED saturation was 3 min~13 min at
750C, much shorter than others have reported for TED
saturation in samples with {311} defects. In our experiment
we looked at the time averaged interstitial concentration
instead of the instantaneous one, so we did not actually
determine the time required for the interstitial
concentration to return to its equilibrium level. However,
we used the extracted Db for the 2600 A deep spike in the
5keV implant and simulated how Zhang et al.'s 4 keV
lxlO-*-^ cm2 implanted boron profile would evolve after 3 min,
15 min and 2 hr and found that TED ended after 15 min, in
agreement with their conclusion. Thus it is proposed that
when {311} defects exist, their dissolution process might
interact with that of mobile boron interstitial pairs, so
that TED continues for much longer periods of time. The type
of defects that supplies the interstitials regulates the
duration of TED.
From Fig. 4.10(c) we can see that, after an anneal of

100
2 hr, i.e. after equilibrium is reached, the interstitial
supersaturation level /Ci* varies little when implant
energy is above 5 keV. The factors that we discussed in the
previous section on why {311} defects do not form in the 5
keV implant can also be used to explain the TED behavior.
These factors include the formation of boron interstitial
clusters, the implant damage, the I-V recombination rate and
the surface recombination effect variations with implant
energy. Since the first three factors are expected to change
gradually with the changing of energy, surface recombination
effect is believed to be the most significant factor that
reduces the amount of diffusion in the 5keV implant. The
energy threshold below which the surface starts to play an
important role lies between 5keV and lOkeV.
In summary, boron 5-doping superlattices have been used
to study the effect of implant energy (5 keV to 40 keV) on
TED of boron implanted silicon. Higher implant energies
( > 5 keV) cause more total diffusion enhancement and longer
TED duration. It is speculated that a threshold exists at an
implant energy between 5keV and lOkeV, below which the
surface recombination effect strongly affects diffusion. It
is also speculated that both mobile boron interstitial pairs
and {311} defects existed in higher energy ( > 5 keV)
implants and they function together to provide interstitials
for TED. Hence the formation of {311} defects appears to
result in a dramatic increase in the duration of TED.

0-2um
(a)
Figure 4.1 Weak beam dark field PTEM (g^) micrographs of 2xlOl4cm'2 B+ implanted Si at an
energy of (a) lOkeV, (b) 20kkeV, (c) 30keV and (d) 40keV.

0.2|im

(C)
Figure 4.1 (Continued)

5min
30min
0.2[im
15min
lhr
o
(d)
Figure 4.1 (Continued)

Density of Interstitials Trapped in {311} Defects (cm'2)
105
Annealing Time (sec)
Figure 4.2. Density of interstitials trapped in {311}
defects in 10keV~40keV B+ implanted Si, annealed at 750C.

Time Constant of {311} Defect Dissolution (sec)
106
Implant Energy (keV)
Figure 4.3. Time constant of the dissolution process of the
interstitials trapped in {311} defects as a function of
implant energy, annealed at 750C.

{311} Defect Density (cm'2)
107
Annealing Time (sec)
Figure 4.4. Density of {311} defects as a function of
annealing time for different implant energies.

Average {311} Defect Size ()
108
Annealing Time (sec)
Figure 4.5. Average size of {311} defects as a function of
annealing time for different implant energies.

Defect Defect Defect Defect
Density (cm 2) Density (cm'2) Density (cm'2) Density (cm'2)
109
1011
1010
109
108
107
60 260 460 660 860 1060 1260 1460
1011 |1 1 ~r i
io10 r
10a
10a
I I 1 I 1 '
10keV 30min
Mi
MKMMl I I I I I
10 60 260 460 660 860 1060 1260 1460
{311} Defect Size ()
(a)
Figure 4.6. {311} defect size distribution for (a)10keV
(b)20keV (c)30keV and (d)40keV implants, annealed at 750C.

Defect Defect Defect Defect
Density (cm ) Density (cm ) Density (cm'2) Density (cm'2)
110
| I I1 I IIi[niIiir
20keV 10min
60 260 460 660 860 1060 1260 1460
1011
1010
109
108
107
60 260 460 660 860 1060 1260 1460
11
10
101 j.
109 j.
108 j.
^07
tIiIIIr
J' l l I '| l ll | l l I
20keV 30min .
60
260 460 660 860 1060 1260 1460
{311} Defect Size (A)
(b)
Figure 4.6. (Continued)

Defect Defect Defect Defect
Density (cm'2) Density (cm'2) Density (cm'2) Density (cm'2)
ill
* i i i i
60 260 460 660 860 1060 1260 1460
1i i ir-r tiIiiiiirr-ii i ir-r i iiiir-|
30keV 15min !
lULi uJ
L
260 460 660 860 1060 1260 1460
7-111-71r111i1-7
30keV 30min
i
111
SSL
L
260 460 660 860 1060 1260 1460
JO11 1 1 1 1-1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 t 1 1 1 1
30keV 1 hri
1010
r

109
r
108
r
107
i
60
{311} Defect Size ()
(c)
Figure 4.6. (Continued)

Defect Defect Defect Defect
Density (cm'2) Density (cm'2) Density (cm'2) Density (cm'2)
112
triiiiii i |i-1 i| i ir-i
40keV 5min =
a ""'"'"""'Tniingiir 11 i i i i i i i .
10 60 260 460 660 860 1060 1260 1460
1011
1010
109
108
107
60 260 460 660 860 1060 1260 1460
1011
101
109
108
107
T 1I 1II 1| 1I 1
40keV 30min
60 260 460 660 860 1060 1260 1460
j oil
IU r 1 TTrrTTl 1"1 I 1 1 1T '1 I 111 I 1 1 1 I 1 1
p 40keV 1 hrj
_ili i.,!)i..i.illi .run rali i
60 260 460 660 860 1060 1260 1460
{311} Defect Size ()
(d)
Figure 4.6. (Continued)

113
r
E
c
o
ro
c
CD
O
c
o
O
c
o
l
o
en
0 1000 2000 3000 4000 5000 6000
Depth ()
(a)
0 1000 2000 3000 4000 5000 6000
Depth ()
(b)
Figure 4.7. SIMS profiles of boron in 2xl014cm-2 B+
Implanted Si superlattices at an energy of (a)5keV, (b)
lOkeV, (c) 20keV and (d) 40keV, annealed at 750C.

Boron Concentration (cm'3) Boron concentration (cm'3)
114
I 1I 1 I II I I 1 I I I II 1IIU I I lifMi 1 i I I i i
0 1000 2000 3000 4000 5000 6000
Depth ()
(c)
Depth ()
(d)
Figure 4.7. (continued)

115
m
m
Q
'a

Q
V
c
Q)
E
CD
O
c
CO
.c
c:
LU
>
'in
3
5=
Depth ()
(a)
*
CD
Q
'a
m
Q
V
-*<
c
CD
E
CD
O
c
CO
sz
c
LU
>
w
3
it
b
Depth ()
(b)
Figure 4.8. Enhancement of B diffusivity /Db* in B+
implanted Si superlattices at an energy of (a) 5keV, (b)
lOkeV, (c) 20keV and (d) 40keV, annealed at 750C.

Diffusivity Enhancement /D
116
Depth ()
(c)
Depth ()
(d)
Figure 4.8. (Continued)

/D
117
1 10 100 1000
Annealing Time (min)
Figure 4.9. Enhancement of B diffusivity /Db* at a depth
of 3600 away from Rp as a function of annealing time for
different implant energies.

118
*
O
v
in
05
CO
3
c
Implant Energy (keV)
(a)
Figure 4.10. Supersaturation of interstitials as a function
of implant energy after an anneal at 750C for (a) 3min and
(b) 2hrs.

CHAPTER 5
THE INFLUENCE OF IMPLANT DOSE ON DEFECT BEHAVIOR AND
TRANSIENT ENHANCED DIFFUSION
5,1 Overview
Implant dose is one of the most important parameters
that strongly affect the form of implantation damage. Since
transient enhanced diffusion is directly related to the
damage created by the implantation, variation of implant dose
is also expected to influence diffusion behavior.
For elements heavier than carbon, at low doses, the
damage is in the form of point defects or small point defect
clusters. At higher doses, extended defects begin to form.
At very high doses, the surface of the substrate can be
amorphized. The amorphization threshold depends on the
implant condition and ion mass. Most of the experiments
investigating TED due to implantation damage have used high
dose amorphizing implants. In this case, the end-of-range
damage can act as both a source and a sink of excess
interstitials. For more information regarding the influence
of implant damage on TED in amorphized sample, please refer
to ref.17,111,131.
In the case of low dose non-amorphizing implants,
Packan130 studied boron diffusion in Si+ implanted silicon
with doses ranging from 1x10^2 cm-2 (-0 2x10^4 cm2 at an
119

120
energy of 200 keV. Since the annealing temperatures were
above 750C and the annealing times were long enough to
remove the {311} defects that would possibly be seen after
low budget anneals, it was claimed that few extended defects
were present in the sample. He studied the effect of varying
Si implant dose on the diffusion behavior of a separately
implanted deeper boron profile and found that the amount of
diffusion increased with increasing dose but doubling the
dose did not double the diffusion coefficient. This sub-
linear dependence suggests a saturation of enhanced diffusion
with increasing implant dose.
Solmi et al.132 studied boron diffusion in 20 keV and
30 keV B+ implanted silicon to doses from 2x10^4 Cm2 to
5x10-L5 cm-2. Thermal treatments were carried out at 800C to
1000C. They observed a reduced displacement of the profile
for the higher dose associated with extended defect
formation. The transient time for the enhanced diffusion was
reduced by a factor of 2 for doses higher than 5x10^-^ cm2.
They attributed this reduction to the formation of extended
defects at the projected range at these higher doses. The
extended defects act as a sink for the interstitials, thus
reducing the interstitial supersaturation level.
In this section we will discuss our study on boron
diffusion after boron implantation at low and medium doses
and anneals at a relatively low temperature (750C).
Different doses give rise to different types of extended
defects, including {311} defects and sub-amorphization (type

