I. Heteroepitaxy on Si. II. Ion implanta

I. Heteroepitaxy on Si. II. Ion implantation in Si and heterostructures - CaltechTHESIS
CaltechTHESIS
A Caltech Library Service
About
Browse
Deposit an Item
Instructions for Students
I. Heteroepitaxy on Si. II. Ion implantation in Si and heterostructures
Citation
Bai, Gang
(1991)
I. Heteroepitaxy on Si. II. Ion implantation in Si and heterostructures.
Dissertation (Ph.D.), California Institute of Technology.
doi:10.7907/v8kr-gm23.
Abstract
The themes of this thesis, heteroepitaxy and ion implantation, are two areas that have been very actively researched in the last two decades.

Heterostructures made of III-V compound semiconductors by MBE and OMVPE have been used extensively in the fabrication of optoelectronics devices such as high-speed transistors and semiconductor lasers. Heterostructures on Si, which is the focus of part I of this thesis, have the advantage of compatibility with Si-based VLSI and promise to have impact on the microelectronics industry. Studies on the structural, elastic, thermal, and electrical properties of heteroepitaxial CoSi2, ReSi2, and GeSi films grown on Si constitute the backbone of this thesis. Some new characteristics of heterostructures were discovered as a result of this investigation. Among them are the observation and modeling of misorientation effects on an epitaxial film grown on a vicinal substrate; the misorientation induced by interfacial misfit dislocation arrays; the experimental measurements and phenomenological analysis of thermal strain, dislocation generation, and strain relaxation; and illustrative measurements of elastic, thermal, and structural properties of epitaxial films.

Ion implantation is an important process in the fabrication of integrated circuits. The second part of this thesis deals with the production and annealing of damage produced by ion implantation in semiconductors. The defect production, stability, microstructure, and the induced strain in implanted bulk Si crystals were quantitatively investigated as a function of ion species, dose, and implantation temperature. Many new features, such as the rapid rise of damage near the amorphization threshold, the correlation between the strain and defect concentration, and the scaling behavior of the damage with ion species and implantation temperature, are revealed.

The last chapter concerns the effects of ion implantation in CoSi2, ReSi2, and GeSi/Si heterostrcutures which is a marriage of heteroepitaxial and of ion implantation studies. Some interesting phenomena, such as the selective damage of the film and the substrate, the superposition of the intrinsic and the induced strain, are observed, and some preliminary results are obtained. Many interesting questions remain, and there are great research opportunities in this relatively unexplored area.
Item Type:
Thesis (Dissertation (Ph.D.))
Degree Grantor:
California Institute of Technology
Division:
Engineering and Applied Science
Major Option:
Applied Physics
Thesis Availability:
Public (worldwide access)
Research Advisor(s):
Nicolet, Marc-Aurele (advisor)
Vreeland, Thad (co-advisor)
Thesis Committee:
Unknown, Unknown
Defense Date:
14 May 1991
Record Number:
CaltechETD:etd-06282007-105319
Persistent URL:
DOI:
10.7907/v8kr-gm23
Default Usage Policy:
No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:
2762
Collection:
CaltechTHESIS
Deposited By:
Imported from ETD-db
Deposited On:
20 Jul 2007
Last Modified:
24 May 2024 18:22
Thesis Files
Preview
PDF (Bai_g_1991.pdf)
- Final Version
See Usage Policy.
7MB
Repository Staff Only:
item control page
CaltechTHESIS is powered by
EPrints 3.3
which is developed by the
School of Electronics and Computer Science
at the University of Southampton.
More information and software credits
I. HETEROEPITAXY ON SI

II. ION IMPLANTATION IN SI AND HETEROSTRUCTURES

Thesis by

Gang Bai

In Partial Fulfillment of the Requirements
for the Degree of
Doctor of Philosophy

California Institute of Technology

Pasadena, California,

1991
(Submitted May 14, 1991)

To my parents

iil
ACKNOWLEDGEMENTS

First and foremost, I wish to express my sincere gratitute to Professor Marc-A. Nicolet for
making my graduate study at Caltech a very rewarding experience. His support, encouragement,
inspiration, and guidance are truly appreciated. J am also deeply indebted to Professor Thad
Vreeland, Jr., whose advice has been extremely valuable.

I would like to thank Professor K.L. Wang at UCLA and Professor J.E. Mahan at CSU. Part of
the work described here was completed by fruitful collaboration with their research groups. Special
thanks go to Dr. Y.C. Kao, C.H. Chern, Q. Ye, and Dr. V. Mii at UCLA, and Dr. K.M. Geib at
CSU, for their efficient cooperation.

Particular acknowledgements are due to Dr. D.N. Jamieson, Dr. K.H. Kim, Dr. S.J. Kim, A.
Venezia, C.J. Tsai, Dr. B. Paine, and Dr. A. Dormman at Caltech, Dr. J.L. Tandon at McDonnel
Douglas Aeronautics Company, and Dr. A. Venezia in Israel, who have also contributed to the
completion of this thesis through collaboration.

I am especially indebted to R. Gorris and B. Stevens for their expert technical assistance, and
to R. Sampley for her outstanding secretarial work.

I would also like to thank members of Professor Nicolet’s group, Dr. E. Ma, T. Workman, J.
Reid, Dr. Q.T. Vu, Dr. E. Kolawa, J.S. Chen, Dr. S. Nieh, WS. Liu, Dr. C.K. Kwok, Dr. P.
Pokela, Dr. J.M. Pereda, Dr. Y.T. Cheng, Dr. F. So, E. Pan, B. Flick, and Dr. X.A. Zhao, for their
generous help and friendship.

Special thanks are due to Dr. A.E. White at Bell Labs, Dr. S. Akbar at IBM, Dr. D.E. Holmes
at. Rockwell, for providing some of the samples. Financial support of this work has been provided

by the Semiconductor Research Corporation and the National Science Foundation.

PREFACE

The themes of this thesis, heteroepitaxy and ion implantation, are two areas that have been
very actively researched in the last two decades.

Heterostructures made of III-V compound semiconductors by MBE and OMVPE have been
used extensively in the fabrication of optoelectronics devices such as high-speed transistors and
semiconductor lasers. Heterostructures on Si, which is the focus of part I of this thesis, have the
advantage of compatibity with Si-based VLSI and promise to have impact on the microelectronics
industry. Studies on the structural, elastic, thermal, and electrical properties of heteroepitaxial
CoSi2, ReSiz, and GeSi films grown on Si constitute the backbone of this thesis. Some new charac-
teristics of heterosetructures were discovered as a result of this investigation. Among them are the
observation and modeling of misorientation effects on an epitaxial film grown on a vicinal substrate;
the misorientation induced by interfacial misfit dislocation arrays; the experimental measurements
and phenomenological analysis of thermal strain, dislocation generation, and strain relaxation; and
illustrative measurements of elastic, thermal, and structural properties of epitaxial films.

Ion implantation is an important process in the fabrication of integrated circuits. The second
part of this thesis deals with the production and annealing of damage produced by ion implanta-
tion in semiconductors. The defect production, stability, microstructure, and the induced strain in
implanted bulk Si crystals were quantitatively investigated as a function of ion species, dose, and
implantation temperature. Many new features, such as the rapid rise of damage near the amorphiza-
tion threshold, the correlation between the strain and defect concentration, and the scaling behavior
of the damage with ion species and implantation temperature, are revealed.

The last chapter concerns the effects of ion implantation in CoSiz, ReSi2, and GeSi/Si hetero-
strcutures, which is a marriage of heteroepitaxial and of ion implantation studies. Some interesting
phenomena, such as the selective damage of the film and the substrate, the superposition of the
intrinsic and the induced strain, are observed, and some preliminary results are obtained. Many
interesting questions remain, and there are great research opportunities in this relatively unexplored

area.

CONTENTS
ACKNOWLEDGEMENTS... 00. n ene eee teens en en ene cues iii
PREFACE . 2000 cc cen en een e ene n een denen bette en beeeens iv
CONTENTS. 2.0... n dene eben e tenet seb benteteenensnrnnnenes Vv
LIST OF TABLES 2.0... ccc eee n nee been eee eeeeeneenbeeneeees vii
LIST OF FIGURES... cn ten en eben eee n cece neces beer enenvenens viii

Part I Heteroepitaxy on Si

Chapter 1 Heteroepitaxy on Si...... 6... cece cence eet e een neceennenannnes 2
Chapter 2 Epitaxial CoSig Films on Si......... 6... ccc ccc eee e nen ne ence 9
2.1 Introduction 2.00. nen een eee e nen beeen en tees enbeveanaennes 9
2.2 Growth and characterization of CoSig on Si(111)... 0... eee cece cence eee eee anes 10
2.3 Growth and characterization of CoSig on vicinal Si(111) ......... 0. cece cece cece cee eee 17
2.4 Critical thickness and strain relaxation .......0.0.0 00.0 cece eee ce cee cee eect eeeteneenens 22
2.5 Thermal strain and its inference for pseudomorphic growth ...............00ececeseeeee 25
2.6 Elastic and thermal properties of mesotaxial CoSig films on Si.................. ccc eee 31
Chapter 3 Epitaxial ReSiy Films on S1........0.... 00.0. cece cece cece ene ee nee veneues 43
3.1 Semiconducting silicides .... 00... cee cece ne ne erence ene tence bene vaees 43
3.2 Growth and characterization of ReSig on Si(100)..... 0... eee cence ce ee eens 44
3.3 Channeling of MeV ions in polyatomic crystals ...........0 00. c ccc ccc c cece eee eeeeneees 49
Chapter 4 GeSi Films on Si(100) 20.0.0. c ccc cee nee eet e ene aeeaes 59
4.1 Introduction 20.0.0... cnn enn ened ede nese ree b en tebe etn tannins 59
4.2 Properties of GeSi films grown at 550°C... 0. cece eee e ene ent e eee enenes 59
4.3. Pseudomorphic GeSi films and superlattices on vicinal Si(100) ....................00005 66
4.4 Strain relaxation of pseudomorphic GeSi films grown at 300°C ....... 00... cee cece 73
4.5 Asymmetrical tilt boundary in GeSi/Si(001) heterostructures ..........0.0 cece eee eee 79
Chapter 5 Porous Si and Its Properties............ 0.0.00. c cc cece ccc cece een eeeeeeeees 86
5.1 Introduction 2.0.00. cn ene eee ee eee eee eee b sb been enenenens 86
5.2 Strain and stress in porous Si(100) 0.0... cee cee recente ee ne ee eae teen eenes 87

5.3 Epitaxial films of GeSi and CoSiz on porous Si... 2... ec ecc ee eee cent ences 96

Part II Ion Implantation in Si and Heterostructures

Chapter 6 Ion-Solid Interaction ........0.0 0.00... e tence nee ceetnneneees 102
Chapter 7 Damage Production and Annealing in Si(100)......................... 105
7.1 Introduction .. 0... cece ccc n ene ence eee b ee tneeecueneaeeetesuntaenes 105
7.2 Damage by self-implantation at room temperature.......... 0... ccc cece ccceceuceeeeues 106
7.3 Damage by '°F, *°Ar, and 13!Xe implantation at room temperature................05. 118
7.4 Damage by ion implantation at liquid nitrogen temperature............0.0ccecceeceees 124
Chapter 8 Ion Implantation in Heterostructures...........00.0 0.00. ccc ccc neseeeecues 130
8.1 Introduction 0.00... e neces eee ee eeseseeeeeevaentntnbanntnens 130
8.2 Damage production and annealing in ?°Si implanted CoSia films...............-.0005. 131
8.3 Radiation damage in ReSig by an MeV *He beam.............0.. 0. ccc ce ceeeeceeeees 141
8.4 Amorphization and recrystallization of epitaxial ReSig films..............0ccecececeeee 148
8.5 Strain induced by ion implantation in GeSi films............0.00 000 cc cece ce ceueeeeeees 154
Appendix I Some Studies of Compound Semiconductors ..............0..0c.0 eee 163

Appendix IT List of Publications ...........0.00.0 00 ccc cence ccc ce cece eee eneneenens 165

vii
LIST OF TABLES

Table 2.6-I Lattice distortion of CoSig films on (100) and (111) oriented Si substrates. Data for
Si/CoSi2/Si samples are from Ref. 49 of Chapter 2 and those for B-type sample are from Ref. 7 of
Chapter 2.0 ccc enn ene bea nee n etn tet b eee eeentneneneeees 33

Table 2.6-II Ratios and elastic constants (in units of GPa) of cubic CoSig from strain and curvature

measurements. Data for Si are from Ref. 51 of Chapter 2 and are listed for comparison. ...... 33

Table 3.3-I Calculated values of the characteristic and critical angles, measured values of half
angles, and the calculated and measured minimum yields, for the [100] axial channeling of 1.4 MeV

4He beam in an epitaxial ReSig film. 2.2.0.0... 0 cece cceeeeceececeeueveceueueceeerannenns 54

Table 4.3-I The strain obtained from rocking curve analyses and the biaxial stress model of the
single layer Ge,Si;_, films, and the superlattice with the average Ge composition of 0.07, grown on

vicinal Si(100) substrates at 550°C. 60. cece nee tet e teen e nee een envananenes 72

vill
LIST OF FIGURES

FIG. 1 An artist’s view of total integration: a multiple heterostructure where the logic is fabricated
on Si, while all other functions are designed on the optimum choice of materials epitaxially grown

on a Si substrate (from Fig. 1 of Ref. 6 of Chapter 1). 000... cece ccc ec c eee ence ent e eens 5

FIG. 2.2-1 2 MeV *He: (a) backscattering spectra of a CoSi2/Si(111) heterostructure for a beam
incident along a random (solid line) and the [111] axial direction (dotted line); (b) an angular scan
about the [111] axis of the same sample; the Si and Co signal in the film were taken at 1.53 and 1.13

MeV, respectively, and the substrate signal was taken at 1.05 MeV. ....... cc. cece eee ee eee 12

FIG. 2.2-2 A Fe Kg, x-ray (A = 0.1936 nm) rocking curve from the (111) symmetrical diffraction

of the same CoSi2/Si(111) sample shown in Fig. 2-2-1. oo. cece eect net e eee e nets 14

FIG. 2.2-3 Transmission electron micrographs of CoSi2/Si(111) samples: (a) a cross-sectional high
resolution lattice image of the interface (from Ref. 12 of Chapter 2); (b) a weak-beam dark-field
plane-view ({022] beam) micrograph showing a hexagonal misfit dislocation network (from Ref. 24
of Chapter, 2); (c) a plane-view morié pattern caused by the lattice mismatch between the layer and

the substrate (from Ref. 12 of Chapter 2). 0.2.0... e cc e eee e eee eee Se ceeeeeee eens 16

FIG. 2.3-1 Fe Kg, x-ray (111) symmetrical diffraction from a 100 nm thick CoSiz film on a vicinal
Si(111) substrate (offset angle 6, = 16°): (a) diffraction geometry, n is the surface normal; (b) the
Bragg peak position of the substrate versus the azimuthal angle; (c) the separation between the

Bragg peaks of the film and the substrate versus the azimuthal angle. ...............0.0eeee es 18

FIG. 2.3-2 The misorientation angle a of (a) the CoSi2/Si(111) samples of various film thickness
versus the substrate offset angle ¢,; (b) the sample shown in Fig. 2.3-1 (¢, = 16°, 100 nm thick)

versus the perpendicular strain. 0... 1. ec cece nee n ene nee ee ent n ent e enn eeeenenes 20

FIG. 2.3-3 (a) Schematical diagram of the interface between a CoSiz film and a vicinal $i(111)
substrate along the [112] tilt direction, and the proposed geometrical model on the misorientation
between the film and the substrate; (b) the geometrical model that predicts a = «+ tang, agrees
excellently with the least-squares fit to the experimental data of epitaxial CoSi films grown on

vicinal Si(111) substrates. 0.000. n eee c ene t nnn e tenn eee e aes 21

FIG. 2.4-1 Perpendicular strain in an epitaxial CoSig film as a function of film thickness for a

CoSi2/Si(111) heterostructure. The solid line is the strain relaxation behavior predicted by Matthews

and Blakeslee’s minimum energy model. The dashed line is the average of the experimentally mea-

sured strain value on samples of various film thickness, substrate offset angle, and growth condition.

FIG. 2.5-1 Perpendicular x-ray strain e+ as a function of sample temperature T for two 100 nm
thick CoSi films on (i) a Si substrate whose surface is offset from the [111] direction by 16° towards
the [110] direction of the substrate (e), and (ii) a Si substrate whose surface is aligned with the
(111) planes (Av). All strain values are reversible below 490°C. When the latter sample was
heated above that temperature up to 650°C and then cooled, a reduced strain was measured (7),
again reversible below 500°C. The solid lines are linear fits to the data (omitting the two highest
temperature points). The strain for a coherent interface is calculated with the Poisson ratio v = 1/3
(lower dashed line). At the temperature at which the samples were grown (~ 600°C), the measured

strain is that predicted for an elastically relaxed film (upper dashed line). ................0005 27

FIG. 2.5-2 Three x-ray rocking curves diffracted from symmetrical (111) planes measured at room
temperature on sample #1 as-grown, vacuum-annealed, and air-annealed. The curve of the as-grown
sample is indistinguishable from that measured after annealing in vacuum at 700°C for 40 min (solid
line). Annealing in air at 650°C induces an irreversible reduction of the strain in the epitaxial CoSig

film as indicated by the shift and the broadening of the diffraction peak (dotted line). ........ 29

FIG, 2.6-1. Fe Kg, x-ray (A = 0.1932 nm) rocking curves of symmetrical (400) and asymmetrical
(311) diffractions from a mesotaxial CoSi2/Si(100) sample. The diffraction geometry and direction

of x-ray incidence are shown in the inset above the corresponding Bragg peaks from the CoSizg film.

FIG. 2.6-2. The lattice mismatch, «+ (square), ¢! (triangle), and f (circle), as a function of
the measurement temperature for both (a) the CoSi2/Si(100) and (b) the CoSip/Si(111) mesotaxial

samples. Open (filled) symbols are for the measurements when the temperature was raised (lowered).

FIG. 2.6-3. Schematics of the proposed model showing how an epitaxial CoSig film relaxes to an
equilibrium strain state at Tp and above, and that misfit dislocations are locked-in below Tr. Ty
is the melting temperature of CoSij, and To is the hypothetical temperature at which the lattice

mismatch between CoSig and Si becomes zero. ... occ ce cence cn ccenenceuaeeuenes 38

FIG. 3.2-1 2 MeV *He minimum channeling yields of Re and Si for ReSiz films grown on Si(100)

by reactive deposition epitaxy as a function of substrate temperature during deposition. ...... 45

FIG, 3.2-2 Epitaxial relationship between a ReSiz film and a Si(100) substrate: (a) transmission
electron diffraction pattern of the Si[{100] zone and the ReSi2[010] zone (from Ref. 14 of Chapter
3). The bright spots are Si diffractions. The faint spots are ReSi diffractions (note their fourfold
symmetry). (b) Schematical drawing of two equivalent common unit meshes (from Ref. 15 of
Chapter 3). The dots are Si atoms in the substrate. The crosses are Re atoms in the film. The

actual atomic positions are unknown and assumed... kee eee cece cent eeeneaes 47

FIG. 3.2-3 Transmission electron micrograph of a 150 nm thick ReSi, film grown on Si(100) at
650°C (from Ref. 15 of Chapter 3): (a) plane-view dark-field image with a beam diffraction vector
of ReSi.(002); (b) cross-sectional bright-field image with a beam incident along Si{022]. ....... 48

FIG. 3.3-1 Backscattering spectra of 1.4 MeV ‘He incident along a random direction (solid line)

and a [100] aligned axial direction (dotted line), 6.0... ok cece cece ccc e ccc e eee eee eeteesenees 51

FIG. 3.3-2. A plot of normalized backscattering yield versus tilt angle. The normalization is
performed with respect to the backscattering yield of a random incident beam. The half angle is |

the half-of the full width of the angular dip. oo... 6. ec c ccc cece eee cece teeevaeevaneees 51:
FIG. 3.3-3 A cross-sectional schematic diagram showing the [100] axial channel of ReSig. ....54

FIG. 4.2-1 The lattice mismatch between epitaxial Ge,Si;_, alloy and Si, f, as a function of Ge

composition. The data agree well with the Vegard law over the entire composition range (0 < x < 1).

FIG. 4.2-2 Ge composition—film thickness plot of epitaxial Ge,Si;_, films on Si(100): dotted line
is the equilibrium critical thickness from Matthews and Blakeslee’s model; A is the measured critical
thickness of films grown at ~ 550°C by Bean et al., and e is from our work. The number associated

with each datum point is the normalized parallel strain el/f. 00... 0.00 cc ccc eceeeeeeeeeee 61

FIG. 4.2-3 Channeling characteristics vs x-ray diffraction of epitaxial GeSi films grown on Si(100) at
550°C: (a) dechanneling probability of an aligned MeV *He beam across the interface as a function
of parallel strain ¢!! of the film; (b) the difference of the minimum channeling yield between a relaxed
and a coherent GeSi film as function of the square of the x-ray peak broadening caused by threading

dislocations in the film. oo... ccc cece cece teen cee eeeceeeeneeeuneetneeeeuneennees 64

FIG. 4.2-4 Thermal properties of a metastable coherent Ge,Si,_z film grown on Si(100) at 550°C:
(a) the lattice mismatch (dotted line) and the strains (solid lines) as a function of temperature;

dashed lines are for a film with rigid interface; (b) Fe K., x-ray rocking curves diffracted from (400)

planes of of sample before (solid line) and after (dotted line) thermal annealing at 630°C for 2h in

ambient air. 0. ccc cece eee eee eevee cee eeeeneseeauneeetteeeueneeetnnbeneess 65

FIG. 4.3-1 (a) The angular position of the (400) peak diffracted from the substrate and (b) the
difference of the peaks from the film and the substrate versus the azimuthal angle of the sample
configuration. (c) Schematics of the results from analysis of the x-ray rocking curve data (a) and

ces 67

FIG. 4.3-2 Schematics of one symmetrical (400) and four asymmetrical {311} diffracting planes of

vicinal samples. The surface normal is taken as Z-axis. 2.0.0... . cee cece cece cece eee eeaennaes 68

FIG. 4.3-3 (a) The mismatch in interplanar spacing and (b) the misorientation angle as a function
of diffracting planes of a Geo.o5Sio.95 film 100 nm thick on a vicinal Si(100). The solid line is the

prediction of the biaxial stress model with e** = «YY = 0, «72 = 0.52%, and ¢, =3.1°....... 70

FIG. 43-4 Fe Ky, (A = 0.1932 nm) x-ray rocking curves diffracted from the (400) symmetrical
planes of the GeSi superlattice on vicinal Si(100): (a) two diffraction geometries corresponding to

a rotation of 180° of the incident x-ray about the surface normal; (b) the corresponding diffraction

0042 eee eee 72

FIG. 4.4-1 The strain relaxation of a highly metastable GeSi strained layer grown on Si(100) at
~ 300°C as a function of temperature upon isochronal annealing for 30 min in vacuum. The strain

relaxes sharply at 370+ 25°C. oo c cee ec ene ene e eben teen teen ee eneneeeees 15

FIG. 4.4-2 Parallel strain ell of 570 nm thick epitaxial Geo.3Sig7 layers grown on Si(100) at
~ 300°C obtained from (400) diffraction x-ray rocking curve measurements vs duration of ex situ

thermal annealing in vacuum (~ 5 x 10~? Torr) at various temperatures. ................0000. 75

FIG. 4.4-3 Relaxation rate measured by the rate of increasing parallel strain with increasing time
at modest relaxation level vs the inverse of the temperature at which the relaxation proceeds (see

Fig. 4.4-2). The data follow an Arrhenius behavior with a slope of 2.140.2eV. ...........0.. 78

FIG. 4.4-4 X-ray peak broadening of (400) diffraction from 570 nm thick epitaxial Geo 3Sio.7 layers

grown at ~ 300°C caused by imperfections in the layers versus the strain relaxation. ......... 78

FIG. 4.5-1 Schematic representation of parallel strain and misorientation in a heterostructure.

FIG. 4.5-2 Four different types of dislocation arrangements at heterointerfaces: (I) mismatch

relieving dislocations; (II) asymmetrical low angle tilt boundary; (III) mixed dislocations with a

Xii

non-zero parallel strain and a zero misorientation angle; (IV) mixed dislocations with both non-zero

net mismatch-relieving and tilt components. 0.0.0.0... ccc c eee cece ee eee e cece netaeeeeas 80

FIG. 4.5-3 Schematic representation of a 60°-mixed dislocation in a GeSi/Si(001) structure.

FIG. 4.5-4 Misorientation angle vs parallel strain of ~ 570 nm thick Gey 3Sig.7 films grown on
Si(001) substrates at 300-500°C. Most samples (A) analyzed have zero misorientation angles re-
gardless of the value of the parallel strain. Samples B and C are initially pseudomorphic and have
a zero misorientation angle. They relax upon thermal annealing and develop finite misorientation
angles. Sample D is relaxed initially and has a non-zero misorientation angle. It relaxes further
upon annealing. After amorphization of the entire film and solid phase epitaxial regrowth, the

Misorientation becomes Zero. 6... eee cece cece ene ueneeenteennateneeeeuneetcneenanes 82

FIG. 5.2-1 Backscattering spectra of porous Si and bulk Si: (a) a 2 MeV *He beam incident along
a near-normal random direction (solid lines) and along the [100] aligned direction (dotted lines); (b)
3.05 MeV oxygen resonance of a bulk Si, of a porous Si after storage of 18 months in air at 23°C,

and of a thermally oxidized Si wafer (~ 1 wm SiO02/Si). 6... cee eee eee eee ete v eee e ness 88

FIG. 5.2-2 Auger electron spectroscopy depth profile of oxygen and silicon in porous Si of the

sample that had been stored in air at room temperature for ~ 18 months after the formation of

POFOUS Si. cence enter tenet e betes nese bbb ta bee ee sae eesbeevanteneusnns 90

FIG. 5.2-3 Perpendicular strain e+ of porous Si aged for 18 months in air at 23°C and then
annealed in vacuum for 30 min, measured immediately after annealing (0), after storage in air at

room temperature for about 1 month (¢) and 7 months (filled triangle), ................00005. 90

FIG. 5.2-4 Decreasing rate of perpendicular strain in porous Si upon vacuum annealing: (a) e+
versus annealing duration at 300, 400, and 600°C; (b) Arrhenius plot of the time constant 7 of the

decrease of the perpendicular strain. ..................00. Skee ence eee e cent eee neneeeennes 92

FIG. 5.2-5 Time evolution of the perpendicular strain e+ of the porous Si stored at room temper-
ature in different ambient gases after being annealed in vacuum at 500°C for 30 min (see the point

marked by star in Fig. 5.2-3). 0.0... ccc ccc ete e ee cence ne tenet ene etn eeeneees 93

FIG. 5.2-6 Schematic representation of porous Si and of the evolution of the perpendicular strain
as a function of the sample history given on the abscissa. The upwards (downwards) arrows in the

silicon rods indicate the positive (negative) perpendicular strain; their length suggests the magnitude

xiii

of the strain. The perpendicular strain evolves as a result of absorption by and their desorption

from the native oxide. . 6. ccccceceee ene n ee eee ee eee eee eee ee ee eb eebeeeeeeeeeees 93

FIG. 5.3-1 Illustration of a heteroepitaxial Ge film on a Si substrate of seed pads with lateral
dimension !. The strain energy profile, w(0,z), for the pad mid-cross section is shown on the left

(from Ref. 24 of Chapter 5). 0... ccc ccc een tence eben ene nee e ent eeneneees 97

FIG. 5.3-2 Schematic drawings of (a) a patterned substrate as used in Luryi and Suhir’s model

(Fig. 5.3-1), and (b) a porous Si substrate (from Ref. 26 of Chapter 5). «02... .. 0... cece ee eee 97

Fig. 5.3-3 Perpendicular strain e+ of CoSig-capped porous Si(111) aged for 18 months in air and
then annealed in vacuum for 30 min, measured immediately after annealing (0), and after storage
in air at room temperature for 1 month (e). The figure has identical scales as those of Fig. 5.2-3 for

COMPATISON, 6. een nn e nese denen tebe tee tet tbe een tees 97

FIG. 7.2-1 Fe Kg, x-ray (wavelength=0.1936 nm) rocking curves diffracted from the symmetrical
(400) planes of the Si(100) samples implanted at room temperature by 230 keV 78Si ions to the
doses of (a) 1x, (b) 3.5x, (c) 4.3x, (d) 4.8 x 10!4/em?. ccc cece cece n nee eeeee ees 107

FIG. 7.2-2 The maximum perpendicular strain (o) extracted from dynamical x-ray diffraction
simulations of the experimental rocking curves as a function of the ?8Si implantation dose. The solid
line is to stress the trend. The filled circles correspond to the samples for which the x-ray rocking

curves are shown in Fig. 72-1. oo ccc cnn en enn n nen e nen e eben eee eneeees 107

FIG. 7.2-3 2 MeV *He backscattering (filled triangle) and channeling spectra (solid line) of the set
of the samples shown in Fig. 7.2-1. Also plotted are the channeling spectra (dotted line) of a virgin

Si sample and a sample implanted 7.7 x 10!4 ?8Si/cm? in which a continuous amorphous layer forms.

FIG. 7.2-4 The maximum defect concentration extracted from channeling spectra such as those of
Fig. 7.2-3 as a function of the 8Si dose. The solid line is to highlight the trend. The filled circles
correspond to the samples (a-d) shown in Fig. 7.2-3. The dashed line is the maximum value in the
concentration profile of the Frenkel pair predicted by the TRIM88 simulation code of 230 keV ?8Si

implantation in an amorphous Si target. 0.0.0... ccc ccc ce nee e nent e tenet eee neneees 109

FIG. 7.2-5 The measured maximum defect concentration as a function of dose (0 of Fig. 7.2-4) is
compared with that predicted from the phenomenological model of the accelerated damage growth
in a predamaged crystal (solid line, see text). The dashed lines are the fraction of the amorphous

zones calculated from Gibbson’s overlap model with various (m,A;) parameters. ............. 112

xiv

FIG. 7.2-6 The depth profile of the Frenkel pair concentration from TRIM88 simulation (dashed
line), the defect concentration from the channeling measurements of the sample (c) (solid line), and
the perpendicular strain from the dynamical x-ray diffraction simulation of the rocking curve (dotted

line). The vertical scale is in an arbitrarily normalized unit. .......... 0. cece cee ee eee ences 112

FIG. 7.2-7 The relationship between the maximum values of the perpendicular strain from x-ray
diffraction measurements and the defect concentration from 2 MeV *He channeling measurements.

The solid line is the least-squares fit of the data (o) to a linear function. ..................0. 115

FIG. 7.2-8 The isochronal annealing characteristics of the perpendicular strain in the implanted
layers as a function of the annealing temperature. All annealings were performed in a vacuum of
~ 7x 107" Torr for a duration of 30 min. The data are from the four samples for which the x-ray

rocking curves are shown in Fig. 7.2-1 and the channeling spectra in Fig. 7.2-3. ............. 115

FIG. 7.3-1 The maximum defect concentration extracted from channeling spectra similar to those
of Fig. .7.2-3 as a function of the 19°F dose. The solid line is to highlight the trend. The dashed
line is the maximum value in the depth profile of the Frenkel pair concentration predicted by the

TRIM88 simulation of 230 keV 19F implantation into an amorphous Si target. ............. 119

FIG. 7.3-2 The relationship between the maximum values of the perpendicular strain from x-ray
diffraction measurements and the defect concentration from 2 MeV *He channeling measurements.

The solid line is the least-squares fit to the data (e) of a linear function. ...............00005 119

FIG. 7.3-3 The maximum perpendicular strain obtained by fitting the dynamical x-ray diffraction
simulations to the experimental rocking curves as a function of the implantation dose for four different

ions. The solid line is to stress the trend. «1... 0c ccc ccc cece cece cee c cent eceuneeunneevevens 121

FIG. 7.3-4 The initial (regime I) slope of the maximum perpendicular strain vs dose as a function

of the Frenkel pair concentration per unit dose for various incident ions. ................000. 121

FIG. 7.4-1 The maximum perpendicular strain extracted from the dynamical x-ray diffraction
simulations of experimental rocking curves as a function of the 250 keV *°Ar dose of implantations
at 100K (7) and at 300K (A) into Si(100). The solid line is to stress the trend. The modified data

points (filled inverse triangle) are obtained by multiplying the dose of the data V7 by a factor of 3.5.

Fig. 74-2 The cross-sectional TEM (from Ref. 40 of Chapter 7) of the sample implanted by 250
keV 5 x 10!5 #°Ar/cm? into Si(100) at 100K (e+, ~ 0.4%, see Fig. 7.4-1). A heavily damaged layer

is located from a depth of 120 nm to 240 nm... cece cece cen cee vaerneees 125

FIG. 8.2-1 2 MeV *He backscattering spectra with a beam incident along a random (e) and a
[111] axial channel orientation of the as-grown CoSi2(50 nm)/Si(111) (solid line); and the samples
implanted at room temperature by 150 keV 78Si to doses of 2x (V7), 5x (0), 30 x 10!4/cm? (A).

The detected He particles exit at an angle of 82° from the line of the incident beam. ....... 132

FIG. 8.2-2 The defect concentration in the CoSig films extracted from the channeling yields of Fig.
8.2-1 vs the Si dose, for the as-implanted samples (¢), and those annealed for 60 min at 400°C (0),
600°C (A), 800°C (square). The Frenkel pair concentration as a function of dose, predicted from

TRIM88, is also shown (dashed line). 02.0... eee eee erent e ene ne tne eee eeeeaes 132

FIG. 8.2-3 Fe Ka, x-ray rocking curves diffracted from the symmetrical (111) planes of (a) as-grown
CoSi2/Si(111); and the samples implanted to doses of (b) 0.5x, (c) 1x, (d) 2x, (e) 5 x 10!4/cm?.
Bragg peak from the bulk Si substrate is 0g = 18°. oo... ccc cece eee e ese n een eaens 135

FIG. 8.2-4 The static atomic displacement induced by the defects in the CoSig films vs the Si dose,
for the as-implanted sample (¢), and for those annealed in vacuum for 60 min at 250°C (v) and
400°C (0). The shaded area represents the error in estimating the displacement of a perfect CoSig

070 005 cc 135

FIG. 8.2-5 The resistivity difference between the implanted and as-grown CoSig films vs the dose,
for the as-implanted sample (¢), and for those annealed in vacuum for 60 min at 250°C (v7), 400°C
(0), 600°C (A), 800°C (square), 6. ee cee tenet t nett e este cnet teneeenes 139

FIG. 8.2-6 The resistivity difference vs the static displacement of the lightly damaged CoSiz films
(ep < 4% or ¢ < 5 x 10!*/cm?). The approximately linear relationship indicates a good correlation

between the concentration of the carrier scatterers and the structural defects in such films. ..139

FIG. 8.3-1 1.4 MeV *He backscattering and channeling spectra of a 150 nm thick epitaxial ReSiy
layer grown on a Si(100) substrate. All four spectra were taken at room temperature and are
plotted by normalizing incident doses to a common value. The solid line is the spectrum for random
incidence. The three [100] channeling spectra are for samples irradiated at room temperature with

doses of (a) ~ 10!4/cm?, ~ 10!7/cm? (b) in a [100] aligned direction and (c) in a random direction.

FIG. 8.3-2 The normalized backscattering yield of the Re signal versus angle of tilt between the

incident beam and the [100] direction of the sample for the three damage stages of Fig. 8.3-1.

Xvi

FIG. 8.3-3 The minimum channeling yields of the Si and Re signals for an epitaxial ReSi film as

a function of the 1.4 MeV *He irradiation dose for both a random and a [100] aligned incidence.

FIG. 8.3-4 Comparison of the measured defect concentration versus irradiation dose in ReSig
produced by 1.4 MeV ‘He ion beams of random (0) and [100] aligned (e) incidence at 300K, and

one calculated (dashed line) by a TRIM88 computer simulation of a beam of random incidence at

0 145

FIG. 8.4-1 X-ray diffraction spectra of epitaxial ReSi2/Si(100) samples: (a) as-grown; and implanted
by 300 keV 78Si to doses of (b) 10'9/cm?, (c) 10'4/cm?, and (d) 5 x 10!4/cm?. A Cu x-ray source

and 9-20 geometry were used. 1. nee teen cent net eens tneteeeeneenenenes 149

FIG. 8.4-2 Resistivity of the samples shown in Fig. 8.4-1, measured at temperatures ranging from

90K to 330K in the dark. .....cc ec cecceccccec cena een eeneesueunnesvenneunnnens 150

FIG. 8.4-3 Changes of the minimum yield for Re upon thermal annealing in vacuum at 500°C,

600°C and 700°C for 30 min for the samples shown in Fig 8.4-1. ......... 0c. cece eeecceeeeees 150

FIG. 8.5-1 X-ray rocking curves diffracted from the (400) symmetrical planes of the as-grown
Geo .09Sio.91/Si(100) sample (solid line), and of those implanted at room temperature with 320 keV
281 to 2x (0), 5x 104 /em? (0). cece cece cece eee e eet tees enbtteeenenenney 156

FIG. 8.5-2 The strain induced by 320 keV ?8Si implantation into pseudomorphic GeSi layers vs
the Ge composition. The dashed lines are the strains predicted from a linear interpolation model of

De CE) ce 156

FIG. 8.5-3 (400) x-ray rocking curves of a pseudomorphic Geo o9Sio.91/Si(100) implanted at room
temperature (RT) by 320 keV 5 x 10/4 78Si/cm? and annealed for 30 min at various temperatures.
The spectrum of the 700°C annealed sample (solid line) is indistinguishable from that of the as-grown

sample (Fig. 8.5-1). 0... 0... cece cece e cence nett eee ee Lene cence eee e eee eee tenn eeee ees 159

FIG. 8.5-4 The strain in the Ge,Si,_, layer vs the annealing temperature. The square, triangle,
and circle are for the samples with z = 0.04, 0.09, and 0.13, respectively. The filled symbols represent
the unimplanted samples. The small (big) open symbols represent the samples implanted by 320
keV 1.2 (2x) 104 8Sifem?. 0. cccecnecec nce en en ee ee se ennbeteceeaunnees 159

Part I

Heteroepitaxy on Si

Chapter 1 Heteroepitaxy on Si

Heteroepitaxy is oriented overgrowth of a thin layer of material A on a substrate of material
B. The key concept of pseudomorphic growth, where the overlayer is in perfect atomic registry with
the substrate, was first introduced by Frank and van der Merwe in 1949.! They also showed that
with a given lattice mismatch there exists a critical layer thickness for pseudomorphic growth.

Heteroepitaxial growth can'be achieved in vapor, liquid, or solid phase. To date, most studies
have focused on vapor phase deposition, such as molecular beam epitaxy (MBE) or organometallic
vapor phase epitaxy (OMVPE). Growth of a thin film by vapor deposition proceeds in three different
modes:?

1. layer growth (Frank-van der Merwe);

2. island growth (Volmer-Weber); |

3. layer plus subsequent island growth (Stranski-Krastanov).

Using molecular dynamics simulation, Grabow and Gilmer? showed that the equilibrium morphology
of a heteroepitaxial layer is that of islands (growth mode 2 or 3). In an actual growth, kinetic
processes such as surface diffusion as well as energetics such as surface energy* determine the growth
mode. Indeed, one of the key problems of heteroepitaxy is how to control the kinetics to promote
layer growth. Another fundamental issue® is what ultimately limits heteroepitaxial growth. The
understanding of such issues and the eventual ability to manipulate heteroepitaxial growth are the
key for realizing heterostructure-based devices.

Silicon microelectronics lies at the heart of information technology. Silicon has many superior
physical properties, such as its almost perfectly passivating oxide, high mechanical strength, thermal
stability and conductivity. Unfortunately, having an indirect bandgap, Si cannot be used for light
sources. Heteroepitaxy on Si opens new possibilities for Si-based optoelectronics. This approach
attempts to combine the best properties of individual materials. Figure 1 is an example of that
“total integration” concept.® The key is the ability to grow functionally desirable materials in high
quality, single crystalline form on a Si substrate.

There exist two broad configurations of heteroepitaxial devices. In one, physical properties
of heterointerfaces determine the device performance. Examples are metal-base transistors and
heterojunction bipolar transitors. The interface has to be free of defects such as misfit dislocations for
good device performance. In another, the Si is used only as a substrate to support a heterostructural
device, such as laser diodes fabricated from GaAs epilayers grown on Si. The interface becomes

irrelevent in this case so long as the near-surface region is of high crystalline quality. However, the

crystalline perfection near the surface is closely related to the defects at the interface. The control

of interfacial defects is therefore important even in this configuration.

Most studies of heteroepitaxy on Si focus on MBE growth where precise atomic control and in
situ monitoring of the growth process are possible. Many materials have been grown on Si substrates
and various properties studied.®” Single crystalline epitaxial silicides of CoSig and NiSiz were suc-
cessfully grown on Si(111) by MBE in 1982.2° Both silicides have cubic CaF, structure. A NiSig
overlayer on Si(111) can have two different orientations, type-A, where the layer is fully aligned with
the substrate, and type-B, where the layer is rotated by 180 degree about the [111] axis with respect
to the substrate.? Tung observed that the Schottky barrier height between NiSiz and Si differs for
type-A and type-B heterostructures.!° A CoSiz overlayer always has type-B orientation.® Transisitor
action has been demonstrated for Si/CoSiz/Si(111) permeable-base transistors.!! Recently, pinhole
free CoSig layers were grown on Si(111) by low temperature deposition and annealing.!? Single crys-
talline CoSig and NiSig layers were also successfully grown on Si(100) and (110) substrates by the
template technique and low temperature deposition.!5 High quality single crystalline buried CoSig
layers were also fabricated by high dose ®°Co implantation into Si(100), (110), and (111) substrates,
followed by thermal annealing.'4

Single crystalline rare-earth metal silicides of YSig_, and ErSig_, were grown on Si(111)

recently.!5:16

Some silicides such as FeSig, CrSig, and ReSig, are narrow-gap semiconductors. They have
potential applications as infrared light sources and detectors. The key is the growth of an epitaxial
silicide/Si heterostructure of high perfection and a defect-free interface. Recently, some progress
has been made in growing epitaxial semiconducting ReSiz,!’ FeSig,'*19 and CrSi2?°—?? films on Si
substrates. .

By employing MBE deposition at relatively low temperature (~ 550°C), pseudomorphic GeSi
alloys over the entire composition range can be grown on Si.” The low temperature growth produces
metastable strained layers because of energy barriers for dislocation generations.?4 The nucleation
and propagation of dislocations have been examined by transmission electron microscopy.?> Mo-
tivated by potential device applications, both the band alignment at the Ge/Si interface*® and
the bandgap of coherently strained GeSi alloys?’ were studied. Devices such as infrared waveg-
uide photodetectors,?* n-channel”? and p-channel®? modulation-doped SiGe/Si field-effect transis-
tors, and Si/GeSi/Si heterojunction bipolar transistors by MBE*! and chemical vapor deposition

(CVD#?) have been demonstrated.

Photonic devices are made from direct bandgap III-V semiconductors. Successful growth of high
quality GaAs*** and InP* layers on Si substrates opens the possibilties for application of Si-based,
optoelectronic integrated circuits in optical interchip connection and lightwave communication.

Current integrated circuits are based on 2-dimensional planar architecture. Growth of single
crystalline insulators such as CaF2, BaF 2 on Si** makes it possible to build 3-dimensional integrated
circuits, which enable one to multiply the device elements in one chip. This approach also provides
an opportunity to fabricate novel electronic devices.>” Single crystalline insulating films can also be
used as buffer layers (e.g., between GaAs epilayers and Si substrates) to relieve stress in epilayers.*®

3C-SiC is a semicondcutor with zinc-blende structure. Its wide bandgap and high thermal
conductivity and stability are ideal for high-temperature and high-power device applications.39*°
Since the growth of single crystalline 3C-SiC films on Si by CVD was demonstrated in 1983,*!
significant progress in improvement of film quality and reduction of defects has been made. Prototype
devices such as p-n junction and field-effect transistors have been demonstrated.39:*

As the device size shrinks to a submicron regime, electrical resistence in interconnections be-
comes the limiting factor for high-speed performance. Superconductors therefore become the ideal
candidate. Growth of high-T¢ superconductor films on Si by various deposition techniques has been
demonstrated.42-44 Highly epitaxial layers with high critical currents were successfully grown on Si
with buffer layers.4®

In the following chapters, we present some results from experimental investigations on epitaxial

films of metallic CoSiz, semiconducting ReSig silicides, and GeSi alloys grown on Si. Many properties

found are generic and apply to other epitaxial films on Si, and to other heterostructures.

GaAs Parallel Photodiodes Visible or far-IR detection
L£2 of Laser ; Lnput/ Output ‘Input ] | array ( HoldTe , InSb
for interconnections fe or pyroelectric polymer

by: EI Tp |

SQ, waveguide

SUKON SUBSTRATE { or superconductor

on a Si substrate (from Fig. 1 of Ref. 6).

so oF YS NS

10.
11.
12.

13.
14.

15.

16.

17.

18.

References

. F.C. Frank and J.H. van der Merwe, Proc. Roy. Soc. A198, 216 (1949).

J.A. Venables, G.D.T. Spiller, and M. Hanbiicken, Rep. Prog. Phys. 47, 399 (1984).

M.H. Grabow and G.H. Gilmer, Mat. Res. Soc. Proc. 56, 13 (1984).

E. Bauer, Z. Krist 110, 372 (1958).

E.G. Bauer, B.W. Dodson, D.J. Ehrlich, L.C. Feldman, C.P. Flynn, M.W. Geis, J.P. Harbison,
R.J. Matyi, P.S. Peercy, P.M. Petroff, J.M. Phillips, G.B. Stringfellow, A. Zangwill, J. Mater.
Res. 5, 852 (1990).

Proc. of the NATO Adv. Res. Workshop on Heterostructures on Silicon, eds. E. Rosencher
and Y. Nissim, Kluwer Academic, Norwell, 1988.

. Mat. Res. Soc. Symp. Proc. 116, Heteroepitaxy on Silicon: Fundamentals, Structures, and

Devices, eds. H.K. Choi, R. Hull, H. Ishiwara, R.J. Nemanich, Materials Research Society,
Pittsbugh, 1988.

R.T. Tung, J.C. Bean, J.M. Gibson, J.M. Poate, and D.C. Jacobson, Appl. Phys. Lett. 40, 684
(1982).

. R.T. Tung, J.M. Gibson, and J.M. Poate, Phys. Rev. Lett. 50, 429 (1983).

R.T. Tung, Phys. Rev. Lett. 52, 461 (1984).

J.C. Hensel, A.F.J. Levi, R.T. Tung, and J.M. Gibson, Appl. Phys. Lett. 47, 151 (1985).

H. von Kanel, J. Henz, M. Ospelt, J. Hugi, E. Miller, N. Onda, and A. Gruhle, Thin Solid
Films 184, 295 (1990).

R.T. Tung, F. Schrey, and $.M. Yalisove, Appl. Phys. Lett. 55, 2005 (1989).

A.E. White, K.T. Short, R.C. Dynes, J.P. Garno, and J.M. Gibson, Appl. Phys. Lett. 50, 95
(1987).

M.P. Siegal, F.H. Kaatz, W.R. Graham, J.J. Santiago, and J.V.D. Spiegel, Appl. Surf. Sci. 38,
162 (1989). |

J.Y. Duboz, P.A. Badoz, A. Perio, J.C. Oberlin, F. Arnaud D’Avitaya, Y. Campidelli, and J.A.
Chroboczek, Appl. Surf. Sci. 38, 171 (1989).

J.E. Mahan, K.M. Geib, G.Y. Robinson, R.G. Long, X.H. Yan, G. Bai, M-A. Nicolet, M. Nathan,
Appl. Phys. Lett. 56, 2439 (1990).

N. Cherief, C.D’Anterroches, R.C. Cinti, T.A. Nguyen Tan, and J. Derrien, Appl. Phys. Lett.
55, 1671 (1989).

19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.

38
39

J.E. Mahan, K.M. Geib, G.Y. Robinson, R.G. Long, X.H. Yan, G. Bai, M-A. Nicolet, M. Nathan,
Appl. Phys. Lett. 56, 2126 (1990).

L. Haderbache, P. Wetzel, C. Pirri, J.C. Peruchetti, P. Bolmont, and G. Gewinner, Surf. Sci.
209, L139 (1989).

R.W. Fathauer, P.J. Grunthaner, T.L. Lin, K.T. Chang, J.H. Mazur, D.N. Jamieson, J. Vac.
Sci. Technol. B6, 708 (1988).

J.E. Mahan, K.M. Geib, G.Y. Robinson, G. Bai, and M-A. Nicolet, J. Vac. Sci. Technol. B9,
64 (1991).

J.C. Bean, L.C. Feldman, A.T. Fiory, S. Nakahara, and I.K. Robinson, J. Vac. Sci. Technol.
A2, 436 (1984).

B.W. Dodson and J.Y. Ysao, Appl. Phys. Lett. 51, 1325 (1987).

R. Hull, J.C. Bean, D.J. Werder, and R.E. Leibenguth, Phys. Rev. B40, 1681 (1989).

C.G. Van de Walle, R.M. Martin, J. Vac. Sci. Technol. B3, 1256 (1985).

R. People, Phys. Rev. B32, 1405 (1985).

H. Temkin, T.P. Pearsall, J.C. Bean, R.A. Logan, and S. Luryi, Appl. Phys. Lett. 48, 963
(1986).

H. Daembkes, H-J. Herzog, H. Jorke, H. Kibbel, and E. Kaspar, IEEE Trans. Electron Devices,
ED-33, 633 (1986).

T.P. Pearsall, J.C. Bean, IEEE Electron Device Lett. EDL-7, 308 (1986).

T. Tatsumi, H. Hirayama, and N. Aizaki, Appl. Phys. Lett. 52, 895 (1988).

C.A. King, J.L. Hoyt, C.M. Gronet, J.F. Gibbons, M.P. Scott, and J. Turner, IEEE Electron
Device Lett. ED-34, 52 (1989).

S.F. Fang, K. Adomi, S. Iyer, H. Morkoc, H. Zabel, C. Choi, and N. Otsuka, J. Appl. Phys.
68, R31 (1990).

D.W. Shaw, in Proc. of the NATO Adv. Res. Workshop on Heterostructures on Silicon (Kluwer
Academic, Norwell, 1988), p. 61. |

A. Yamamoto and M. Yamaguchi, Mat. Res. Soc. Symp. Proc. 116, 285 (1988).

J.M. Phillips, Mat. Res. Soc. Symp. Proc. 71, 97 (1986).

L.J. Schowalter, R.W. Fathauer, F.A. Ponce, G. Anderson, and S. Hashimoto, Mat. Res. Soc.
Symp. Proc. 67, 125 (1986).

. H. Zogg, Appl. Phys. Lett. 49, 933 (1986).

. H. Matsunami, Mat. Res. Soc. Symp. Proc. 116, 325 (1988).

40.
41.
42.

43.
44,
45.

R.F. Davis, Thin Solid Films 181, 1 (1989).

S. Nishino, J.A. Powell, and H.A. Will, Appl. Phys. Lett. 42, 460 (1983).

T. Venkatesan, E.W. Chase, X.D. Wu, A. Inam, C.C. Chang, F.K. Shokoohi, Appl. Phys. Lett.
53, 243 (1988).

R.M. Silver, A.B. Berezin, M. Wendman, A.L. de Lozanne, Appl. Phys. Lett. 52, 2147 (1988).
W.Y. Lee, J. Salem, V. Lee, T. Huang, R. Savoy, V. Deline, J. Duran, 52, 2263 (1988).

D.K. Fork, D.B. Fenner, R.W. Barton, J.M. Phillips, G.A.N. Connell, J.B. Boyce, and T.H.
Geballe, Appl. Phys. Lett. 57, 1161 (1990).

Chapter 2 Epitaxial CoSi, Films on Si

2.1 Introduction

Thin films of transistion-metal silicides on silicon have important applications in metallization
of Si-based VLSI as contacts and interconnections.!” Silicide thin films can be readily formed by
solid state reaction between deposited metal films and silicon substrates at temperatures well below
the melting point.

CoSiz has some distinct properties among silicides: high thermal stability and low resistivity.
Together with NiSic, it has a cubic CaF», structure with a dlose lattice match to Si. The lattice
mismatch equals —1.2% at room temperature and decreases with rising temperature because of the
larger thermal expansion of CoSig than that of Si. This small lattice mismatch is important for the
growth of high quality epilayers on Si.

Highly oriented CoSiz film can be formed on Si by conventional solid phase epitaxy (SPE)—
deposition of thick cobalt film (> 1 nm) on Si substrate and subsequent thermal annealing.? The
Co2Si phase nucleates first at the Co/Si interface upon thermal annealing and grows uniformly until
complete consumption of cobalt. The CoSi phase then nucleates and grows at the expense of the
Co2Si phase. Finally the highly oriented CoSi2 phase is formed at ~ 700°C.3

Single crystalline CoSie films can be grown on atomically clean Si(111) substrates in an ultrahigh
vacuum.’ The crystallographic orientation of the layer is rotated by 180° about the [111] axis with
respect to the substrate (type-B).* The interface is highly ordered with a cobalt coordination number
of 5 (cobalt coordination number in bulk CoSig is 8).> The critical thickness for coherent growth is 3
nm.° For films thicker than the critical thickness, misfit dislocations of pure edge type are generated
with Burger’s vector 1/6 < 112 >.* The average dislocation spacing of all thick films (> 10 nm) is
roughly the same, about 30 nm.”® This universal lateral mismatch measured at room temperature
is a combined result of strain relaxation at the growth temperature and dislocation locking upon
cooling.”® The layers grown on vicinal Si(111) are misoriented with respect to the substrates.? The
misorientation angle is proportional to the offset angle and the perpendicular lattice mismatch.?
The growth of CoSig layers with a smooth surface morphology and with no pinholes was recently
achieved by a low temperature deposition process.!°~!* This approach provides the possibilty for Si
overgrowth and fabrication of Si/silicide superlattices.1° High quality layers were also successfully
grown on Si(100), (110) by the template technique.!3

Following the successful growth of single crystalline CoSiz layers on Si(111) substrates by molec-

ular beam epitaxy (MBE), A.E. White and her colleagues!* demonstrated that such layers can also
be formed by high dose implantation of °°Co into Si substrates and by subsequent thermal annealing.
This “mesotaxy” technique has several advantages over the vacuum deposition. The best mesotaxial
layers have a residual resistivity of ~ 142 cm, half of the value of the best MBE-grown films.!5 The
mesotaxial layers formed on Si(111) are mostly A-type,'® while the layers grown on Si(111) by MBE
deposition are B-type.

One’s ability to fabricate single crystalline metal/semiconductor heterostructures provides op-
portunities for fundamental research on metal-semiconductor interfaces such as the Schottky barrier
formation!” and for novel device applications such as metal- or permeable-base transistors.'® The-
oretical calculations of the energy-band structure and related electronic properties of CoSig show
that ballistic electron transmission through CoSi2/Si interfaces is possible for (100) and (110), but
not for (111) orientations.!®

In the following sections, we will focus on the structural, elastic, and thermal properties of single
layer CoSig/Si, discuss experimental results from transmission electron microscopy, double crystal

x-ray diffractometry, and MeV backscattering spectrometry, and develop phenomenological models

to understand some of the results.

2.2 Growth and characterization of CoSi, on Si(111)

Single crystalline CoSiz films of type-B and 20-200 nm thick were grown on Si(111) substrates
by molecular beam epitaxy (MBE) at University of California at Los Angeles (UCLA).?° Cho and
Arthur?! showed that an ultrahigh vacuum environment and an atomically clean wafer surface are
paramount for growth of high quality films. The Si wafers used for our epitaxial CoSig films were
cleaned by the Shiraki method,?? which consists of repeated oxidation and etching to remove carbon
and oxygen (two major contaminations), and then immediately loaded into the MBE chamber (base
pressure ~ 1071° Torr). The protective oxide layer was stripped off by flash heating to 900°C.?°
CoSiz films were grown mostly by codeposition of cobalt (flux rate ~ 0.1 nm/s) and silicon (~ 0.2
nm/s) on Si substrates at ~ 600°C (growth pressure ~ 10-° Torr).2° Some samples were also
made by solid phase epitaxy of room-temperature-codeposited stoichiometric Co/Si mixtures?3 or
by cobalt deposition onto hot Si substrates.

Epitaxial CoSiz/Si(111) samples were characterized at Caltech 7~9:%4 principally by three ana-
lytical techniques: backscattering spectrometry (BS),?° x-ray double crystal diffractometry (DCD),?6

and transmission electron microscopy (TEM).”’ MeV *He BS analysis shows that the cobalt and sil-

icon concentrations are uniform through the entire film with correct stoichiometry (Co:Si=1:240.2)
for all samples.?4 Channeling?® measurements indicate that most films are of reasonable epitaxial
quality, with a minimum channeling yield, X¥min, of ~ 2— 10%.*4 Fig. 2.2-1(a) displays both 2 MeV
He backscattering (solid line) and [111] axial channeling (dotted line) spectrum of a demonstration
sample composed of a 100 nm thick single layer of CoSig on a Si(111) substrate. The surface peaks
(at 1.53 MeV and 1.13 MeV for Co and Si, respectively) of the channeling spectrum are due to direct
backscattering of an aligned incident beam from the CoSi2 surface. The minimum channeling yield
(counts of the channeling spectrum measured immediately beneath the surface peak normalized
against those of the backscattering spectrum at the same energy)?® is ~ 2% for both Si and Co in
the film of the demonstration sample, indicating that the film is highly epitaxial. Energetic ions can
be steered by the crystal potential to follow a channel within an angular divergence measured by
a channeling half angle, /2-28 Fig. 2.2-1(b) plots the normalized backscattering yield versus the
angular deviation of the incident He beam direction from the [111] channel of the sample. The half
angle 1/2 is the same for both Si (e) and Co (0) in the film, ~ 1.2°, and is larger than that of Si
(m) in the substrate, ~ 0.9°. All these features can be understood by the phenomenological model of
channeling phenomena in polyatomic crystals that we developed recently.2° The dechanneling near
~ 1.06 MeV may be caused by extended defects such as dislocations at the interface.28 However, in
a heterostructure, the channeling half angle of the film differs from that of the substrate (see Fig.
2.2-1(b)). This also contributes to the change of the channeling yield across the interface.2° The
measured dechanneling is a combined contribution of both factors. Therefore, care must exercised
in extracting a defect density from a dechanneling yield.

A thin epitaxial film is usually under uniform strain because of its lattice mismatch to a sub-
strate. DCD is routinely used to measure the lattice dimension of a film and hence its strain.?®
In a heterostructure, one uses lattice mismatch, f, perpendicular and parallel strain, e+ and ell, to
describe the lattice dimension of a film. If both film and substrate have cubic structure, the above

quantities are defined according to

at —Gs
f= a , (2.2 — 1)

where ay (a) is the lattice constant of an unstrained film (substrate);

4 1
1_ oF —4,

er =o
“Lt 5)
d;

(2.2 — 2)

where dy (d}) is the interplanar spacing of a film (substrate) along the surface normal, and ell is

defined similarly. Both e+ and ell can be extracted from x-ray rocking curve measurements.2° The

2 MeV ‘He : 100 nm CoSig on Si(111)

5000 Co

random

4000

3000

Counts

2000

1000

1.0 1.2 1.4 1.6
Energy ( MeV )

2 MeV *He : 100 nm CoSig on Si(111)

1.0 # : substrate
- e : film (Si)
a r °o: film (Co)
he L
ao} L
vO
| 0.5-
g L
bo
e)
cS 5

gut '

Tilt Angle ( ° )

FIG, 2.2-1 2 MeV *He: (a) backscattering spectra of a CoSi2/Si(111) heterostructure for a beam
incident along a random (solid line) and the [111] axial direction (dotted line); (b) an angular scan
about the [111] axis of the same sample; the Si and Co signal in the film were taken at 1.53 and 1.13

MeV, respectively, and the substrate signal was taken at 1.05 MeV.

Bragg peak separation between a substrate and a film, A@z, is related to e+ and ell by
—Abg = kyet + kell, (2.2 — 3)

where k, and kz are numerical coefficients determined by diffraction geometry.?° For a symmetrical

diffraction, one has

ki =tan@g and k, =0,
where @g is the Bragg angle of the substrate. Equation (2.2-3) hence simplifies to
e+ = ~—cotOpAOz. (2.2 — 4)

The Fe Ka, x-ray rocking curve from the (111) symmetrical diffraction (9g = 18°) of our demonstra-
tion sample is shown in Fig. 2.2-2. The Bragg peak separation, Aég = 0.33°, gives a perpendicular
strain of «+ = —1.7% according to Eq. (2.2-4). This is the value for all thick films (> 10 nm).7®
Additional information can be extracted from an x-ray rocking curve.® A small-amplitude oscillation
on either side of the Bragg peak is caused by the finite thickness of the film?® and is clearly visible
in Fig. 2.2-2. This indicates that the film is elastically uniform with few extended defects inside the
film, and most dislocations are confined to the interface. The periodicity of the oscillation, (56)o,
measures the thickness of the film, t 1

t= Sos 93° G2-§)
where A is the x-ray wavelength (0.1936 nm). The sample has a periodicity of (50)o = 0.06° (see
Fig. 2.2-2), which gives the film thickness of ty = 100 nm according to Eq.(2.2-5). The result agrees
well with that obtained from the BS meaurement (see Fig. 2.2-1(a)). The finite film thickness also
broadens the diffraction peak, with a full-width at half-maximum intensity (FWHM), Ws, given by

the Scherrer equation 3+

0.942

Ws = 2t, cos OB

(2.2 - 6)

For a perfect single crystalline film, the measured FWHM, Ws, is the same as Ws. For a film

containing extended defects that produce an imhomogeneous strain, the diffraction peak broadens,
Wo? =Ws*?+Wp’,

where Wp is the broadening by defects. For a dislocated crystal with a threading dislocation density,

p, the defect broadening can be estimated by®?

Wp? = 9b’p, . (2.2-7)

symmetrical x-ray diffraction

substrate

x-ray rocking curve: (111) diffraction
10°

TF Tirtr
Dob pits

100 nm CoSig on Si(111)

rare
wre

TT TTTITTT
po il

Reflecting Power ( % )
°o
T TT PVIUTy
cc
i ii ris!

107" (50)o E
Lat” @y=18° "\
10" 0 93 | Od
da (°)

FIG. 2.2-2. A Fe Kg, x-ray (A = 0.1936 nm) rocking curve from the (111) symmetrical diffraction

of the same CoSip/Si(111) sample shown in Fig. 2-2-1.

where b is Burger’s vector. One could therefore get an estimate of the threading dislocation density
in a film from measured FWHM.® The measured FWHM of our demonstration sample (see Fig. 2.2-
2) is about the same as that caused by finite film thickness, meaning that the threading dislocation
density in the film is below the detection limit (< 10°/cm?).

Microstructure and extended defects such as dislocations in a crystal can be revealed by TEM.2”
In particular, high-resolution transmission electron microscopy (HRTEM) is capable of direct lattice
imaging and has been used to study the atomic structure at the interface in a heterostructure.>3
Fig. 2.2-3(a) is a cross-sectional HRTEM image of the interface for an MBE-grown CoSi2/Si(111)
sample,!? which clearly shows an atomically sharp interface and type-B orientation. Fig. 2.2-3(b)
is a weak-beam dark-field plane-view TEM micrograph of a 50 nm thick CoSi2/Si(111) sample
prepared by MBE,”* which illustrates the hexagonal dislocation network at the interface. These
misfit dislocations have been identified to be pure edge type with Burger’s vertor b = 1/6 < 112 >.4
The average spacing between misfit dislocations, p, is about 30 nm. It is the same for all thick

samples (> 10 nm).® The parallel strain ell is related to dislocation spacing and Burger’s vector,

ell = —-. 2.2~-8
D ( )

For the sample shown in Fig. 2.2-3(b), one has
ell = —0.7%.

Fig. 2.2-3(c) is the plane-view TEM morié pattern of a 10 nm thick CoSi2/Si(111) sample.!? The
regular morié fringes indicate that the film is uniformly strained. The periodicity of ~ 30 nm (see
Fig. 2.2-3(c)) means that the parallel strain of this heterostructure is about —0.6%, about the
same as that obtained above for the sample shown in Fig. 2.2-3(b). The termination of the lines
(arrowed in Fig. 2.2-3(c)) indicates that a dislocation threads into the film and emerges at the
surface. The average spacing between threading dislocations is about 100 nm (see Fig. 2.2-3(c)),
meaning that the areal density of threading dislocations in this sample is 10!°/cm?. This density
is more than 10* times greater than that obtained for our demonstration sample, while the average
spacing between misfit dislocations is about the same for both samples. This fact means that the
threading dislocations in CoSig films are not closely related to the misfit dislocations at the interface,
and suggests that relaxed epitaxial CoSig films free of threading dislocations can be grown.

In this section, we discussed some general properties of epitaxial CoSig films on Si(111) and
introduced three analytical tools to characterize heteroepitaxial structures. The following sections

detail various aspects on the strain state of CoSiz films.

ati “a . ane is CoSi5 oe
. ae 7 he . Sih

SUESEES Mj y
peddeed POTS PO? f° edd:
RESSSLESLS 9 HEALS @oeoge
oo POPE re eHr Meee

ve CPP rPveores °
oon TeeerFveorve Ft PFE WUE BAGH y 04

erere OV ECCHT HTH VESE EOL & 8 ose’ Si eee
(OPSPOU CE TE VON Te DOE ro ewe ve roeeres U1 neve
a era om ee 2- ~ee eee mew e eee *eoowe

FIG, 2.2-3 Transmission electron micrographs of CoSi2/Si(111) samples: (a) a cross-sectional high
resolution lattice image of the interface (from Ref. 12); (b) a weak-beam dark-field plane-view ([022]
beam) micrograph showing a hexagonal misfit dislocation network (from Ref. 24); (c) a plane-view

morié pattern caused by the lattice mismatch between the layer and the substrate (from Ref. 12).

2.3 Growth and characterization of CoSi2 on vicinal Si(111)

In this section we study the properties of CoSi, films grown on Si wafers whose surfaces were
tilted away from (111) planes toward [112] direction by an offset angle, ¢,, ranging from 0° to 16°.9
The surface of these vicinal Si(111) substrates consists of wide (111) ledges, and clusters of steps
with height d(i11) and edges parallel to [110].34 Epitaxial B-type CoSiz films 10-200 nm thick were
grown by MBE at UCLA.?° We used BS and channeling to characterize the stoichiometry, thickness,
and epitaxial quality of these films. Some samples were also analyzed by both plane-view and cross-
sectional TEM to reveal pinholes, misfit dislocations and interface structure.2%24 Back-reflection
Laue patterns were used to measure the substrate offset angles ¢,. DCD shows that there is a
misorientation angle, a, between the [111] orientation of the film and of the substrate for the sample
with a non-zero offset angle.®

All x-ray rocking curves were taken from symmetrical (111) diffractions. For a vicinal substrate
with an offset angle ¢,, the Bragg peak position, Op, is related to the Bragg angle 0g of the
(111) plane and the azimuthal angle, 7 (see Fig. 2.3-1(a)). The relationship can be approximately

expressed as?°

Op = 0p + ¢, cosy. (2.3 — 1)

In particular, at two extreme sample configurations of » = 0° or 180° (see Fig. 2-3-1(a)), the relation

is exact and simplifies to

Op =On+ 6, or Op = Op — 4,.

Fig. 2.3-1(b) shows the measured (111) Bragg peak position 6p of a vicinal Si(111) substrate with
¢@, = 16° (@g = 18°) as a function of the sample azimuthal angle w. It closely follows the relation
dictated by Eq. (2.3-1). Furthermore, the Bragg peak separation between the film (~ 100 nm thick)
and the substrate, A@p, has a similar functional dependence on # (see Fig. 2.3-1(c)). This clearly
demonstrates that (a) there is a misorientation angle, a, between the {111] orientation of the film
and the substrate; (b) the [111] orientation of the film and the substrate, and the surface normal lie
in the same plane.? Under such circumstances, only two rocking curve measurements corresponding
to ~ = 0° and 180° are required to extract the misorientation angle a, and the difference of Bragg

angles between the film and the substrate Aéz,?

ga DoPd EH, (2.3 ~ 2a)
Ap = Aépr + Oop ir (2.3 — 2b),

2 3

Aép (°)

FIG. 2.3-1 Fe Kg, x-ray (111) symmetrical diffraction from a 100 nm thick CoSig film on a vicinal
Si(111) substrate (offset angle ¢, = 16°): (a) diffraction geometry, n is the surface normal; (b) the

Bragg peak position of the substrate versus the azimuthal angle; (c) the separation between the

Op (°)

incident
x-ray

sample surface

111],
111],

100 nm CoSig on vicinal Si(111)

6,=18°, $,=16°

0.6

0.4

0.2

A@g=0.31°, Ap=—0.27°

100

200
y(°)

300

Bragg peaks of the film and the substrate versus the azimuthal angle.

where Aép; (A@p zr) is the Bragg peak separation obtained for the % = 0° (180°) configuration.

For the sample shown in Fig. 2.3-1, we obtain
@ = —0.27° and Aég = 0.31°,

according to Eq. (2.3-2). The perpendicular strain (defined here as the strain along the [111]
direction) e+ is therefore ~ —1.7% according to Eq. (2.2-4).

Our results from back-reflection Laue and DCD analyses of more than 20 samples can be
summarized as follows:°
(1) The [111] directions of the substrate and of the film, and the surface normal lie in the same (110)
plane (Fig. 2.3-1(a)).° The [111] direction of the film lies between the other two directions, and very
close to the [111] direction of the substrate at a misorientation angle a whose magnitude is much
less (about two orders of magnitude) than that of the offset angle itself.