121
I) dislocation loops. Our aim is to investigate the effect
of varying dose on the defect behavior as well as on
transient enhanced diffusion.
5.2 Implant Dose Effect on Defect Behavior
5.2.1 Experimental Procedure
To study the dose effect on defect evolution, we
selected a single row from the implant matrix (see section
2.2). The implant energy was kept as 20 keV and the dose was
varied from 5x10-'-^ cm" 2 to 1x10^-^ cm" 2. Subsequent furnace
anneals were carried out in a nitrogen ambient for times
between 5 min and 2 hr at 750C. PTEM micrographs were taken
under weak beam dark field imaging condition using g220 beam.
The {311} defect density, average defect size, areal density
of interstitials trapped by {311} defects and the defect
distribution after annealing were measured from the TEM
micrographs.
5.2.2 Defect Microstructure and Dissolution Process
PTEM micrographs of the defects after boron implantation
at 20 keV to doses of 1x10-*-^ cm"2, 5x10^-^ cm"2, and
lxlO15 cm2 and anneals at 750C from 5 min to 2 hr are shown
in Fig. 5.1 (a)-(c). No defects were observed for the lowest
dose, 5x10^3 cm2, and therefore the pictures of these
samples are not included. The 1x1014 cm-2 implant contains
only a small number of {311} defects after 5 min and 10 min

122
anneals and when annealing time is increased to 15 min, these
defects completely dissolve. Defect behavior of the
2xl014 cm 2 implant was shown in the last chapter (Fig. 4.1
(b)) and is left out here. A majority of {311} defects
together with just a few dislocation loops are observed in
the 5xl0^-4 cm-2 sample and the loop density increases for the
lxiol- cm-2 implant. For each of the implants showing {311}
defects, the number of the defects decreases with increasing
annealing time. This dissolution process is much more rapid
for a lower implant dose.
The areal density of interstitials trapped in {311}
defects during the annealing process was quantified from the
defect measurements and the results are shown in Fig. 5.2.
Results for the lxlO14 cm-2 implant are not included because
the defect density is too low even after a 5min anneal. The
emission of interstitials from {311} defects again shows an
exponential time-dependence, which is much stronger for lower
implant dose. After a 30 min anneal, {311} defects in the
2x101* cm-2 implant dissolved to a very low density. The
number of remaining trapped interstitials is about
7.7xl0H cm-2. If the dose is increased to 5xl0l4 cm-2,
after an anneal of lhr, there is still an areal interstitial
density of about 1.7xlC)l3 cm-2. If we compare Fig. 5.2 with
Fig. 4.2, we will find that doubling the dose decreased the
dissolution rate much more than doubling the energy. Since
the number of displaced atoms per implanted ion is
proportional to the implant energy, such an observation might

123
support the idea of "plus 1" model over the idea of collision
cascade as the source of the interstitials. However, this
time constant increase with increasing dose seems to become
less pronounced when the dose is increased further. From
2xl014 cm 2 to 5xl0^-4 cm-2, the time constant changes from
424 sec to 3029 sec. But from 5x10-*-4 cm-2 to 1x10^-^ cm-2,
the time constant only changes from 3029 sec to 4604 sec. As
we discussed in Chapter 2, the threshold dose for the
formation of stable sub-amorphization (type I) dislocation
loops lies somewhere between 2xl0^-4 cm-2 and 5xl0-*-4 cm-2 at
20 keV (see Table 2.2). The presence of loops seems to
expedite the dissolution process of {311} defects.
Interstitials emitted from {311} defects will feed the
dislocation loops and the sink effect of loops makes the
{311} dissolution faster.
By measuring the loop size on the PTEM micrographs, the
concentration of interstitials trapped by dislocation loops
was obtained. By assuming a circular loop, the radius of
each loop or partial loop was measured along its longest axis
and the corresponding loop area was calculated. The
concentration of atoms bound by loops could be estimated by
multiplying the fraction of loop area by the atomic density
of atoms on the {111} plane, which is 1.6xl015 cm-2.
The density of interstitials trapped in {311} defects
and dislocation loops after anneals for 5 min and 2 hr at
750C as a function of implant dose is shown in Fig. 5.4. We
can see that the density of interstitial in {311} defects or

124
loops is much less than the implant dose. Assuming that the
net number of excess interstitials is roughly equal to the
dose (plus 1 model) and the missing interstitials are all
trapped in sub-microscopic boron interstitial complexes,
including mobile B-I pairs and immobile B-I clusters, then
the percentage of the interstitials trapped in these boron
interstitial complexes after a 5 min anneal at 750C is
87.5%, 87.8% and 88.6% for the 2xl014 cm"2, 5xl014 cm-2 and
lxlO45 cm-2 implant respectively. Within the experimental
error range we can say that this percentage does not change
with implant dose. This is different from the energy
dependence which we discussed in Chapter 4. The percentage
of interstitials trapped in boron interstitial complexes
decreases with increasing energy because of the lower boron
peak concentration due to the spreading out of the dopant
profile. Here, when implant dose is increased, both the
boron peak concentration and the implant damage increase
proportionally. As a result more boron interstitial
complexes and more {311} defects/loops are formed and the
percentage of interstitials trapped in B-I complexes versus
{311}s/loops stays constant. This argument, however, cannot
be used to predict the phenomenon in a sample without {311}
defects or loops.
It is interesting to notice from Fig. 5.4 how
interstitials re-distribute themselves during the annealing.
The {311} defects for the 2xl014 cm-2 implant dissolve almost
completely after 2 hr anneal at 750C. The trapped

125
interstitials either diffuse to the surface or into the bulk.
For the 5xl0l4 cm2 implant, the number of interstitials
released from {311} defects between 5 min to 2 hr is higher
than that trapped in the loops. Because of the small density
of the loops they do not act as a very strong interstitial
sink. While for the 1x10-*-^ cm~2 implant, loops trap more
interstitials than those released from {311} defects.
Although the number of interstitials introduced into the bulk
is the highest for this dose, there is no dramatic increase
in transient diffusion compared with a lower dose (e.g.
2x10-1-4 cm-2) lue to the presence of larger density of
interstitial traps--dislocation loops. This will be
discussed in detail later.
5.2.3 Defect Density. Size and Distribution during Annealing
The {311} defect density of different implants as a
function of annealing time measured from PTEM images is shown
in Fig. 5.5. For all annealing times, the defect density
increases with increasing dose. After a 5 min anneal, the
{311} defect density is about 3.3x10-1-0 cm-2 for the
2x10-1-4 cm 2
implant,
and about
1.1x10-1-1 cm 2
for
the
1x10-1-5 cm-2
implant.
For each
implant dose,
the
defect
density decreases with increasing annealing time, with the
decrease occurring faster for a lower dose. The reason for
the strong dose dependence of defect density might be that in
a sample implanted with higher dose, the interstitial
supersaturation level is higher. As a result, the number of

126
small defect nuclei is much larger during the initial defect
formation process. Subsequently more defects are grown from
these nuclei and therefore more defects are observed in a
higher dose implant.
The average {311} defect size as a function of annealing
time for different implant doses is shown in Fig. 5.6. The
average defect size increases with annealing time. For a
particular time, the defect size is larger for a higher dose.
After an anneal of 15 min, the 2x10-*-^ cm2 implant shows an
average defect size of about 517 A, while for the 1x10^5 cm-2
implant, it is about 613 A. Unlike the defect density
dependence, the size dependence on dose is not very strong.
This is not surprising because the total number of
interstitials introduced into the substrate represents the
product of the defect density and the defect size. Given a
fixed amount of total interstitials, a strong dose dependence
of the defect density must be accompanied by a weak dose
dependence of the defect size.
The size distribution of {311} defects after anneals at
750C for various times for 5xlC)14 cm2 and 1x10-*-^ cm"2
implants is shown in Fig. 5.7 (a) and (b). For each implant
we can see the same trend as we observed in the previous
chapters: a shift of the average {311} defect size to a
larger value and a decrease of the defect density at a
particular size with increasing annealing time. The
decreasing defect density and the increasing defect size are
the characteristics of the Ostwald ripening process. The

127
decreasing interstitial density (Fig. 5.2) and decreasing
defect density are the characteristics of defect dissolution
process. The two coupled processes give rise to the defect
distribution shown in Fig. 5.7. If we compare Fig. 5.7(a)
and (b) with Fig. 4.6(b) (the size distribution of
2xl014 cm-2 implant), we see that the standard deviation of
the size distribution increases as dose increases. This
again might be due to the larger number of defect nuclei in
the sample implanted with higher dose. Some nuclei might
grow faster than the others which results in more possible
defect sizes than the sample with fewer nuclei.
As a summary, the {311} defect microstructure and
annealing behavior as a function of implant dose (5xl013 cm-2
~ lxlC)15 cm-2) have been studied using TEM. After anneals at
750C for times ranging from 5 min to 2 hr, no {311} defects
are observed in the 5x10-*--^ cm-2 implant, a few in the
lxlO-*-4 cm-2 implant and more in the 2xl0-*-4 cm-2 implant. A
majority of {311} defects and a few dislocation loops are
present at the higher doses. Assuming a valid "plus 1"
model, the number of interstitials bound by boron
interstitial complexes is estimated to be about 88% of the
implant dose for 2xl044 cm-2, 5xl0^-4 cm-2 and lxlO-4^ cm-2
implants after an anneal for 5 min. The {311} defect
dissolution process shows a smaller time constant for a lower
dose. Dislocation loops expedite {311} defect dissolution at
doses > 5xl014citT2 by acting as an effective interstitial
sink and therefore reducing the time constant increase rate