(2) The misorientation angle a is proportional to the substrate offset angle ¢, (Fig. 2.3-2(a)). The
slope is the same (~ —1.7%) for all samples investigated and is independent of the thickness of the
film and the growth processes.

(3) The perpendicular strain €+ is essentially a constant (~ —1.7%) at room temperature for all the
samples,

These last two facts suggest that the increase of the misorientation angle with an increasing
substrate offset angle is given by the perpendicular strain, a = e+¢,.

To check the validity of this observation, experiments were undertaken to measure the change of
the misorientation angle, as the perpendicular strain varies while keeping the substrate offset angle
the same. One technique to accomplish this is to perform x-ray rocking curve measurements at
different temperatures on the same sample. The strain decreases as the temperature rises because of
the difference of thermal expansion coefficients between the film and the substrate,”® but the offset
angle does not change. The experimental result of the sample shown in Fig. 2.3-1 (¢, = 16° and
film thickness=100 nm) is plotted in Fig. 2.3-2(b). It clearly verifies that the misorientation angle
is proportional to the perpendicular strain.

On the basis of the experimental findings and current knowledge about the interfacial structure
of CoSiz films on vicinal Si(111),3° we propose a simple geometrical model to relate the geometrical
quantities of a film and a substrate. For a CoSi film on a vicinal Si(111) substrate, the interface
consists of evenly distributed, approximately parallel steps of single atomic height di1i1) = 0.314 nm

(see Fig. 2.3-3(a)).°° In the ideal case of a coherent interface, by imposing “length matching” across

CoSig films on vicinal Si(111) substrates

oL T qT U T T T T if T ¥ T T T t T T "]
_-0.1b slope=—(1.7 0.1)% |
fo)
rad i |
-0.2- 4
a a OO
0 5 10 15
o,(°)
—0.18 r
-0.20- =
—~-0.22+ |
fo]
—0.24- slope=(15.5+ 0.5)°
—0.26}+ 4
L L 1 I a |
-1.6 —1.4 _ =4.2
e+ (%)

FIG. 2.3-2 The misorientation angle a of (a) the CoSi2/Si(111) samples of various film thickness

versus the substrate offset angle ¢,; (b) the sample shown in Fig. 2.3-1 (¢, = 16°, 100 nm thick)

versus the perpendicular strain.

—___| CoSiz (111)

Si(111)
0 af *) 12 16
0.3 T— T | -0.3
CoSig on vicinal Si(111)
8 0.2 4-0.2 ~
be} °
fy + 2 —
~ 3
4 least—s fit
% ob st—squares tod
3 slope=1.001 0.02
0 \ I ‘ “t ' 0
¢) 0.1 0.2 0.3
tan,

FIG. 2.3-3 (a) Schematical diagram of the interface between a CoSiz film and a vicinal Si(111)
substrate along the [112] tilt direction, and the proposed geometrical model on the misorientation
between the film and the substrate; (b) the geometrical model that predicts a = e+ tang, agrees
excellently with the least-squares fit to the experimental data of epitaxial CoSig films grown on

vicinal Si(111) substrates.

the interface (Fig. 2.3-3(a)), we obtain the following relations,

all dl dt dt
=—/f = =] Ijp=2—b = =],. 3
ty cos¢, cosd, ~ and ly sing, sind, ° (2.3 — 3)

After simple algebra, one has

a=ettand, (2.3 — 4a)

and

ell = ~e+ tan? 4,, (2.3 — 5a)

to the first order in a. Non-zero (but very small) ¢!! from Eq. (2.3-5a) for a coherent interface is
caused by misorientation between planes such as (112) of the film and the substrate. Along [170]
direction (normal to the paper face in Fig. 2.3-3(a)), there is no misorientation, and hence el vio)
equals 0 for a coherent interface. In a general case where there exist misfit dislocation arrays at the

interface, similar results can be derived,

a=e+tand, (2.3 — 46)

and
cll = ~§~— + tan’ ¢,, (2.3 — 5b)

where 6 is the strain relaxation from the misfit dislocations. This model agrees excellently with the
experimental results (see Fig. 2.3-3(b)). A similar model has been proposed by Nagai to explain the
observed misorientation effect between a film and a vicinal substrate in a compound semiconductor

heterostructure.?”

2.4 Critical thickness and strain relaxation

We investigate here the (elastic) energetics and (meta-) stability of a CoSi2/Si(111) heterostruc-
ture. The material is treated as an elastic continuum. The (elastic) energy is composed of two parts:
uniform strain energy and dislocation energy. The equilibrium state of this system corresponds to
that of the minimum elastic energy.°* For a very thin film, the film is strained to match the lattice
constants of the substrate. As the film thickens, it becomes energetically favorable to generate misfit
dislocations at the interface to relieve the lattice mismatch f between the film and the substrate.
Using the minimum energy criterion, Matthews and Blakeslee®® estimated the critical thickness of a

film, ter, the maximum film thickness above which generation of misfit dislocations is energetically

favored, for a given lattice mismatch f,

t 1 t
Se Jy) AH
6 Bt ot) (2.4— 1)

where v is Poisson’s ratio of the film. For a CoSi2/Si(111) heterostructure, the lattice mismatch at
room temperature is f = —1.2%. The critical thickness obtained from Eq. (2.4-1) is t., = 1.7 nm,
using v = 1/3 for CoSig on Si(111).8 The experimental critical thickness determined by plane-view
TEM at room temperature is ~ 3 nm,® about 80% larger than that from the theoretical estimate.
We will explain this discrepency in Ch. 2.6.

For a film thicker than the critical thickness (t; > ter), misfit dislocations are generated and
the elastic strain relaxes. The strain of the film in the equilibrium state is given by®

Qu ter Ints/b+1
—vy ty, Int.-/b+1

che flts ). (2.4 ~2)

As an example, we considered a CoSig/Si(111) heterostructure at room temperature, and computed
the perpendicular strain e+ as a function of film thickness according to Eq. (2.4-2) (solid line in
Fig. 2.4-1). We used the measured ¢,, in the computation (using theoretical t., gives a similar
result beyond critical thickness). The data points in the same figure are from all the CoSiz/Si(111)
samples we have analyzed (see Ch. 2.2 and 2.3), grown by MBE or SPE, with film thickness from 10
nm to 230 nm, and with a substrate offset angle from 0° to 16°. It clearly shows that the strain is
roughly a constant, «+ ~ —1.7%, over the thickness range from 10 nm to 230 nm (Fig. 2.4-1). And
the film is more strained than that predicted by Matthews and Blakeslee’s model, meaning that the
CoSi2/Si(111) samples are metastable at room temperature.

To study the effect of thermal processing on such metastable heterostructures, we annealed the
samples in a vacuum (~ 5 x 10~" Torr) at 650, 750, 850°C for 30 min and analyzed the structural
change of the samples at room temperature after such postgrowth thermal processing. 2 MeV *He
backscattering spectra of both random and aligned incidence remain the same before and after
thermal processing, meaning that no significant diffusion or structural reordering occurs.”?° Misfit
dislocation spacings from plan view TEM remain unchanged, ~ 30 nm,”° so does the perpendicular
strain from DCD measurements, ~ —1.7%.":?° These results demonstrate that the CoSiz/Si(111)
structures do not change upon annealing at 850°C for 30 min in a vacuum.

To gain insight on such elastic metastability of CoSig/Si(111) heterostructures at room tem-
perature, we conducted DCD measurements at elevated temperatures on some of these metastable

samples. As shown in the next section, we discovered that at their growth temperature, the samples

—1.0
| 2 decoupled CoSig/Si(111)
i A
Matthews—Blakeslee model
_~ —1.5 ¢ average strain ~ —1.7%
x wag ete ecg .
~" ee e e
i)
—2.0
coherent
ee a a a SD
0 50 100 150 200

Film Thickness ( nm )

FIG. 2.4-1 Perpendicular strain in an epitaxial CoSia film as a function of film thickness for a
CoSi2/Si(111) heterostructure. The solid line is the strain relaxation behavior predicted by Matthews
and Blakeslee’s minimum energy model. The dashed line is the average of the experimentally mea-

sured strain value on samples of various film thickness, substrate offset angle, and growth condition.

are relaxed. The strain measured at room temperature is caused by different thermal contractions
of CoSig and Si upon cooling while the misfit dislocations created during the growth do not shear

and no new ones are nucleated.’

2.5 Thermal strain and its inference for pseudomorphic growth

We report here a DCD measurement of the perpendicular strain in epitaxial CoSiz films ~ 100
nm thick (which is much thicker than t,,) formed on Si substrates by codeposition of Co and Si
at either ~ 600°C”° or at room temperature followed by in situ annealing.?° The surfaces of the
substrates, are tilted from the (111) planes of the Si lattice towards the [112] direction by offset
angles ¢, ranging from 0° to 16°.

A sample was mounted on a heating stage capable of reaching 650°C in air.” The perpendicular

strain at temperature, T,,
d* (T) — d*,(T)
d(T)

was extracted from Fe Kg, x-ray rocking curve diffracted from the symmetrical (111) diffraction

_ peaks. The parallel strain ll defined analogously to ¢+ is related to the lattice mismatch f and e+,”

l+v 2v
a _ H, 5
€ (TF (= (2.5 — 2)
The measured perpendicular strain at room temperature is «+(24°C) = ~—1.66 + 0.01%, a

typical value for all thick films (ty > t.,-).”% That strain is between that of a fully relaxed film
for which €+ decoupted(24°C) = f(24°C) = —1.23%, and that of a pseudomorphic film for which
€+ coherent(24°C) = —2.46%, obtained from Eq. (2.5-2) with ell = 0 for a coherent film and v = 1/3
for a CoSiz film on Si(111).’ Thus, the thick CoSig films (t+ > ter) are elastically strained at room
temperature with a common perpendicular elastic strain, e+, of e+(24°C) = e+(24°C) — f(24°C) =
—0.43%.

Measurements of the perpendicular strain at elevated temperatures provide a clue as to why
e+ (24°C) is always nearly —1.66%. Figure 2.5-1 shows two sets of experimental data (e and 4A)
for two samples that have slightly different perpendicular strains at room temperature. As the
temperature rises, the strain decreases, because CoSiy expands faster than Si does upon heating.
The slopes, (1.34 0.1) x 10-5/°C, are twice as large as the difference of the bulk thermal expansion
coefficients between CoSiz (ay = 0.94 x 10-°/°C!) and Si (a, = 0.29 x 10-5/°C%9), ay — a, =

0.65 x 10-°/°C. All strain values remain reversible after ~ 2 Ar annealing in air up to 490°C.

This temperature dependence can be explained if it is assumed that the lateral change in the
CoSi,g lattice is constrained to follow that of the Si substrate. The slope of the perpendicular strain

versus temperature is then given by

6 + on rat 1
€ a ained _ 5 ~*) . (ay _ Qs). (2.5 —.3)

Substituting vy = 1/3,”* the model predicts a slope of 1.3 x 10-5/°C, in excellent agreement with
the measured value of (1.3 + 0.1) x 10-°/°C. We thus conclude that no new misfit dislocations are
created by thermal cycling in air up to 490°.

Figure 2.5-1 also contains the predicted perpendicular strains versus temperature for a fully
relaxed (stress-free) film, labeled “decoupled,” and a pseudomorphic film, labeled “coherent.” At
24°C, these lines have known values of —1.23% (from Eq. (2.5-2) with ell = f, m) and —2.46% (from
Eq. (2.5-2) with ell = 0, open square). The slope for the fully relaxed film is given by

“decoupled _ Of(T
be decoupled LM) = (ay - a4), (2.5 — 4)

while that for the coherent film is given by Eq. (2.5-3). The line for the fully relaxed film intercepts
the experimental curves near 600°C, which is the nominal growth temperature of the films. This
fact says that within the uncertainties of the experimental data, the epitarial CoSig films grow
elastically unstrained at the temperature of their formation. This fact, combined with the observation
that no new misfit dislocations are generated upon subsequent cooling of the sample, explains the
common value of strain observed at room temperature for all thick films (ty > t,-). The observed
elastic strains originate after the films are formed and are the consequence of the dissimilar thermal
contraction of the film and substrate upon cooling.

As a corollary of the above discussion, one can obtain the parallel strain of the CoSi, films
at room temperature in two ways. For a rigid interface, the parallel strain at room temperature
must be the same as that at the growth temperature. At the growth temperature, the CoSi film
is fully relaxed, and hence cubic. The parallel strain ¢l! at the growth temperature thus equals the

perpendicular strain e+ at that temperature,
e!l(24°C) = ell(600°C) = «+(600°C) = —0.86%, (2.5 — 5)

where the numerical value is obtained from Fig. 2.5-1. On the other hand, ell can be obtained from
Eq. (2.5-2). With «1(24°C) = —1.66%, f(24°C) = —1.23%, and v = 1/3, one finds «ll(24°C) =
~—0.80%. This consistency supports the notion that the interface does not shear, and the parallel

strain is independent of temperature with zero elastic strain at the growth temperature.

100 nm thick CoSig films on Si(111)

0 200 400 600

T( °° )

FIG. 2.5-1

thick CoSi films on (i) a Si substrate whose surface is offset from the [111] direction by 16° towards

the [110] direction of the substrate (e), and (ii) a Si substrate whose surface is aligned with the

(111) planes (Av). All

heated above that temperature up to 650°C and then cooled, a reduced strain was measured (v7),
again reversible below 500°C. The solid lines are linear fits to the data (omitting the two highest
temperature points). The strain for a coherent interface is calculated with the Poisson ratio v = 1/3

(lower dashed line). At the temperature at which the samples were grown (~ 600°C), the measured

Perpendicular x-ray strain ¢€

strain values are reversible below 490°C. When the latter sample was

strain is that predicted for an elastically relaxed film (upper dashed line).

as a function of sample temperature T for two 100 nm

The estimated average spacing between misfit dislocations from the parallel strain «ll,

p=—- w28 nm,

lel

agrees well with the result obtained from the plan view TEM (~ 30 nm) of the samples produced
by the same deposition process (see Fig. 2.2-3(b)).

Furthermore, sample #1 was raised to 600°C and subsequently to 650°C in air, and rocking
curves were measured in situ. The perpendicular x-ray strains derived from these measurements
indicate that the sample is fully relaxed at these temperatures (Fig. 2.5-1). Cooling this sample to
room temperature now yields a slightly reduced perpendicular strain (—1.61% instead of —1.66%).
Reheating the sample in air then traces a new line for the perpendicular x-ray strain vs. temperature
(points marked 7). That line has the same slope as that measured initially (points marked A). This
new curve is reproducible up to 550°C. This result shows that maintaining the sample in air near or
above 650°C for a sufficient length of time (~ 40 min) modifies the elastic strain in the film. The
x-ray rocking curve of this modified sample (point 7 in Fig. 2.5-1) is shown in Fig. 2.5-2 (dotted
line). One sees that the interaction of the sample with air at ~ 650°C has reduced the perpendicular
x-ray strain and has broadened the diffraction peak from the film when compared with the rocking
curve of the as-grown sample (solid line). We conclude therefore that even though the interface does
not shear at temperatures below ~ 600°C (the growth temperature) in air, the interfacial registry
and the film quality are altered by annealing above ~ 600°C in air.

As an additional experiment, a piece of the same sample #1 was thermally annealed in vacuum
at 700°C for 40 min. The rocking curve of the sample was then remeasured at room temperature
(solid line in Fig. 2.5-2). The vacuum treatment produces no detectable change, consistent with
previous results of Ch. 2.4. This indicates that the strain state of an epitaxial CoSig film is not
altered by an annealing at temperature higher than the CoSi formation temperature in vacuum.
The change after annealing in air at temperatures above ~ 600°C then must be the result of reaction
of the epitaxial film with the ambient-air. Auger electron spectroscopy of the ambient-air-annealed
sample shows that a thin oxide of ~ 10 nm is present on the surface of the silicide but that is
absent in the vacuum-annealed sample. An oxidation of the CoSig at its surface induces atomic
rearrangements at the silicide/silicon interface.*® All three samples of as-grown, vacuum-annealed,
and air-annealed were also analyzed by 2 MeV “He backscattering and channeling. No significant
differences are noted (see Fig. 2.2-1 as an example of the spectra).

The findings reported here have interesting implications that may have very general validity.

10* 3

F 100 nm CoSig on Si(111) ;

ye 10! = — as-grown |
~~ Er 3
ts 5 — 700°C anneal in vacuum J
o LE _— ) 40min
ZF i199 b * 650°C anneal in air a
[°) E E |
am E 5
e :
@ 1077 / 3
P F Ug 4
‘3 _, q
[aa] 2 , 4
10-7F aie

E ! ! j R ! _ VAS

0 0.2 0.4 0.6

film as indicated by the shift and the broadening of the diffraction peak (dotted line).

We note that for an epitaxial GaAs film grown on a Si(100) substrate, findings like ours have recently
been reported.*° There too, the strain observed at room temperature is explainable in terms of the
differential thermal contraction of the film and substrate upon cooling after the epitaxial structure
was created at an elevated temperature. When two such dissimilar systems behave so similarly, the
observation may apply to a large class of epitaxial systems. Generalizing accordingly, we extract the
following rules:

— Differential thermal expansion does not introduce new misfit dislocations in

epitaztal CoSiz films on Si(111), even when the film is thicker than the critical

thickness. This statement probably applies to all systems where non-ionic bonding

dominates,

— For pseudomorphic growth in such systems, it is important to minimize the lattice

mismatch at the growth temperature.

Tabulated lattice constants are usually measured at room temperature. It thus follows that
thermal expansion data are also important to consider in optimizing conditions for pseudomorphic
growth. The data are, however, often unavailable. The importance of a lattice match at the growth
temperature for pseudomorphic growth has already been pointed out in the literature.4!-42

We note further that when an epitaxial CoSiz film is exposed to a reactive medium (air), ir-
reversible changes take place in the strain state of the film that do not arise in the absence of a
reaction. Similar effects have been reported for different systems. A polycrystalline film of Pd2Si
on an epitaxial Pd2Si film raised to 275°C in vacuum is stable. The same film undergoes an epi-
taxial reordering when additional polycrystalline Pd2Si is formed by a reaction with an overlaid Pd
film.*? The disordering of III-V heterostructures by impurity diffusion is another example where
a defect-generating process (here group II lattice vacancies generated by the diffusion of an impu-
rity) destabilizes a structure that is quite (meta)stable in the absence of this process.44 Yet another
example is the epitaxial arrangement of a polycrystalline Ge film on a Si single crystal substrate
induced by thermal annealing in oxygen.*° The feature common to all these examples is the presence
of a defect-generating process (diffusion, reaction, irradiation). The defects destabilize a metastable
state. We are thus led to conclude that

— To favor pseudomorphic growth, processes that induce atomic rearrangement

should be minimized,
Procedures advocated in the literature for the successful growth of epitaxial layers are consistent

with that conclusion. For example, a successful procedure to grow high quality epitaxial layers

consists of first growing a very thin ( ¢ < t., ) epitaxial film by solid phase epitaxy, followed by
codeposition of a thick layer ( t > tep ) at low temperature and subsequently inducing epitaxial
rearrangement of the full layer at an elevated temperature.*® This template procedure actually
implements the idea of minimizing atomic rearrangement. One also understands why codeposition
of a compound in the correct atomic ratio is more likely to result in a pseudomorphic structure than
its formation by solid phase reaction if the epitaxial film is strained at the formation temperature,
or why deposition at a low temperature followed by high temperature annealing is more likely to
result in a pseudomorphic structure than deposition at a high temperature.*”

The three rules for pseudomorphic growth enunciated here are in the nature of guidelines, as
is evident from the way they are deduced. They are nonetheless useful, because they emphasize
considerations not previously stressed in the literature (lattice matching at the growth temperature,
with the concomitant relevance of knowing thermal expansion behavior for the materials involved;

minimizing atomic relocation processes during the film formation).

2.6 Elastic and thermal properties of mesotaxial CoSi, films on Si

Thin CoSiz films formed by high dose *°Co implantation are A-type,!4 enabling one to make
a high-precision determination of both parallel and perpendicular strains by x-ray rocking curves8
With B-type films, the Bragg peaks from asymmetrical diffraction of the films are widely separated
from those of the substrates, precluding high-precision measurements of ¢ll. Recognizing this op-
portunity that mesotaxial A-type CoSiz films on Si(111) offer, we measured both the perpendicular
and the parallel strain in such films, as well as those of mesotaxial films formed on Si(100). These
two measurements enable us to extract two ratios of the three independent elastic constants of cubic
single crystal CoSig. We also measured the curvature of one sample to estimate the biaxial stress
in the film. These three measurements yield the absolute values of the three elastic constants of
CoSiz. We repeated similar measurements up to ~ 500°C. Assuming that the elastic constants do
not change between 20°C and 500°C, we are able to extract the linear thermal expansion coefficient
for single crystalline CoSig.
A. Sample Preparation

Single crystalline buried CoSiz films 110 nm thick were formed at AT&T Bell Laboratories by
200 keV 3 x 10'’/cm? *°Co implantation at ~ 400°C into Si substrates of both (100) and (111)

orientation, followed by vacuum annealing at 600°C for 60 min and 1000°C for 30 min.'+ The top Si

layers were then removed by reactive ion etching. Cross-sectional transmission electron microscopy
shows that the interfaces between the films and substrates are flat and atomically sharp.'4 MeV *He
backscattering and channeling spectrometry indicate that the films are stoichiometric and highly
oriented, with a minimum yield of ~ 3%.14
B. Lattice Mismatch and Misfit Dislocations

We used DCD to measure both e+ and ell between the CoSiz film and the Si substrate. Figure
2.6-1 shows the Fe Ky, x-ray rocking curves from the symmetrical (400) and asymmetrical (311)
diffraction planes of the CoSiz/Si(100) sample. The two curves diffracted from the same (311) planes
(A and B in Fig. 2.6-1) correspond to the x-ray incidence of opposite directions. The strain e+ and
ell were extracted from the angular separations of the Bragg peaks between the film and the substrate
shown in Fig. 2.6-1. The results are listed in the first column of Table 2.6-I. They are very close
to those measured for buried CoSig mesotaxial layers in the second column of Table 2.6-1.49 This
agreement means that the Si capping layer has little effect on the strain state of the buried CoSig
layer. Unequal e+ and ell means that the CoSia film is distorted tetragonally under the tensile stress
imposed by the Si substrate. The relative volume expansion, AV/V, is ~ 0.2%, more than 3 times
less than the average linear dilatation, AL/L, (~ 0.7%, see Table 2.6-I). This means that the volume
is almost conserved under biaxial stress.

The non-zero ¢!l means that there exist misfit dislocations at the interface to relax strain.
Burger’s vector of the dislocations for epitaxial CoSig films on Si(100) substrates is b = 1/4 <
111 >.5° The average spacing p between the misfit dislocations is therefore

be 0.19 nm

p= jal = 762% = 3l nm,

where b, is the edge component of Burger’s vector projected onto the interface plane. This is roughly

the same as that of MBE-grown thick (> 10 nm) B-type CoSi2/Si(111) samples (~ 30 nm).4
Single crystalline CoSi2 has three independent elastic constants, C11, Ci2, C44. Measurements

of the lattice distortion of CoSis films on Si substrates of two different orientations enable one to

extract two ratios, C2/Ci, and C44/Ci1. From the definition of the lattice mismatch and the elastic

strain, e+ and ell, one has the following relationship,

et e+-f

o= 9H (2.6 — 1)
Assuming that the film is under biaxial stress in the (100) plane, the relation®!

et 2C 12

T=-= (2.6 — 2)

10!
r 110 nm CoSig on Si(100)

a b a
10° £ —
2 Fé (311), (311)g_ (400)
> | ! Y
3 1071 Lag
2 cS a
oO
pe

10-#

a9 (°)

FIG. 2.6-1. Fe Kg, x-ray (A = 0.1932 nm) rocking curves of symmetrical (400) and asymmetrical
(311) diffractions from a mesotaxial CoSi2/Si(100) sample. The diffraction geometry and direction

of x-ray incidence are shown in the inset above the corresponding Bragg peaks from the CoSig film.

Table 2.6-I. Lattice distortion of CoSiz films on (100) and (111) oriented Si substrates. Data for
Si/CoSi2/Si samples are from Ref. 49 and that for B-type sample is from Ref. 7.

f =-1.22% — CoSiz/Si(100) Si/CoSig CoSi2/Si(111) Si/CoSiz B-CoSiz/Si(111)
/Si(100) /Si(111)

e+ (%) -2.18 -2.14 -1.69 -1.74 -1.61

ell (%) -0.62 -0.66 -0.72 -0.66 -0.80

AL/L (%) 0.7 0.7 0.5 0.6 0.4

AV/V (%) 0.2 0.2 0.5 0.6 0.5

Table 2.6-II. Ratios and elastic constants (in units of GPa) of cubic CoSi, from strain and curvature

measurements. Data for Si are from Ref. 51 and are listed for comparison.

100) (111) Cy2/Ci Caa/Crur Cun Cia C44

CoSig 0.44 0.32 0.80 0.36 277 222 100
Si 0.28 0.18 0.39 0.48 166 64 80

holds in the linear elasticity theory. From the measured strain (Table 2.6-I) and Eqs. (2.6-1)&(2.6-2),
the ratio C12/Ci1 is obtained (Table 2.6-II). This value (0.80) is about twice that of silicon (0.39).5°

For later convenience, we define a Poisson ratio, v, for thin films under biaxial stress, according to

et —2y

ell ~ 1l-—v

. (2.6 — 3)

This yields 1190) = 0.44 for the CoSig film on Si(100) substrate (Table 2.6-II).

Similarly, symmetrical (111) and asymmetrical (311) x-ray rocking curves were also recorded
for the CoSi2/Si(111) sample. The perpendicular and parallel strain were extracted from the Bragg
peak separations. The results are given in Table 2.6-I, which again agrees well with those for buried
CoSiz films (et = —1.74% and el = —0.66%).*° Furthermore, they are also about the same as those
for MBE-deposited B-type CoSiz films on Si(111) substrates (e+ = —1.61% and «ll = —0.80%).”
This shows that the strain state of thick (> 10nm) epitaxial CoSi films on Si(111) substrates is
independent of the process by which the silicide films are formed, and whether the films are type-A
or type-B. Burger’s vector of the misfit dislocations is b = 1/6 < 112 > for both type-A CoSig
formed by ®°Co implantation*® and type-B films by MBE‘ on Si(111) substrates. The average misfit
dislocation spacing is therefore p = 31 nm, obtained from Table 2.6-I. This is the same as that on
Si(100), implying that the dislocation spacing is independent of substrate orientation.

The areal density p of imperfections such as threading dislocations in epitaxial CoSi, films can

be estimated from the measured x-ray peak broadening Ws using Equation (2.2-7),

2 2
p= we Ws" (2.6 — 4)

9b?
The size broadening Ws can be obtained from the Scherrer Equation (2.2-6). The imperfection
density estimated from Eq. (2.6-4) varies from ~ 2 x 10°/cm? for the (100) and (111) CoSiz films
formed by *°Co implantation (see the peak broadening in Fig. 2.6-1), to < 10°/cm? for the best
MBE-grown B-type CoSiz film on Si(111) that we have measured (see Fig. 2.2-2). However, the
average misfit dislocation spacing is about the same (~ 30 nm) for all samples. This means that
the strain relaxation and the imperfections in the film are unrelated, suggesting that the misfit
dislocations nucleate at interfacial defects such as atomic steps rather than on the surface. We
therefore speculate that specular Si surfaces free of any surface defects such as atomic steps are
needed to grow metastable pseudormorphic CoSig films (> 10 nm). The inference then is that high
dose *°Co-implantation will not produce metastable pseudormorphic CoSig layers because defects

like atomic steps are always present at the silicide/silicon interface in this case. This is unlike the

relaxation of epitaxial GeSi films on Si, where the strain relaxation necessarily yields to threading
dislocations in the film because misfit dislocations nucleate at the surface and glide down to the
interface.5?

In summary, all these observations suggest that the strain relaxation of thick (> 10 nm) epitaxial
CoSiy films is intrinsic to the silicide, and insensitive to the type of the film (A or B), the silicide
formation process (high dose implantation or vacuum deposition), the orientation of the substrate,
the imperfections in the film, and the thickness of the films.® This is in contrast with epitaxial GeSi
films grown on Si substrates, where the misfit dislocation spacing is very sensitive to the growth
temperature and film thickness for a fixed lattice mismatch.®3

The perpendicular strain e+ of the CoSiz film is distinctly smaller on Si(111) than on Si(100)
(Table 2.6-1), showing that single-crystalline CoSig films are elastically anisotropic. This means that
the bond strength between (111) planes is stronger than that between (100) planes. This result is
similar to that of silicon where the covalent bond along < 111 > direction gives rise to the strongest
bond between {111} planes. On Si(111), the relative volume expansion of the CoSiz film is ~ 0.5%,
the same as the average linear dilatation (~ 0.5%, Table 2.6-I).

To extract the second ratio C,4/C; from the measurements on the (111) sample, the procedure
outlined for the (100) case was repeated with Eq. (2.6-1) and a suitably modified Eq. (2.6-2),5!

ew - Cag — (C11 + 2C 2) /2
ell C44 + (Cia + 2C12)/4°

The result is given in Table 2.6-II. This ratio (0.36) is less than that of silicon (0.48).>! The Poisson

(2.6 — 5)

ratio is 1111) = 0.32, obtained from Eqs. (2.6-1)&(2.6-3) and Table 2.6-I. It is the same as that for
MBE-grown B-type CoSiz films on Si(111) substrates (~ 1/3).”
C. Stress and Sample Bending

To obtain the absolute values of the elastic constants, the biaxial tensile stress in the CoSi.
film, of, was estimated by measuring the bending of the CoSi2/Si(100) sample. The stress is related

to the tensile strain in the plane according to Hooke’s law in the linear elasticity,
oy = Byell = By (ell — f), (2.6 — 6)

where By is the biaxial elastic constant of the film. The stress causes the sample to bend with a
concave radius of curvature, R. In the case where the thickness of the substrate, t,, is much larger
than that of the film ¢; and is smaller than the lateral dimension of the sample, the following

relationship holds,**
_ B,t?

t= BRE’ (2.6 — 7)

where B, is the biaxial elastic constant of the substrate. Combining Eqs. (2.6-6) and (2.6-7), one

has ,

Bo I (2.6 — 8)
The radius R was obtained by measuring the angular difference of the (400) Bragg peaks diffracted
from the substrate at two different spots of the sample separated by 4 mm, using a double crystal
diffractometer equipped with a translational stage. Substituting appropriate parameters for the
aforementioned CoSiz/Si(100) sample, we obtain the ratio By/B, = 0.8 from Eq. (2.6-8). Knowing
B, = 180 GPa for Si(100),51 we obtain By; = 144 GPa for CoSi2(100). This value agrees well with
that extracted from thermal stress measurement by van Ommen et al. (140 GPa).®5 It is slightly
larger than the measured biaxial elastic constants of several transition-metal disilicide films (Ti, Ta,
Mo, W) on Si(100) substrates (~ 110 GPa®®). The biaxial elastic constant of (100) oriented films
equals®!

C12 Cie

B=Cn(l + G~ - UG-)’). (2.6 — 9)

From the the measured values of B and C2/C\, for the CoSi2(100) film, the absolute value of C11
can be obtained from Eq. (2.6-9). We thus have all three elastic constants of single crystalline CoSi,
(Table 2.6-II). Lambrecht et al.5” studied theoretically the electronic band structure of CoSig, using
the linear muffin-tin orbital method and calculated the bulk modulus of CoSiz to be 190 GPa. In
comparison, we used the elastic constants in Table 2.6-II and obtained the bulk modulus of 240 GPa,
about 25% larger than this theoretical estimate.
D. Dislocation Locking and Thermal Stress

To extract the linear thermal expansion coefficient of CoSig and to study the thermal stress, we
measured the parallel and perpendicular strains between CoSig films and Si substrates up to 500°C.

The lattice mismatch f between stress-free CoSiz and Si equals

l-—v

fa (Get +(e

from Eqs. (2.6-1) and (2.6-3) (also from Eq. (2.5-2)). Assuming that the Poisson ratio v does
not change with temperature, f can then be extracted from the v obtained at room temperature
(Table 2.6-II) and the measured e+ and el! at various temperatures (Fig. 2.6-2). f decreases linearly
with rising temperature up to 500°C (open and filled circles in Fig. 2.6-2). The slope yields the
difference between the linear thermal expansion coefficients of CoSig and Si. The slope has the same
value, within the experimental error, for both the (100) and (111) samples (Fig. 2.6-2(a) and (b)),

which averages (6.5 + 0.6) x 10-°/°C. This result shows that the thermal expansion coefficient of

CoSi, is isotropic, in accord with the fact that the unit cell of stress-free CoSiz is cubic. The linear
thermal expansion coefficient of bulk Si is known to be 2.9 x 1076/°C between 23°C and 500°C.39
The coefficient for CoSig films is therefore 9.4 x 10-°/°C, in good agreement with that reported for
bulk CoSiz polycrystalline samples (9.4 x 10~°/°C).1 It is smaller than the linear thermal expansion
coefficients of several transition-metal disilicides (Ti, Ta, Mo, W) (~ 15 x 10-§/°C*S).

The parallel strain ¢!! of CoSig films on both Si(100) and Si(111) substrates does not change up
to 500°C (open and filled triangles in Fig. 2.6-2(a) and (b)). This means that the misfit dislocations
do not glide up to 500°C. By extrapolating ¢! and e+ to higher temperatures, we found that they
meet (and consequently f also) at 825°C, for both (100) and (111) samples (Fig. 2.6-2(a) and (b)).
This indicates that the CoSiz film is fully relaxed at ~ 800°C.