128
of defect dissolution. The interstitial sink effect in the
lxlO-*-^ cm2 implant is more prominent than that in the
5xl014 cm"2 implant due to the higher loop density resulted
from the higher dose. Both defect density and average defect
size increases when implant dose increases. The evolution of
{311} defects indicates an Ostwald ripening process and a
defect dissolution process. The standard deviation of defect
size distribution is larger for higher doses.
5.3 Implant Dose Effect on Transient Enhanced Diffusion
5.3.1 Experimental Procedure
It is critical to understand how all the defect
evolutionary changes affect the transient enhanced diffusion
process. In order to study this again we separated the
implant profile from the diffusion marker layer. A thin 8-
doping boron marker layer was used to monitor boron diffusion
behavior after ion implantation and annealing. The layer was
grown via atmospheric pressure chemical vapor deposition
(APCVD) on p-type (100) Czochralski-grown silicon substrate,
followed by growth of an overlayer of approximately 7000 A of
undoped silicon. The full width at half maximum (FWHM) of
the boron spike is about 500 A and peak concentration is
about 2xl0l8 cm"3. This low concentration allows dopant
diffusion to remain in the intrinsic regime at our processing
temperature and also reduces boron interstitial cluster
formation in the doping spike. The as-grown wafers were pre-

129
annealed in a nitrogen ambient at 800C for lhr to remove the
excess point defects introduced by the APCVD process. The
wafers were then implanted with 20 keV boron ions to doses of
5xl0l3 cm-2, 2xl0-'-4 cm-2 and 1x10^-^ cm-2. Subsequent anneals
were performed in nitrogen at 750C for 15 min, 30 min, 1 hr,
2 hr and 4 hr. SIMS analysis was performed on the CAMECA
IMS4f system. The parameters used during the SIMS analysis
were the same as discussed in section 3.3.1. The dopant
profiles were then imported into FLOOPS and the diffusivity
enhancement was extracted for each anneal.
5.3.2 Dopant Diffusion Behavior
SIMS profiles of boron in the as-grown and pre-annealed
samples as well as in the implanted and annealed samples are
shown in Fig. 5.8(a)~(d). The small displacement of the
profile of the pre-annealed sample with respect to that of
the as-grown sample is believed to be due to boron intrinsic
diffusion. The buried boron spike in the 5xlC)l2 cm-2
implanted sample shows an enhanced diffusion but the profile
displacement is smaller compared with those at higher doses.
The 30min annealed profile of the 2x10^^ cm-2 implanted
sample shows an enhanced diffusion to a similar degree as
that of the lxlO^-5 cm-2 implant. But the profile after a 4hr
anneal shows a smaller enhancement than that at lxlO15 cm-2.
Besides the buried boron spike, the tail of implanted boron
profile also shows an enhanced diffusion in each sample. The
peak of this implanted profile, however, is clustered and

130
therefore stays immobile for the two higher doses. The
breaking point between the mobile and immobile part is at a
concentration of about 3x10-*-^ cirT^, the same for both
implants and for every anneal.
The SIMS data of the control and implanted samples were
imported into FLOOPS. Diffusivity enhancement /Db* was
obtained by matching the initial profile with the target
profile assuming an intrinsic diffusivity with the form
Db*=0.757exp(-3.46eV/kT) (eq.(3.8)). We extracted the
enhancement factor for control sample 1 (as-grown plus 800C
1 hr anneal), control sample 2 (pre-anneal plus 750C 1 hr
anneal) and control sample 3 (pre-anneal plus 750C 4 hr
anneal). The purpose of control sample 1 was to examine how
much enhancement would come from the excess interstitials
induced by the APCVD growth process. The purpose of control
samples 2 and 3 was to test whether the nitrogen ambient
indeed functions as an inert ambient. The enhancement factor
for these three samples was 4.1, 4.9 and 4.7 respectively.
This small enhancement might be due to the residue oxygen in
the nitrogen ambient. However, it might also just result
from the experimental errors, including the SIMS depth error
range of 5% and the annealing temperature accuracy of 10C.
The number of excess interstitials introduced by the APCVD
process is obviously not high enough to cause much enhanced
diffusion.
Fig.5.9 is a plot of the time-averaged diffusivity
enhancement /Db* of the buried boron spike as a function

131
of annealing time for different implant doses after each
anneal. Before the simulation, dose normalization was
conducted on the spike for the as-implanted and annealed
samples regardless of the small influx of boron from the
surface implant into the spike region which actually violates
the dose conservation. From Fig.5.9 we can see that the
value of /Db* decreases with increasing annealing time
for each implant. At all annealing times, /Db* is lower
for the 5x10^3 cm-2 implant than the two higher doses. After
an anneal at 750C for 4 hr, /Db* increases from 409 for
the 5xl0l3c m2 implant to 854 for the 2x10^4 Cm"2 implant.
For B diffusion, /Db* is proportional to the interstitial
supersaturation level /Ci*. If the "plus 1" model is
valid, the difference in /Db* simply should be four
times. Since only a two times increase is seen yet all the
{311} defects have dissolved, stable B-I cluster formation in
the 2xl0l4 cm2 implant could account for the small increase
in /Db*.
5.3.3 TED Saturation at Higher Dose
It is interesting to notice that /Db* is close for
the two higher doses, 2x10-'-^ cm-2 and lxlO-^5 cm-^. As we
discussed before, the formation threshold for the sub-
amorphization dislocation loops lies between doses of
2x10-1-4 cm2 and 5x10-*-^ cm-^. When the dose reaches
lxlO15 cm 2f quite a few loops form after annealing. These
loops can act as a sink for interstitials released from

132
either {311} defects or invisible mobile boron interstitial
pairs. Besides this trapping effect of interstitials by-
loops, more interstitials might be trapped in the boron
interstitial clusters in the implant damage peak region
because more clusters are formed in the higher dose implant.
As a result of both trapping effects, the number of
interstitials available for TED is reduced and so is the
enhancement of diffusion. This can explain why we do not
observe a decrease in /Db* for times < 1 hr even though
the dose is increased four times.
Another explanation of the TED saturation behavior with
increasing dose was proposed by Griffin et al133. They
claimed that this saturation can occur if most of the dopant
atoms are already pushed into a mobile state so that more
damage produces no increase in diffusivity. They used a
simple chemical reaction model to explain the phenomena. The
interaction of a substitutional dopant atom Bs with a free
interstitial I to produce a mobile Bx can be described by
Bs + I <-> Bj (3.5)
If this reaction is in equilibrium, then we have
Bi KeqBsI (5.1)
where Keq is the equilibrium reaction constant. The
substitutional Bs and mobile BI are related to the total
dopant by
BT = Bs + Bj (5.2)
Using eq.(5.1) and (5.2) we have

133
Bt = Bt KeqI/(1+KeqI) (5.3)
The above equation shows that for large damage dose, Br
approaches a maximum value of BT. Further increase in the
dose will not produce more B!. Effectively all the B is
paired and further increase in I do not enhance Db.
At this time we do not know which mechanism, the
loop/cluster trapping of interstitials or dopant interstitial
pair saturation, is valid in our case. Quantification of the
interstitials in the clustered region will shed light on this
issue. If the extra interstitials due to the dose increase
from 2x1()14 cm2 to lxlO^^ cm2 are all trapped in the
loops/clusters, then the first mechanism might dominate.
Otherwise, we should also consider the second mechanism. For
our APCVD grown boron spike, the area integration under the
spike shows a dose of lxlC)13 cm_2. Because this is only 5%
of the 2x10^4 cm"2 dose and 1% of the lxlO^^ cm"2 dose, it is
quite possible that dopant interstitial pair saturation is
reached. But we also observed more clustering for the higher
dose in Fig. 5.7 and the trap of interstitials by loops in
Fig. 5.4. Both mechanisms might function together and either
one of them could give rise to the current diffusion
behavior.
The diffusion length of the buried boron spike V(t)
as a function of annealing time for each implant is shown in
Fig. 5.10. The dots are the data points and the lines are
the fits in the form of aV (1-exp (-t/T)) where a and x are
constants and t is the annealing time (please refer to

134
Chapter 3 for the derivation). Another data point (t=0,
V(t)=0) is added on each set of data for the convenience
of fitting. It is obvious that TED saturates after a shorter
time (around 16 min) at the low dose of 5x10^-^ cm" 2 than at
higher doses (37 min at 2x10^-^ cm"2 and 4 min at
lxlC)15 cm-2). The diffusion length at 2x10-*-^ cm"2 dose is
very close to that at 1x10-*-^ cm"2 dose due to the factors we
discussed above.
A similar dose dependent diffusion behavior was observed
by Packan130. As was mentioned in the overview section in
this chapter, the diffusion of a buried implanted boron
profile was studied by using Si+ implanted silicon with doses
ranging from 1x10-*-^ cm"2 to 2x10-*-^ cm"^ at an energy of 200
keV. He presented the diffusion length versus annealing time
plots for doses of 5x10-^^ cm"2, 1x10-*-4 cm"2 and 2x10-*-^ cm" 2
after anneals at 750C, 800C and 850C for various times.
The diffusion length for the 5x10^-^ cm"2 implant is the
lowest for all the anneals. The diffusion lengths for the
1x1014 cm-2 and the 2x10^4 cm"2 implants stay about the same
for t < 330 min at 750C, t < 60 min at 800C and t < lOmin
at 850C. After these times higher dose gives rise to larger
diffusion length. Comparing our results with Packan*s, it is
reasonable to believe that for 200 keV Si+ implantation, the
threshold dose for the formation of sub-amorphization
dislocation loops is between lxlO^ Cm"2 and 2x10^^ cm"2.
During the early stage of annealing these loops absorb
interstitials and reduce the enhancement of boron diffusion.