E. Synthesis and Model

Given the above results, we propose the following model: (1) the strain in epitaxial CoSi, films
on Si substrates reaches the equilibrium value at a relaxation temperature, Tp; (2) the misfit dislo-
cations do not shear below Tp. According to Matthews and Blakeslee’s strain relaxation model,3®

the equilibrium critical thickness t,, for a pseudomorphic layer is

b ter

For a layer of thickness ty larger than t,,, the equilibrium parallel strain ell, equals?®

ter Int; /b+1
i ~ 2, fi ~
they =f (2.6 — 10)

We apply these predictions to a 110 nm thick CoSiz film on a Si(111) substrate. Assuming Tp =
700°C, the lattice mismatch equals f = —0.78% at this relaxation temperature (Fig. 2.6-3), and
hence the equilibrium critical thickness is 3 nm (b=1/6 < 112 > and vy = 1/3). This value agrees
well with the measured critical thickness of B-type CoSiz grown on Si(111) by MBE at ~ 650°C
(~ 3 nm).® For that same 110 nm thick CoSiz at Tr = 700°C, the equilibrium parallel strain equals
ell, = 0.95f = —0.74% from Eq. (2.6-10), and the perpendicular one equals e+.9 = —0.82% (Fig.
2.6-3). Above Tp, misfit dislocations are generated by either nucleation or multiplication, or both,
to minimize the strain energy so that the equilibrium state maintains (Fig. 2.6-3). Below Tp, the
misfit. dislocations are locked in and e!l remains constant (Fig. 2.6-3). Thermal strain and stress
are generated by the different thermal expansions between the film and the substrate. At room

temperature, ¢!l remains the same (—0.74%) and «+ decreases to —1.70% (Fig. 2.6-3). These

~0.5 er
> a aia. pote cee
x of "825°C
~1.0F a
Eas (a)
g 7h 110 nm CoSiz on Si(100)
o difference in thermal
2 expansion coefficients
% —2.0 =(7.140.6)x1078/°C
\ a i \ ! a | 1
0 200 400 600 800 1000
T(°% )
~0.5 -
ye Pare ettinmeens 202220 oo oss efrrl.,
~-10b 0 gee 7.77" 825°C
§ aa a
3) -
2 f a
SI sar (b)
42 -1.5 110 nm CoSig on Si(111)
A difference in thermal
8 r expansion coefficients
% 2.05 =(5.9+0.6)x107-8/°C
eS eT
0 200 400 600 800 1000
T (°C )

FIG. 2.6-2._ The lattice mismatch, e+ (square), el! (triangle), and f (circle), as a function of
the measurement temperature for both (a) the CoSiz/Si(100) and (b) the CoSip/Si(111) mesotaxial

samples. Open (filled) symbols are for the measurements when the temperature was raised (lowered).

~ | 100 nm CoSi2/Si(111) To

ne L

~ -O.5F

r) L

o L

fel L

q L

2 -1.0-

b| L

> [

2 r

ae} L

9 -1.5+

oy L
Co Ph te
0 500 1000 1500 2000

T( °C)

FIG. 2.6-3. Schematics of the proposed model showing how an epitaxial CoSis film relaxes to an
equilibrium strain state at Tp and above, and that misfit dislocations are locked-in below Tr. Tu
is the melting temperature of CoSiz, and To is the hypothetical temperature at which the lattice

mismatch between CoSiy and Si becomes zero.

estimates agree well with experimental observations (Table 2.6-I). The exact value of ell at room
temperature depends on the relaxation temperature Tr. An increase of Tp from 600°C to 800°C
causes a corresponding increase of f from —0.84% to —0.71%. This change raises ¢!l from —0.80%
to —0.67% according to Eq. (2.6-10). This shows that the lateral lattice mismatch is not sensitive
to the change in Tg and explains the observed apparent universal ell at room temperature (Table
2.6-I).

The relaxation temperature Tip depends on many factors such as the formation process of the
silicides. It varies from ~ 600°C for MBE-grown CoSiz on Si at ~ 600°C to ~ 800°C for the sample
formed by high dose °°Co implantation followed by 1000°C vacuum annealing for 30 min. Fig.
2.6-2(a) also shows that the thermal strain in the layer relaxes slightly after heating in ambient
air at ~ 500°C for ~ 2 hrs, even if the sample had been annealed in vacuum at 1000°C for 30
min. This suggests that thermal annealing in ambient air lowers the relaxation temperature Tp.
This phenomenon is similar to what we observed for MBE-grown B-type CoSig layers on Si(111)
substrates. There the thermal stress also relaxes slightly after thermal annealing in ambient air at
~ 600°C for ~ 2 hrs, but remains unchanged after vacuum annealing at ~ 800°C for 1 hr.” These
observations indicate that atomic transport at the silicide/silicon interface lowers the relaxation
temperature TR (see also discussion in Ch. 2.5).

F. Conclusion

We obtained three elastic constants of cubic CoSig (C11=277, Ci2=222, Cy4=100 GPa) by
measuring the strain and stress in CoSiz films on Si substrates at room temperature using DCD.
X-ray rocking curves were also used to measure the lattice mismatch between the film and substrate
at elevated temperatures up to 500°C. A linear thermal expansion coefficient of 9.4 x 10~°/°C was
derived for CoSiz. The parallel strain at room temperature is about the same (~ —0.7%) for all
the samples, regardless of the silicide formation process and the substrate orientation. It does not
change with temperature up to 500°C. The universal el! was explained by the model that CoSig
films reach an equilibrium strain state at a relaxation temperature Tr (~ 600-800°C) by generation
of misfit dislocations, and the dislocations are locked-in below Tr. We proposed that atomic flux
across the silicide/silicon interface lowers Tr . We also speculate that perfectly flat Si surfaces free

of defects such as atomic steps are needed for the growth of metastable pseudomorphic CoSig films

(> 10 nm).

10.

11.
12.

13.
14,

15.
16.

17.
18.
19.
20

21.

References

M-A.Nicolet, $.S.Lau, in VLSI Electronics, Vol.6, edited by N.G. Einspruch and G.B. Larrabee
(Academic Press, New York, 1983), p. 330.

S.P. Murarka, Silicides for VLSI Applications, Academic, New York, 1983.

G.J. Van Gurp and C. Langereis, J. Appl Phys. 46, 4301 (1975).

. R.T. Tung, J.C. Bean, J.M. Gibson, J.M. Poate, and D.C. Jacobson, Appl. Phys. Lett. 40, 684

(1982).
J.M. Gibson, J.C. Bean, J.M. Poate, and R.T. Tung, Appl. Phys. Lett. 41, 818 (1982).

. J.M. Phillips, J.L. Batstone, J.C. Hensel, and M. Cerullo, Appl. Phys. Lett. 51, 1895 (1987).
. G. Bai, M-A. Nicolet, T. Vreeland, Jr., Q. Ye, and K.L. Wang, Appl. Phys. Lett. 55, 1874

(1989).

. G. Bai, M-A. Nicolet, T. Vreeland Jr., J. Appl. Phys., May 1, 1991.
. G.Bai, D.N. Jamieson, T.Vreeland Jr., and M-A.Nicolet, Mat. Res. Soc. Symp. Proc. 102, 259

(1988).

H. von Kanel, J. Henz, M. Ospelt, J. Hugi, E. Miller, N. Onda, and A. Gruhle, Thin Solid
Films 184, 295 (1990).

R.T. Tung and J.L. Batstone, Appl. Phys. Lett. 52, 648 (1988).

T.L. Lin, R.W. Fathauer, P.J. Grunthaner, and C. d’Anterroches, Appl. Phys. Lett. 52, 804
(1988).

R.T. Tung, F. Schrey, and S.M. Yalisove, Appl. Phys. Lett. 55, 2005 (1989).

A.E. White, K.T. Short, R.C. Dynes, J.P. Garno, and J.M. Gibson, Appl. Phys. Lett. 50, 95
(1987).

J.C. Hensel, R.T. Tung, J.M. Poate, and F.C. Unterwald, Appl. Phys. Lett. 44, 913 (1984).
C.W.T. Bulle-Lieuwma, A.H. van Ommen, and L.J. van IJzendoorn, Appl. Phys. Lett. 54, 244
(1989).

R.T. Tung, Phys. Rev. Lett. 52, 461 (1984).

J.C. Hensel, A.F.J. Levi, R.T. Tung, and J.M. Gibson, Appl. Phys. Lett. 47, 151 (1985).

L.F. Mattheiss and D.R. Hamann, Phys. Rev. B37, 10623 (1988).

Y.C.Kao, K.L.Wang, E de Fresart, R.Hull, G.Bai, D.N.Jamieson, and M-A. Nicolet, J. Vac.
Sci. Technol. B6, 745 (1987).

A.Y. Cho and J.R. Arthur, in Progress in Solid State Chemistry, Vol. 10, eds. G. Somorjai and

22.

23.
24.

25.

26.
27.

28.

29.
30.
31.
32.
33.

34,

35.
36.
37.
38.
39.
40.
41.
42,
43.

44.

J. McCaldin (Pergammon, New York, 1975), p. 157.

A. Ishizaka, K. Nakagawa, and Y. Shiraki, in Proc. of Conf. on Molecular Beam Epitaxy and
Clean Surface Techniques, (Japanese Society of Applied Physics, Tokyo, 1982) p. 183.

Q.Ye, T.W. Kang, K.L. Wang, G. Bai, and M-A. Nicolet, Thin Solid Films 184, 269 (1990).
D.N. Jamieson, G. Bai, Y.C. Kao, C.W. Nieh, M-A. Nicolet, and K.L. Wang, Mat. Res. Soc.
Symp. Proc. 91, 479 (1987).

W.K. Chu, J.W. Mayer, and M-A. Nicolet, Backscattering Spectrometry, Academic, New York,
1978.

J. Hornstra and W.J. Bartels, J. Cryst. Growth 44, 513 (1978).

P. Hirsch, A. Howie, R.B. Nicholson, and M.J. Whelan, Electron Microscopy of Thin Crystals,
Krieger Publishing Inc., Malarbar, Florida, 1977.

L.C. Feldman, J.W. Mayer, and S.T. Picraux, Materials Analysis by Ion Channeling, Academic,
New York, 1982.

G. Bai, M-A. Nicolet, J.E. Mahan, and K.M. Geib, Phys. Rev. B 41, 8603 (1990).

G. Bai and M-A. Nicolet, unpublished.

B.E. Warren, X-ray Diffraction (Dover, New York, 1990), p. 253.

P. Gay, P.B. Hirsch, and A. Kelly, Acta Metal. 1, 315 (1953). |

D. Cherns, in Analytical Techniques for Thin Films, eds. K.N. Tu and R. Rosenberg (Academic,
San Diego, 1988), p. 297.

N.C. Bartelt, E.D. Williams, R.J. Phaneuf, Y. Yang, and S. Das Sarma, J. Vac. Sci. Technol.
A7, 1898 (1989).

G. Bai and M-A, Nicolet, unpublished.

R.T. Tung and F. Schrey, Phys. Rev. Lett. 63, 1277 (1989).

H. Nagai, J. Appl. Phys. 45, 3789 (1974).

J.W. Matthews and A.E. Blakeslee, J. Cryst. Growth 27, 118 (1974).

H.F. Wolf, Silicon Semiconductor Data, (Pergamon Press, New York, 1969) p. 101.

H.Lucas, H.Zabel, H.Morkoc and H.Unlu, Appl. Phys. Lett. 52, 2117 (1988).

K.Nakajima, S.Yamazaki,S.Komiya, and K.Akita, J. Appl. Phys., 52, 4575 (1981).
G.A.Rozgonyi, P.M.Petroff, and M.B.Panish, J. Appl. Phys., 64, 1173 (1988).

C.M. Comrie and J.M. Egan, J. Appl. Phys. 64, 1173 (1988); C.M. Comrie, J.C. Liu, L.S.
Huang and J.W. Mayer, J. Appl. Phys. 63, 2403 (1988).

D.G. Deppe and N. Holonyak, J. Appl. Phys., 64, 1293(1988) and references therein.

45.
46.
47.

48.
49.

50.
51.
52.
53.

54.

55.

56.
57.

S.M.Prokes, W.F.Tseng, and A.Christou, Appl. Phys. Lett. 53, 2483 (1988).

R.T.Tung, J.M.Gibson, and J.M.Poate, Appl. Phys. Lett. 42, 888 (1983).

§.S.lyer, J.C.Tsang, M.W.Copel, P.R.Pukite, and R.M.Tromp, Appl. Phys. Lett. 54, 219
(1989).

M. Bartur and M-A. Nicolet, Appl. Phys. A29, 69 (1982).

J.M. Vandenberg, A.E. White, R. Hull, K.T. Short, and S.M. Yalisove, J. Appl. Phys. 67, 787
(1990).

S.M. Yalisove, R.T. Tung, and D. Loretto, J. Vac. Sci. Technol. A7, 1472 (1989).

W.A. Brantley, J. Appl. Phys. 44, 534 (1973).

R. Hull, J.C. Bean, D.J. Werder, and R.E. Leibenguth, Appl. Phys. Lett. 52, 1605 (1988).
J.C. Bean, L.C. Feldman, A.T. Fiory, S. Nakahara, and I.K. Robinson, J. Vac. Sci. Technol.
A2, 436 (1984).

R.F.S. Hearmon, An Introduction to Applied Anisotropic Elasticity (Oxford University, London,
1961), Chap. VII.

A.H. van Ommen, C.W.T. Bulle-Lieuwma, and C. Langereis, J. Appl. Phys. 64, 2706 (1988).
T.F. Retajezyk, Jr. and A.K. Sinha, Thin Solid Films, 70, 241 (1980).

W.R.L. Lambrecht, N.E. Christensen, and P. Bléchl, Phys. Rev. B36, 2493 (1987).

Chapter 3 Epitaxial ReSi, Films on Si

3.1 Semiconducting silicides

Several metal silicides have been identified to be semiconductors.! These semiconducting silicides
have potential applications as infrared light sources and detectors in Si-based microelectronics. Some
optical and electrical properties of such silicides such as FeSig,?3 CrSiz,4® and ReSi2®” have been
characterized. The compatibility of silicides with Si process makes semiconducting silicides the ideal
materials for novel Si-based optoelectronic integrated circuits, equalled in their simplicity only by
GeSi and SiC.

The epitaxial growth of FeSis,8-1° CrSiz,!!-!3 and ReSi,!415 films on clean Si substrates in
ultrahigh vacuum have been investigated. Iron silicides of FegSi, FeSi, and FeSig (both a- and /-
phases) can be formed by solid-state reaction of iron films on silicon substrates at 200-1100°C.1®7
§-FeSig is thermodynamically stable at < 900°C(a-FeSiz at > 900°C). It is a direct gap (0.87 eV)
semiconductor.” A thin film of 3-FeSig formed by vacuum deposition without intentional doping is
extrinsic at room temperature with p-type conduction. Resisitivity, hole concentration and mobil-
ity, are ~ 2 Qem, ~ 10!8/cm? and ~ 3 cm?/Vs.?> G-FeSi2 is orthorhombic with lattice constants
of a=0.986 nm, b=0.779 nm, c=0.783 nm, and has a structure like that of CoSiz.! The epitaxial
growth of @-FeSi, films on clean $i(100) and Si(111) substrates in ultrahigh vacuum has been ana-
lyzed by reflection high energy electron diffraction, Auger electron microscopy, transmission electron
microscopy, x-ray diffraction, and backscattering spectrometry.3~!° On Si(111), two epitaxial rela-
tionships were observed:® (I) $-FeSi2(101)//Si(111) and (II) @-FeSi2(110)//Si(111). Furthermore,
Si(111) has threefold symmetry, while the @-FeSiz matching plane has only twofold symmetry. There-
fore, three equivalent and equiprobable azimuthal orientations are expected and were observed.® The
azimuthal relationship is @-FeSi[010](or [001])//Si< 011 >. On Si(100), epitaxial relationship is
B-FeSi2(100)//Si(100).9:1° However, two competing, inequivalent azimuthal orientations which differ
by a rotation of 45° were observed.’° Type-A orientation has FeSi2[010]//Si< 011 > and predomi-
nates at 300-550°C.° Type-B orientation has FeSi2[010]//Si< 001 > and occurs at lower (200-300°C)
ot higher (550-700°C) temperatures.1°

CrSiz has an indirect gap of 0.35 eV.° Unintentionally doped CrSig is a degenerate p-type semi-
condcutor at room temperature.** Resisitivity, hole concentration and mobility, are ~ 0.01 Qem,
~ 4x 10!9/cm3 and ~ 15 cm?/Vs.4> CrSiz has a hexagonal structure with a=0.443 nm and c=0.636

nm.’ Epitaxial CrSiz films were grown on Si(111) in ultrahigh vacuum.!!-13 The matching face

relationship is CrSi2(0001)/Si(111).!1~-1% Two competing, inequivalent azimuthal orientations were
observed: type-A with CrSig{1010]//Si[101] and type-B with CrSi2[1120]//Si[{101].!! Type-B is ob-
tained by a rotation of 30° of type-A about the [111] axis. Recently, the growth of purely type-A
epitaxial layers was demonstrated by means of chromium and silicon codeposition onto Si(111) at
450°C.1?

ReSi. has an indirect bandgap of 0.12 eV, and is a p-type semiconductor without inten-
tional doping.*'” The electrical resistivity, hole concentration and mobility at room temperature
are ~ 0.02 Qem, ~ 4x 101°/cm? and ~ 10 cm?/Vs.° ReSiz has a body-centered orthorhombic lattice
with lattice constants of a=0.313 nm, b=0.314 nm, c=0.768 nm, and closely resembles the tetrag-
onal MoSiz structure type.!® Localized epitaxy of ReSig was observed upon vacuum annealing of
evaporated rhenium films on Si(111) and (100) substrates at ~ 1000°C.!° Recently, highly epitaxial
ReSig films 150 nm thick were grown on Si({100) by rhenium deposition onto a hot substrate at

650°C.1415 The next section focuses on the growth and characterization of such films.

3.2 Growth and characterization of ReSi, on $i(100)

Thin films of ReSig were grown on Si(100) substrates at Colorado State University.!45 The Si
surface was cleaned by etching a Si wafer for 30s in buffered HF solution, and the wafer was then
loaded into an ultrahigh vacuum chamber (base pressure 10~1! Torr).14 A “silicon beam clean” 2°
was then applied to remove the residual oxide, which amounts to exposure of the surface to a silicon
flux corresponding to a deposition rate of 1 nm/min for 4min at 700°C. Rhenium (Pure Tech, Inc.,
99.99%) from an electron beam evaporation source was deposited onto a hot silicon substrate at
400-1000°C at a pressure in the mid-10~° Torr range (“reactive deposition epitaxy,” or RDE).!4

The atomic composition and thickness of the formed silicide films were characterized by MeV
backscattering spectrometry (see Ch. 2.2). All films (~65-530 nm thick) are of uniform composition
through their entire depth and have the correct Re:Si1:2 stoichiometry. The epitaxial quality of
films was analyzed by channeling spectrometry (see Ch. 2.2).2) The minimum channeling yields
for both the silicon and rhenium signals of ReSi2 films as a function of the deposition temperature
are plotted in Fig. 3.2-1, which shows that highly epitaxial ReSig films can be grown on Si(100) at
600-700°C. The data at 650°C are for films of three different thicknesses, 65, 150, and 530 nm. The
150 nm thick ReSiz film grown at the optimum deposition temperature of 650°C has a minimum
channeling yield of 2% for rhenium (Fig. 3.2-1). This yield is comparable with that of the best

epitaxial silicide films (see Ch. 2.2) and means that the film is in perfect alignment with the

ReSig films on Si(100)

100 =
r o -Si (film)
80/- A —Re (film)
ye 60
# aol
20
O | r J i | l |
400 600 800 1000

Deposition Temperature ( °C )

FIG. 3.2-1 2 MeV *He minimum channeling yields of Re and Si for ReSi films grown on Si(100)

by reactive deposition epitaxy as a function of substrate temperature during deposition.

substrate. We also noted that the minimum channeling yield of silicon is always larger than that of
thenium for the same sample. This does not mean that the silicon sublattice is less ordered than that
of rhenium, but is a feature of ion channeling in a polyatomic crystal.?? We discuss this phenomenon
in the next section.

Bragg-Brentano x-ray diffraction of ReSiz films grown at 650°C shows only the (020) peak from
ReSiz, meaning that the film is highly oriented, with a matching face of ReSi2(010)//Si(100).15 The
same matching relationship was observed for epitaxial tetragonal MoSiz films grown on Si(100) at
650°C by RDE in ultrahigh vacuum.?? For films grown at higher temperature, other matching faces,
such as ReSi2(011), appear.'®

The S$i(100) plane has 4-fold symmetry, while that of ReSi2(010) is only 2-fold. Two distinct
but equivalent azimuthal alignments are therefore equally probable, so that an epitaxial ReSi» film
is expected to have twins that differ by a rotation of 90 degrees about the ReSi[010] axis. Fig. 3.2-
2(a) is an electron diffraction pattern of the ReSi2[010] zone.!® It has 4-fold symmetry, although the
reciprocal lattice of the ReSig[010] zone has only 2-fold symmetry. This clearly verifies the presence
of a twin structure. The relationship between the patterns of ReSig and Si shows that the azimuthal
alignment is ReSi2[001]//Si[0T1] or (011].!° The twinning and azimuthal relationship observed here
are again identical to those of tetragonal MoSiz on Si(100).?? Twinning is inherent in the epitaxial
growth of ReSiz on Si(100) because substrate surface has a higher symmetry than that of the ReSig
matching face. One way to overcome this symmetry mismatch may be to use a vicinal Si(100) to
lower the symmetry of the substrate surface.

Given above results, Mahan et al.!° proposed a common unit mesh for the ReSi2/Si(100) het-
erostructure shown in Fig. 3.2-2(b). The lattice match at room temperature is excellent along the
c-axis (—0.04%), and fair along the a-axis (1.8%). The mismatch along the a-axis is better than
that along the b-axis (2.3%), which probably is the reason that the matching face is ReSi(010),
not (100).1° All these three lattice mismatches become positive and worsen at the growth temper-
ature, because of the larger thermal expansion coefficient of ReSiy (6.6x10-°/°C') than that of Si
(2.9x10-§/°C).

Transmission electron microscopy (TEM) was used to characterize the microstructure of the
ReSi2 films grown at 650°C. Fig. 3.2-3(a) is a plan-view TEM dark-field image, which reveals
domains with an average size of 10 nm. Cross-sectional TEM (Fig. 3.2-3(b)) shows that the ReSig

film consists of tall columns of ~ 10 nm in diameter.

Reflection high energy electron diffraction patterns indicate that the ReSi2 films grown by RDE

A LATTICE MATCHING FOR
EPITAXIAL - 7+, ReSi,(010)/Si(001)

2 +

FIG. 3.2-2 Epitaxial relationship between a ReSiz film and a Si(100) substrate: (a) Transmission
electron diffraction pattern of the Si{100] zone and the ReSi2[010] zone (from Ref. 14). The bright
spots are Si diffractions. The faint spots are ReSig diffractions (note their fourfold symmetry). (b)
Schematical drawing of two equivalent common unit meshes (from Ref. 15). The dots are Si atoms

in the substrate. The crosses are Re atoms in the film. The actual atomic positions are unknown

and assumed.

FIG. 3.2-3 Transmission electron micrograph of a 150 nm thick ReSiz film grown on Si(100) at
650°C (from Ref. 15): (a) plane-view dark-field image with a beam diffraction vector of ReSi2(002);

(b) cross-sectional bright-field image with a beam incident along Si[022].

have rough surfaces.!® The surface morpholgy might be improved by using more sophisticated growth
procedures involving “template” and codeposition. Another possible technique to fabricate ReSi2

films is by high dose implantation of Re into Si substrate and thermal annealing.

3.3 Channeling of MeV ions in polyatomic crystals

Channeling phenomena of MeV ions in crystals and their applications to materials characteriza-
tion have been extensively reviewed.?!:24 Most experimental results on channeling can be understood
in the framework of the continuum model established by Lindhard.?®> Two fundamental channeling
parameters, the critical angle, ,, and the minimum yield, ymin, have been throughly studied for
various ions (e.g., He and H) over a wide energy range (0.1-10 MeV) in numerous monoatomic
crystals. For example, in diamond-type crystals, Picraux et al?® found that the experimental half
angle (measured from angular scan), 4, /2, has the same functional dependence on parameters such
as ion energy and atomic number as the critical angle ~, of the continuum model. Except for a
few perfect crystals such as Si and Ge,?® the measured minimum yields are usually larger than the
predicted ones because of the existence of imperfections. The sensitivity of ymin to the defects in
crystals makes it a good indicator of crystal perfection.

Channeling in polyatomic crystals has distinctive characteristics owing to the existence of dis-
tinct sublattices. The differences in channeling behavior from different sublattices are pronounced
when the atoms occupying the sublattices have very different atomic numbers.2*:2° While these
phenomena are of fundamental interest, for practical reasons, there have been relatively few experi-
mental investigations of channeling in polyatomic crystals. For example, for bulk single crystals, it
is difficult with backscattering analysis alone to detect the signal from light elements in the presence
of heavy ones. One remedy is to measure the close encounter probability of an incident beam with
the light element by detecting the characteristic x-ray production or nuclear reaction products from
the light element in the channeling mode, while measuring the backscattered beam from the heavy
element.?’~3° Another remedy is to use high quality epitaxial thin films so that the signals from light
and heavy elements do not overlap. We report here on a comprehensive and revealing experimental
study of this type.

We have channeled a *He beam of 1.4—2.7 MeV into an epitaxial ReSia film 150 nm thick grown
on a Si(100) substrate at 650°C by RDE.'* Since the difference between the a- and b-axes of ReSiz
is only ~ 0.5%, we will hereafter assume for convenience that ReSiz is tetragonal with a=0.313 nm.

For a sufficiently thin film, the backscattering signal of the heavy (Re) element does not overlap

in energy with that of the light (Si) element, which enables us to separate clearly the Si signal
from the Re signal in the backscattering spectrum. Thus the half angles and the minimum yields
of both elements can be obtained from the backscattering measurements alone. This allows us to
compare directly the channeling characteristics of different components of a polyatomic crystal. The
experimental results are discussed in the framework of the continuum model, suitably extended for
polyatomic crystals. The agreement is found to be good for the critical angles but only fair for the

predicted minimum yields.

The channeling measurements were performed at room temperature with an x-y rotation and
x-y translation goniometer. The vertical y rotation axis is fixed in space and is perpendicular to the
horizontal *He beam. The x rotation axis lies in a horizontal plane and moves with the y rotation.
The x-y translation directions are parallel to the corresponding rotation axes. First, the MeV *He
beam was aligned with the [100] axial channel of ReSiy by finding the minimum backscattering
yield in a spectral window extending from beneath the surface peak of the Re signal to the energy
corresponding to ~ 100 nm in depth. The same [100] axial channel was found when the window
was placed in the Si signal of the silicide or of the Si substrate instead of in the Re signal. This
concurrence means that the [100] directions of the Si substrate and the ReSig film are exactly aligned.
Once the [100] channel had been identified, a channeling spectrum was then taken by translating
the sample to a virgin spot (beam size ~ 0.4 x 0.4cm?, sample size ~ 2 x 2cm?) to minimize the
effect of radiation damage.*! Figure 3.3-1 shows the [100] axial channeling spectrum for a 1.4 MeV
4He beam. One sees that in the ReSi, film, the minimum yield of Si much exceeds that of Re. The
factor is seven—14% vs. 2%. Channeling spectra were taken with several energies (1.4, 2.0, 2.4, 2.7
MeV) and the minimum yields are the same. This result is consistent with the continuum model,
which predicts that the minimum yield is independent of beam energy and is only a property of the

host crystal.

The half angles of the Re and the Si in the film andthe Si in the substrate were obtained from
angular scan measurements, which were performed by tilting the sample about either the x or y
rotation axis and recording the backscattering yields within a spectral window corresponding to a
thickness of ~ 100 nm placed in the proper region. The half angles measured with a single channel
just beneath the surface peak are the same as those measured with a spectral window, to within
experimental error. To minimize the effect of radiation damage, we used a spectral window instead
of a single channel for the angular scan measurements, and we started each measurement on a virgin

sample spot. The half angles obtained by either x or y rotation are the same. Figure 3.3-2 shows

1000 1.4 MeV *He

800

600

Counts

400

200

FIG. 3.3-1 Backscattering spectra of 1.4 MeV ‘He incident along a random direction (solid line)

and a [100]-aligned axial direction (dotted line).

r 150 nm ReSig/Si(100)
| eee [100] aligned

- ——- random

1.0
Energy ( MeV )

1.4 MeV “*He: 150 nm ReSig/Si(100)

1.0
cs)
GC 0.8
Co
% 0.6 half angle (°)
= L © Si (sub.) Si (sub.) 0.62 |
£0.4- ¢ Si (film) Si (film) 0.65 _]
hy
S | o Re (film) Re (film) 1.54 4
0.2 4
i 1 I L | 1 | L ]
-6 —4 ~2 0 4 6

FIG. 3.3-2 A plot of normalized backscattering yield versus tilt angle. The normalization is

performed with respect to the backscattering yield of a random incident beam. The half angle is

Tilt Angle ( °

the half of the full width of the angular dip.

the [100] angular scans for a 1.4 MeV *He beam. Two results immediately follow: (1) the half angle
of Re is ~ 2.3 times larger than that of Si in ReSig; (2) the half angles of Si in the film and in the
substrate are about the same. The half angles were measured for several beam energies (1.4, 2.0, 2.7
MeV) and the same conclusions were derived. Furthermore, all half angles decrease with increasing
energy, proportionally to 1/VE , as predicted by the continuum model.

The continuum model of axial channeling by Lindhard?® represents the atomic chains along the
channel of an elemental host crystal as continuous columns of radius rmin and electrostatic potential
U(r), where r is the distance from the axis of the column. rmin is the minimum distance a channeled
ion may approach for the continuum model still to validly describe the gentle interaction between
the ion and host crystal that keeps the ion channeled. There are two characteristic lengths: (1)
the Thomas-Fermi screening distance, a,,; and (2) the thermal vibrational amplitude of the host
crystal, p. Both are of the order of 0.01 nm in a typical channeling experiment. Lindhard has taken
Tmin? to be the sum of a,” and p?.

The meaning of rmin is that an ion of the incident beam aligned with a channel will be dechan-
neled if it impinges on the ends of the columns at the surface. Therefore, the minimum yield ymin
measured immediately beneath the surface peak for a perfect crystal is just a fraction of the surface
area occupied by the columns,

Xmin = ORT min’ (3.3 - 1)

where o is the areal density of surface atoms.
The existence of rmin also implies that there is a critical angle , between an ion’s trajectory
and a channel above which the ion will be dechanneled,”®
Pmin
ve = a(——) - v1, (3.3 — 2)
Gop

where the numerical coefficient w has a typical value between 0.6 and 1.676. The characteristic angle,

_ 2Z,Z2e7 °
wey Bq (3.3 — 3)

where Z; and 22 are the atomic numbers of the ion and the host atom, E is the ion energy and d

1, equals

is the atomic spacing in the channel direction.

In general, axial channeling in a diatomic crystal is complex because more than one type of
columns exist. However, there are special directions where channeling is simple. The simplest
channeling direction has only one type of column, which consists of both elements (e.g., the [001]

channel in ReSig for which we have no data). Channeling can then be viewed as being equivalent to

that in a monoatomic crystal with an average atomic number and spacing.?° The minimum yields
and the critical angles of the two elements are then the same. The next simplest channeling direction
has two types of columns such as the [100] channel in ReSiz which we discuss in detail and for which
we do present data.

The [100] channel consists of two types of columns (Fig. 3.3-3); one type contains only Si
atoms (Si columns) and another only Re atoms (Re columns). We assign a minimum distance of
approach for each type of column, ri, and r®¢_, which is determined by the characteristics of each
corresponding column. This is the single column potential approximation.

For quantitative evaluations of minimum yields and critical angles, it is necessary to estimate

the values of rmin in terms of channeling parameters of the ion and the host crystal. In general,

Tmin May depend on both the Thomas-Fermi screening distance?”

Opp = 0.047(/Z, + VZ2)-? nm, (3.3 — 4)

and the thermal vibrational amplitudes of the host crystal. For an incident He (Z, = 2) ion scat-
tered by Si (Z2 = 14) and Re (Z, = 75) atoms, the screening distances are 0.016 nm and 0.011
nm respectively. For elemental Si and Re crystals, the known Debye temperatures 8, St = 550K
and 9 pt = 300K yield thermal vibrational amplitudes of 0.011 nm and 0.008 nm, respectively,
at room temperature, in the Debye approximation. These numbers are close to the values of the
screening distances. To our knowledge, there are no reliable data on the thermal vibrational ampli-
tudes in ReSiz. Therefore, we shall approximate ryin by a,, in computing the critical angles and
the minimum yields.

In our single column potential approximation with rmin = @,,, the critical angles can be directly
calculated from eq. (3.3-2). The numerical coefficient a equals 0.8 for both Re and Si columns using
the “standard potential.”*! The ratio of the critical angle of the Re columns, ,"*, to the critical

angle of the Si columns, we, thus becomes

be [be = OF / bi = V/Zre/Zsi = 2.3, (3.3 — 5)

which agrees with the experimental data (see Fig. 3.3-2). Table 3.3-1 lists the values of the charac-
teristic angle ~, the calculated critical angle y,. and the measured half angle 1/2 for Si and Re in
the ReSiz film. The measured half angles agree well with the critical angles.

The minimum yield describes the fraction of an aligned incident beam that is dechanneled by

atomic columns. An aligned beam impinging on ReSiz in the [100] direction will be scattered by

ReSig [100] axial channel

ReO O—9

1.4 MeV Sie . °

“He O—9

c=0.767 nm 2 9 ———

FIG. 3.3-3 A cross-sectional schematic diagram showing the [100] axial channel of ReSig.

Table 3.3-I Calculated values of the characteristic and critical angles, measured values of half
angles, and the calculated and measured minimum yields, for the [100] axial channeling of 1.4 MeV

4He beam in an epitaxial ReSi film.

‘¢1(°) ve(?) v172 (°) Xmin,1(%) Xmin,2(%) Xmin(%) Xmin(%)
Si 0.78 0.65 0.65 3 10 13 14
Re 1.81 1.52 1.54 1 0 1 2

both Si and Re columns. The incident ion will be channeled by both Si and Re columns if the
scattering angle, 7%, is less than the critical angles of both the Si columns and the Re columns. For
the incident ion that is close to a Si column, the ion will be dechanneled by the Si columns but will
still be channeled by the Re columns if the scattering angle ~ is greater than the critical angle of the
Si columns #,** but is still less than the critical angle of the Re columns ¥,”°. That dechanneling
process thus does not contribute to the minimum yield of Re. Conversely, an incident ion that is
close to a Re column will be channeled by Re columns but not by the Si columns if the scattering
angle w is between the critical angles of the Si columns and the Re columns (i.e., weit >y> we?).
That channeling process thus contributes to the minimum yield of Si. The minimum yield of the Si
columns is therefore larger than that of the Re columns.

According to the previous reasoning, the minimum yield of Re columns, xFe, consists of two
contributions, backscattering from Re atoms originating from the dechanneling when incident ions
impinge on Re columns, y*¢,, 1) and backscattering from Re atoms originating from the dechanneling

when incident ions impinge on Si columns, y*é,, ,

Xinin = Xin +XxBbn o- (3.3 — 6)

The values of these minimum yields can be estimated from appropriate modification of Eq. (3.3-
1). Experimental results of channeling in elemental crystals indicate that Eq. (3.3-1) underestimates
the value of the minimum yields.?! Furthermore, Monte Carlo simulation of channeling phenomena
by Barrett** indicates that the minimum yield extracted from computer simulation is about 3 times
greater than that estimated from Eq. (3.3-1) and is in good agreement with the experiemntal data
for perfect crystals such as Si and Ge.?! We therefore modify Eq. (3.3-1) by a multiplying factor of
3 in order to get a better numerical estimation of the minimum yields. pena thus can be estimated
by
Xmnin,t = 30ReMmin,Re *s | (3.3 —7)

where rmin,re is the minimum distance of approach to Re columns. To get xh, 9, we need to know

the approaching distance of the incident ion to the Si columns, rs;, when the scattering angle ~

equals the critical angle of the Re columns,
b(rsi) = ve. (3.3 — 8)

Knowing rs;, we obtain

Xinin 2 = 305; r2;. (3.3 — 9)

The approaching distance rs; obtained from Eq. (3.3-8) with the “standard potential”?! is much
smaller than rmin,re (~ 10-> nm vs. ~ 107? nm), and hence xRn is negligible compared to

xR en (see Table 3.3-1). Therefore, one has
xBe, MXR, | Sone TO pe 1%. (3.3 — 10)

This predicted value of y®¢, is only about half of the measured value. One explanation is that
the ReSiy sample contains imperfections. It is not a single crystalline film. Another is that the
approximation of rin by a,, is inaccurate.