135
Later on these trapped interstitials are released from the
loops and make contribution to boron diffusion.
Interstitials are easier to escape the trap of dislocation
loops at higher temperatures and therefore it takes shorter
time to observe a larger diffusion at a higher dose.
Our results are also comparable with those of Solmi et
al.132. They used 20 keV to 30 keV B+ implantations to doses
from 2xl044 cm-2 to 5x10-*-^ cm-2 to study the diffusion of the
B implant itself. They observed a less displacement in the
profile tail region for the higher doses (> 5xl014 cm-2
compared with < 5xl044 cm-2) within a time range of 5min to
2hrs at 800C. The displacements of the 2xl015 cm-2 and the
5x10-*-5 cm-2 implants are very close. They attributed the
reduction of diffusion to the extended defects which might
act as a sink of the free interstitials.
In summary, 8-doping boron marker layer was used to
study dopant diffusion behavior after B+ implantation at 20
keV to various doses. This APCVD grown layer acted as a
sensitive diffusion monitor for diffusion because of the
steep dopant gradient. The doses chosen cover the transition
regions from sample with no {311} defects (5x10-*--^ cm-2), to
that with {311} defects (2xl014 cm-2), then to that with both
dislocation loops and {311} defects (1x10^-^ cm-2). The
lowest dose shows the lowest diffusion enhancement and the
smallest diffusion length for all the anneals. The diffusion
enhancement and the diffusion length saturate for doses above
2xl014 cm-2. There are two mechanisms which can both explain

136
this saturation effect. One is that dislocation loops and
more B-I clusters form at the highest dose, which act as a
sink of interstitials for a limited period of time, during
which dopant diffusion is reduced, and after which
interstitials are emitted again and give rise to more
diffusion. The other is that all the boron atoms are paired
with interstitials and more interstitials would not cause any
further diffusion. Our results are comparable to those
reported by Packan, where lower dose Si+ implantation was
used to examine the dose dependence of dopant diffusion, and
to those of Solmi et al., where higher dose B+ implantation
was used and the formation of dislocation loops was claimed
to result in a reduction in TED.

Figure 5.1. Weak beam dark field PTEM (g^) micrographs of 20keVcni2 B+ implanted Si to a dose
of (a) lxl014cm'2, (b) 5xl014cm'2 and (c) lxl015cm'2.

0.2|im
(b)
Figure 5.1 (Continued)

0.2(im
Figure 5.1 (Continued)

Density of Interstitials Trapped in {311} Defects (cm'2)
140
Annealing time (sec)
Figure 5.2. The exponential decay of interstitials trapped
in {311} defects in 20keV B+ implanted Si to various doses.

Time Constant of {311} Dissolution Process (sec)
141
Implant Dose (cm'2)
Figure 5.3. Time constant of {311} dissolution process as a
function of implant dose, annealed at 750C.

Areal Density of Interstitials
Trapped in {311} Defects or Loops (cm'2)
142
0 10 2 1014 4 1014 6 1014 8 1014 1 1015 1.2 1015
Implant Dose (cm'2)
Figure 5.4. Areal density of interstitials trapped in {311}
defects and dislocation loops as a function of dose.

{311} Defect Density (cm 2)
143
Annealing Time (sec)
Figure 5.5. {311} defect density as a function of annealing
time in 20keV B+ implanted Si for different doses.

144
Annealing Time (sec)
Figure 5.6. Average {311} defect size as a function of
annealing time in 20keV B+ implanted Si for different doses.

Defect Defect Defect Defect
Density (cm'2) Density (cm'2) Density (cm'2) Density (cm2)
145
1011
1010
109
108
107
60 260 460 660 860 1060 1260 1460
11
10
10
10
10s
10£
a rJ I 111,, 1^11 i 1 1 1 L1-X..-1Li-Jj I 1 I I L
10 60 260 460 660 860 1060 1260 1460
T|III | III|IIII'" I I1|
5e14 1hr:
60 260 460 660 860 1060 1260 1460
1011
101
109
108
107
i tt I i i i | n | i i i | i i i | n i i iiiinr"|
5e14 2hrs -
60 260 460 660 860 1060 1260 1460
{311} Defect Size ()
(a)
Figure 5.7. {311} defect size distribution for (a)5xl014cm-2
and (b)lxl015cm-2 implants, annealed at 750C.

Defect Defect Defect Defect
Density (cm'2) Density (cm'2) Density (cm'2) Density (cm2)
146
tii|ii1\ iin
1e15 bmin -
J I I I L
60 260 460 660 860 1060 1260 1460
1011
1010
109
108
107
60 260 460 660 860 1060 1260 1460
11
10
10
10
10s
10
10
r
8 '
7
60
i | 1 1 1 I 1 1 1 l 1 1 1 I 1 1 1 I 1 1 1 I 1 1 1 !
1e15 1hr
260 460 660 860 1060 1260 1460
1011 rr"r 1 i 1 1 1 i 1 1 1 i l"1 1 i 1 1 1 i 1 1 i 1 1 i 1 1 i
", q1o i- 1e15 *hrs -I
60 260 460 660 860 1060 1260 1460
{311} Defect Size ()
(b)
Figure 5.7. (Continued)

147
i
E
o
c
o
cC
c
0
o
c
o
O
c
o
o
CQ
0 2000 4000 6000 8000 1 104
Depth ()
(a)
Depth ()
(b)
Figure 5.8. SIMS profiles of boron in 20keV B+ implanted Si
superlattice at a dose of (b)5xl013cm-2, (c) 2xl014cm-2, (d)
lxl0-*-^cm-2. (a) shows the control sample profiles.

Boron Concentration (cm'3) Boron Concentration (cm'3)
148
0 2000 4000 6000 8000 1 104 1.2 104
Depth ()
(c)
0 2000 4000 6000 8000 1 104 1.2 104
Depth ()
(d)
Figure 5.8. (Continued)

149
Annealing Time (min)
Figure 5.9. Time-averaged boron diffusivity enhancement
/Db* as a function of annealing time for different
implant doses.

Diffusion Length (t)0-5 ()
B
150
Figure 5.10. Boron diffusion length vs. annealing time for
different implant doses.

CHAPTER 6
SUMMARY AND FUTURE WORK
6.1 Summary
Transient enhanced diffusion presents a major obstacle
for the optimization of shallow junction devices. The excess
interstitials introduced by an ion implantation process are
the driving force for TED. In order to gain better control
of dopant diffusion, it is crucial to understand the behavior
of ion implantation induced defects and their correlation
with dopant diffusion. This work represents an effort
towards meeting this goal. A brief summary of this work is
given below.
A low energy B implant matrix has been employed in this
study. The formation threshold doses of both {311} rod-like
defects and [110] sub-amorphization loops were examined using
TEM. Both threshold doses increase with decreasing implant
energy and the one for {311} defect formation is lower than
that for loop formation. Displaced atom density calculated
from TRIM seems to be able to predict the formation threshold
very well.
One particular sample in the matrix was used to study
the effect of annealing on {311} defect evolution and TED
behavior. {311} defects dissolve rapidly during annealing.
151

152
The interstitials trapped in {311} defects decay
exponentially with annealing time with a characteristic time
constant smaller at higher temperature. The activation
energy of {311} defect dissolution process was determined to
be 3.8eV. The defect density and the average defect size
were also quantified. The size distribution of {311} defects
demonstrates an Ostwald ripening process along with a defect
dissolution process. Boron diffusion was measured through
SIMS and the diffusivity enhancement /Db* was extracted
from FLOOPS simulations. /Db* decreases with increasing
annealing time. The diffusion length increases initially
during the annealing and then reaches a saturation point when
it starts to plateau out. The time constant decreases
dramatically when annealing temperature increases. The
activation energy for the TED saturation process was
determined to be 1.6eV. Both mobile boron interstitial pair
complexes and {311} defects are believed to be the source of
interstitials driving TED. The static implant profile peak
is believed to be due to the Bs-Bx-I clustering effect. The
breaking point between the mobile tail region and the
immobile peak region Cenh moves to higher concentration when
temperature increases due to a decreasing interstitial
supersaturation level and thus less cluster formation.
The influence of implant energy on {311} defect
evolution was investigated by studying samples implanted to a
single dose (2x10-*-^ cm-^) at various energies (5~40 keV) .
{311} defects were observed at energies greater than 5 keV.

153
The lack of {311} defects in the 5 keV implant might be due
to a combination of factors including surface recombination
effect, large density of boron interstitial clusters and less
implant damage and high I-V recombination. The time constant
for the {311} dissolution process is larger for higher
energy. Boron 8-doping superlattices were used to monitor
the diffusion behavior for each implant energy. /Db*
decays exponentially with depth with a decay length larger at
shorter annealing time before equilibrium is reached. TED
lasts longer and the amount of TED is larger for E>5 keV than
the 5keV implant.
The influence of implant dose on {311} defect evolution
was studied by using samples implanted at the same energy
(20keV) to various doses (5xlC)13~lxlO-*-5 cm2). {311} defects
dissolve much slower at higher dose. The ratio of the
interstitials trapped in {311} defects and loops to the
implant dose stays about the same (12%) after a 5min anneal
at 750C regardless of the dose change. Both defect density
and average defect size increase with increasing dose.
Wafers with a buried boron 8-doping layer were implanted to
the same doses at the same energy to study the diffusion
behavior. The amount of diffusion increases with increasing
dose before dislocation loops are formed. Further increase
in the dose showed no additional increase in TED presumably
due to the trap of interstitials by the loops or the
saturation of dopant atoms paired with interstitials.

154
This work supports the following models for {311} defect
evolution and for interstitial storage and release during
annealing in B+ implanted silicon.
Rafferty et al.134 proposed the following theory to
explain the {311} defect formation and dissolution. The
model can be represented by
n r
^ = 4jcaaD,I(x,t)C(x,t) Cf exp(h~)
at a kT
(6.1)
where a is the capture radius of an interstitial by a {311}
defect expressed in units of the average interatomic spacing
a. D^Doexp (-Em/kT) is the interstitial diffusivity. I(x,t)
is the concentration of free interstitials and C(x,t) is the
concentration of interstitials trapped in {311} defects. The
first term on the right side of eq.(6.1) is the {311} defect
formation term. A defect grow whenever diffusing
interstitials are trapped by it. The second term is the
dissolution term. An interstitial in a {311} defect may
escape at a rate given by the interstitial hopping frequency
and the binding energy to the defect. At steady state, the
first term is balanced by the second. The interstitial
supersaturation Ie/I* is given by
/ 1 Eh Ef
~ ;exP(
I 4nar
kT
(6.2)
where T is a dimensionless prefactor that arises from the
entropy of formation of an interstitial and Ef is the
formation energy of an interstitial. The time to dissolve
the {311} defects is given by

155
T =
4p*,6 +
exp(-
)
(6.3)
where Rp is the projected range, Q is the implant dose,
Ns=a3 is the number of lattice sites and Em is the migration
energy of an interstitial. This model predicts that the
dissolution time of {311} defects is proportional to the
projected range, and therefore to the implant energy (see
Fig.2.4), and is proportional to the same degree to the
implant dose. A strong temperature dependence is shown by
the exponential term. This agrees with our results, which
show an increase in T with increasing energy and dose and with
decreasing temperature. However, we observed that doubling
the dose increases T more than doubling the energy. This can
not be explained by the current model.
In Chapter 3, we used the variation of the energy of an
interstitial trapped in {311} defects to explain the Ostwald
ripening process. In fact the experimentally observed
Ostwald ripening process is also in agreement with the model
discussed above. Since larger {311} defects are more stable
than smaller ones, Eb must be larger for a larger defect.
From eq.(6.2) we can see that Ie/I* is lower for larger Eb-
This lower interstitial supersaturation is the driving force
for the interstitial flow from smaller defects to larger
ones, i.e., the Ostwald ripening process.
Recently a homogeneous nucleation mechanism for the
{311} defects was proposed by Chao et al.135. They claimed
that the diffusion enhancement, which is proportional to the

156
interstitial supersaturation level Ie/I* in eq. (6.2), is
related to the defect size (due to the fact that larger
defects have larger Efc,), and they observed that various doses
and energies all show similar enhancement within t < 2 min.
It thus appears that the defect size is independent of the
implant conditions, only their number being different. This
suggests a homogeneous nucleation mechanism, rather than a
mechanism which is limited by the number of available
nucleation sites. Our results on {311} defect size variation
with implant energy and dose can not directly prove this
model since we used furnace anneals with thermal budgets
large enough to pass the nucleation stage of the {311}
defects.
Regarding the interstitial storage and release during
annealing, the following model is suggested based on our
observations. At very low energies and doses where no {311}
defects or dislocation loops are formed, interstitials are
stored in the immobile boron clusters if the implant profile
peak concentration is above a critical value (Cenh>ni), or in
the mobile boron interstitial pairs. During annealing, the
clustered peak with B concentration greater than Cenh is
static and electrically inactive, while the B atoms paired
with interstitials show enhanced diffusion in the profile
tail region with concentration less than Cenh- Interstitials
are released from the B-I pairs once the B atoms return to
substitutional sites. The first saturation of TED occurs
when the interstitial supersaturation level reaches

157
equilibrium. When the annealing temperature is high enough
and the time is long enough to break up the immobile B
clusters (e.g. 800C >4hrs), more interstitials are released
from these clusters and a second burst of TED will be
observed. Only this time the breaking point between the
mobile and immobile regions moves up from Cenh to Cs, the
solid solubility limit of B in silicon.
When the implant dose and/or energy increase so that
{311} defects form during annealing, interstitials are not
only trapped in the B-I clusters and mobile B-I pairs as
discussed above, but also trapped in {311} defects. {311}
defects dissolve rapidly during annealing and emit
interstitials. These interstitials together with those
emitted by the B-I pairs are available to drive dopant
diffusion. Because of the co-existence of the two sources of
interstitials, the activation energy of TED saturation
process is lower than that in Si+ implanted samples with only
{311} defects and higher than that in low energy low dose B+
implanted sample without {311} defects but presumably with B-
I pairs.
When the implant dose and/or energy increase further,
stable dislocation loops form in addition to {311} defects.
Unlike the {311} defects, the loops dissolve relatively slow.
They may trap a small or a large amount of interstitials,
depending on the density of the loops. As a result of the
interstitial sink effect the loops present, the total number
of interstitials available to drive TED is reduced. When the

158
annealing temperature is high enough or the time is long
enough (e.g. 800C > 1 hr), the loops start to dissolve,
release interstitials and give rise to more diffusion. The
dissolution of loops occurs earlier than the de-clustering of
the immobile profile peak, but later than the first
saturation of TED.
Assuming the presence of immobile B-I clusters and
mobile B-I pairs, a summary of different types of defects
formed in samples implanted with various energies and doses
based on our implant matrix is shown in Fig.6.1. and a
schematic of different sources of TED during an annealing
sequence is shown in Fig.6.2. Quantification of how the
interstitials partition themselves into B-I clusters or B-I
pairs will allow us to gain better understanding on dopant
diffusion.
6.2 Future Work
There is still a significant amount of research to be
done in the area of defect and diffusion behavior after ion
implantation. The following are some suggestions.
1. High annealing temperature regime
An activation energy of 1.6eV for the TED saturation
process was obtained through the diffusion study of 20keV
2xl014cm-2 B+ implanted silicon after furnace anneals at
650~800C for various times (see Chapter 3). The sample
contains {311} defects only. A much higher activation energy
of 4.3eV has been reported by Michel for 60keV 2xl0l4cm-2 B+

159
implant after furnace and rapid thermal anneals at
800~1000C. His sample, according to our defect formation
threshold study, should contain both {311} defects and
dislocation loops. It would be interesting to anneal our
samples in that high temperature regime to investigate
possible difference in diffusion mechanism and the effect of
loops on TED.
2. Electrical activity measurements
Through spreading resistance profiling (SRP)
measurements we can determine the electrically active portion
of the implanted boron atoms. This can help us gain
information regarding the clustering/de-clustering motion of
the boron peak region and its effect on TED in the tail
region.
3. Effect of implant species, dose rate and implant
temperature
Our group's preliminary results show that P implants
have higher formation threshold than Si and B implant.
Implant dose rate, if changing within one order of magnitude,
does not affect TED behavior in crystalline silicon, however,
higher dose rate does result in faster {311} defect
dissolution under the same annealing conditions.
Implantation temperature, if changing between 5~40C, does
not affect TED behavior either. Further studies would be
necessary to investigate the effects of these variables on
defect evolution and dopant diffusion.
4. Effect of background doping

160
Boron doped silicon has been found to show less {311}
defects when the doping level is increased. The doping of
other species, such as P, may also influence the defect
behavior.
5. Surface recombination effect
We have discussed that the effect of surface
recombination coupled with other factors which altogether
result in the lack of {311} defects in the 5keV implant and a
smaller enhanced diffusion than higher energy implants. It
would be interesting if the surface recombination effect can
be quantified and sorted out from the others.

161
Surface Recombination, B-B-I Clustering,
I-V Recombination

Dose
Implant Damage, {311}/Loop Formation
Energy-
Figure 6.1. A summary of different types of defects that
presumably exist in B+ implanted Si, effects of implant
energies and doses.
Implant Damage, B-B-I Clustering,
{311}/Loop Formation

Q)
ON
to
Figure 6.2. A schematic of different sources of interstitials for TED that presumably
exist in B+ implanted Si.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
LIST OF REFERENCES
C. M. Osburn, J. Electron. Mater. 19. 67 (1980) .
H. Jiang, C. M. Osburn, P. Smith, Z.-G. Xiao, D. Griffis,
G. McGuire, and G. A. Rozgonyi, J. Electrochem. Soc. 139.
196 (1992).
S. N. Hong, G. A. Ruggles, J. J. Wortman, and M. C.
Ozturk, IEEE Trans. Electron Devices ED-38. 476 (1991) .
R. H. Dennard, F. H. Gaensslen, N. H. Yu, V. L. Rideout,
E. Bassous, and A. R. LeBlanc, IEEE J. Solid-State
Circuits SC-9. 256 (1974) .
J. R. Brews, W. Fichtner, E. H. Nicollian, and S. M. Sze,
IEEE Electron Device Lett, EDL-lf 2 (1980) .
S. N. Hong, G. A. Ruggles, J. J. Paulos, J. J. Wortman,
and M. C. Ozturk, Appl. Phvs, Lett. 53. 1741 (1988).
E. Ling, P. D. Maguire, H. S. Gamble, and B. M.
Armstrong, IEEE Electron Device Lett. EDL-8. 96 (1987) .
K. T. Kim and C. K. Kim, IEEE Electron Device Lett. EDL-
£, 569 (1987) .
H. J. Bohm, H. Wendt, H. Oppolzer, K. Masseli, and R.
Kassing, J. Appl. Phvs. 62. 2784 (1987) .
M. Horiuchi and K. Yamaguchi, IEEE Trans. Electron
Devices ED-33. 260 (1986).
P. G. Carey, T. W. Sigmon, R. L. Press, and T. S. Fahlen,
IEEE Electron Device Lett. EDL-6. 291 (1985) .
Y. El-Mansy, IEEE.Trans. Electron Devices ED-29. 567
(1982) .
Y. Kim, H. Z. Massoud, and R. B. Fair, AppI. Phvs. Lett.
3, 2197 (1988).
S. Guimaraes, E. Landi, and S. Solmi, Phvs. Stat. Sol
a) 95. 589 (1986).
163