Similarly, the minimum yield of the Si columns, y5%,,, can be obtained,

Xmin = Xinin,r + Xprin,g = 30S:T min 2s; + 30 Re TT Re. (3.3 — 11)
However, here the rg, obtained from the equation
Uerre) = Ve, (3.3 — 12)

is larger than rmin,s; SO Ximin,2 is the major contribution to the minimum yield of the Si columns.
Table 3.3-I lists the calculated and the measured values of the minimum yields for both Si and Re
columns. Compared to the Re column, the relatively good agreement between the measured value
and the estimated value of the minimum yield for the Si column is probably due to its relative
high value and hence its relative insensitivity to the imperfections in the film. This suggests that
one should use only the minimum yield for the heavy (Re) element as an indicator of crystalline
perfection, in particular, in the case of low defects density.

The continuum model, as extended here to polyatomic crystals, is able to explain the channeling
phenomena that took place along directions for which there are more than one type of column as
we have observed in ReSig. The critical angle of each column is determined only by the parameters
of that column. The minimum yields are determined by the parameters of all columns, but are
dominated by the dechanneling from the column with the largest average atomic number. These
findings are readily generalized to channeling in polyatomic crystals with a number of different
types of elemental columns, and from there to columns with different average atomic numbers and
spacings. From the point of view of applying channelling to characterize the crystalline perfection of
a polyatomic crystal, an important corollary is that high crystalline quality is not synonymous with
a low minimum yield for the light element. In particular, a high value of the minimum yield for the

light element does not necessarily mean that the sublattice of the light element is disordered.

10.

11.

12.

13.

14.

15.
16.
17.
18.
19.
20.
21.

References

. M-A. Nicolet and $.S. Lau, in VESI Electronics: Microstructure Science vol. 6, eds. N. Einspruch

and G. Larrabee (Academic, New York, 1983), p. 391.
M.C. Bost and J.E. Mahan, J. Appl. Phys. 58, 2696 (1985).
K. Lefki, P. Muret, N. Cherief, and R.C. Cinti, J. Appl. Phys. 69, 352 (1991).

. D. Shinoda, S. Asanabe, and Y. Sasaki, J. Phys. Soc. Jpn. 19, 269 (1964).

M.C. Bost and J.E. Mahan, J. Appl. Phys. 63, 839 (1987).
C. Krontiras, L. Gronbero, I. Suni, F.M. D’Heurle, T. Tersoff, I. Engstrom, B. Karlsson, C.S.
Petersson, Thin Solid Films 69, 119 (1988).

. R.G. Long, M.C. Bost, and J.E. Mahan, Thin Solid Films 162, 29 (1988).

N. Cherief, C.D’Anterroches, R.C. Cinti, T.A. Nguyen Tan, and J. Derrien, Appl. Phys. Lett.
55, 1671 (1989).

. J.E. Mahan, K.M. Geib, G.Y. Robinson, R.G. Long, X.H. Yan, G. Bai, M-A. Nicolet, M. Nathan,

Appl. Phys. Lett. 56, 2126 (1990).

K.M. Geib, J.E. Mahan, R.G. Long, M. Nathan, G. Bai, and M-A. Nicolet, J. Appl. Phys. (in
press).

R.W. Fathauer, P.J. Grunthaner, T.L. Lin, K.T. Chang, J.H. Mazur, D.N. Jamieson, J. Vac.
Sci. Technol. B6, 708 (1988).

L. Haderbache, P. Wetzel, C. Pirri, J.C. Peruchetti, P. Bolmont, and G. Gewinner, Surf. Sci.
209, L139 (1989).

J. Mahan, K.M. Geib, G.Y. Robinson, G. Bai, M-A. Nicolet, and M. Nathan, J. Vac. Sci.
Technol. B (in press).

J.E. Mahan, K.M. Geib, G.Y. Robinson, R.G. Long, X.H. Yan, G. Bai, M-A. Nicolet, M. Nathan,
Appl. Phys. Lett. 56, 2439 (1990).

J.E. Mahan, G. Bai, M-A. Nicolet, R.G. Long, and K.M. Geib (unpublished).

S.S. Lau, J.S-Y. Feng, J.O. Olowolafe, and M-A. Nicolet, Thin Solid Films 25, 415 (1975).
H.C. Cheng, T.R. Yew, and L.J. Chen, J. Appl. Phys. 57, 5246 (1985).

T. Siegrist, F. Hulliger, and G. Travaglini, J. Less-Comm. Metals 92, 119 (1983).

J.J. Chu, L.J. Chen and K.N. Tu, J. Appl. Phys, 62, 461 (1987).

M. Tabe, Jpn. J. Appl. Phys. 21, 534 (1982).

L.C. Feldman, J.W. Mayer, S.T. Picraux, Materials Analysis by Ion Channeling, Academic
Press, New York, 1982.

22
23

24.
25.
26.

27.
28.
29.
30.
31.

32.
33.

34,

. G. Bai, M-A. Nicolet, John E. Mahan and Kent M. Geib, Phys. Rev. B 41, 8603 (1990).

. A. Perio, J. Torres, G. Bomchil, F. Arnaud d’Avitaya, and R. Pantel, Appl. Phys. Lett. 45, 857
(1984).

D.S. Gemmell, Rev. Mod. Phys. 46, 129 (1974)

J. Lindhard, Dansk. Vid. Selsk. Mat. Fys. Medd., 34, 14 (1965).

S.T. Picraux, J.A. Davies, L. Eriksson, N.G.E. Johansson, and J.W. Mayer, Phys. Rev., 180,
873 (1969).

D.S. Gemmell and R.C. Mikkelson, Phys. Rev. B6, 1613 (1972).

L. Eriksson and J.A. Davies, Ark. Fys. 39, 439 (1972).

F.W. Clinard Jr. and W.M. Sanders, J. Appl. Pyhs. 43, 4937 (1972).

O. Meyer, Nucl. Instr. Meth. 149, 377 (1978).

G. Bai, M-A. Nicolet, J.E. Mahan, K.M. Geib, and G.Y. Robinson, Appl. Phys. Lett. 57, 1657
(1990).

O.B. Firsov, Soviet Phys. JETP 6, 534 (1957).

International Tables for X-ray Crystallographs, Vol. II (Kynoch Press, Birmingham, England,
1959), pp. 234, 239.

J.H. Barrett, Phys. Rev., 3, 1527 (1971).

Chapter 4 GeSi Films on Si(100)

4.1 Introduction

Silicon and germanium both have a diamond cubic structure, and form a solid solution over the
entire composition. The lattice mismatch is 4.2% at room temperature.! The difference between the
thermal expansion coefficient of Ge and Si is ~ 3 x 10-°/°C.! The lattice mismatch at 550°C (a
typical growth temperature) is therefore 4.4%, not much different from the value at room temperature
(4.2%): High quality epitaxial GeSi films over the entire Ge composition range were successfully
grown on Si(100) substrates at ~ 550°C by MBE in 1984.? The films are metastable-strained,
meaning that (1) the measured critical thickness for pseudomorphic films is larger than that predicted
by Matthews and Blakeslee’s equlibrium model,? or (2) the strain relaxation in a film thicker than
the measured critical thickness is smaller than that predicted by the model. This is due to the fact
that generation of misfit dislocations is a thermally activated process.4 The strain state in a film is
determined by kinetics, not by thermodynamics.* This also suggests that by lowering the growth
temperature (e.g., from 550°C to 300°C), the critical thickness can be increased.»

The strain in a GeSi film influences its bandgap” and its band alignment with the Si substrate.®
By controlling the strain state of a heterostructure, one could engineer the band gap and band offset
to make novel heterostructure devices.? One example is a graded SiGe-base heterojunction bipolar
transistor, the fastest transistors achieved to date in Si technology with a unity-current gain cutoff
frequency of 75 GHz.!°

Defects such as dislocations in a relaxed film degrade device performance. An understanding of
the mechanisms of dislocation generation upon postthermal processing in metastable heterostruc-
tures is therefore important to implement novel device structures successfully. In the following
section, we first summarize our results for GeSi films grown on both $i(100)!+ and vicinal Si(100)!?
at 550°C. We then investigate the kinetics of strain relaxation of highly metastable films grown at

300°C.® Some properties of misfit dislocations and their effects will also be discussed.!3

4.2 Properties of GeSi films grown at 550°C

In this section, we focus on characterization of epitaxial GeSi films on Si(100) by BS/channeling .

and DCD. The structural, elastic, and thermal properties will be discussed.

A. Sample Preparation

The silicon wafer surface was cleaned by the procedure described in Ch. 2.2. The GeSi alloy film
was grown on Si(100) at 550°C with a rate of ~ 0.2 nm/s by codeposition of silicon and germanium
in ultrahigh vacuum at UCLA. The Ge composition of films ranges from 0.05 to 1, and the thickness
from 30 nm to 7 pm.

B. Lattice Mismatch and Vegard’s Law

The lattice mismatch between GeSi alloy and Si, f, can be estimated according to Vegard’s
law, which states that the lattice constant of an alloy varies in proportion to the composition of the
alloy linearly between the lattice constants of the constituent elements. Knowing that the lattice

mismatch between Si and Ge is 4.2%, one obtains
f =4.2%e, (4.2 ~ 1)

where x is the Ge composition. To verify the validity of Vegard’s law in the case of thin film GeSi,
we independently measured the Ge composition and the lattice mismatch of the samples by BS and
DCD, and compared the data with the prediction.

The lattice mismatch between a GeSi film and a Si substrate was extracted from the x-ray
rocking curve measurements. Perpendicular and parallel strains, e+ and ell, were obtained from the
peak separations of the rocking curves diffracted from (400) symmetrical and (311) asymmetrical

planes of the sample (see Ch. 2.2). The lattice mismatch equals
l-—v 2v

f= +v l+v

The Poisson ratio of the film, v, is extracted from the Poisson ratio for Si (0.28) and Ge (0.27)!4

jer + ell,
using linear interpolation. The result is plotted in Fig. 4.2-1, showing that Vegard’s law is applicable
to epitaxial GeSi films.
C. Critical Thickness

For a given lattice mismatch of a heteroepitaxial structure such as GeSi/Si, there is an equi-
librium critical thickness below which an epitaxial film that is in perfect registry with a substrate
has a minimum energy. For GeSi on Si(100), the equilibrium critical thickness from Matthews and
Blakeslee’s model® is plotted as a function of Ge composition in F ig. 4.2-2 (dotted line). The mea-
sured critical thickness of GeSi films grown at 550°C by Bean et al. is larger than that predicted
by equilibrium model (A in Fig. 4.2-2). The results from the DCD and BS measurements of our
samples are summarized in the same composition-thickness plot (@ in Fig. 4.2-2), where the number
associated with each datum point is the normalized parallel strain, ¢ll/f, which equals 0 (1) for a
pseudomorphic (decoupled) film. Our data are consistent with those of Bean et al.? These results

indicate that metastable-strained pseudomorphic films can be grown at low temperature.

4} epitaxial Ge,Si,_, films on Si(100)
} 4
3b _
on, 4

nae 2 least-squares fit
slope = ( 4.2+0.1 ) % 5
1 -

O ! ——i. ! n

0.4 0.6 0.8 1.0

FIG. 4.2-1 The lattice mismatch between epitaxial Ge,Si;_, alloy and Si, f, as a function of Ge

composition. The data agree well with the Vegard law over the entire composition range (0 < z < 1).

critical thickness of Ge,Si,_, films

1.0 @1.01

dotted line: equilibrium model
F A : 550°C (Bean et al.)

¢ : 550°C (our data)

@0.89
* O.5- © 0.82
i @0.81
. . 0.95
“.@0 @0 %e
. a eo: 0.88
ae eo |.
i rest 1 i Lui wal Ll 1 Lit wut _ j _— Aarne
10! 10° 108 104

Film Thickness ( nm )

FIG. 4.2-2 Ge composition—film thickness plot of epitaxial Ge,Si,_, films on Si(100): dotted line
is the equilibrium critical thickness from Matthews and Blakeslee’s model; A is the measured critical
thickness of films grown at ~ 550°C by Bean et al., and e is from our work. The number associated

with each datum point is the normalized parallel strain ¢ll/f.

It is known that the measured critical thickness differs when different analytical tools are used
because of different resolution of dislocation detection.'> The common techniques (DCD, channeling,
TEM) which were used in both Bean et al.’s and our results have a resolution of about 10-4,
corresponding to a dislocation spacing of ~ lum. A heterostructure with a dislocation spacing >~
lum is therefore regarded as pseudomorphic in this context and subsequent discussion. Techniques
capable of resolving a single dislocation in a wafer (e.g., etch pits) reveal that the measured critical
thickness of the samples grown at 550°C is smaller than that obtained by the common methods, but
is still larger than the equilibrium one.?®
D. Dislocations and Strain Relxation

Strain in the film thicker than the measured critical thickness (samples in the upper right of
the solid line in Fig. 4.2-2) relax via generation of misfit dislocations. The amount of the measured
relaxation (number associated with each point in Fig. 4.2-2) is less than that predicted by Matthews
and Blakeslee’s equilibrium model.?

TEM analysis shows that the dominant extended defects in a relaxed GeSi/Si(100) heterostruc-
ture are threading dislocations in the epitaxial film and misfit dislocations at the interface.!? The
misfit dislocations are aligned with < 011 > direction with Burger’s vector b= 1/2 < 110 >.!”
Both 60°-mixed and 90°-edge type are observed.” The spacing between misfit dislocations, p, can

be estimated from the parallel strain from

— bm
p= al’
where b, ~ 0.1 nm is the edge component of Burger’s vector on the interfacial plane.
The x-ray diffraction peak from a relaxed film is broader than that caused by the finite film
thickness, because of the threading dislocations in the film. This additional peak broadening, Wp,
can be used to estimate the threading dislocation density, p; (see Ch 2.2),

_ We
~ 9b?"

Our results show that the larger the parallel strain el! is, the broader the x-ray diffraction peak from
the film becomes. This suggests that the generation of misfit dislocations also produces threading
dislocations.

While channeling spectrum of a pseudomorphic GeSi/Si structure shows no dechanneling at
the interface, that of a relaxed one has a measurable interfacial dechanneling caused by misfit

dislocations. The dechanneling probability, Pp, is proportional to the misfit dislocation density at

the interface, pm,

Pp =opDPpm, (4.2 - 2)

where gp is the dechanneling cross section.'® The linear misfit dislocation density pm is defined as

the total length of misfit dislocations per unit area at the interface, and equals

_ 2 2el
pm bn
The above relationships show that the dechanneling probability Pp across the interface is propor-

tional to the parallel strain ell ,

Pp = 220 al (4.2 — 3)

‘m
Pp can be extracted from channeling measurements (see Ch. 7.2 for detailed discussion), and
increases linearly with ell with a slope of 20 (Fig. 4.2-3(a)). This observation agrees with Eq. (4.2-3)
and results in the dechanneling cross section of gp = 2 nm. This value is qualitatively consistent
with that from the theoretical estimates (~ 1 — 5 nm).!8
The minimum channeling yield of a pseudomorphic film is about 4% for both the Ge and Si
signal, and is about the value for a perfect single crystal. It becomes large for a relaxed film,
indicating a dislocated crystal. The difference, Axmin, is proportional to the threading dislocation
density p; in the film,!®
AXmin « pr «x Wh. (4.2 — 4)

This relationship agrees with our experimental results, where the measured minimum channeling
yield increases as the x-ray diffraction peak for the film broadens (Fig. 4.2-3(b)).
E. Thermal Strain

The thermal expansion coefficient of Ge (6 x 10-°/°C) is greater than that of Si (3 x 10-°/°C),
which generates a thermal mismatch of ~ 0.2% between room temperature and 550°C.! This mag-
nitude is insignificant compared to the lattice mismatch of 4.2%, and suggests that thermal strain in
a Ge/Si structure is not important, unlike in CoSi2/Si (see Ch 2.5). DCD measurements of epitaxial
Ge films on Si(100) at 20-500°C confirm that thermal strain is small, and also indicate that the
interface does not shear below about 500°C. For a GeSi alloy, the lattice mismatch is determined by
Vegard’s law. If we assume that the thermal expansion coefficient can also be estimated by linear
interpolation, thermal strain in an alloy film is also negligible compared to the lattice mismatch. Fig.
4.2-4(a) shows the evolution of the lattice mismatch and strain with temperature for a metastable

coherent GeSi film on Si(100). The small slope means that the thermal strain is very small. The

40 CT T T T T T T T T T T T T T T T T
dechanneling vs parallel strain

155 r T 7 T t T
r minimum yield vs peak broadening O;

AXmin ( a )

0 2 4 6
Wp” ( 107° )

FIG. 4.2-3 Channeling characteristics vs x-ray diffraction of epitaxial GeSi films grown on Si(100) at
550°C: (a) dechanneling probability of an aligned MeV *He beam across the interface as a function
of parallel strain el! of the film; (b) the difference of the minimum channeling yield between a relaxed
and a coherent GeSi film as function of the square of the x-ray peak broadening caused by threading

dislocations in the film.

pseudomorphic Gepg 95Sig g5/Si(100)

0.4
Rogl 8 aI -
G a
oO
5 08 ce One
A + f coherent
5 y el!
i I | lL | L ]
0 200 400 600
T (°C)
2.0

——— as—grown

Reflectivity ( % )

FIG. 4.2-4 Thermal properties of a metastable coherent Ge,Si,_, film grown on Si(100) at 550°C:
(a) the lattice mismatch (dotted line) and the strains (solid lines) as a function of temperature;
dashed line is for a film with rigid interface; (b) Fe K, x-ray rocking curves diffracted from (400)
planes of of sample before (solid line) and after (dotted line) thermal annealing at 630°C for 2h in

ambient air.

figure also indicates that there is a small amount of strain relaxation after heating at 630°C in air
for 2h. The rocking curve of the annealed sample (dotted line Fig. 4.2-4(b)) has a broader peak
compared to that of the as-grown one (solid line Fig. 4.2-4(b)), indicating a degraded film. On
the other hand, the sample annealed in vacuum at 630°C for 2 h shows no detectable change. This
means that the strain relaxation is induced by the sample interaction with the ambient. GeSi films
of low Ge content are known to react with oxygen by selectively oxidizing Si. We hence conclude
that interactions promote strain relaxation of metastable GeSi film. A similar effect is observed for
epitaxial CoSiz films on Si(111) (see Ch. 2-5). ‘The phenomenon is quite probably a very general

one.

4.3 Pseudomorphic GeSi films and superlattices on vicinal Si(100)

Some properties of GeSi films on vicinal Si(100) substrates are discussed in this section. The
substrates used in this study are vicinal Si(100) wafers, with their parallel front and back surface
normals tilted from [100] towards [011] by an offset angle ¢, of 3.12 and 6°. GeSi films 100 nm thick
with Ge composition from 0.06 to 0.16, and a 30 period superlattice of Geo .2Si9.g(8.4 nm)/Si(15.6
nm) were grown at 550°C by MBE at UCLA. Film thicknesses and compositions were determined
by BS, and epitaxial quality by channeling measurements. The strain in the films was characterized

by DCD.

A. Single Layer GeSi Films

All films are-in the lower-left side of Bean et al.’s critical thickness curve in the thickness-
composition plot of Fig. 4.2-2, and are therefore pseudomorphic. In the following, we describe in
detail the methods we have used to determine the strain state in the Geo ogSio.94 100 nm thick film
on a vicinal Si(100) substrate with an offset angle of 3.1° (as determined by the back-reflection Laue
pattern) by DCD.

First, a series of x-ray rocking curves diffracted fromi symmetrical (400) planes of the sample in
various azimuthal configuration were recorded (see Ch 2.3). The substrate peak position, 6p, and
the difference between the film and the substrate peak position, A@p, were plotted as a function
of the azimuthal angle in Fig. 4.3-1(a) and (b). That both 6p and Aép have the same azimuthal
dependence means that the [100] directions of the substrate and of the film are misoriented against
the surface normal in the same direction, and that the [100] direction of the film is farther away from
the surface normal than that of the substrate (Fig. 4.3-1(c)). The solid lines are the least-squares

fit of a cos function to the data (see Eq. (2.3-1) and related discussion in Ch 2.3). Fig. 4.3-1{b)

FIG. 4.3-1

Op (°)

-0.23

100 nm Geo ogSig.g4 on vicinal Si(100)

6y=45.5°, $,=3.1

200

w(°)
To" oh
a A@p=—0.247°, Ap=0.018° a
a \ | ! __|
0 100 200 300
v(°)
Nn [100],
Hf fLioo}g
et
GeSi
Si

(a) the angular position of the (400) peak diffracted from the substrate and (b) the

difference of the peaks from the film and the substrate versus the azimuthal angle of the sample

configuration. (c) schematics of the results from analysis of the x-ray rocking curve data (a) and

(b).

[ oy
X=[011]——

FIG. 4.3-2 Schematics of one symmetrical (400) and four asymmetrical {311} diffracting planes of

vicinal samples. The surface normal is taken as Z-axis.

shows that for the (400) plane, the misorientation angle, Ad, is 0.018°, and the interplanar spacing
mismatch, €¢, is 0.43%, where ¢ is the angle between the surface normal and the substrate diffracting
plane.

Unlike symmetrical diffraction from the (400) plane, there are four sets of asymmetrical diffrac-
tion planes such as (311), (311), (311), and (311) (see Fig. 4.3-2). X-ray rocking curves diffracted
from all four such aymmetrical planes of the sample were also recorded. Both the misorientation
angles A@ and the strains eg for each set of planes were extracted.

To simplify the analysis, we made two assumptions: (1) the film is under biaxial stress in the x-y
plane (parallel to the surface), and the principal axes of stress are therefore along x-, y-, z-directions;
(2) the film is elastically isotropic; the principal axes of the strain tensor hence coincide with those

of the stress tensor. If the principal strains are e**, eYY , «72, the strain along the direction in the

y-z plane (Fig. 4.3-2(a)) is?9

eg = €77 cos? b + YY sin? hd = c47 — (4% — YY) sin? 6, (4.3 — 1)

and the misorientation angle is
2Z_ YY

5 sin 2¢. (4.3 — 2)

Ad = (44 —e¥Y) cos dsing = c

We hence plot the strain eg obtained from DCD measurements of the diffracting planes (e.g., those
shown in Fig. 4.3-2(a)) as a function of sin? ¢ in Fig. 4.3-3(a). The solid line is the least-squares
fit of Eq. (4.3-1) to the data (o). The principal perpendicular and parallel strain extracted from
the fitting are «7% = 0.42 + 0.02% and «YY = 0 + 0.02% (see Table 4.3-1). Similarly, we plot the
misorientation angle as a function of sin 2¢ in Fig. 4.3-4(b). The solid line is obtained from Eq.
(4.3-2) using previously determined «7% and e¥¥. The good agreement obtained in Fig. 4.3-4(b)
between the data (o) and the predicted value means that the misorientation observed here can be
explained by the biaxial stress model and results from the deviation of the diffracting plane from
the principal strain axes.

This present misorientation model differs from the misorientation model based on geometrical
matching at the interface, developed in Ch. 2.3 for CoSiz on vicinal Si(111). The misorientation angle
predicted from Eq. (4.3-2) for the CoSiz films on vicinal Si(111) discussed in Ch. 2.3 («74 ~ —1.7%
and e*Y w —0.8%) is

Aé = —0.9% sin d cos d & —0.9%d

for small ¢. This prediction does not agree with the experimental data: A¢ = —1.7%@. In particular,

100 nm Geg ogSig.g4 on vicinal Si(100)

f ' T u T
(400)

sin”>

0.10- 4

0.05 }- 4

Ae (° )

. (400) J
l Lt !

0.4 0.6 0.8
sineo

0 0.2

FIG. 4.3-3 (a) The mismatch in interplanar spacing and (b) the misorientation angle as a function
of diffracting planes of a Geo.os5Sio.95 film 100 nm thick on a vicinal Si(100). The solid line is the

prediction of the biaxial stress model with e** = «YY = 0, «7% = 0.52%, and o, = 3.1°.

for a stress-free film, «77=«Y ¥ =0, the present biaxial stress model yields Ag = 0, in contrast with
the experimental observation for CoSig (Ch 2.3). This means that the biaxial stress model cannot
explain the misorientation in the case of CoSig on vicinal Si(111).

The results obtained by analyzing the rocking curves diffracted from asymmetrical planes such
as (311) and (311) (see Fig. 4.3-2(b)) yield the strain along the direction in the z-x plane. A
similar analysis shows that e** also equals 0 (Table 4.3-I). We therefore conclude that the film
is pseudomorphic within the experimental resolution, and that the strain is isotropic in x-y plane
(eXX =eY¥ =).

Other samples with different Ge composition and offset angles were analyzed in the same frame-
work. The results are summarized in Table 4.3-I. All the films are pseudomorphic, and the misori-
entation between diffracting planes of the film and the substrate can be explained by the biaxial

stress model.

B. GeSi/Si Strained-Layer Superlattice

A superlattice 720 nm thick was grown on a vicinal Si(100) with an offset angle of 3.1°. Chan-
neling spectrometry shows that the film is highly epitaxial with a minimum yield of ~ 3%, and
pseudomorphic without detectable dechanneling at the interface. X-ray rocking curves from sym-
metrical (400) diffraction planes of the sample are shown in Fig. 4.3-4. The solid (0) line in Fig.
4.3-4(b) corresponds to the diffraction geometry I (II) in Fig. 4.3-4(a).

The overall elastical properties of a superlattice are equivalent to those of a uniform alloy with
an average composition.’® The Oth order superlattice peak of two rocking curves are at different
angular positions, meaning that the [100] axis of the superlattice is misoriented against that of the
substrate. The average strain and misorientation angle can be extracted from the peak separations
betweem the Oth order superlattice peak and the substrate peak, A@,; and A@7;, according to the
previous analysis for a single layer film.!° The results are summarized in Table 4.3-I, which shows
that the superlattice is pseudomorphic and that the misorientation angle can be explained by the
biaxial stress model.

The angular oscillation periodicity, 667, of the superlattice rocking curve for the diffraction
geometry I also differs from that for the geometry II, 66;7. This effect is due to the different
outgoing angle, @,, for the two x-ray paths. For a superlattice whose period has a thickness, tp, the
angular periodicity equals!®

Asin 6,

% )
Oo

107!

Reflecting Power (

nan

30 period SigGe2/Si on vicinal Si(100)

T T T T T

T PUTTiney

T VT TENT

T PV TTeny

period=24 nm |
offset angle=3.1°

T T T T

posal

piriiul

FIG. 4.3-4 Fe Ka, (A = 0.1932 nm) x-ray rocking curves diffracted from the (400) symmetrical

planes of the GeSi superlattice on vicinal Si(100): (a) two diffraction geometries corresponding to

a rotation of 180° of the incident x-ray about the surface normal; (b) the corresponding diffraction

spectra.

Table 4.3-I

The strain obtained from rocking curve analyses and the biaxial stress model of the

single layer Ge,Si,_z films and the superlattice with the average Ge composition of 0.07 grown on

vicinal $i(100) substrates at 550°C.

Sample t (nm) z ds eXX(%) eX (%) 644%)
single layer 100 0.06 3.1 0 0 0.42
single layer 100 0.14 3.1 0 0 1.07
single layer 100 0.16 6.0 0 0 1.22
superlattice 720 0.07 3.1 0 0 0.53

For (400) symetrical diffraction and samll offset angle, ¢, < 1, we obtain

66, 4+ 6677 _ r

66) = —— = Ti cosy’ (4.3 — 4a)
and
50; — 60
a aa = (66, cot Og) bs. (4.3 — 4b)

From the measured periodicity 60; and 60;;, one has the superlattice period tp = 24 nm from Eq.
(4.3-4a) and the offset angle ¢, = 3° from Eq. (4.3-4b). These results agree well with those obtained
by other techniques, confirming that the apparent different angular periodicity 60; and 60 rr 18 caused

by the inclination of the diffracting plane from the surface.

4.4 Strain relaxation of pseudomorphic GeSi films grown at 300°C

The growth and stability of coherently strained epitaxial GeSi layers on Si substrates are key
issues for the fabrication and operation of Si/GeSi/Si heterojunction bipolar transistors. It is known
that the strain state of epitaxial GeSi layers on Si substrates depends critically on the layer thickness,
the lattice mismatch, and the growth temperature.’ In the case of growth by molecular beam epi-
taxy at low temperature (< half of the melting point), a metastable structure is usually obtained.25.
The metastability comes from the kinetic barriers that hinder the generation of misfit dislocations
to relax the elastic strain.t Such a structure relaxes upon subsequent thermal processing at higher
temperatures.”:© Dodson and Tsao* modeled the kinetics of strain relaxation in terms of dislocation
nucleation and multiplication driven by the excess stress in the metastable structure. This model
agrees well with experimental results on the strain relaxation of GeSi/Si heterostructures, and pro-
vides a framework to interpret data. There exist a handful of detailed quantitative experimetal
investigations on strain relaxation.»»°,?°-?4 However, a full data set covering a wide range of layer
metastability and relaxation temperature is still lacking.

We present here some quantitative results on strain relaxation of metastable pseudomorphic
Geo.3Sio.7 layers 570 nm thick on Si(100). The layers were grown at 300°C at a rate of ~0.1 nm/s
by co-deposition of silicon and germanium in an ultrahigh vacuum of a base pressure ~ 10~!° Torr.

Double crystal x-ray diffractometry and MeV *He channeling spectrometry were used to charac-
terize the strain and the crystalline quality of the epitaxial layer. The composition and the thickness
of the layer were confirmed by MeV *He backscattering spectrometry. [100] axial channeling of the
sample shows that the layer is of high quality with a minimum yield of ~ 5% for both Ge and Si.

The interface between the layer and the substrate has no detectable dechanneling centers.

The x-ray rocking curve from (400) symmetrical diffraction shows a sharp, intense diffraction
peak from the layer. The full-width at half-maximum of the peak, Wa, is only ~ 3 x 107? radian.
It equals the value expected for the peak broadening that is due to the finite thickness (570 nm)
of the layer, Ws. This means that the layer is of high crystalline perfection without measurable
defects. The peak separations between the layer and substrate for both (400) symmetrical and (311)
asymmetrical diffractions show that the layer is pseudomorphic with a non-measurable parallel
strain, ¢ll< 0.01%, and a perpendicular strain, e+, of 2.1%. According to Vegard’s law, the lattice
mismatch, f, between a stress-free Geo 3Sio.7 layer and Si substrate is 1.2%. In linear elasticity

theory, the parallel and perpendicular strains are related to the lattice mismatch by

a_ityv
l—vp

(4.4 — 1)

— ell
where v is the Poisson ratio. Using v = 0.273,!* the perpendicular strain for pseudomorphic (cll=0)
Geo.3Sio.7 layers on Si (f = 1.2%) is e+ = 2.1%, in excellent agreement with the experimental results.

Houghton et al.'® used an etching technique to map the onset of dislocations in epitaxial GeSi
layers grown on Si(100) substrates and showed that the equilibrium critical thickness of the layer
agrees with that predicted by Matthews and Blakeslee’s model.? According to this model, the equi-
librium critical thickness of epitaxial Geo 3Sig.7 layers on Si(100) is only 8 nm (see Fig. 4.2-2), and
the equilibrium strain state of 570 nm thick layers is almost decoupled, with ell ~ e+ x f = 1.2%.
This means that 570 nm pseudomorphic Geo .3Sig.7 layers on 5i(100) are highly metastable. We also
note that the experimentally measured critical thickness of Geo 3Sio7 layers grown on Si(100) at
550°C by molecular beam epitaxy is 40 nm (see Fig. 4.2-2), more than 10 times smaller than that
of the layers grown at 300°C. This shows that a low growth temperature can greatly enhance the
critical thickness of epitaxial GeSi layers.

The strain relaxation of GeSi layers proceeds by nucleation!7?526 and multiplication2”:® of
misfit dislocations. We monitored by x-ray rocking curves taken at room temperature the strain
relaxation of 570 nm thick pseudomorphic Geg.3Sio.7 layers upon thermal annealing in a vacuum
(5x 10-7 torr). The number of misfit dislocations per unit length at the interface, pm, is proportional
to the parallel strain ell,

ell

Pm = Fs

(4.4 — 2)

where 6,, = 0.2 nm is the edge component of Burger’s vector in the interfacial plane. We obtained
e+ from (400) diffractions for all the samples and then extracted ell according to Eq. (4.4-1) with
known y = 0.273 and f = 1.2%. The parallel strain ¢!! was also measured directly from combined

(311) and (400) diffractions for selected samples. The result agrees with that extracted from

570 nm Geg3Sig7 on Si(100)

ra decoupled

1.Q9- isochronal annealing
| (30 min)

| L l 1
400 600

T ( °C )

0 200

FIG. 4.4-1_ The strain relaxation of a highly metastable GeSi strained layer grown on Si(100) at

~ 300°C as a function of temperature upon isochronal annealing for 30 min in vacuum. The strain

relaxes sharply at 375 + 25°C.

570 nm Geg3Sig.7 on Si(100)

decoupled

Time Duration ( min )

FIG. 4.4-2 Parallel strain ell of 570 nm thick epitaxial Geo 35io.7 layers grown on Si(100) at
~ 300°C obtained from (400) diffraction x-ray rocking curve measurements vs duration of ez situ

thermal annealing in vacuum (~ 5 x 1077 Torr) at various temperatures.

Eq. (4.4-1). As the metastable Geo 3Si9.7 layer relaxes upon annealing, ell increases from < 0.01%
for coherent layers to ~ 1.2% for relaxed ones (see Fig. 4.4-1). The misfit dislocation density p,,
thus obtained from Eq. (4.4-2) increases correspondingly from < 1/um to ~ 50/um. The transition
occurs sharply at 375 + 25°C. At 300°C, little relaxation is detectable for annealing lasting up to
180 min. At 500°C, the strain reaches the equilibrium value in less than 10 min. Figure 4.4-2 shows
that the parallel strain increases slowly at a low relaxation level («ll< 0.1%; see in particular data of
T < 365°C, where this initial phase is well resolved), then increases roughly linearly at a modestly
relaxed level (0.1% < ell < 0.6%), and gradually saturates towards nearly complete relaxation
(cll > 0.6%). On the basis of this observation, we envision the following strain relaxation process:
The initial sluggish relaxation is likely associated with a nucleation barrier and low density of misfit
dislocations; the approximately linear relaxation at modest level is dominated by multiplication; and
the final gradual saturation is related to the dislocation interaction, which impedes their motion,
and to a decrease of excess stress, which reduces the driving force for strain relaxation.