15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
164
F. F. Morehead and R. T. Hodgson, Mat. Res. Soc. Svxno.
Proc. 35. 341 (1985).
J, F. Marchiando and J. Albers, J. Appl. Phvs. 61. 1380
(1987) .
T. 0. Sedgwick, A. E. Michel, V. R. Deline, S. A. Cohen,
and J. B. Lasky, J. AppI. Phvs. 63. 1452 (1988).
S. Solmi, R. Angelucci, F. Cembali, M. Servidori, and M.
Anderle, AppI. Phvs. Lett. 51. 331 (1987).
S. Nishikawa, A. Tanaka, and T. Yamaji, AppI. Phvs. Lett.
£0# 2270 (1992).
Q. Guo, X. Bao, J. Hong, Y. Yan, and D. Feng, AppI. Phvs.
Lett. 54. 1433 (1989).
C. S. Rafferty, H.-H. Vuong, S. A. Eshragi, M. D. Giles,
M. R. Pinto, and S. J. Hillenius, IEDM Technical Digest.
311 (1993).
M. C. Ozturk and J. J. Wortman, AppI. Phvs. Lett. 52. 963
(1988) .
M. Kase, M. Kimura, H. Mori, and T. Ogawa, AppI. Phvs.
Lett. 56. 1231 (1990).
A. Bousetta, J. A. van den Berg, D. G. Amour, and P. C.
Zalm, AppI. Phvs. Lett. 58. 1626 (1991).
0. M. Holland and J. Narayan, Nucl. Inst. Meth, B 40/41,
537 (1989).
R. G. Wilson, J. AppI. Phvs. 54. 6879 (1983).
E. Ganin and A. Marwick, Mat. Res. Soc. 147. 13 (1989) .
A. C. Ajmera and G. A. Rozgonyi, AppI. Phvs. Lett. 49.
1269 (1986).
L. Laanab, C. Bergaud, C. Bonafos, A. Martinez, and A.
Claverie, Nucl. Inst. Meth. B 96. 236 (1995) .
R. G. Taylor, C. Andre, T. Salama, and P. Ratnam, IEDM
Tech. Dig., p.644 (1988).
D. C. Chen, T. R. Cass, J. E. Turner, P. P. Merchant, and
K. Y. Chiu, IEEE Trans. Electron Devices ED-33. 1463
(1986) .

32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
165
E. K. Broadbent, R. F. Inrani, A. E. Morgan, and P.
Maillot, IEEE Trans. Electron Devices ED-36, 2440 (1989).
J. Lin, S. Barnerjee, J. Lee, and C. Teng, IEEE Electron
Device EDL-11. 191 (1990).
V. Ilderem and R. Reif, J. Electrochem. Soc. 136, 2989
(1989).
R. Liu, D. S. Williams, and W. T. Lynch, J. Aool. Phvs.
£2, 1990 (1988).
V. Probst, H. Schaber, A. Mitwalski, H. Kabza, L. Van den
hove, and K. Maex, J. Appl. Phvs. 70. 708 (1991) .
M. H. Juang and H. C. Cheng, J. Aool. Phvs. 71. 1271
(1992).
S. B. Herner, K. S. Jones, H.-J. Gossman, R. T. Tung, J.
M. Poate, and H. S. Luftman, in The effect of TiSi2 film
thickness and growth on the point defect perturbance in
Si, 1996 (Electrochemical Society), p.337-347.
D.-S. Wen, P. Smith, C. M. Osburn, and G. A. Rozgonyi, J.
Electrochem. Soc. 136, 466 (1989) .
K. H. Weiner, P. G. Garey, A. M. McCarthy, and T. W.
Sigmon, IEEE Electron Device Lett. EDL-13. 369 (1991).
J. F. Gibbons, Proc. of the IEEE 60. 1062-1096 (1972).
K. S. Jones, S. Prussin, and E. R. Weber, AppI. Phvs. A
45, 1 (1988) .
K. S. Jones and J. Gyulai, Ion Implantation Science and
Technology (Ion Implantation Technology Co., Yorktown,
NY, 1996).
J. Liu, M. E. Law, and K. S. Jones, Solid-State
Electronics 38. 1305 (1995).
D. Venables and K. S. Jones, Nucl. Inst. Meth. Phv, Res.
B59/60. 1019 (1991).
L. Laanab, C. Bergaud, M. M. Faye, J. Faure, A. Martinez,
and A. Claverie, Mat. Res. Soc. Svmp. Proc. 279. 381
(1993) .
O. Dokumaci, P. Rousseau, S. Luning, V. Krishnamoorthy,
K. S. Jones, and M. E. Law, J. AppI. Phvs. 78. 828
(1995) .

48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
166
K. Krishnamoorthy, D. Venables, and K. S. Jones, in Proc.
Ion Implantation Technology, Austin, TX, 1996.
R. G. Elliman, J. S. Williams, D. M. Maher, and W. L.
Brown, Mat. Res. Soc. Svmp. Proc. 51 (1986).
F. F. Komarov, A. P. Novikov, T. T. Samoilyuk, V. S.
Solov'yev, and S. U. Shiryaev, Rad. Eff. 90. 307 (1985).
J. Linnros and G. Holmes, Mat. Res. Soc. Svmp. Proc. 51
(1986).
I. G. Salisbury and M. H. Loretto, Phil. Maa. 39. 317
(1979) .
I. G. Salisbury and M. H. Loretto, Rad. Eff. 59. 59
(1981).
Y. Tsubokawa, M. Kuwabara, H. Endoh, and H. Hashimoto, J.
Electron Microscopy 35 Suool, 953 (1986).
M. Pasemann, D. Hoehl, A. L. Aseev, and 0. P. Pchelyakov,
Phvs. Status Solidi a80. 135 (1983).
T. Y. Tan, Phil. Maa. 44. 101 (1981).
P. K. Madden and S. M. Davidson, Radiation Effects 14.
271 (1972).
K. Seshan and J. Washburn, Phvs. Stat. Sol. 26. 345-352
(1974).
C. A. Ferreira Lima and A. Howie, Phil. Mag. 34. 1057
(1976).
S. Takeda, Jpn, J, AppI. Phvs. 30. L639 (1991) .
K. S. Jones, J. Liu, L. Zhang, V. Krishnamoorthy, and R.
T. DeHoff, Nucl. Inst. Meth. B 106. 227 (1995) .
J. A. Lambert and P. S. Dodson, Phil. Maa. 44. 1043
(1981) .
W. Wu and J. Washburn, J. Appl. Phvs. 48. 3742 (1977) .
J. J. Comer, Radiation Effects 36. 57 (1978).
H. Foell, T. Y. Tan, and W. Krakow, Mat. Res. Soc. Svmp.
Proc. 2. 173 (1981).

167
66 F. Cembali, L. Dori, R. Galloni, M. Servidori, and F.
Zignani, Radiation Effects 36. Ill (1978).
67 K. Seshan and J. Washburn, Rad. Eff. 26. 31 (1975) .
68 K. Seshan and J. Washburn, Radiation Effects 14. 267
(1972) .
69 I. W. Wu and L. J. Chen, J. Appl. Phvs. 58, 3032 (1985).
70 E. Nes and J. Washburn, J, Appl. Phvs. 42. 3559 (1971).
71 J. Narayan and J. Fletcher, MRS Svmo. Proc. 2. 191
(1981) .
72 M. D. Matthews and S. J. Ashby, Phil. Mag. 27. 1313
(1973).
73 S. M. Davidson and G. R. Booker, Rad. Effects 6. 33
(1970) .
74 M. Tamura, Appl, Phvs. Lett. 23# 651 (1973).
75 L. T. Chadderton and F. H. Eisen, Radiation Effects 7.
129 (1971).
76 A. Parisini and A. Bourret, Phil. Mag. A 67. 605 (1992).
77 T. Y. Tan and U. Gosele, J. AppI. Phvs. 37. 1 (1985).
78 S. Wolf and R. N. Tauber, Silicon Processing for the VLSI
Era. Vol. 1 (Proc. Tech. Latt. Press, Sunset Beach, FL,
1983).
79 W. Frank, U. Gosele, H. Maher, and A. Seeger, Diffusion
in Si and Ge (Academic Press, Orlando, FL, 1983).
80 S. K. Ghandhi, VLSI Fabrication Principles. John Wiley &
Sons, Inc. (1982).
81 C. M. Hsieh and D. M. Maher, J. AppI. Phvs. 44. 1302
(1973) .
82 S. M. Hu, J. AppI. Phvs. 45. 1567 (1974).
83 T. Y. Tan and U. Gosele, Appl. Phvs. Lett. 39. 86 (1981).
84 D. A. Antoniadis and I. Moskowitz, J. AppI. Phvs. 53.
6788 (1982).
85 P. A. Packan and J. D. Plummer, J. AppI. Phvs. 68. 4327
(1990) .

168
86 D. J. Roth, Y. S. Huang, J. D. Plummer, and R. W. Dutton,
AppI. Phvs. Lett. 62, 2498 (1993).
87 S. Mizuo, T. Kusaka, A. Shintani, M. Nanba, and H.
Higuchi, J. AppI. Phv. 54. 3860 (1983).
88 S. Mizuo and H. Higuchi, Japn, J, AppI. PhYS, 21, 281
(1982) .
89 P. Fahey, R. W. Dutton, and M. Moslehi, AppI, Phvs. Lett.
43, 683 (1983).
90 P. Fahey, G. Barbuscia, M. Moslehi, and R. W. Dutton,
AppI. Phvs. Lett. 46. 784 (1985).
91 M. Witter and K. N. Tu, Phvs. Rev. B 29. 2010 (1984).
92 P. Fahey and M. Wittmer, Mat. Res. Soc. Svmp. Proc. 163.
529 (1990).
93 D. S. Wen, P. L. Smith, C. M. Osburn, and G. A. Rozgonyi,
AppI. Phvs, Lett. 51, 1182 (1987).
94 F. M. d'Heurle and P. Gas, J. Mater. Res. 1. 205 (1986).
95 C. M. Osburn, J. Electrochem. Soc. 19. 67 (1990).
96 H. Jiang, C. M. Osburn, P. Smith, Z.-G. Xiao, G. McGuire,
and G. A. Rozgonyi, J. Electrochem. Soc. 139. 211 (1992).
97 N. R. Wu, D. K. Sadana, and J. Washburn, AppI. Phvs.
Lett. 44. 782 (1984).
98 P. A. Stolk, H. J. Gossmann, D. J. Eaglesham, D. C.
Jacobson, J. M. Poate, and H. S. Luftman, AppI. Phvs.
Let.iL,-66, 568 (1995).
99 R. Angelucci, G. Celotti, D. Nobili, and S. Solmi, J.
Electrochem. Soc. 132. 2726 (1985).
100 C. van Opdorp, L. J. van IJzendoorn, C. W. Fredriksz, and
D. J. Gravesteijn, J. AppI. Phvs. 72. 4047 (1992).
101 R. Angelucci, P. Negrini, and S. Solmi, AppI. Phvs. Lett.
4# 1468 (1986).
102 M. D. Giles, AppI. Phvs. Lett. 62. 1940 (1993).
103N. E. B. Cowern, G. F. A. van de Walle, P. C. Zalm, and
D. W. E. Vandenhoudt, AppI. Phvs. Lett. 65. 2981 (1994) .