We extracted the strain relaxation rate at a modest relaxation level from Fig. 4.4-2 by fitting a
linear increase to ell in the range of 0.1% to 0.6% (see dashed line in Fig. 4.4-2). Figure 4.4-3 plots
that rate as a function of the inverse annealing temperature. The data clearly show an Arrhenius
dependence and the slope gives an activation energy of 2.1+ 0.2 eV. This demonstrates that the
strain relaxation of metastable GeSi layers at modest relaxation levels (between the pseudomorphic
and the equilibrium strain state) is a thermally activated process.

Dodson and Tsao* modeled the strain relaxation by plastic flow, which can be described by a

non-linear differential equation, written here in a slightly different form,

dell _ (ell — ell 4 )?(ell + ll.)

i : (4.4 —- 3)

where elleg. is the equilibrium parallel strain and 7 is a time constant. The parameter «ll, is used to
simulate the “background” dislocation density or dislocation nucleation, which is very small (< 0.1%)
for initially pseudomorphic layers. The time constant + determines how fast the relaxation proceeds
and has an exponential temperature dependence, r x exp(Eq/kT). The activation energy, E4, is
related to the kinetic barriers of dislocation glide. When the parallel strain is taken at a fixed value
corresponding to a modest relaxation level, as was done for Fig. 4.4-3, the relaxation rate of Eq.
(4.4-3) becomes

i= al I ( Eu
rate = dt «T & erpl— TR)

Therefore, the activation energy of 2.140.2 eV from Fig. 4.4-3 for the rate of strain relaxation is the

(4.44)

same as that for dislocation glide in Dodson and Tsao’s model (E,). This value is the same as that

for dislocation glide in bulk Geo.3Sio.7, ~ 2 eV, obtained by linearly interpolating those of Ge (~ 1.6
eV) and Si (~ 2.2 eV).?9 It also agress reasonably well with that of Tuppen and Gibbings?? and
Houghton 74 (~ 2.2 eV), and Timbrell et al.?? (~ 1.9 eV), but is about twice that of Hull et al.22!
(~ 1.1 eV). Dodson and Tsao*° suggest that the excess stress in the layer reduces the activation
energy for dislocation glide. Our data show that this reduction is not important in our 570 nm thick
epitaxial Geo 3Sio,7 strained layers.

The broadening of the x-ray diffraction peak is a measure of crystalline imperfection. The
dominant imperfection in epitaxial GeSi layers grown on Si is threading dislocations, which cause a
peak broadening, Wp. The areal density of threading dislocations, p;, in the epitaxial layer, can be

estimated (see Ch 2.2) by

Wp?

where 6 = 0.4 nm is Burger’s vector of threading dislocations. We found that the x-ray (400)
diffraction peak from the Geg.3Sio.7 layer broadens as the strain relaxes upon thermal annealing of
the sample. Assuming that both the finite layer thickness and the threading dislocations produce
a Gaussian broadening of the x-ray peak, one has Wp? = Wo? — Ws? (Ws, and Wg are defined
in Ch. 2.2). The areal density p; thus obtained from Eq. (4.4-5) increases from < 10®/cm? for
pseudomorphic layers to ~ 3 x 10°/cm? for relaxed ones.

Furthermore, Fig. 4.4-4 shows that Wp increases linearly as ll increases. We recall that pm
and p; are proportional to ell and Wp, respectively, according to Eqs. (4.4-2) and (4.4-5). The areal
density of threading dislocations in Geo 3Sig.7 layers is therefore proportional to the linear density of
misfit dislocations at the interface. This means that threading and misfit dislocations arise together,
as is the case if they are different segments of one dislocation.

In that case, the number of misfit dislocations per unit length pm at the interface is proportional

to the areal density of threading dislocations p; in the layer,

cL
Pm = 4? (4.4 — 6)

where L is the average length of misfit dislocation in the interfacial plane. Combining Eqs. (4.4-6),

(4.4-2) and (4.4-5), one obtains

-4
£ 10°F Ba=2.140.2 eV 7
© i 1
& L. least-squares fit .
B 49-5
g 10°F 7
8 ' as
“” L 1
C { rT | 1 7

1.5 1.6
1/T ( 1/1000K )

FIG. 4.4-3 Relaxation rate measured by the rate of increasing parallel strain with increasing time
at modest relaxation level vs the inverse of the temperature at which the relaxation proceeds (see

Fig. 4.4-2). The data follow an Arrhenius behavior with a slope of 2.1 +0.2 eV.

6 4
° L 4
O 4b 4
a | 2 @)east-squares fit 3 1
sol slope=6.3+0.5 o
a 8
| 9 o 4
0 [_ 4 | L rT 1 1 | i /
0 0.5 1.0
ell x107* )

FIG. 4.4-4 X-ray peak broadening of (400) diffraction from 570 nm thick epitaxial Geo 3Sig.7 layers

grown at ~ 300°C that is due to imperfections in the layers versus the strain relaxation.

according to Fig. 4.4-4. This value agrees roughly with what Hull et al.?° obtained from transmission
electron microscopy. The well-behaved linear relationship between p; and pm (Fig. 4.4-4) means that
the average length of misfit dislocations remains a constant (~ 4.6 wm) during the strain relaxation.
The results suggest that the dislocations are nucleated at the surface and glide down to the interface
in slip planes. The strain relaxation at the interface therefore necessarily results in the existence
of threading dislocations in the layer. Equation (4.4-6) also shows that increasing the length L of
misfit dislocations in the interfacial plane reduces the threading dislocation density p; in the layer
for a given strain relaxation level py.

We also monitored the strain relaxation of pseudomorphic Geo3Sio,7 layers by MeV *He {100]
axial channeling spectrometry. As the strain relaxes upon thermal annealing, dechanneling at the
interface occurs, confirming that misfit dislocations are generated. At the same time, the minimum
yield of the layer for [100] axial channeling increases from ~ 5% for pseudomorphic layers to ~ 12%
for relaxed ones. This means that imperfections are also generated within the layer as it relaxes,
and supports the previous assertion that relaxation produces threading dislocations.

In summary, the strain in the pseudomorphic Geo 3Si9.7 layers 570 nm thick grown on Si( 100)
at 300°C relaxes sharply at (375 + 25)°C, and reaches the thermal equilibrium value after 60 min
at 400°C. The rate of strain relaxation and misfit dislocation generation as a function of inverse
temperature follows an Arrhenius behavior with an activation energy of 2.1+0.2 eV. As the strain
relaxes, the defect density in the layer increases proportionally. This fact suggests that the generation

of misfit dislocations necessarily results in the occurrence of threading dislocations.

4.5 Asymmetrical tilt boundary in GeSi/Si(001) heterostructures

In a relaxed heterostructure with a non-zero ell, there exist misfit dislocations at the hetero-
interface to relieve the lattice mismatch between the film and the substrate (Fig. 4.5-1). These
interfacial dislocations could also produce a misorientation, J, between the corresponding low index
plane of the film and the substrate (Fig. 4.5-1), if the Burger vector of the dislocations has a
component along the direction of the surface normal.*!

To analyze the effect of an interfacial dislocation array, we first summarize the results of simple
ones shown in Fig. 4.5-2. The case (I) corresponds to the common misfit dislocation arrays, where
the Burger vector is perpendicular to the surface normal. They generate a non-zero parallel strain
but a zero misorientation angle. The case (II) is a symmetrical low angle tilt boundary,?! where the

Burger vector is parallel to the surface normal. In case (III), the Burger vector has components

Al Lo

FIG. 4.5-1 Schematic representation of parallel strain and misorientation in a heterostructure.

(I) Misfit Dislocations film
r TT
e'=b/p, ¥=0 substrate

(II) Tilt Boundary po
e'=0, B=b/p p

(III) Mixed Dislocations an ann
e'=b!/p, v=0

(IV) Mixed Dislocations Le
e!=p!/p, %=b'/p

FIG. 4.5-2 Four different types of dislocation arrangements at heterointerfaces: (I) mismatch
relieving dislocations; (II) asymmetrical low angle tilt boundary; (III) mixed dislocations with a

non-zero parallel strain and a zero misorientation angle; (IV) mixed dislocations with both non-zero

net mismatch relieving and tilt components.

both perpendicular and parallel to the surface normal, but the net sum of the Burger vector of the
dislocation arrays parallel to the surface normal is zero; the net effect of such dislocation arrays is
thus similar to case (I). In case (IV), the dislocation is in fully aligned arrangement; both components
of the Burger vector of each dislocation parallel and perpendicular to the surface normal simply add

up, producing a non-zero parallel strain and misorientation angle.

In a GeSi/Si(001) structure, the interfacial dislocations consist of square arrays running along
< 110 > directions with both 60°-mixed and 90?-edge type characters.!” The 60°-mixed dislocations
can be nucleated at the surface in the {111} slip plane and glide down to the interface to relieve the
elastic strain.®?°?® The 90° edge dislocations relieve the elastic strain most efficiently and move by
glide and climb.!” At low temperature (< 600°C), the climb motion is very difficult; one therefore
expects that the interfacial dislocations are predominantly of 60°-mixed type.!77628 The Burger
vector of the 60°-mixed dislocation can be decomposed into screw and edge components (Fig. 4.5-
3). The interfacial dislocations of square arrays with a net screw component form a twist boundary,?!
which produces a rotation of the film about the surface normal with respect to the substrate. The
edge component resembles the mixed dislocation shown in cases (III) and (IV) of Fig. 4.5-2. It
can be further decomposed into the components parallel and perpendicular to the surface normal
(see Fig. 4.5-3). The parallel component relieves the elastic strain in the film. The perpendicular
component produces a misorientation between the film and the substrate. In this section, we will

focus on the misorientation effect by interfacial dislocations.!3

A set of Geo 3Sig.7 films ~ 570 nm thick were grown on Si(001) at ~ 300 — 500°C by MBE at
UCLA. About 30-40 samples were made by various deposition-temperature sequences with varying
growth pressure from high 10-!° Torr to low 10~® Torr. The critical thickness, stability, and strain
relaxation upon thermal processing of such samples have been discussed in a previous section. We
investigate here the misorientation between the film and the substrate of such samples (both as-
grown and after annealing). To simplify the analysis, we will assume in the following discussion that

all the interfacial dislocations are of the 60°-mixed type.

DCD was used to extract the parallel strain ell and the misorientation angle J of each sample.
The experiemntal results are summarized in Fig. 4.5-4. The thick solid line labeled A in the figure
represents the data from the majority of the samples analyzed. Some samples in this set are initially
pseudomorphic with ell=0. Upon thermal annealing in the vacuum, the structures relax and develop
a non-zero ell. The parallel strain increases with more interfacial dislocations being generated to

relieve the elastic strain till it reaches the equilibrium value. Other samples in this set are initially

[001] AN
Zid ‘
aN
yo .
/ | \
he Lr
Sf pm TOL Ss 5
/ 7 a
oe
a= = 1100]

FIG. 4.5-3 Schematic representation of a 60°-mixed dislocation in a GeSi/Si(001) structure.

570 nm Geg 3Sig7 on Si(001)

0.4- “ OH
0.3L _
°0.2b _
ea Lf |
O.1b _
OF 7 < 5, ‘ .
Poy a a a, oN a "~N a
0 0.5 1.0
el (2%)

FIG. 4.5-4 Misorientation angle vs parallel strain of ~ 570 nm thick Geo.3Si9.7 films grown on
Si(001) substrates at 300-500°C. Most samples (A) analyzed have zero misorientation angles re-
gardless of the value of the parallel strain. Samples B and C are initially pseudomorphic and have
zero misorientation angle. They relax upon thermal annealing and develop finite misorientation
angles. Sample D is relaxed initially and has a non-zero misorientation angle. It relaxes further

upon annealing. After amorphization of the entire film and solid phase epitaxial regrowth, the

misorientation becomes zero.

383

partially relaxed with a non-zero el. The common feature of the samples in this set is that there is
no misorientation between the film and the substrate. This indicates that the arrangement of the
interfacial dislocations in this set of samples corresponds to case (III) of Fig. 4.5-2. The sample B
in the figure is initially pseudomorphic with both ell and 8 equal to zero. Upon thermal annealing
(300-700°C for 30 min), cll increases. At the same time, J increases proportionally. The sample
C is similar to the sample B except that the slope is different. The sample D is initially partially
relaxed, and has a non-zero el! and J. Upon annealing, it relaxes further and the misorientation also
increases further, correspondingly. In addition, the as-grown sample D was implanted with 380 keV
1015 ?8Si/cm? at room temperature and the entire film was completely amorphized.3? Solid phase
epitaxy was then initiated by vacuum annealing at 600 and 700°C for 30 min. The parallel strain of
the regrown film equals the value of the equilibrium strain state. The misorientation angle becomes
zero. This result clearly shows that the misorientation is caused by the aligned arrangement of the
interfacial dislocations, not by the surface morphology of the substrate like that discussed in Ch.
4.3. The misorientation angle of all the samples studied here is smaller than the maximum value
expected for an arrangement where all the dislocations are exactly aligned (dotted line in Fig. 4.5-4).

For a given number of interfacial dislocations, N, the parallel strain is the same. The misori-
entation can vary from zero of case (III) to a maximum of case (IV), depending on the particular
arrangement. There is an elastic energy associated with a finite misorientation angle. This suggests
that the arrangement giving a zero misorientation angle has minimum energy. This could explain
the fact that most samples have a zero misorientation angle. However, it is known that kinetics de-
termines the process of the dislocation generation and strain relaxation. The dislocation generation
proceeds by initial nucleation and then multiplication. The orientation of initially nucleated dislo-
cations is equally probable among possible orientations, since these initially nucleated dislocations
are far apart so that interaction energy among them is negligible. The further relaxation proceeds
by the multiplication of these dislocations. The misorientation angle increases proportionally. The
slope of the misorientation angle versus parallel strain is hence determined by the arrangement of
very few initially nucleated dislocations. As the initial arrangement is governed by probability, the
misorientation angle for any given parallel strain is hence also determined by probability The prob-
abilty for the arrangement of the zero misorientation angle is the highest. Therefore, most samples
have a zero misorientation angle. Furthermore, there is finite probability for the arrangement where

the misorientation angle is not zero. This explains the misorientation observed for some samples.

References

10.

11.
12.
13.
14.
15.
16.

17.
18.

19.
20.
21.
22.

23.
24.
25.

. E. Kasper and H.-J. Herzog, Thin Solid Films 44, 357 (1977).

. J.C. Bean, L.C. Feldman, A.T. Fiory, S. Nakahara, and I.K. Robinson, J. Vac. Sci. Technol. A2,
436 (1984).

J.W. Matthews and A.E. Blakeslee, J. Cryst. Growth 27, 118 (1974).

. B.W. Dodson and J.Y. Tsao, Appl. Phys. Lett. 51, 1325 (1987).

R.J. Hauenstein, R.H. Miles, E.T. Croke, and T.C. McGill, Thin Solid Films 183, 79 (1989).

. G. Bai, M-A. Nicolet, C.H. Chern, and K.L. Wang, unpublished.

C.G. Van de Walle and R.M. Martin, J. Vac. Sci. Technol. B3, 1256 (1985).

D.V. Lang, R. People, J.C. Bean, and A.M. Sergent, Appl. Phys. Lett. 47, 1333 (1985).

. See, e.g., R. People, IEEE QE-22, 1696 (1986).

G.L. Patton, J.H. Comfort, B.S. Meyerson, E.F. Crabbé, G.J. Scilla, E.D. Frésart, J.M.C. Stork,
J.Y.-C, Sun, D.L. Harame, J.N. Burghartz, IEEE Electron Device Lett. 4, 171 (1990).

G. Bai and M-A. Nicolet, unpublished.

G. Bai, M-A. Nicolet, C.H. Chern, and K.L. Wang, unpublished.

W.A. Brantley, J. Appl. Phys. 44, 534 (1973).

I.J. Fritz, Appl. Phys. Lett. 51, 1080 (1987).

D.C. Houghton, C.J. Gibbings, CG. Tuppen, M.H. Lyons, and M.A.G. Halliwell, Appl. Phys.
Lett. 56, 460 (1990).

R. Hull and J.C. Bean, J. Vac. Sci. Technol. A7, 2580 (1989).

L.C. Feldman, J.W. Mayer, S.T. Picraux, Materials Analysis by Ion Channeling, Academic,
New York, 1982.

V.S. Speriosu and T. Vreeland, Jr., J. Appl. Phys. 56, 1591 (1984).

R. Hull, J.C. Bean, D.J. Werder, and R.E. Leibenguth, Phys. Rev. B40, 1681 (1989).

R. Hull, J.C. Bean, and C. Buescher, J. Appl. Phys. 66, 5837 (1989).

P.Y. Timbrell, J.-M. Baribeau, D.J. Lockwood, and J.P. McCaffrey, J. Appl. Phys. 67, 6292
(1990).

C.G. Tuppen and C.J. Gibbings, J. Appl. Phys. 68, 1526 (1990).

D.C. Houghton, Appl. Phys. Lett. 57, 2124 (1990).

C.J. Gibbings, C.G. Tuppen, and M. Hockly, Appl. Phys. Lett. 54, 148 (1989).

26. D.J. Eaglesham, E.P. Kvam, D.M. Maher, C.J. Humphreys, and J.C. Bean, Philos. Mag. A59,
1059 (1989).

27. W. Hagen and H. Strunk. Appl. Phys. 17, 85 (1978).

28. K. Rajan and M. Denhoff, J. Appl. Phys. 62, 1710 (1987).

29. H. Alexander and P. Hassen, Solid State Phys. 22, 27 (1968).

30. B.W. Dodson and J.Y. Tsao, Appl. Phys. Lett. 53, 2498 (1988).

31. D. Hull and D.J. Bacon, Introduction to Dislocations, 3rd Edition, Pergamon, Oxford, 1984.

32. G. Bai and M-A. Nicolet, unpublished.

Chapter 5 Porous Si and Its Properties

5.1 Introduction

Porous Si technology has potential applications in dielectric isolation of integrated circuits,! and
in epitaxial growth of heterostructures on Si substrates.? Recent work has demonstrated efficient
visible light emission from porous Si pumped with a green light,3 which could open the door to

Si-based optoelectronics. Since the formation of porous Si by anodic dissolution of single crystalline

Si in a concentrated HF solution was first reported by Uhlirt and Turner,® various properties of

porous Si have been extensively studied.

Porous Si has a very large surface area (surface here means the physical boundary between solid
and external ambient) and a complex surface morphology. One therefore anticipates that many prop-
erties of porous Si are dominated by surface effects. The morphology and the microstrcuture, such
as porosity, specific surface area, and pore size, have been carefully characterized by gas adsorption
isotherms® and transmission electron microscopy.’ X-ray diffraction was used to characterize the
strain and the elastic properties of porous Si.3° Barla et al.5 reported that as-prepared porous Si
has a larger lattice constant than that of bulk Si substrate in the direction normal to the surface ,
and is less stiff. Young et al. attributed the lattice expansion to the formation of the native oxide
layer on the pore surface. Thermal annealing behavior of porous Si has also been investigated.1°-!?
The composition of porous Si was measured by ion beam analysis, Auger electron spectroscopy, and
x-ray photoelectron spectroscopy.!*!4 High concentration of light impurity atoms of H, O, C, F were
detected (up to a few tens %).!5:14

The study described in the following sections was undetaken to further clarify the nature of
strain in porous Si and to explain the experimental findings on the influences of temperature and
ambient gases on the strain in porous Si.!>1® What resulted from this investigation is a model for
porous Si, where the large internal surface is the primary feature of this material.15 We will discuss
briefly heteroepitaxial films of CoSig (Ref. 2) and GeSi alloy!” grown on porous Si substrates and

the effects of such capping layers on the strain change in porous Si.!®

5.2 Strain and stress in porous Si(100)

In this section, we study the strain in porous Si formed by anodization of Si(100) wafers, and

the change of that strain in various environments.!5:16

A. Formation of Porous Si Layers

Single crystalline p*+-Si(100) wafers of 2-in diameter with a resistivity of ~ 0.01 Qcm were used
for the formation of porous Si layers. Si wafers were first cleaned by repeatedly etching in boiling
nitric acid and dipping in HF solution. Porous Si layers of ~ 2um were then formed by anodization
of the pt-Si wafers in a 30% HF electrolyte. The porosity of the porous Si layers thus formed is

about 30%. After the formation of porous Si, the samples were kept at room temperature in air.

B. Experimental Results

Double crystal x-ray diffractometry was used to measure the strain of porous Si (the relative
difference between the lattice constants of the layer and the substrate). The perpendicular strain,
e+, was extracted from the x-ray rocking curves diffracted from (400) planes. The parallel strain, ell,
was extracted from (311) asymmetrical diffraction planes. The strain measured several days after
the porous Si was made is e+ ~ 0.1% and e!! < 0.01%. When porous Si is kept at room temperature
in air, the perpendicular strain e+ increases from ~ 0.1% to ~ 0.3% in ~ 18 months, and the parallel
strain ell remains zero. Since it is unlikely that the microstructure of porous Si changes at room
temperature, the increase of strain must be caused by other sources, such as absorption of gas from
air.

Figure 5.2-1(a) shows the 2 MeV “He backscattering and [100] axial channeling spectra of as-
prepared porous Si and bulk Si. The backscattering yield from Si atoms in porous Si is ~ 25%
smaller than that from bulk Si. This indicates that porous Si contains a high concentration of
light impurities. The backscattering yield rises by ~ 10% at ~ 0.7 MeV in Fig. 5.2-1(a). This
can be attributed to the backscattering from O atoms in porous Si. The ratio of the yields from
Si to that from O gives a relative concentration of Si to O of about 1:0.3. The minimum yield of
porous Si (~ 30%) is much larger than that of bulk crystalline Si (~ 3%), and increases rapidly
as the aligned beam penetrates into the porous Si. This means that porous Si contains a high
concentration of dechanneling scattering centers. However, x-ray rocking curves indicate that the Si
lattices in porous Si layers are highly ordered. The large minimum yield may therefore be caused
by the surface dechanneling of an aligned incident beam, since porous Si has a very large specific

surface area ( ~ 200 m?/cm3; see, e.g., Ref. 6). Furthermore, the spectra of the samples kept in air

2 MeV *He backscattering spectra

O Si

1500 bulk Si ff
n — e
: 1000} Wade,
9 } Neg ZorpU Si
Ss) r ° J

500 - random
| ©ee[100] aligned
bulk Si
0 |
0.6 0.8 1.0 1.2
Energy (MeV)
3.05 MeV ‘He oxygen resonance spectra

4000+

3000+
gt O
2000
8 L

1000

porous Si

l I.
0.8 1.0 1.2 1.4 1.6 1.8
Energy (MeV)

FIG. 5.2-1 Backscattering spectra of porous Si and bulk Si: (a) a 2 MeV *He beam incident along
a near-normal random direction (solid lines) and along the [100]-aligned direction (dotted lines); (b)
3.05 MeV oxygen resonance of a bulk Si, of a porous Si after storage of 18 months in air at 23°C,

and of a thermally oxidized Si wafer (~ 1 pm SiO2/Si).

at room temperature for ~ 18 months are the same as those shown in Fig. 5.2-1(a) for the as-
prepared one. This implies that the overall structure of porous Si changes little upon storage in air
at room temperature.

We also used the 3.05 MeV !®O(a,a)!®O resonance backscattering!® to measure the oxygen
concentration in porous Si samples that had been kept at room temperature in air for ~ 18 months
(Fig. 5.2-1(b)). By comparing the oxygen resonance peaks from porous Si and from thermally grown
SiO2, we estimat the oxygen concentration of porous Si to be Si:O ~1:0.3, the same as the value
obtained before. By tuning the incident *He ion energy so that the resonance scattering occurs at a
different depth, an oxygen depth profile was obtained. The oxygen composition is roughly the same
throughout the porous Si layer, indicating that there exists a substantial fraction of oxidized Si over
the entire porous Si layer. The oxygen fraction of ~ 23 at.% is consistent with the assumption that
the pores of the porous Si with a specific surface area of ~ 200 m?/cm® and a porosity of ~ 30%
are covered by a native oxide layer of ~ 1 nm thick.

Auger electron spectroscopy (AES) equipped with Ar+ ion sputtering was used for measuring
the elemental depth profile of porous Si. The detection limit of AES is typically ~ 0.1% of volume
concentration and the depth resolution is about Inm. The detailed shape of the Siz, ,vv Auger
line was recorded to monitor the chemical state of Si in the porous layer. Figure 5.2-2 shows
the AES depth profiles for porous Si stored at room temperature in air for ~ 18 months. The
oxygen concentration is nearly constant throughout the porous layer and equals about 22 to 25

at.%, calculated from the empirical equation!®

_ To/So
To/So + Is: /Ssi

So and Sg; are the inverse Auger sensitivity factors of oxygen and silicon, respectively, determined

from the ratio of the peak-to-peak height of the Ox rr and the Sizyvy signal from a standard SiO»
sample. Io and Is; are the peak-to-peak amplitude of the oxygen and silicon Auger spectrum in the
derivative mode. The oxygen concentration obtained from the Auger spectrum agrees well with that
obtained from '°O(a,a)'®O resonance. The detailed line shape of the Siz, ,vv derivative Auger
spectrum shows that the bonding state of the Si orbitals in the porous layer is mainly mixed sp? and
Si-O instead of pure Si-O bonding. This is consistent with the idea that only the specific surface of
the pores are covered with a native oxide.

To better understand the possible causes for the observed strain change in the porous Si layer,
we conducted a set of experiments to monitor the influence on the strain of external variables such

as temperature and ambient gas.

AES depth profile of porous Si

100
x 80- 4
3 A
o L . 4
—_ Si
[e}
se = 4
2°70 0 :
& 20; 7
O i 4
0 n i 1 i r
0 20 40

FIG. 5.2-2 Auger electron spectroscopy depth profile of oxygen and silicon in porous Si of the

sample that had been stored in air at room temperature for ~ 18 months after the formation of

porous Si.

0.4

‘Sputtering Time ( min )

0.2

-O0.2-

—0.4F

er (%)

7 months later
stored at 23°C in air

1 month later
tored at 23°C in

right after annealing
in vacuum for 30 min

FIG. 5.2-3 Perpendicular strain «+ of porous Si aged for 18 months in air at 23°C and then

annealed in vacuum for 30 min, measured immediately after annealing (0), after storage in air at

400 600

T (°C )

room temperature for about 1 month (¢) and 7 months (filled triangle).

800

Samples that had been stored for ~ 18 months in air were annealed in vacuum (~ 5x 10-7 torr)
between 200°C and 800°C for 30 min. X-ray rocking curves were taken at room temperature in air
immediately after the annealing was completed. The measured perpendicular strain «+ decreases
monotonically from ~ 0.3% with increasing annealing temperature (0 in Fig. 5.2-3) and saturates
at ~ —0.6% above 600°C. The measured parallel strain «ll remains zero in all cases. These results
suggest that the decrease of strain is caused by desorption of gas from porous Si during vacuum

annealing.

To reveal kinetics and energetics of gas desorption from porous Si, we monitored the decreases
of «+ as a function of annealing duration in vacuum at 300°C, 400°C and 600°C (Fig. 5.2-4(a)).
The time dependence of e+ is characteristic to thermally activated desorption.2° The strain decreases
exponentially with a time constant, r, which decreases with increasing annealing temperature from
about ~ 9 min at 300°C to ~ 4 min at 600°C. Figure 5.2-4(b) is an Arrhenius plot of 7 as a function
of inverse temperature, 1/T. The slope yields the activation energy, Eq + 0.1 eV, of desorption
of molecules from porous Si. It is about the same as the typical binding energy of physisorbed
molecules on solid surfaces.?! This fact convincingly shows that the decrease of strain upon thermal

annealing is caused by the desorption of physisorbed molecules from porous Si.

When the annealed porous Si samples are stored at room temperature in air, the perpendicular
strain increases again. Figure 5.2-3 also shows the perpendicular strain of the annealed porous Si
after storage at room temperature for ~ 1 month (e) and for ~ 7 months (filled triangle). These
facts suggest that the increase of strain in the porous Si stored in ambient air is caused by the

adsorption of impurity atoms (molecules) from air.

To identify the molecular species that alter the strain, we annealed samples at 400°C in vacuum
for 30 min and then kept them in different ambient gases (vacuum, forming gas, Oz, air, moisture)
at room temperature. The perpendicular strain «+ increases with time approximately as a com-
plementary exponential (Fig. 5.2-5). Little change of strain was detected for the sample kept in
vacuum. The observed change is attributable to air exposure during the measurements. The strain
changed most for the sample kept in moisture (water-saturated air). This shows that the observed

increase of strain at room temperature is caused mainly by the absorption of water molecules.

The non-resonant backscattering and channeling spectra for all of the samples discussed above
are indistinguishable within experimental uncertainty. This indicates that thermal annealing up to
800°C in vacuum does not cause significant structural or compositional change in porous Si. The

sensitive 3.05 MeV oxygen resonance backscattering confirms that the oxygen concentration

0.4
L porous Si(100) in vacuum ses J
0.24 5-
L 300°C ;
oo OF 4
IN i id o
~ 400°C
4 -—0.2 ZZ ”
—0.4- 4
L 600°C 4
0.65) *s i 1 aT
0 20 40 60
Annealing Duration ( min )
10}
| Arrhenius plot of time constant ]
E al
! Pr l os !

1.2 1.4 1.6
1/T ( 1/10°K )

FIG. 5.2-4 Decreasing rate of perpendicular strain in porous Si upon vacuum annealing: (a) e+
versus annealing duration at 300, 400, and 600°C; (b) Arrhenius plot of the time constant 7 of the

decrease of the perpendicular strain.

| porous Si(100) at 23°C
after at 500°C in vacuum for 30 min
-O.1F 4
~~ ' moisture
X 9 2b
ny | oxygen A
~0.3- ° air
. forming gas a
-0.44 C= 77 7 Feet |

0 2 4 6 8 10
Storage Duration ( days )

FIG. 5.2-5 Time evolution of the perpendicular strain e+ of the porous Si stored at room temper-

ature in different ambient gases after being annealed in vacuum at 500°C for 30 min (see the point

marked by star in Fig. 5.2-3).

adsorption desorption

[AVN NoY

—t— oxide
silicon
0 1 wk 15y 30min 1month
RT RT 400°C RT
air air vac. air

LL | | | |

FIG. 5.2-6 Schematic representation of porous Si and of the evolution of the perpendicular strain
as a function of the sample history given on the abscissa. The upwards (downwards) arrows in the
silicon rods indicate the positive (negative) perpendicular strain; their length suggests the magnitude

of the strain. The perpendicular strain evolves as a result of absorption by and their desorption

from the native oxide.

decreases by only a small fraction (<~ 10%) of its original value after thermal annealing in vacuum.

The native oxide thus remains present in porous Si after annealing in vacuum up to 800°C.

C. Discussion

The increase of strain in porous Si at room temperature in ambient air has been discussed by
Young et al.? who observed that the strain of porous Si increases at room temperature when stored
in ambient air for a few weeks after the formation of the porous Si. They attribute the increase to
slow thickening of the native oxide layer. This model is inconsistent with our observations where we
also find the increase of strain of the porous Si stored in air at room temperature, but the oxygen
content in the porous Si remains constant.

Decreases of strain in porous Si upon thermal annealing have been discussed by others.!!)12
Labunov et al.!! observed that the perpendicular strain of porous Si layer decreases from ~ 4 x 1074
to ~ —1 x 10~* upon thermal annealing in Hz at ~ 1000°C and they attributed it to the collapse of
porous structure to form closed cavities. The change we observe upon annealing in vacuum is about
20 times greater (~ —1% vs ~ —0.05%), and occurs at lower temperature (600°C vs 1000°C) where
the porous Si structure is known not to change significantly.!° It is therefore clear that the decrease
of strain at low temperature (e.g., < 800°C) is not related to the overall structural change of porous
Si. Sugiyama and Nittono!? also found that the perpendicular strain decreases from ~ 4 x 1074
to ~ —2 x 10-3 upon annealing at ~ 450°C in vacuum. At the same time, H» concentration also
decreases. These authors therefore proposed that the decrease of strain is related to the formation of
a clean pore surface by desorption of chemisorped H» at an elevated temperature. This interpretation
is inconsistent with our experimental findings where we observe a similar decrease of strain, but the
oxide layer covering the pore surface remains. To better clarify the influence of H» on strain in a
porous Si layer, we annealed a sample at 400°C for 30 minutes in pressurized Hy (~ 45 psi) gas.
The strain decreases, as it does in vacuum. For comparison, we also heated a sample at 400°C for
30 minutes in air. The strain increases by a small amount. ‘These results clearly demonstrate that

H2 is not the major cause for the change of strain.

D. Model

On the basis of the above experimental results, we propose the following model for the strain
in porous Si (see Fig. 5.2-6 for illustration). In simplified terms, a porous Si layer may be modeled
as consisting of many long Si rods of a small average diameter d (~ 10 nm), standing on the bulk Si

substrate. The surface tension between the rods and the ambient gas, 7, tends to reduce the surface

area by decreasing the size of the Si rods. An equilibrium state is reached when the surface tension
is balanced by the volume strain such that the total energy of surface and volume is minimized. The
resultant compressive strain is ~ —7y/pd (p is elastic modulus). It has a value of ~ —10~3, using
typical values of y ~ 107 erg/cm? and yp ~ 101? dyn/cm?. On the other hand, the formation of
a native oxide of thickness djz (~ 1 nm) on the pore surface produces a compressive stress of oor
(~ —10'° dyn/cm?) in the oxide layer.?? This gives rise to a tensile stress of ¢ ~ —Oo2doz/d~ 10°
dyn/cm? in the Si rods, inducing a tensile strain of ~ o/p ~ 10-3. The compressive strain from
surface tension and the tensile strain from oxidation compensate each other. The constraint imposed
by the substrate on the porous Si layer confines the parallel strain in the layer to be zero. The
perpendicular strain thus estimated is consistent with the measured values of ~ 10-4-10~3 (see also
data in Refs 8,9,11,12). If the sample is kept in air, the oxide layer in porous Si will absorb gas
molecules such as H2O from air. This causes its volume to expand,?9 inducing a tensile strain in the
Si rods. When the sample is heated in vacuum, the absorbed gas molecules such as HzO will escape
from the oxide layer. This causes volume contraction in the native oxide, inducing a compressive
strain in the Si rods. This model qualitatively explains observations made here and in the literature.
A quantitative estimation is, however, obscured by the complexity of the microstructure of porous
Si and of the native oxide.
E. Summary

The perpendicular strain in the as-formed porous Si layer is ~ 10-3. The parallel strain is not
measurable (< 10-*). Upon annealing in vacuum at an elevated temperature, the perpendicular
strain decreases, while the parallel strain remains zero. The perpendicular strain of the annealed
sample increases with time at room temperature by amounts that depend on the ambient gas (vac-
uum, forming gas, Oo, air, moisture). The pores in the porous Si are covered by a native oxide layer.
The observed changes of strain in porous Si layers are caused mainly by modification of stress in
the oxide layer that is due to desorption and absorption of gas. The measured strain in porous Si
is the net result of stress induced by surface tension, oxide formation, and exchange of gas between

the oxide and the ambient gas.

5.3 Epitaxial films of GeSi and CoSiz on porous Si

Luryi and Suhir?* studied the elastic system of a heteroepitaxial film grown on a patterned
substrate of small seed pads with a lateral dimension, / (Fig. 5.3-1). They showed that for a given
lattice mismatch between the film and the substrate, f, the entire elastic stress in the film can be
accommodated without dislocations when the lateral dimension I is sufficiently small, | < Imin (f).
For example, in the case of a Ge film on Si (f = 4%), Imin is ~ 10 nm.?4 Porous Si provides a

possible candidate for such a substrate.

A. Properties of films grown on porous Si

We have analyzed epitaxial GeSi alloy and CoSiz films grown on porous Si substrates by MBE at

UCLA,?"!7 and have compared the properties of these films with those grown on regular Si substrates.
GeSi on porous Si(100)

Epitaxial GeSi alloy films 8 to 200 nm thick with Ge compositions of 20%, 50%, and 100% grown
on both porous and regular Si(100) substrates were characterized by backscattering spectrometry
(BS), x-ray double crystal diffractometry (DCD), and transmission electron microscopy (TEM).?”
Channeling spectra of films on porous and regular substrates are the same, meaning that the films
on porous Si are also highly epitaxial. The strain in the films on porous and regular substrates is
also the same, meaning that there is no reduction of misfit dislocations in the film on porous Si, in
contrast to what one would expect. However, the x-ray diffraction peak from the film on the porous
substrate is broader than that from the corresponding film on the regular one, suggesting that the
strain distribution in a film grown on a porous substrate is very inhomogeneous. Plane view TEM
confirmed that the average dislocation spacing in films on porous and regular Si is the same, and
that the strain in films on porous Si is inhomogeneous. !":75

The apparent discrepancy between Luryi and Suhir’s model and the experimental observations
may be due to the fact that seed pads in the model are isolated while the rods in porous Si are
interconnected (Fig. 5.3-2)?5

CoSig: on porous Si(111)

Epitaxial CoSiz films of various thickness grown on regular and porous Si(111) substrates were
also analyzed by BS, DCD, and TEM.? The same results as those discussed previously for GeSi films
were obtained. The room temperature strain in a CoSi, film on porous Si is again caused by the

difference of thermal expansion coefficient. between the film and the substrate (see also Ch. 2.5 and

2.6).

‘Az

w(0,z) | 0 | L_

Si SUBSTRATE )

FIG. 5.3-1 Illustration of a heteroepitaxial Ge film on a Si substrate of seed pads with lateral

dimension |. The strain energy profile, w(0,z), for the pad mid-cross section is shown on the left

(from Ref. 24).

(a) _ Cb)

FIG. 5.38-2 Schematic drawings of (a) a patterned substrate as used in Luryi and Suhir’s model

(Fig. 5.3-1), and (b) a porous Si substrate (from Ref. 26).

0.4
0.2 1 month later
L stored at 293°C in air
~ OF
x - right after annealing
~ 9.2L in vacuum for 30 min
b L
—~0.4/ epitaxial CoSig capping layer
L. grown on porous Si
-0.6b formed on Si(111) substrate
1 L \ ! : ! ‘ |
0 200 400 600 800
T( °C)

Fig. 5.3-3 Perpendicular strain e+ of CoSig-capped porous Si(111) aged for 18 months in air and
then annealed in vacuum for 30 min, measured immediately after annealing (0), and after storage

in air at room temperature for 1 month (e). The figure has identical scales as those of Fig. 5.2-3 for

comparison.

In summary, we conclude that heteroepitaxial growths on porous Si are similar to those on
regular Si.

B. Properties of porous Si with caps

To confirm our model!® that the exchange of gaseous species between porous Si and the ambient
gas causes the observed strain changes (see Ch. 5.2), we analyzed the strain change in porous Si
with caps.!§

GeSi cap/porous Si/Si(100)

The perpendicular strain of porous Si(100) immediately after the GeSi films were grown by
MBE is ~ —0.1%. This decrease of strain from that of the as-grown porous Si (~ 0.1%, see Chp.
5.2) was caused by desorption of impurity atoms from porous Si when the sample was heated to
~ 600°C in the MBE system prior to film growth. The perpendicular strain does not change at
room temperature in ambient air in ~ 18 months, regardless of the different Ge composition and
cap thickness. Furthermore, the perpendicular strain remains unchanged upon thermal annealing
at 600°C for 30 min in vacuum. These facts show that epitaxial GeSi films are perfect seals that
prevent any exchange of molecules between the porous Si and the ambient gas. The result verifies
that the change of strain in uncapped porous Si is caused by gas desorption and adsorption.

CoSig cap/porous Si/Si(111)

The perpendicular strain of the porous Si(111) increases from ~ 0.2% to ~ 0.3% at room
temperature in ambient air in ~ 18 months. We attribute this change to the presence of pinholes in
CoSiz films (pinhole density ~ 10°/cm?, Ref. 2). The perpendicular strain of the porous Si(111) with
a different cap thickness is the same within experimental uncertainty. Upon thermal annealing in
vacuum, the perpendicular strain decreases monotonically with increasing temperature and saturates
at ~ 600°C (o in Fig. 5.3-3). When stored in air at room temperature, the perpendicular strain
of the annealed porous Si increases a little in 1 month (e in Fig. 5.3-3). These results show that
the epitaxial CoSiz caps do not fully prevent the exchange of gas molecules between the porous Si
and the ambient gas. The result also suggests that one could measure the strain change of capped

porous Si to probe for the presence of pinholes in the capping layer.

References
1. K. Imai and H. Unno, IEEE Transactions on Electron Devices, ED-31, 297 (1984).
2. See, e.g., Y. C. Kao, K. L. Wang, B.J. Wu, T.L. Lin, C.W. Nieh, D. Jamieson and G. Bai, Appl.
Phys. Lett. 51, 1809 (1987).
3. L.T. Canham, Appl. Phys. Lett. 57, 1046 (1990).
4. A. Uhlir, Bell. Syst. Tech. J. 35, 333 (1956).
5. D. R. Turner, J. Electrochem. Soc. 105, 402 (1958).
6. R. Herino, G. Bomchil, K. Barla, C. Bertrand and J. L. Ginoux, J. Electrochem. Soc. 134, 1994
(1987).
7. M.LJ. Beale, N.G. Chew, M.J. Uren, A.G. Cullis, and J.D. Benjamin, Appl. Phys. Lett., 46, 86
(1985).
8. K. Barla, R. Herino, G. Bomchil, J. C. Pfister and A. Freund, J. Cryst. Growth, 68, 727 (1984).
9. I. M. Young, M. I. J. Beale and J. D. Benjamin, Appl. Phys. Lett. 46, 1133 (1985).
10. R. Herino, A. Perio, K. Barla and G. Bomchil, Mater. Lett., 2, 519 (1984).
11. V. Labunov, V. Bondarenko, L. Gurenko, A. Dorofeev and L. Tabulina, Thin Solid Films, 137,
123 (1986).
12. H. Sugiyama and O. Nittono, Jpn. J. App]. Phys. 28, £2013 (1989).
13. L.G. Earwaker, J.P.G. Farr, P.E. Grzeszczyk, I. Sturland and J.M. Keen, Nucl. Instr. and Meth.,
B9, 317 (1985).
14. R.W. Hardeman, M.I.J. Beale, D.B. Gasson, J.M. Keen, C. Pickering and D.J. Robbins, Surface
Sci., 152, 1051 (1985).
15. G. Bai, Kun Ho Kim and M-A. Nicolet, Appl. Phys. Lett. 57, (1990).
16. K.H. Kim, G. Bai, and M-A. Nicolet, J. Appl. Phys. 69, 2201 (1991).
17. C.H. Chern, ¥.C. Kao, C.W. Nieh, G. Bai, K.L. Wang, and M-A. Nicolet, in 2nd Int. Symp. on
Si-MBE, (Honolulu, Hawaii, 1987).
18. W. K. Chu, J. W. Mayer and M-A. Nicolet, Backscattering Spectrometry (Academic Press,
New York, 1978), p. 210.
19. C. C. Chang, Characterization of Solid Surfaces, Plenum, New York, 1974.
20. A. V. Kiselov, Surface Chemistry of Oxides, Discussions of the Faraday Society, No. 52 (Faraday
Society, London, 1971), p. 14.
21. H. J. Krenzer and Z. W. Gortel, Physisorption Kinetics (Springer Verlag, Berlin, 1985), p. 8.
22. E.P. EerNisse, Appl. Phys. Lett. 35, 8 (1979).

100

23. I. Blech and U. Cohen, J. Appl. Phys. 53, 4202(1982).
24, S. Luryi and E. Suhir, Appl. Phys. Lett. 49, 140 (1986).
25. Y.H. Xie and J.C. Bean, J. Vac. Sci. Technol. B8, 227 (1990).

101

Part IT

Ion Implantation in Si and Heterostructures

102

Chapter 6 Ion-Solid Interaction

When an energetic ion penetrates into a solid, it transfers energy to the target atoms, slows
down and finally stops inside the solid. This transient process takes place in ~ 10-1! — 10-13s.
The energy loss can be seperated into two parts, electronic and nuclear stopping. On the basis
of the pioneer works by Bohr, Bethe and Bloch, Fermi and Teller, and others on the interaction
between incident ions and target atoms, Lindhard, Scharff, and Schiott (LSS theory) developed a
unified approach to calculate the stopping of low energy (10 keV-1 MeV) ions and their range in
solids. The theoretical prediction agrees excellently with experimental results (within ~ 5%). A
target nucleus that acquires energy exceeding the displacement threshold energy (~ 15 eV) leaves
its lattice site, and generates secondary collision. A collision cascade is thus formed in a solid.
One effect of the cascade is sputtering—removal of target atoms from the surface of a solid.? The
cascade also generates defects in the solid, such as interstitials and vacancies. Kinchin and Pease?
developed a simple model to estimate the amount of such Frenkel pairs. However, the model applies
only to the case of light (low atomic number) ion implantation at low temperature where individual
defects are well seperated and are not mobile. In the case of heavy (high atomic number) ion
implantation, high energy density is deposited around the track of incident ions, and a spike may
form.4 Winterbon, Sigmund, and Sanders (WSS theory),® established a model to calculate the
cascade energy density. Dense cascade (or spike) causes enhanced production of defects compared
to that predicted by the Kinchin-Pease model.® Recently, Cheng et al.” applied the fractal concepts
to the collision cascade and proposed “space filling” as a criterion of thermal spike formation. In
the case of ion implantation at an elevated temperature, individual defects are mobile, and damage
production becomes a dynamic process where defects interact, cluster, and recombine. This process
results in complex defect structures. In both cases, the final microstructure and concentration of
defects produced by ion implantation are almost impossible to predict.

One key application of ion implantation technology is semiconductor doping in integrated circuit
fabrication, because of precise controllability and high reproducibility of implanted dopant profiles.
Associated with the introduction of dopants is the damage produced in a target. The annealing
of damage and the recovery of crystalline perfection of the target is critical to activating dopants
and achieving the desired electrical properties. There are many books and reviews on damage
production and annealing in ion implanted Si.>? Defects produced by ion implantation also act as
dilatational centers, which induce strain.'°? In most semiconductors such as Si, Ge, and GaAs, the

strain is positive, meaning that the damage causes volume expansion.’ One known exception is

103

MeV ion-implanted InP, where the negative strain is induced.!? Ion implantation can also produce
an amorphous phase, and provides an excellent vehicle for studying thermodynamics and kinetics
of phase transformation among crystalline, amorphous, and liquid Si. An amorphous Si layer on a
crystalline Si substrate recrystallizes by solid phase epitaxy upon thermal annealing.!3 The rate of
regrowth follows an Arrehnius behavior with an activation energy of 2.7 eV.'* Solid phase epitaxy
can also be induced by a deeply penetrating ion beam, with lower activation energy.!®

In recent years, ion implantation technology has been extended to other fields of materials
research. One example is the formation of mesotaxial silicides such as CoSi2!® and CrSig!” by high
dose implantation followed by thermal annealing. Formation of epitaxial GeSi alloy layers by high
dose Ge implantation into Si substrates and solid phase epitaxy was also demonstrated./® This
technique has the potential of being used to make heterostructures that are difficult to grow by
vacuum deposition. Another application is the fabrication of amorphous-crystalline superlattices
from epitaxial ones by using ion implantation to induce selective amorphization.!®

In the next two chapters, we will discuss some aspects of ion-solid interaction. The damage and
induced strain in the implanted layers are emphasized. Chapter 7 focuses on damage production
and annealing in bulk Si crystals, while Chapter 8 concentrates on the effects of ion implantation

into silicides/Si and GeSi/Si heterostructures.

104

References
1. J. Lindhard, M.E. Scharff, and H.E. Schiott, Kgl. Dan. Vid. Selsk. Mat. Fys. Medd. 13, 14
(1963).
2. P. Sigmund, Appl. Phys. Lett. 14, 114 (1969).
3. G.H. Kinchin and R.S. Pease, Rep. Prog. Phys. 18, 1 (1955).
4. P. Sigmund, Appl. Phys. Lett. 25, 169 (1974).
5. K.B. Winterbon, P. Sigmund, and J.B. Sanders, Kgl. Dan. Vid. Selsk. Mat. Fys. Medd. 37,
14 (1970).
6. D.A. Thompson and R.S. Walker, Rad. Effects 36, 91 (1978).
7. Y.-T. Cheng, M-A. Nicolet, and W.L. Johnson, Phys. Rev. Lett. 58, 2083 (1987).
8. J.F. Gibbons, Proc. IEEE 60, 1062 (1972).
9. J.W. Mayer, L. Eriksson, and J.A. Davies, Ion Implantation in Semiconductors, Academic, New
York, 1970.
10. G. Bai and M-A. Nicolet, J. Appl. Phys., July 15, 1991.
11. V.S. Speriosu, B.M. Paine, M-A. Nicolet, and H.L. Glass, Appl. Phys. Lett. 40, 604 (1982).
12. C.R. Wie, T. Jones, T.A. Tombrello, T. Vreeland, Jr., F.L. Xiong, Z. Zhou, G. Burns, and F.H.
Dacol, Mat. Res. Soc. Symp. Proc. 74, 517 (1987).
13. L. Cespregi, E.F. Kennedy, S.S. Lau, J.W. Mayer, and T.W. Sigmon, Appl. Phys. Lett. 29,
645 (1976).
14. G.L. Olson, Mat. Res. Soc. Symp. Proc. 35, 25 (1985).
15. J. Linnors and G. Holmen, Phys. Rev. B32, 2770 (1985).
16. A.E. White, K.T. Short, R.C. Dynes, J.P. Garno, and J.M. Gibson, Appl. Phys. Lett. 50, 95
(1987).
17. A.E. White, K.T. Short, and D.J. Eaglesham, Appl. Phys. Lett. 56, 1260 (1990).
18. D.C. Paine, D.J. Howard, N.G. Stoffel, J.A. Horton, J. Mater. Res. 5, 1023 (1990).
19. D.J. Eaglesham, J.M. Poate, D.C. Jacobson, M. Cerullo, L.N. Pfeiffer, and K. West, Appl.

Phys. Lett. 58, 523 (1991).

105

Chapter 7 Damage Production and Annealing in Si(100)

7.1 Introduction

Radiation damage in semiconductors produced by energetic electrons or neutrons attracted
much attension in the 60’s.! A great amount of information about the nature of simple defects and
their annealing characteristics has been assembled with techniques such as electron paramagnetic
resonance” and infrared optical absorption.? Since the 70’s, ion implantation and ion-solid interac-
tion became the focus of investigation because of their technological importance. Unlike electron-
produced damage, which consists mainly of isolated interstitials and vacancies, defects produced by
ion implantation are complex. Additional techniques such as channeling spectrometry®, transmis-
sion electron microscopy’, optical reflection spectroscopy®, and double crystal diffractometry®, have
been used to reveal various aspects of the defect structures of ion-implanted semiconductors and of
Si in particular.1° However, a detailed picture of the nature of defect production and their stability
is still lacking. In recent years, ion implantation technology has found new applications in areas such
as ion-beam-induced epitaxial growth!! and the synthesis of buried heterostructure.!? The critical
role of point defects produced by ion implantation in enhanced diffusion of dopants upon thermal
annealing has been recognized.!3 An improved understanding of ion-induced defect production and
annealing promises deepened insights into these phenomena.

In the following sections, we employ both x-ray double crystal diffractometry and MeV *He
channeling spectrometry to analyze quantitatively the damage in a Si(100) crystal produced by im-
plantation of 200—600 keV ions at both room and liquid nitrogen temperature. Some perspectives on
the nature of defects produced by the implantation, and their stability and annealing characteristics

are discussed in the light of the experimental results.

106

7.2 Damage by self-implantation at room temperature

Recognizing the important role that defects play in an implanted material, we undertook a
careful analysis of damage produced by implantation of 230 keV 28Si in a Si(100) crystal at room
temperature. Double crystal diffractometry and channeling spectrometry were used to characterize

the strain and the defect concentration.

A. Sample preparation

The samples were prepared by implantating 230 keV 78Si ions into ~ 1 Qcem n-type Si(100)
wafers at room temperature in high vacumm (~ 10-7 Torr). The wafers were chemically cleaned
before loaded into the implanter. The surface normal was misoriented by 7° from the incidence line
of the beam to avoid channeling. The beam current was limited to be small (~ 0.2 4A/cm?) to

minimize beam heating. The implantation dose ranges from 10!4/cm? to 10!5/cm?.

B. Double crystal diffractometry

Double crystal diffractometry was used to monitor the strain in the implanted layer. X-ray
rocking curves of both symmetrical (400) and asymmetrical (311) diffractions were taken at room
temperature in air 1 hour after implantation and 6 months later. The strain profile as a function of
depth is extracted by simulating the experimental rocking curve using the dynamical x-ray diffraction
theory.!* The parallel strain, ell, of all samples considered here is zero within experimental sensitivity
(~ 10-*), meaning that the lateral lattice spacing in the implanted layer is constrained to be equal
to that in the substrate. The perpendicular strain, «+, is positive and increases with dose, meaning
that the perpendicular lattice spacing in the implanted layer is larger than that of the substrate and
increases with the damage in the layer. The expansion of the lattice in the implanted layer implies
that the strain contribution is dominated by the interstitial-like defects.15 The magnitude of the
perpendicular strain relaxes linearly with the logarithm of time at room temperature. The defects
are almost stabilized at room temperature in less than 1 hr after implantation. Subsequently to
that, relaxation proceeds very slowly. We shall neglect this subsequent time dependence thereafter
because it is minute (~ 0.02% in 6 months).

Figure 7.2-1 shows a set of selected x-ray rocking curves from symmtrical (400) diffraction. The
x-ray diffraction peak intensity from the implanted layer decreases rapidly when the dose increases
(notice the ordinate scales). At the same time, the largest angular separation between the diffraction
peaks of the implanted layer and the substrate increases. These facts mean that both the damage

and the strain rise rapidly as the dose increases (notice the abscissa scales). Figure 7.2-2 is a plot of

107

3 0.6
—~ + (a) 1x10!4/cm? ~0.5+(c) 4.8x10!4/em?
Sab *0.4b
Bt 20.3b
n n L
g 1b § o.2b
rE eT
mF on46° = 0.1 i 9=45°

Leccclocesbeomabeamal pou i | |

-8 76 —0.05 0 S35 02-01 0
1.0 0.3

~al.(b) 3.5x10!4/em? L (d) 4.8x10!4/cm?
0.8

x BX
l 0.21

0.6 ~

> Ps

wy 0.4- n

8 - 3 O.1-

&0.2F A: L
L 8=45° 9=45°
sroute i Porites p41 t I L L 1 |

0.20 0.15 -0.10 0.05 0 0.05 -¥5 0.4 —0.8 -0.2 -01 0
Ae (°) da (°)

FIG. 7.2-1. Fe Kg, x-ray (wavelength=0.1936 nm) rocking curves diffracted from the symmetrical
(400) planes of the Si(100) samples implanted at room temperature by 230 keV ?8Si ions to the
doses of (a) 1x, (b) 3.5x, (c) 4.3x, (d) 4.8 x 10!4/cm?.

+ 230 kev 783i
into Si(100)

1.0}-

max (% )

0 2 4 6 8 10
¢ ( 10'*/cm® )

FIG. 7.2-2 The maximum perpendicular strain (o) extracted from dynamical x-ray diffraction
simulations of the experimental rocking curves as a function of the Si implantation dose. The solid
line is to stress the trend. The filled circles correspond to the samples for which the x-ray rocking

curves are shown in Fig. 7.2-1.

108

the maximum value ¢+ 4, from the perpendicular strain profile as a function of implantation dose.
There exist three distinct regimes in this room-temperature implanted Si layer: (1) the strain builds
up slowly to a critical dose of ~ 4 x 10!4/cm? ; (II) the strain then rises rapidly within a very narrow
dose range, and (III) finally saturates beyond a dose of ~ 4.5x 1014/cm?. Dynamical x-ray diffraction
simulation also gives an estimate of the static Debye-Waller factor, (6r),ms, (the root mean-square
of the atomic displacement from the lattice site). At any depth, the static displacement (6r)pms
is approximately proportional to the perpendicular strain. The maximum displacement is small in
regime I (~ 0.01 nm), increases to ~ 0.06 nm in regime II and saturates at this value in regime
III. One can obtain an “equivalent temperature,” T.,, corresponding to the atomic displacement of
(6r)rms, by using the Debye-Waller formula,

2. Qh’ T,
(67) pms = ike? (7.2 _ 1)

where M is the mass of silicon atoms and © (=645K!®) is the Debye temperature of crystalline Si.
The maximum “equivalent temperature” thus obtained for the implanted Si layer is about 0.2 Ty,
(Tm = 1685K is the melting temperature) in the low damage regime (I), and reaches about 4T7,, in
regime (III). The very high “equivelent temperature” associated with regime (III) suggests that the

heavily damaged state is far from equilibrium and hence is highly metastable.

C. Backscattering and channeling spectrometry

Backscattering and channeling spectrometry were used to measure defect concentration in the
implanted layer. Figure 7.2-3 shows 2 MeV *He [100] axial channeling spectra of selected samples.
The defect concentrations were extracted from the channeling spectra according to the following
model. The channeling yield, yp, of the damaged layer consists of two contributions: normal
backscattering from the dechanneled (“random”) fraction of an aligned incident beam, yr, and

direct backscattering of a channeled beam from the defect of concentration, cp,!”

XD =xr+(1— xr)ep. (7.2 — 2)
The dechanneling of an aligned incident beam in the crystal equals

Xr = Pp +(1—Pp)xv, (7.2 — 3)

where Pp is the probability that a channeled incident ion is dechanneled by the defect and yy is

the channeling yield of a virgin crystal. Combining Eqs. (7.2-2) and (7.2-3) gives

YD =cp +(l1—cp)Pp, (7.2 — 4)

109

10000

| 230 keV *8Si implanted into Si(100) at 23°C |
AA stan at
8000+ ame 4

random late

))

6000} . |

7.7x10'4/em?!

Te weet eV

Counts

4000

2000

et oat Natt mre
Nee ON ere aey peat ts as

(100) aligned virgin
V6 0.8 1.0 1.2
Energy ( MeV )

Si sample and a sample implanted 7.7 x 10!4 ?8Si/cm? in which a continuous amorphous layer forms.

100- 230 kev 78si

| into Si(100)

L. at 23°C a
80 (a)
60 4

Cp™®* ( %, )

TRIM88,-”
40+ a

0 2 4 6. 8 10
¢ ( 10!*/cm?® )

FIG. 7.2-4 The maximum defect concentration extracted from channeling spectra such as those

of Fig. 7.2-3 as a function of the Si dose. The solid line is to highlight the trend. The filled circle

corresponds to the samples (a-d) shown in Fig. 7.2-3. The dashed line is the maximum value in the

concentration profile of the Frenkel pair predicted by the TRIM88 simulation code of 230 keV ?8Si

implantation in an amorphous Si target.

110

where the dechanneling parameter, yp, is defined by
yp = 2 € 0,1). (7.2 — 5)
l—yv

Yp can be obtained directly from the channeling spectra of the virgin and the damaged crystals.
To extract the defect concentration cp from Eq. (7.2-4), one needs to know how the dechanneling
probability Pp is related to cp. In the single scattering approximation, one can define a dechanneling

cross section, op, by

dP
= =nopcp, (7.2 — 6a)
or
Pp(z) = | nopep(z')dz"’, (7.2 — 6b)

where z is the depth from the surface and n is the atomic density of the crystal. From Eqs. (7.2-4)
and (7.2-6b), one finally obtains

yp(t) = cp(t) + (1—- ep(t)) [ cp(t')dt’, (7.2 - 7)

where
t=nopz (7.2 — 8)

is a dimensionless depth scale. Eq. (7.2-7) is a nonlinear integral equation for the defect concentra-
tion cp. The dechanneling parameter yp(t) can be extracted from channeling measurements, the
defect concentration cp(t) is hence obtained by solving Eq. (7.2-7). We solved Eq. (7.2-7) numer-
ically by adjusting the parameter op to satisfy the boundary condition that Pp equals yp beyond
the damaged layer (where cp = 0), and the defect concentration profile cp(x) was then obtained.
The best-fitted dechanneling cross section op obtained according to the above procedure for the
samples implanted to various doses is about the same. The average value is ~ 7 x 10-!%cm?. This
is significantly smaller than the cross-sectional area of a'‘channel (~ 10-}5 cm?).

The results from channeling measurements are summarized in Fig. 7.2-4, a plot of the maximum
value cf** from the defect concentration profile cp(z) as a function of dose. Although derived from
quite different experimental inputs, the dependence closely resembles that of strain-dose relationship
and also exhibits the three distinct damage regimes with the same critical dose (~ 4 x 10!4/cm?).

To elucidate what the amount of the measured damage means, we computed the maximum
Frenkel pair concentrations produced by 230 keV ?8Si implantation into an amorphous Si target at 0

K using the TRIM88 simulation program!® (dashed line in Fig. 7.2-4). We used the binding energy

111

of 1 eV and the displacement threshold energy of 15 eV as input parameters in the simulation.
One sees that the measured damage in the low dose regime (I) is only ~ 0.2 of that predicted.
This suggests that the majority of initially created defects anneals out at room temperature. That
result is consistent with the observation that single vacancies and interstitials are mobile at room
temperature.!? We thus conclude that the majority of initially created defects in regime I are in the
form of simple Frenkel pairs, which are mobile and readily recombine at room temperature. The
measured stable defects are therefore di-vacancies, di-interstials, and their clusters, formed during
the migration of the point defects.19

In regime II, the damage increases with dose much faster (~ 8 times) than the production of
Frenkel pairs calculated from TRIM88. This indicates that the defect production in the predamaged
crystal is more efficient than that in a virgin crystal, suggesting that the newly produced defects can
destabilize the predamaged crystal and cause the formation of disordered zones of increased size.
Such a process produces about 8 times more displacements than can be directly generated by nuclear
collisions (see Fig. 7.2-4). In other words, in a predamaged crystal with a defect concentration
> 15%, the effective threshold energy of atomic displacement is reduced from ~ 15 eV® in the virgin
crystal to ~ 2 eV, which is approximately the formation energy of point defects in a solid by thermal
activation.® Stated differently, the damage production depends on the interaction with existing
defects beyond the critical defect concentration (~ 15%). The effect of dose is no longer simply
additive. Guided by our observation of the accelerated growth of damage, we model the damage
build-up phenomologically by assuming that the production rate of stable defects is proportional to
the concentration of existing defects. We also take into account the fact that the defects produced
within already existing damaged regions do not increase defect concentration. Combining these two

factors, we obtain the net rate of the production of stable defects,

dep _ tp +Co
Wo ‘(L- ep), (7.2 —9)

where c, and @, are fitting parameters. The solution is' obtainable by direct integration and gives
the growth of defect concentration as a function of dose. The best fit with experimental data (0)
gives co ~ 2x 10-° and ¢, & 3 x 10!3/cm? (solid line in Fig. 7.2-5). The small value of c,
refiects the difficulty in producing stable defects in a virgin crystal. This simple model fits the
data reasonably well. Thompson and Walker?° also observed that the effective threshold energy
for atomic displacement decreases as the energy density deposited in the nuclear collision increases.
They attribute the enhancement to the thermal spike phenomenon. That phenomenon differs from

what we observe in that the thermal spike also occurs in a virgin crystal. The enhancement seen

112

100+ 230 keV *®si —~O—-—
L into Si(100) (20,48) “7 were 4
¢ (518) 2-77 eee
__ 80 at 23°C ao eta) vet a
ne r a A)=(0,.2 nm’) 4
~ 60h ‘e 4
x L dashed line: overlap model 4
a 40- of amorphization |
20 i a solid line: proposed model ]
wo of defect growth |

0 2 -Q = an n | 1 i 1 I

i) 2 4 6 8 10

¢ ( 10!*/cm? )

FIG. 7.2-5 The measured maximum defect concentration as a function of dose (0 of Fig. 7.2-4) is

compared with that predicted from the phenomenological model of the accelerated damage growth

in a predamaged crystal (solid line, see text). The dashed lines are the fraction of the amorphous

zones calculated from Gibbson’s overlap model with various (m,A;) parameters.

2 + dashed line: Crp. 230 kev *8si_ J
5 1.0 solid line: Cp . . -
oS dotted line: e+ .-~" * . |
3 ae
‘S L z
g ,
fy ‘ata -q
RF 0.5- 4
‘v L .
a0 ‘
cs i J
is] e
s [ e 7

0 0.1 0.2 0.3 0.4

Depth ( um )

FIG. 7.2-6 The depth profile of the Frenkel pair concentration from TRIM88 simulation (dashed

line), the defect concentration from the channeling measurements of the sample (c) (solid line), and

the perpendicular strain from the dynamical x-ray diffraction simulation of the rocking curve (dotted

line). The vertical scale is in an arbitrarily normalized unit.

113

here is caused by the existence of predamage, not the dense casade produced by an incident ion.
In regime ITI, a continuous amorphous layer is known to form at the maximum damage location
when the yield for channeled beam incidence becomes the same as that for random beam incidence.®
The dose for the onset of the formation of a continouous amorphous layer is ~ 5 x 10!4/cm? (see
Fig. 7.2-4), which from TRIM88 simulation corresponds to a maximum energy density deposited in
nuclear collision of ~ 1074 eV/cm*. This value agrees with the prediction of critical energy density
criterion for amorphization (~ 1024 eV/cm).° Further implantation only causes the widening of the
amorphous layer. To gain some insight into the mechanism of amorphization, we apply Gibbons’
overlap model® to fit the measured maximum defect concentration (dashed line in Fig. 7.2-5).
Assuming that each incident ion creats a cylindrical zone of damage of cross section A; and the
formation of an amorphous region is caused by the m-tuple overlap of damage zones, the fraction of

the amorphous regions, f,4, is given by®

fazl— oD Ae ena, (7.2 ~ 10)

Figure 7.2-5 shows several f4 curves with various parameters of (m, Aj). It is evident that the fitting
improves as the m parameter increases (A; increases correspondingly). This strongly suggests that
the direct impact amorphization”! by implanted ions does not occur in self-implanted Si at room
temperature and that amorphous zones are formed because of the overlap of defected regions.??
Futhermore, it is necessary to invoke very many overlaps (m > 20) to fit the rapid growth of
damage. We thus hypothesize that amorphization may occur spontaneously as a result of collaps of

heavily defected crystal.

D. Depth profile of “damage”

The depth profiles of the Frenkel pair concentration (cr.p., dashed line) calculated from TRIM88
simulations is compared with the defect concentration (ep, solid line) extracted from channeling
measurements of the sample (c) in Fig. 72-3, To reveal the difference in the shapes of the damage-
depth profiles, each profile is plotted by normalizing its peak value to unity. The measured profile
is steeper than the calculated one, indicating that defects associated with smaller Frenkel pair
concentration anneal out at room temperature, consistent with the previous discussion. In particular,
the measured defect concentration near the surface is much depleted compared to the simulated one,
indicating that the surface is a very efficient sink for defects. Figure 7.2-6 also shows the depth profile

of the perpendicular strain extracted from the simulation of x-ray rocking curves (e+, dotted line),

114

which closely follows that of the measured defect concentration cp. The strain and the defect

concentration are seen to be proportional to each other.
Figure 7.2-7 shows the measured maximum strain in the implanted Si layer versus the measured

maximum defect concentration. The two quantities are linearly related,
e+ mar = Bes, (7.2 ~ 11)

with a slope of B ~ 0.013. This slope is a constant, valid over the entire range of strain and damage

(regimes I, II, III).

E. Relationship between strain and defect concentration

We apply continuum elasticity theory to estimate the order of the magnitude of the ratio
between the strain and defect concentration. For simplicity of book-keeping, we treat the material
as an elastically isotropic medium, although crystalline Si is anisotropic. The lattice expansion can
be extracted from the strain according to linear elasticity,

fa lava, Qu dll
a 1+yp l+v

where v is the Poisson ratio. Using the measured value of ¢ll=0 and v = 0.28 of bulk Si, we obtain

the relationship between the lattice expansion and the defect concentration,
(==) = 0.56 e+ = 0.007ep. (7.2 — 12)

The last relation is derived from Fig. 7.2-7.

Eshelby? showed that the lattice dilatation induced by point dilatation centers of concentration

c, and “strength” k, equals
Aa

=k-c. . (7.2 — 13)

The “strength” of a dilatation center of radius, R, is given by??

p= TY) | p26n, _ (72-14)
l+yv

where 6 is the the lattice displacement at R, and n is the atomic density.

In fec metals such as Cu, k, ~ —0.2 for single vacancies and k; ~ 1.5 for interstitials.1524 For
a crystal containing an equal number of single vacancies and interstitials, the “strength” becomes
k ~ 1.5—0.2 = 1.3. This number is more than 100 times larger than the coefficient in Eq. (7.2-12).

The smallness of the “strength” in Si could be due to the open structure of the diamond

115

F 230 keV *8si into Si(100) at 23°C 1
1.04 4
. amorphous |
x oe |
4 a “
wQ L 4
¢) [ 1 l 1 l n l 1 | - 1 |
0 20 40 60 80 100
Cp™e* ( % )

FIG. 7.2-7 The relationship between the maximum values of the perpendicular strain from x-ray
diffraction measurements and the defect concentration from 2 MeV *He channeling measurements.

The solid line is the least-squares fit of the data (0) to a linear function.