169
104 G. Hobler and S. Selberherr, IEEE Trans, on Comp. Aid.
Des. 7. 174 (1988).
105 J. K. Listebarger, K. S. Jones, and J. A. Slinkman J.
AppI. Phvs. 73. 4815 (1993).
106 S. Prussin, J. AppI. Phvs. 43. 2850 (1972).
107 C. M. Drum and W. van Gelder, J. AppI. Phvs. 43. 4465
(1972) .
108 O. L. Krivanek and D. M. Maher, AppI. Phvs. Lett. 32. 451
(1978) .
109 H. L. Meng, S. Prussin, M. E. Law, and K. S. Jones, J.
AppI. Phvs. 73. 955 (1993).
110 M. Servidori, Z. Sourek, and S. Solmi, J. AppI. Phvs. 62.
1723 (1987).
111 H. Chao, P. Griffin, and J. Plummer, AppI. Phvs. Lett.
M# 3570 (1996).
112 D. J. Eaglesham, P. A. Stolk, H.-J. Gossman, and J. M.
Poate, AppI. Phvs. Lett. 65. 2305 (1994).
113 P. A. Stolk, H.-J. Gossmann, D. J. Eaglesham, D. C.
Jacobson, H. S. Luftman, and J. M. Poate, Mat. Res. Soc.
Sypip. Prog. 354 (1995) .
114 L. H. Zhang, K. S. Jones, P. H. Chi, and D. S. Simons,
AppI. Phvs. Lett. 67. 2025 (1995).
115 Encyclopedia of Materials Characterization (Butterworth-
Heinemann, Stoneham, MA 02180, 1992).
116 Y. M. Kim, G. Q. Lo, H. Kinoshita, D. L. Kwong, H. H.
Tseng, and R. Hance, J. Electrochem. Soc. 138. 1122
(1991).
117 H. Kinoshita, G. Q. Lo, and D. L. Kwong, J. Electrochem.
Soc. 140(1). 248 (1993).
118 H. U. Jager, J. AppI. Phvs. 78. 176 (1995).
119 M. Tamura, N. Natsuaki, Y. Wada, and E. Mitani, Nucl.
Inst. Meth. B 21. 438 (1987).
120J. F. Gibbons, Proc. IEEE 60. 1062 (1972).

170
121 V. Raineri, R. J. Schreutelkamp, F. W. Saris, R. E. Kaim,
and K. T. F. Janssen, NIMPR B59/60. 1056 (1991).
122 R. J. Schreutelkamp, J. S. Custer, J. R. Liefting, W. X.
Lu, and F. W. Saris, Mat. Sci. Reports 6. 275 (1991).
123 N. E. B. Cowern, A. Cacciato, J. S. Custer, F. W. Saris,
and W. Vandervorst, AdpI. Phvs. Lett. 68. 1150 (1996).
124 M. D. Giles, J. Electrochem. Soc. 138. 1160-1165 (1991).
125 D. J. Eaglesham, P. A. Stolk, H. J. Gossmann, T. E.
Haynes, and J. M. Poate, Nucl. Inst. Meth. B 106. 191
(1995) .
126 A. E. Michel, W. Rausch, P. A. Ronsheim, and R. H. Kastl,
AppI. Phvs. Lett. 50. 416 (1987).
127 R. B. Fair, J. of Electrochem. Soc. 137. 667 (1990) .
128 N. E. B. Cowern, K. T. F. Janssen, and H. F. F. Jos, J.
Appl. Phvs. 68, 6191 (1990).
129 R. B. Fair, in Impurity Doping Processes in Silicon,
edited by F. F. Y. Wang (North Holland, New York, 1981),
p.968, 315.
130 P. Packan, Ph.D. Dissertation. Stanford University
(1989).
131 K. S. Jones, V. Krishnamoorthy, L. H. Zhang, M. Law, D.
S. Simons, P. H. Chi, L. Rubin, and R. G. Elliman, AppI.
Phvs. Lett. 68. 2672 (1996).
132 S. Solmi, F. Baruffaldi, and R. Canteri, J. AppI. Phvs.
£2, 2135 (1991).
133 P. B. Griffin, R. F. Lever, P. A. Packan, and J. D.
Plummer, Appl. Phvs. Lett. 64. 1242 (1994).
134 C. S. Rafferty, G. H. Gilmer, M. Jaraiz, D. Eaglesham,
and H.-J. Gossmann, AppI. Phvs. Lett. 68. 2395 (1996).
135H. S. Chao, C. S. Rafferty, P. B. Griffin, and J. D.
Plummer, unpublished paper.

BIOGRAPHICAL SKETCH
The author was born on September 30, 1968, in Shenyang,
China. She entered Wuhan University in China at the age of
17 and obtained a BS in Physics degree with honors in July
1989. She came to the United States in August 1991 as a
graduate student in the Physics Department at Tulane
University, New Orleans. In July 1992 she transferred to the
Department of Materials Science and Engineering at the
University of Florida, specializing in electronic materials
under the supervision of Dr. Kevin Jones. She received her
MS degree in August 1994 and will complete her Ph.D. study in
August 1996.
171

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
KevinJones?Chair
Associate Professor of
Materials Science and
Engineeing
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Professor of Materials
Science and Engineeing
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Science and Engineeing
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
a dissertation for the degree of Doctor of Philosophy.
fey V1
Rajiv Singh
Associate Professor of
Materials Science and
Engineeing
as

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Mark Law
Associate Professor of
Electrical Engineering
This dissertation was submitted to the Graduate Faculty
of the College of Engineering and to the Graduate School and
was accepted as partial fulfillment of the requirements for
the degree of Doctor of Philosophy.
December, 1996
Winfred M. Phillips
Dean, College of Engineering
Karen A. Holbrook
Dean, Graduate School

LD
1780
199/
.L7S35
UNIVERSITY OF FLORIDA
3 1262 08554 9607



37
Rp. Compared with the SIMS profiles which will be shown
later in Chapter 4, the depth of the defect layer was found
to correspond to Rp. Both the distance between the defect
layer and the surface and the width of the defect layer
decrease as implant energy decreases. This decrease of
distance between the damage region and the surface may induce
a more prominent recombination effect at the surface, as will
be discussed later.
2.4 Criteria for the formation of (311) Defects and Sub-
amorohization Loops
Various suggestions have been made for the criterion for
secondary defect formation. Tamura et al.119 have stated that
implant dose can be used as such a criterion. They studied
defect formation in 1~2 MeV B, P and As implanted silicon and
found that secondary defects form if the dose range of
2xlO-*-3~lxlC)14 cm2 is exceeded, independent of the implant
species. Other experiments, however, show that a higher dose
is required for defect formation if the ion mass is smaller.
Gibbons120 studied B, Al, N, P, Sb, Ne and Si implanted
silicon at energies from 40 keV to 100 keV. The threshold
for small loop formation for Sb implant is about 5x10-1-3 cm~2
while for B implant it is above lxlO14 cm2. This implies
that the implant dose cannot be the sole criterion for loop
formation. Furthermore, the results of studies on channeling
versus random implants show that the defect formation is not
dependent only on impurity dose but on the damage introduced


55
that in the vicinity of small defects. As a result,
interstitials are emitted from the small defects and move
toward the large ones. The large defects then absorb these
interstitials and grow at the expense of the shrinkage of the
small defects. This is the so-called Ostwald ripening
process. Meanwhile, since the interstitial concentration in
the area surrounding a defect is higher than that in the
background environment, interstitials tend to leave the
defect and move into the bulk. As a result, defect shrink
and system approaches equilibrium condition. This is the
defect dissolution process. During the Ostwald ripening
process, defect density is expected to decrease, defect size
is expected to increase and the total number of interstitials
trapped in the defects is expected to remain as a constant.
On the other hand, during the dissolution process, all three
quantities are expected to decrease. In our experiment, we
observed a decreasing interstitial concentration, an
increasing defect size and a decreasing defect density with
increasing annealing time. This indicates an Ostwald
ripening process along with dissolution process. The
distribution plots for longer time anneals (e.g., 800C 13
min) show considerable scatter rather than a regular
distribution because the total number of defects is small and
the counting results are not as representative as those of
the short time anneals.
The driving force for the Ostwald ripening process is
the self-energy difference between the defects with different


65
so that /Db* is higher. It requires longer time for the
system to reach equilibrium at a lower temperature. However,
the time point at which the equilibrium is reached cannot be
read directly from the /Db* graph mainly because they are
the time-averaged values. This time will be determined from
the diffusion length vs. time plot as will be presented in
the following section, from which we will derive the
activation energy for the diffusion process.
3.3.3 The Activation Eneren/ for the Diffusion Saturation
Process
The diffusion length can be calculated as \(t) from
the diffusivity enhancement, the intrinsic diffusivity Db*
and the annealing time at each temperature. Figure 3.10
summarizes the results. The dots in Fig. 3.10 are the actual
data points and the lines are the fitting lines in the form
of aV ( 1-exp (-t/T)) where a and T are constants and t is the
annealing time. This is obtained by assuming the
instantaneous value of Db is proportional to exp(-t/T), and
the diffusion length L of boron is the square root of the
time integration of Db, i.e.,
I
DB{t)~ex (3.9)
L = Jt=i]j'DB(t,)dtl
r-r (3-10)
=^[0 'i, -Vd-i >
The diffusion length increases initially with time and then
plateaus out. The time till the "knee" region, i.e., just
before it plateaus out is 2T. If we choose 2T as the time


QJum
-j
o
Figure 3.1.
(Continued)


Diffusion Length (t)0-5 ()
B
150
Figure 5.10. Boron diffusion length vs. annealing time for
different implant doses.


dedicated to my parents


3
diffusion, and to provide an experimental basis for the
improvements of process development, process simulators and
ion implanters. These improvements would be of paramount
importance to optimizing the device structure.
1.2 Background
1.2.1 Shallow Junctions
A cross section of a complementary metal oxide
semiconductor (CMOS) structure is shown in Fig. 1.1. The
performance of this device is controlled by the gate
electrode, the gate dielectric, the channel and well region
and the source/drain junctions.
Gate
Figure 1.1. Schematic of MOS transistor showing how
interstitials are released from source/drain and LDD implants
and affect channel voltage.


98
is much higher than that for larger depth and as a result the
boron peaks near the implant damage region broaden much more
than the deeper boron peaks. As time increases, this
difference becomes less marked since the longer time allows
for the excess interstitials to diffuse to deeper regions.
For the 3min anneal, the overall trend shows /Ci*
decreasing with increasing depth and increasing with
increasing implant energy. For the 2 hr anneal, the depth
and energy dependencies of /Cj* become much less
prominent except for the lowest energy, 5 keV. The reason
for the much shorter duration in TED and a less amount of
diffusion enhancement at 5 keV may be related to several
factors, including the source of the interstitials, as will
be discussed later.
As we discussed in the first section of this chapter,
there are no {311} defects in the 5 keV implant but they are
the only type of visible defects in the 10 and 20 keV
implants and comprise the majority of the defects in 30 and
40 keV implants. It is important to note that although there
are no {311} defects in the 5 keV implant, TED of boron does
occur. For this 5 keV sample annealed at 750C, for times of
3 min to 2 hr, the diffusivity enhancement reduces from 980
to 17 for the second shallowest spike and from 100 to 11 for
the deepest one. We believe TED occurs because interstitials
are released from mobile boron interstitial pairs during
annealing. Because of the relative proximity of the implant
damage to the surface at this low energy, a large part of the


{311} Defect Density (cm 2)
143
Annealing Time (sec)
Figure 5.5. {311} defect density as a function of annealing
time in 20keV B+ implanted Si for different doses.


49
possibility that there was more than one storage mechanism
for excess interstitials that contribute to TED.
The purpose of this section is to discuss the effect of
annealing on defect behavior and TED in order to investigate
the correlation between the two.
3.2 Effect of Annealing on Defect Behavior
3.2.1 Experimental Procedure
In order to study the dissolution kinetics of {311}
defects from B+ implantation without the complication of
stable loop formation, we picked one sample (20 keV
2x10-*-^ cm2) from the implant matrix discussed in Chapter 2
(refer to section 2.2). Subsequent furnace anneals were
performed in a nitrogen ambient using fast pushes and pulls.
The annealing temperatures and times were chosen as 650C for
2 hr, 6 hr, 14 hr, 24 hr, 36 hr, 48 hr; 700C for 30 min, 1
hr, 2 hr, 4 hr, 8 hr; 750C for 5 min, 20 min, 30 min, 1 hr,
1.5 hr and 800C for 3 min, 5 min, 8 min, 10 min, 13 min.
PTEM micrographs were taken under weak beam dark field
imaging condition using a g220 reflection. The density and
length of {311} rod-like defects were measured directly from
the TEM micrographs. The areal interstitial density bound by
{311} defects after each anneal was calculated by multiplying
the defect length by the density of interstitials per unit
length. The time constant for the interstitial decay process
at each temperature was plotted in an Arrhenius graph as a


156
interstitial supersaturation level Ie/I* in eq. (6.2), is
related to the defect size (due to the fact that larger
defects have larger Efc,), and they observed that various doses
and energies all show similar enhancement within t < 2 min.
It thus appears that the defect size is independent of the
implant conditions, only their number being different. This
suggests a homogeneous nucleation mechanism, rather than a
mechanism which is limited by the number of available
nucleation sites. Our results on {311} defect size variation
with implant energy and dose can not directly prove this
model since we used furnace anneals with thermal budgets
large enough to pass the nucleation stage of the {311}
defects.
Regarding the interstitial storage and release during
annealing, the following model is suggested based on our
observations. At very low energies and doses where no {311}
defects or dislocation loops are formed, interstitials are
stored in the immobile boron clusters if the implant profile
peak concentration is above a critical value (Cenh>ni), or in
the mobile boron interstitial pairs. During annealing, the
clustered peak with B concentration greater than Cenh is
static and electrically inactive, while the B atoms paired
with interstitials show enhanced diffusion in the profile
tail region with concentration less than Cenh- Interstitials
are released from the B-I pairs once the B atoms return to
substitutional sites. The first saturation of TED occurs
when the interstitial supersaturation level reaches


99
interstitials could be attracted to the surface so that the
supersaturation of remaining interstitials is relatively low,
however these remaining interstitials can still diffuse into
the bulk and enhance dopant diffusion. Zhang et al.114 have
previously reported that TED was observed in 4keV lxlo!4cm-2
B+ implanted Si without the existence of {311} defects. They
attribute the source of interstitials driving TED to sub-
microscopic clusters. They determined the duration of TED by
noting when TED stops within the bounds of the SIMS
resolution, and their TED saturation was 3 min~13 min at
750C, much shorter than others have reported for TED
saturation in samples with {311} defects. In our experiment
we looked at the time averaged interstitial concentration
instead of the instantaneous one, so we did not actually
determine the time required for the interstitial
concentration to return to its equilibrium level. However,
we used the extracted Db for the 2600 A deep spike in the
5keV implant and simulated how Zhang et al.'s 4 keV
lxlO-*-^ cm2 implanted boron profile would evolve after 3 min,
15 min and 2 hr and found that TED ended after 15 min, in
agreement with their conclusion. Thus it is proposed that
when {311} defects exist, their dissolution process might
interact with that of mobile boron interstitial pairs, so
that TED continues for much longer periods of time. The type
of defects that supplies the interstitials regulates the
duration of TED.
From Fig. 4.10(c) we can see that, after an anneal of


73
Temperature (C)
Figure 3.3. Rate of {311} dissolution process and the
activation energy.


I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
KevinJones?Chair
Associate Professor of
Materials Science and
Engineeing
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Professor of Materials
Science and Engineeing
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Science and Engineeing
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
a dissertation for the degree of Doctor of Philosophy.
fey V1
Rajiv Singh
Associate Professor of
Materials Science and
Engineeing
as


82
10000/T (K)
Figure 3.8. The breaking point between the mobile profile
tail and the immobile profile peak Cenh as a function of
inverse temperature.


Defect Defect Defect Defect
Density (cm'2) Density (cm'2) Density (cm'2) Density (cm'2)
77
10
11
10
10
109
108
107
700C30min
1 1 i 1
i i i
60 260 460 660 860 1060 1260 1460
Min
1.,
700C60min
60 260 46(
) 660
860 1060 1260 1460
tail
All
LJ
700C4hr
60 260 46(
3 660
860 1
060 1260 1460
r!!!
700C8hr
Average {311} Size ()
(b)
Figure 3.6. (Continued)


168
86 D. J. Roth, Y. S. Huang, J. D. Plummer, and R. W. Dutton,
AppI. Phvs. Lett. 62, 2498 (1993).
87 S. Mizuo, T. Kusaka, A. Shintani, M. Nanba, and H.
Higuchi, J. AppI. Phv. 54. 3860 (1983).
88 S. Mizuo and H. Higuchi, Japn, J, AppI. PhYS, 21, 281
(1982) .
89 P. Fahey, R. W. Dutton, and M. Moslehi, AppI, Phvs. Lett.
43, 683 (1983).
90 P. Fahey, G. Barbuscia, M. Moslehi, and R. W. Dutton,
AppI. Phvs. Lett. 46. 784 (1985).
91 M. Witter and K. N. Tu, Phvs. Rev. B 29. 2010 (1984).
92 P. Fahey and M. Wittmer, Mat. Res. Soc. Svmp. Proc. 163.
529 (1990).
93 D. S. Wen, P. L. Smith, C. M. Osburn, and G. A. Rozgonyi,
AppI. Phvs, Lett. 51, 1182 (1987).
94 F. M. d'Heurle and P. Gas, J. Mater. Res. 1. 205 (1986).
95 C. M. Osburn, J. Electrochem. Soc. 19. 67 (1990).
96 H. Jiang, C. M. Osburn, P. Smith, Z.-G. Xiao, G. McGuire,
and G. A. Rozgonyi, J. Electrochem. Soc. 139. 211 (1992).
97 N. R. Wu, D. K. Sadana, and J. Washburn, AppI. Phvs.
Lett. 44. 782 (1984).
98 P. A. Stolk, H. J. Gossmann, D. J. Eaglesham, D. C.
Jacobson, J. M. Poate, and H. S. Luftman, AppI. Phvs.
Let.iL,-66, 568 (1995).
99 R. Angelucci, G. Celotti, D. Nobili, and S. Solmi, J.
Electrochem. Soc. 132. 2726 (1985).
100 C. van Opdorp, L. J. van IJzendoorn, C. W. Fredriksz, and
D. J. Gravesteijn, J. AppI. Phvs. 72. 4047 (1992).
101 R. Angelucci, P. Negrini, and S. Solmi, AppI. Phvs. Lett.
4# 1468 (1986).
102 M. D. Giles, AppI. Phvs. Lett. 62. 1940 (1993).
103N. E. B. Cowern, G. F. A. van de Walle, P. C. Zalm, and
D. W. E. Vandenhoudt, AppI. Phvs. Lett. 65. 2981 (1994) .