1.0
mer 230 keV 78sj /
at 23°C ]
R 0.6F (c) 4
ae 5 30 min in vacuum 4
4 £0.4- 4
“ i =
0.2+ (b) “
gu4é=—= ee
0) 200 400 600
T ( °C )/

FIG. 7.2-8 The isochronal annealing characteristics of the perpendicular strain in the implanted
layers as a function of the annealing temperature. All annealings were performed in a vacuum of
~ 7x 10-7 Torr for a duration of 30 min. The data are from the four samples for which the x-ray

rocking curves are shown in Fig. 7.2-1 and the channeling spectra in Fig. 7.2-3.

116

lattice compared to the closed-packed fcc lattice.?> However, it is unlikely that this can explain the
difference of more than two orders of magnitude.

Another explanation is that the defect in the room-temperature-implanted Si is in the form of
aggregates of vacancies and interstitials. To simplify the analysis, we will assume that the defect
is in the form of vacancy and interstial clusters containing an average of p lattice sites. We treat
each cluster as an individual dilatation center of “strength” kp. The lattice dilatation induced by

the clusters of concentration cp is

The “strength” k, for the cluster can be obtained from Eq. (7.2-14),
4a(1—v) 1» 2/3(4t(1—v) 10 2/3
be = Ty Rpdon = (Ridin) = pl? ks, (7.2 — 16)

because Ry = p'/3R,, and the displacement 6 of a dilatation center is not sensitive to the size

6, & 67°). For a given defect concentration of ep, the cluster concentration is
P &
Cp = pv) ‘Cp. (7.2 ~ 17)

Substituting Eqs. (7.2-16) and (7.2-17) into Eq. (7.2-15) gives

So = p78. ky ep. (7.2 — 18)

Using the typical value of ky ~ 1 for a single vacancy-interstitial pair in a fcc crystal,?” we have

ae w pul3.¢

D. (7.2 — 19)

The coefficient in Eq. (7.2-19) becomes ~ 0.007 if the cluster contains p ~ 3 x 10° lattice sites.
The diameter, D, of such a cluster is about 50 nm. It is 10 times larger than that in a similarly
implanted Si observed by transmission electron microscopy.”®?° Using the value from transmission

electron microscopy, D ~ 5 nm, Eq. (7.2-18) becomes

ae ~ 0.07k, «ep. (7.2 — 20)

The coefficient in Eq. (7.2-20) becomes 0.007 if the “strength” of a single vacancy-interstitial pair
k; in Si equals 0.1. This is about 10 times less than that in fcc metals.

This oversimplified model demonstrates that the observed smallness of the coefficient can be
attributed to the combined effect of defect clustering (p > 1) and the small single vacancy-interstitial

“strength” (ki < 1). This is consistent with the fact that single vacancies and interstitials are

117

mobile!® at room temperature, and Si has an open structure. The above analysis can be readily
generalized to a more realistic defect structure than that discussed here, where a distribution of

different sizes of clusters exists.

F. Thermal annealing and damage recovery

To further reveal the nature of the defect in self-implanted Si, we conducted experiments to
investigate the effect of thermal annealing on the strain in implanted Si. It is known that single inter-
stitials and vacancies anneal out at temperatures much less than room temperature,!® di-interstitials
anneal at ~ 150°C,%° di-vacancies anneal at ~ 100 — 250°C,*!, small damage clusters (< 10 nm) an-
neal between 100°C and 400°C,?? and a continuous amorphous layer starts to regrow by solid phase
epitaxy with an appreciable rate at ~ 550°C.° Isochronal annealings of the samples of Figs. 7.2-1
and 7.2-3 with different damage levels were performed in high vacuum (5 x 10~” Torr) for 30 min at
temperatures from 200°C to 700°C. The parallel strain remains zero after annealing. The annealing
behavior of the perpendicular strain is shown in Fig. 7.2-8. The temperature for significant recov-
ery increases as the damage increases, indicating that different damage levels have different defect
structures. Regime (I) with low damage (samples (a) and (b)) consists of relatively simple defects
such as di-interstitials and di-vacancies. The intermediate damage regime (II) (samples (c) and (d))
contains more complex defects such as clusters or small disordered zones. As the damage increases
further, a continuous amorphous layer forms because of the overlap of the damaged regions. This
picture differs from Vook and Stein’s** where there are only two distinct annealing stages at ~ 250°C
and ~ 550°C, which they associate with the annealing of di-vacancies and epitaxial regrowth of an
amorphous layer, respectively. They accordingly proposed that amorphization is controlled by di-
vacancy annealing.**:34 Our results are based on data taken with fine increments of the dose near

the amorphization threshold (Fig. 7.2-2 and 7.2-4) and reveal a complex defect hierarchy.

G. Conclusion

In the light of the above experimental results and discussion, we propose the following model
for the damage build-up and amorphization of self-implanted Si at room temperature. Initially, the
majority of defects produced by incident ions in a virgin crystal are Frenkel pairs, which are mobile
at room temperature. The migration of these interstitials and vacancies results in recombination and
clustering to form stable defects such as di-interstitials and di-vacancies. Only a small fraction of
initially created defects remains at room temperature. They generate perpendicular strain. As the

damage rises to the critical value (~ 10 — 20%), a large amount of energy is stored in the damaged

118

layer. Additionally created defects cause the collapse of the damaged region into large disordered
zones. This mechanism produces more damage than that generated directly in a virgin crystal by
nuclear displacements, giving rise to an enhanced production of damage and strain. The larger the
damage, the more complex the defect structure becomes. Amorphization occurs spontaneously as a

cooperative process because of the overlap of heavily defected crystalline regions.

7.3 Damage by '°F, *°Ar, and }*!Xe implantation at room temperature

In this section, we extend the investigation of defects production and their stability in room

temperature implanted Si(100) to ions of vastly different nuclear charges and masses.35

A. Experimental procedures

Ions of either 230 keV !9F, 250 keV *°Ar, or 570 keV 15!Xe were implanted into Si(100) at
room temperature in high vacuum (~ 10-* Torr). The beam flux was limited to < 0.5 A/cm?
to minimize the sample heating. The ranges of implanted ions in Si(100), calculated from TRIM
simulation,!® vary from ~ 0.2 ym for 570 keV '3!Xe to ~ 0.4 ym for 230 keV 19°F. The doses were
chosen to produce a spectrum that covers the entire range of damage, from a lightly damaged crystal
to one with a buried continuous amorphous layer.

2 MeV *He channeling measurements were used to extract the defect concentration in the
implanted layer, according to the procedures described in the previous section.*® The strain induced
_ by the damage was obtained by fitting the measured x-ray rocking curves to the calculated one from

the dynamical diffraction model.!4

B. Results of 230 keV }°F implantation

Channeling spectra of 230 keV }°F implanted Si(100) samples resemble those of 230 keV ?8Si
implanted ones shown in Fig. 7.2-3. Implantation to a dose of 5x 10!4 }9F /cm? produces a marginally
detectable dechanneling yield above the background yield of a virgin sample (labeled by virgin in
Fig. 7.2-3), while implantation to a dose of 2 x 10!° !9F/cm? produces a region in the channeling
spectrum where the channeling yield equals the random one (similar to the dotted line spectrum
labeled by 7.7 x 10'4/cm? in Fig. 7.2-3). We will refer to the corresponding region in the sample as
a continous amorphous layer for the rest of this chapter. The depth profile of defect concentration
cp(z) is also similar to that obtained before for °8Si (Fig. 7.2-6). The defect concentration near the

surface is again depleted compared to that predicted from TRIM88 simulation. The maximum

100

S60

ep™* (

FIG. 7.3-1 The maximum defect concentration extracted from channeling spectra similar to those
of Fig. 7.2-3 as a function of the °F dose. The solid line is to highlight the trend. The dashed

line is the maximum value in the depth profile of the Frenkel pair concentration predicted by the

119

;——¢ e

230 keV !°F into Si(100)

2 MeV *He channeling data

! | ! ! L |

20 30 40 50
¢ ( 10!4/cm? )

TRIM88 simulation of 230 keV '°F implantation into an amorphous Si target.

FIG. 7.3-2 The relationship between the maximum values of the perpendicular strain from x-Tay

diffraction measurements and the defect concentration from 2 MeV *He channeling measurements.

least-squares fit |
slope=0.012

| i ! 1 ! 1 ! !

20 40 60 80 100
cp™a* ( % )

The solid line is the least-squares fit to the data (¢) of a linear function.

120

value of the defect concentration profile is plotted as a function of the }9F dose in Fig. 7.3-1. It is clear
that the dependence is highly nonlinear and similar to that observed in the ?8Si implanted Si(100)
(see Fig. 7.2-4). The damage here can also be categorized into three regimes I, II, III, corresponding
to the lightly damaged, enhanced damage production, and formation of a continuous amorphous
layer, respectively.** In particular, we notice that the transition from I to II occurs at a defect
concentration of ~ 10%, the same damage level as that in the ?8Si implanted samples. This suggests
that the general character of damage production in Si(100) by room-temperature implantation is
insensitive to the implanted ions. The critical dose (~ 8 x 10!4/cm?) of the !°F implanted sample
is, however, more than twice larger than that of the 78Si implanted one (~ 3 x 10!4/cm?). This is
partly due to the fact that each !°F ion produces less Frenkel pairs than each 28Si ion does. The
dose required to produce the same amount of Frenkel pairs is hence greater for }9F than for ?8Si.

The strain profile extracted from x-ray rocking curves has the same shape as the defect profile.
This result is again similar to that of the ?°Si implanted samples (see Fig. 7.2-6). The maximum value
in the perpendicular strain profile rises nonlinearly with the 1°F dose.*® It has the same dependence
as that of the defect concentration of Fig. 7.3-1, with the same critical dose, ~ 8 x 10!4/cm?. The
transition from regime I to IH occurs at a strain value of ~ 0.2%. All these results are similar to
those obtained for the 2°Si implanted samples.

Furthermore, we discover that as for the ?8Si implanted samples, the strain in the damaged

layer of the ‘°F implanted samples is also proportional to the defect concentration over the whole

range of strain and damage (Fig. 7.3-2),
e+ = 0.012ep. (7.3 - 1)

The coefficient is also the same (compared with Fig. 7.2-7). Thus, we have again that e+ = Bep.
This relationship between the defect and its induced strain is thus insensitive to the implanted ions,

but is an intrinsic property of the Si target.

C. Generalization to any ions

X-ray rocking curve analyses of 250 keV *°Ar and 570 keV '3!Xe implanted samples were also
performed. The strain profiles all have the similar shape. In particular, the strain near the surface
is always depleted in comparison with that predicted by TRIM88.°° The maximum perpendicular
strain as a function of dose is plotted in Fig. 7.3-3 for all ions. Firstly, they all show similarly
nonlinear dose dependences. The nonlinearity is strongest for 19F and weakest for !2!Xe. This

tendency is in accord with our previous proposition that the nonlinearity arises from the

121

28g; re) 197

__I _ I I ty ty ft
6 0 5 10 0 10 “20 30

¢ ( 10'4/cm? )

FIG. 7.3-3 The maximum perpendicular strain obtained by fitting the dynamical x-ray diffraction

simulations to the experimental rocking curves as a function of the implantation dose for four different

ions. The solid line is to stress the trend.

to-2L ion implantation into Si(100) _~@
%) j 4
& : ’
Oo - 4
x bee =
oO ae?
~ F _--°7 slope=1 :
am . J
1074 , 1 I L r 1 1 nN a |

Sopp ( 1/10'*/cm? )

FIG. 7.3-4 The initial (regime I) slope of the maximum perpendicular strain vs dose as a function

of the Frenkel pair concentration per unit dose for various incident ions.

122

recombination of simple vacancy-interstitial defects at room temperature.*®* Light ions produce de-
fects that are sparesly distributed within a large cascade volume and consist mainly of isolated
interstitials and vacancies so that few stable complexes are formed and most defects recombine. On
the other hand, heavy ions produce defects that distribute densely in a small cascade volume so
that defect complexes and clusters that are stable at room temperature are generated. Secondly, the
transition from regime I to II occurs at about the same strain value of ~ 0.1 — 0.2%, independent of
the ion species, while the critical dose decreases from ~ 8 x 10!4/cm? for the light !9F to ~ 10!3/cm?
for the heavy !*!Xe.

Given the results on the linear relationship between the strain and the defect concentration
obtained in the °F and 78Si implanted samples (Fig. 7.3-2 and 7.2-7), we assume that this same
linear relationship also applies to other ions. The common critical perpendicular strain e+,, of Fig.
7.3-3 for various ions thus means that the critical defect concentration is also the same (~ 10 ~ 20%)
for all ions. Above this damage level, the defected crystal becomes unstable and the damage rises
rapidly (regime II) till a continuous amorphous layer forms (regime III).

In the lightly damaged regime I, the strain (and hence defect concentration) increases approxi-
mately linearly with dose,

et= S19, (7.3 — 2)
where S.4 is the slope of €*maz vs ¢. S_1 increases from ~ 0.02%/(10'4/cm?) for °F to ~
1.2%/(1014/cm?) for 13!Xe.

To put these values into perspective, we computed the maximum of the Frenkel pair concentra-
tion per unit dose of incident ions, S.,», by TRIM88.!8 S,,. , measures the amount of displacement
produced by each incident ion. For a given implantation dose ¢, the maximum Frenkel concentration
cr p. equals S,,, times the dose ¢. The heavier an incident ion is, the more damage it produces, and
hence the larger S,,, becomes. It increases from ~ 8%/(10!4/cm?) for !°F to ~ 74%/(10!4/cm?)
for 131Xe,36

Figure 7.3-4 shows the measured slope S41 asa function of the calculated Sepp, from TRIM88
for four ions (¢). The solid line represents a quadratic dependence of the slope on S,,,, and the
dashed line, a linear dependence. It is evident from the figure that the slope 5, 1 has an approximate

quadratic dependence on S,,. , ,

Sx $2, 5 (7.3 — 3)
We also plot the critical dose as a function of S,,, for the four ions, and discover that
ber « Sz, (7.3 — 4)

123

Combining these results, we find that the critical strain defined by
et = SL ¢er = 0.17%, (7.3 — 5)

is a constant independent of incident ions. This result agrees with that obtained previously from
Fig. 7.3-3.

Paine and Speriosu®’ observed that the strain in a lightly damaged GaAs also increases linearly
with ion doses. However, the slope of strain versus dose is a linear function of the Frenkel pairs per
unit dose of incident ions, rather than a quadratic one.

We already know that the defect concentration cp in an implanted Si is proportional to the
strain e+ ,6 and have therefore the following relation according to Fig. 7.3-4,

Sep « S? (7.3 — 6)

CF.P,?

where S,,, is the slope of the defect concentration from the channeling analysis of the implanted
samples in regime I versus dose. It states that the stable defect concentration rises quadratically as
the Frenkel pair concentration per unit ion dose increases. This fact means that the stable defects
produced by room-temperature implantation in Si(100) cannot be predicted by the linear cascade
model. It supports our previous hypothesis that the simple vacancy-interstitial defects are not stable
at room temperature. They recombine or form defect complexes and clusters, and it is these that
are stable at room temperature. The fraction of the various microstructural defects depends on the
density of the Frenkel pairs initially produced by an incident ion in a cascade volume. A low density
in a large volume (for light ion) results in a large fraction of simple defects. While a high density
in a small volume (for heavy ion) results in a large fraction of defect complexes and clusters. This

could explain qualitatively the nonlinear dependence of S,,, on Se, p,.*°

124

7.4 Damage by ion implantation at liquid nitrogen temperature

We report here the results on the damage produced by ion implantation at liquid nitrogen
temperature and analyzed at room temperature,®® and compare them with those obtained by room-

temperature implantation .3>

A. Experimental procedures

Ions of either 230 keV *8Si, 250 keV *°Ar, or 570 keV 19!Xe were implanted into Si(100) at
~ 100K in high vacuum (~ 10~’ Torr). The beam flux was kept low to minimize the sample heating
during implantation. The samples gradually warmed up to room temperature in ~ 1 hr after the
completion of implantation. X-ray rocking curves were taken at room temperature in ambient air
immediately afterwards. The samples were then stored in ambient air at room temperature. DCD
measurements of selected samples were also made at room temperature after about one day, one

week, one month, and one year to monitor the time evolution of the rocking curves.

B. Results of 250 keV *° Ar implantation at 100K

The strain in the damaged layer induced by the implantation of 250 keV *°Ar ions into Si(100)
at 100K to doses from 10'#/cm? to 10'4/cm? was extracted by fitting the experimental rocking
curves to the calculated ones based on the dynamical diffraction model.!* The parallel strain is
always zero. The depth profile of the perpendicualr strain has a shape similar to that of the room-
temperature implanted samples (see Fig. 7.2-6).2° The strain decreases to about 90% of the initial
value (measured 1 hr after implantation) in about 1 month and is almost constant thereafter. The
amount of decrease observed here is smaller than that seen in similarly implanted GaAs(100) samples
( where a decrease of ~ 50% is observed).°° We discuss only the final strain measured after one month
or more in the rest of this section.

The maximum perpendicular strain as a function of the *°Ar dose is shown in Fig. 7.4-1 (7).
It is evident that for a given dose, the final strain produced by implantation at 100K is greater
than that generated at 300K. This means that implantation at low temperature followed by room
temperature annealing is not eqiuvalent to that at room temperature. This dissimilarity is a well-
known fact to the Si implantation community and applies also to other semiconductors. This fact
shows that the interaction of defects and the dynamical annealing during the implantation are a
crucial mechanisim in determining the final defect microstructure and damage level. On the other
hand, the dose dependence of the strain is similar to that of the room-temperature-implanted samples

(A in Fig. 7.4-1). One can again identify three distinct regimes of damage production.°® The strain

125

1.0-

e max (2%)

250 keV *°Ar into Si(100)

! 1 ! ; l \ I R |
0 1 2 3 4

¢ ( 10!4/cm? )

points (filled inverse triangle) are obtained by multiplying the dose of the data o by a factor of 3.5.

Fig. 7.4-2 The cross-sectional TEM (from Ref. 40) of the sample implanted by 250 keV 5 x
10'8 *°Ar/cm? into $i(100) at 100K (€+maz ~ 0.4%, see Fig. 7.4-1). A heavily damaged layer is

located from a depth of 120 nm to 240 nm.

126

rises approximately linearly with dose in regime I, then rapidly increases in regime II, and finally
saturates in regime III (see discussion in Ch. 7.2). Furthermore, by multiplying the dose of each
data point corresponding to the 100K-implanted sample (V7 in Fig. 7.4-1) by 3.5, one obtains a
modified set of data points (filled inverse triangle in Fig. 7.4-1) that are the same as those of the
300K-implanted samples (A in Fig. 7.4-1). The damage produced by 250 keV ?°Ar implantation at
100K to a dose ¢ followed by room temperature annealing is thus the same as that by implantation
at 300K to a dose 3.5¢.°8 This result can be explained by assuming that the relationship between e+
and cp is invariant with temperature (i.e., B is constant), but that the production of stable defects
rises with falling temperature.

Fig. 7.4-2 shows a cross-sectional TEM micrograph*° of the sample implanted at 100K to a
dose of 5 x 1013 4°Ar/cm? (e+ maz ~ 0.4%, see Fig. 7.4-1). There is a heavily damaged layer near the
depth at which the maximum Frenkel pair concentration predicted by TRIM simulation!® occurs.
The damage consists of clusters. The average size of each cluster is about 5 nm.*! This is the
same size as that observed in room-temperature implanted Si samples (see Ch. 7.2).2%?9 Since the
strain induced by the defects is determined by their microstrcuture, and their size in particular (see
Ch 7.2), the same cluster size in the liquid-nitrogen- and room-temperature implanted Si explains
why the relationship between the strain and defect concentration obtained previously for the room-

temperature implanted samples*® also applies to low-temperature implanted samples, i.e.,

et= Bep x 0.0lep. (7.4 ~ 1)

C. Generalization to any low temperature of implantation

The strain in the samples implanted at 100K by 230 keV 28Si or 570 keV !*!Xe ions was also
obtained from the rocking curve analysis.** The parallel strain is again zero. The perpendicular
strain also decreases to about 90% of the initial value in about one month and saturates afterwards.
The strain rises nonlinearly with dose for all ions. The transition occurs at a strain value of about
~ 0.1 — 0.2%, the same as that of the room-temperature implanted samples. This means that the
transition is not sensitive to the implantation ions and temperature, but is determined by the intrinsic
damage level in the implanted layer. For a given dose, the final damage produced by implantation
at 100K is always larger than that at 300K, for all ions. Furthermore, as for the 250 keV *Ar
implanted samples discussed above, the strain or defect concentration produced by implantation at

300K can be obtained from that at 100K by a scaling of dose for all ions,38

230 keV **Si: — e+ (100K, 6) = e+(300K,5.5¢) or ecp(100K,¢) =cp(300K,5.5¢);

127

250 keV *°Ar: e+(100K, ¢) = e+(300K,3.5¢) or cp(100K, ¢) = cp(300K, 3.5¢);
570 keV '1Xe: «+ (100K, ¢) = e+ (300K, 2.76) or ep(100K, 4) = cp(300K,2.7¢).

The scaling factor decreases from 5.5 for 78Si ions to 2.7 for '5!Xe ions, meaning that the difference
of the measured damage between implantation at room and low temperature is most pronounced
for light ?8Si ions and least pronouced for heavy !°!Xe ions. This again supports the notion that
the measured damage at room temperature is related to the stability of defects produced during
ion implantation. The light ?8Si ions produce some simple vacancy-interstitial defects that are
unstable and mostly recombine. The heavy !3!Xe ions produce more defect complexes and clusters
that are stable.

Given the above results, we propose the following generalization of (i) the linear relationship

between e+ and cp,

e+ (T, ¢) = Bep(T, ¢), (7.4 —2)

where B is a constant for a given target (~ 0.01 for Si); and (ii) the scaling behavior,
et (T, $) = e+(To, $) or CD (T, ¢) =cp (To, A), (7.4 _- 3)

where A is a dimensionless parameter that depends on two temperatures T and T,. This enables one
to compute the damage produced by implantation to a dose of ¢ at temperature JT and measured
at an elevated temperature T, from the results obtained by implantation at T,, if one knows the
scaling parameter A. We further assume that scales as the ratio of two temperatures based on the

dimensional arguments,

NT ,T.) « Cone (7.4 ~ 4)

where ( is the “critical exponent,” which depends on the incident ions. From the above results
for ?8Si, #°Ar, and 15!Xe ions, we see that # has a value close to 1. By investigating implantation
damage at various low temperatures (< 300K), one could test the proposed scaling behavior and

extract the @ parameter.

128

References

10.
11,
12.

13.
14.
15.
16.

17.
18.
19.
20.
21.
22.
23.

24,

25.

. F.L. Vook, ed. Radiation Effects in Semiconductors, Plenum Press, New York, 1968.

. G.D. Watkins, J. Phys. Soc. Japan 18, Suppl. II, 22 (1963).

L.J. Cheng, J.C. Corelli, J. W. Corbett, and G.D. Watkins, Phys. Rev. 152, 761 (1966).

L. Chadderton and F. Eisen, eds. Proc. Ist Int. Conf. on Ion implantation, Gordon and
Breach, New York, 1971.

J.F. Gibbons, Proc. IEEE, Vol. 60, 1062 (1972).

J.W. Mayer, L. Eriksson, $.T. Picraux, and J.A. Davis, Can. J. Phys. 46, 663 (1968).

. M.L. Swanson, J.R. Parsons, and C.W. Hoelke, Radiation Effects in Semiconductors, eds. J.W.

Corbett and G.D. Watkins (Gordon and Breach, New York, 1971), p. 359.

. 8. Kurtin, G.A. Shifrin, and T.C. McGill, Appl. Phys. Lett. 14, 223 (1969).
. V.S. Speriosu, B.M. Paine, M-A. Nicolet, and H.L. Glass, Appl. Phys. Lett. 40, 604 (1982).

E.Glaser, G. Gotz, N. Sobolev, and W. Wesch, Phys. Stat. Sol. A69, 603 (1982).

J.S. Williams, R.G. Elliman, W.L. Brown, and T.E. Seidel, Phys. Rev. Lett. 55, 1482 (1985).
A.E. White, K.T. Short, R.C. Dynes, J.P. Garno, and J.M. Gibson, Appl. Phys. Lett. 50, 95
(1987).

P.A. Packan and J.D. Plummer, Appl. Phys. Lett. 56, 1787 (1990).

C.R.Wie, T.A. Tombrello, and T. Vreeland, Jr., J. Appl. Phys. 59, 3743 (1986).

L. Tewordt, Phys. Rev. 109, 61 (1958).

C. Kittel, Introduction to Solid State Physics, 5'® ed. (John Wiley & Sons, New York, 1976)
p. 126.

E. Bogh, Can. J. Phys. 46, 653 (1968).

J.P. Biersack and L.G. Haggmark, Nucl. Instr. and Meth. 174, 257 (1980).

J.W. Corbett, J.P. Karins, T.Y. Tan, Nucl. Instr. and Meth. 182/183, 457 (1981).

D.A. Thompson and R.S. Walker, Nucl. Instr. and Meth. 132, 281 (1976).

F.F, Morehead, Jr. and B.L. Crowder, Rad. Eff. 6, 27 (1970).

J.R. Dennis and E.B. Hale, J. Appl. Phys. 49, 1119 (1978).

J.D. Eshelby, Solid State Physics, F. Seitz and D. Turnbull, eds. (Academic Press, New York,
1956) p. 79.

H.G. Haubold and D. Martinsen, J. Nucl. Mater. 69/70, 644 (1978).

F.L. Vook, Phys. Rev. 125, 855 (1962).

26.

27.
28.
29.
30.
31.
32.

33.
34.
35.
36.
37.
38.
39.

40.
41.

129

M.W. Thompson, Defects and Radiation Damage in Metals (University Press, Cambridge,
1969), p. 16.

R.O. Simmons and R.W. Balluffi, J. Appl. Phys. 30, 1249 (1959).

D.J. Mazey, R.S. Nelson, and R.S. Barnes, Phil. Mag. 17, 1145 (1968).

M.O. Ruault, J. Chaumont, and H. Bernas, Nucl. Instr. and Meth. 209/210, 351 (1983).

W. Jung and G.S. Newell, Phys. Rev. 132, 648 (1963).

H.J. Stein, F.L. Vook, and J.A. Borders, Appl. Phys. Lett. 14, 328 (1969).

L.M. Howe, M.H. Rainville, H.K. Haugen, and D.A. Thompson, Nucl. Instr. and Meth. 170,
419 (1980).

F.L. Vook and H.J. Stein, Rad. Eff. 2, 23 (1969).

O.W. Holland, $.J. Pennycook, and G.L. Albert, Appl. Phys. Lett. 55, 2503 (1989).

G. Bai and M-A. Nicolet, submitted to J. App]. Phys.

G. Bai and M-A. Nicolet, J. Appl. Phys., July 1, 1991.

B.M. Paine and V.S. Speriosu, J. Appl. Phys. 62, 1704 (1987).

G. Bai and M-A. Nicolet, unpublished.

G. Bai, D.N. Jamieson, M-A. Nicolet, T. Vreeland, Jr., Mat. Res. Soc. Symp. Proc. 93, 67
(1987).

C.J. Tsai, A. Dommann, M-A. Nicolet, and T. Vreeland, Jr., J. Appl. Phys. 69, 2076 (1991).

C.J. Tsai, private communication.

130

Chapter 8 Ion Implantation in Heterostructures

8.1 Introduction

Epitaxial heterostructures are the building blocks of novel electronic and photonic devices. Some
effects of ion beams on these structures have been actively explored in the past decade from the view-
points of both fundamental materials research and potential applications.!~?? Most earlier studies
focused on III-V compound semiconductor superlattices.'-> Many interesting phenomena pertaining
to interaction of ion beams with superlattices were observed, such as compositional disordering and
mixing of AlAs/GaAs superlattices by ion implantation and subsequent annealing,! superposition
of strain induced by damage and built-in (intrinsic) strain in AlGaAs/GaAs? and GaAsP/GaP?
superlattices, and selective amorphization of AlAs/GaAs heterostructures and superlattices. Ion
implantation can also induce the generation of misfit dislocations and strain relaxation in strained-
layer superlattices.© Radiation damage in some silicides was also explored in the past decade.’—!7
Epitaxial NiSig and CoSiz films on Si can be amorphized by ion implantation, and solid phase epi-
taxial regrowth proceeds in a layer-by-layer manner.°-!? The resistivity of these metallic silicides
increases upon implantation.1*4 Compared to metallic silicides, semiconducting silicides such as
ReSiz are more susceptible to radiation damage,!'® and the resistvity of semiconducting ReSi2'®
and CrSi2!” decreases upon ion implantation, which is opposite to what metallic silicides do. Re-
cently, there has been a surge of interest in GeSi/Si heterostructures, because of the successful
demonstration of high-speed heterojunction bipolar transistors. Some properties of GeSi/Si het-
erostructures under ion beams, such as amorphization of GeSi by implantation and solid phase
epitaxial regrowth,'®° selective damage production and amorphization in GeSi/Si heterostructures
and superlattices,”'?° enhanced strain relaxation of implanted GeSi/Si,?! and strain modification in
ion-assisted grown GeSi layers,?” have been studied.

In the following sections, we describe some of our results on the response of epitaxial CoSi2,?3
ReSiz,'>!6 and GeSi4 films to ion implantation. Properties related to damage production and
annealing, stability of strained layers under implantation and subsequent annealing, and strain

induced by defects, are the focus of our investigation.

131

8.2 Damage production and annealing in 7°Si implanted CoSi, films

Transistion-metal silicides have applications as contacting layers in Si- MOSFET integrated cir-
cuits. Compared to Si, the metallic silicides are highly resistive to radiation damage.’—!? Most
previous studies focused on the amorphization of the silicides by implantation and recrystallization
by subsequent thermal annealing.8~!2 The amorphized CoSig layer on a crystalline seed recrystal-
lize in a layer-by-layer manner by solid phase epitaxy.°-!” Hensel et al.!3 studied the effect of ion
implantation on carrier transport of CoSiy films and found that the resistivity increases with dose.
In this section, we report some studies on defects production in room-temperature 78Si implanted
CoSiz films on Si(111), and concentrate on the strain and its relationship with the defects in the

implanted films.?? Damage recovery upon vacuum annealing is also investigated.

A. Experimental Procedures

An epitaxial CoSig film 50 nm thick was grown on Si(111) at ~ 600°C by MBE at UCLA
(Ch. 2.2). The samples were implanted at room temperature in vacuum (~ 10-7 Torr) by 150
keV 8Si ions to doses from 5 x 10'3/cm? to 3 x 10'°/cm?.. The maximum damage locates inside
the Si substrate at a depth of ~ 150 nm beneath the interface, according to TRIM88 simulation.?®
The damage in the implanted layers was characterized by 2 MeV ‘He ion channeling, Fe Ka, x-ray
(wavelength A = 0.1936 nm) double crystal diffraction, and electrical resistivity measurements.

The implanted samples were annealed afterwards in vacuum (~ 5 x 10" Torr) at 250-800°C

for 60 min, and the damage recovery was monitored.

B. Results and Discussion
(i) damage production

2 MeV *He [111] axial channeling with a glancing exit angle (82°) was used to measure the
damage in the implanted layers. The spectra of the implanted samples show that the channeling
yields of both the film and the substrate rise with the, dose (Fig. 8.2-1). For a given dose, the
damage level in the CoSi, film is much smaller than that in the Si substrate, meaning that CoSi,
is more radiation resistant than Si. The high radiation resistance of CoSiz agrees with the results
of room temperature implanted Pd2Si, NiSi2,”° and is consistent with the metallic nature of these
silicides.

The damage build-up in the Si substrate as a function of the dose is very similar to that observed
in the self-implanted Si samples (see Ch. 7.2). The damage rises slowly as the dose increases from

5 x 10'8/cm? to 2 x 10'4/cm?. It then rapidly rises to the level of an amorphous Si at 5 x 10!4/em?.

132

150 keV *8Si into CoSie(40 nm) /Si(111)

T T T | T T T t T T T

2 MeV *He

nie Co

ae! mn |

Counts

[111] aligned , ” ne
1.0 1.5
Energy ( MeV )

FIG. 8.2-1 2 MeV *He backscattering spectra with a beam incident along a random (e) and a
[111] axial channel orientation of the as-grown CoSi2(50 nm)/Si(111) (solid line); and the samples
implanted at room temperature by 150 keV *8Si to doses of 2x (V7), 5x (0), 30 x 10!4/cem? (A).

The detected *He particles exit at an angle of 82° from the line of the incident beam.

150 keV *8si into CoSig(40 nm)/Si(111)

L | :
TRIM88
40k! -
~~ {
ef :
a 400°C/60 min
oegk ! 4
600°C /60 min
800°C/60 min
Or | ; \ ! |

0 10 20 30
@ ( 10!4/cm? )

FIG. 8.2-2 The defect concentration in the CoSiy films extracted from the channeling yields of F ig.
8.2-1 vs the 8Si dose, for the as-implanted samples (e), and those annealed for 60 min at 400°C (0),

600°C (A), 800°C (square). The Frenkel pair concentration as a function of dose, predicted from

TRIM88, is also shown (dashed line).

133

As the dose increases further, the maximum damage level saturates, and the damage region widens.

On the other hand, the damage build-up in the CoSi, film has very different dose dependence.
The channeling yield in the film implanted to < 2 x 10!4/cm? is the same as that of the as-grown
one (~ 5%, Fig. 8.2-1). It increases to ~ 9% at 5 x 10!4/cm?, and rises to ~ 56% at 3 x 10!5/cm?.
Maex et al.!? also found that the channeling yield of CoSiy films implanted at room temperature
by 200 keV 78Si to 2 x 10'5/cm? is ~ 50%, in good agreement with our results. They also observed
that the damage in CoSiz is sensitive to the implantation temperature, as is in Si.2° Hewett et al.°
demonstrated that CoSiz can be amorphized by 40-200 keV 78Si implantation to ~ 2 x 10!5/cm? at
liquid nitrogen temperature.°~!?

We apply the procedure outlined in Ch. 7.2 to estimate the defect concentration cp in the CoSig
film from the channeling yield. To simplify the analysis, we assume that the fraction of displaced Si

atoms in the film is the same as that of displaced Co atoms. The dechanneling factor yp (defined

in terms of channeling yields, see Ch. 7.2), then equals?”
Yo =¢p + Po(1— ep). (8.2 — 1)

TRIM88 simulation shows that the damage is roughly a constant through the entire film. The
channeling spectra of Fig. 8.2-1 suggest the same conclusion. We therefore assume that cp is

depth-independent, so that the dechanneling probability Pp at depth x, becomes
Pp = | nopcpdz = (nopx_)cp, (8.2 — 2)

where n is the atomic density of CoSi2, and op is the average dechanneling cross section of Si and
Co atoms. We compute yp from the minimum channeling yield of the Co signal at the energy
immediately beneath the surface peak (~ 1.65 MeV in Fig. 8.2-1). The corresponding depth is x, %
10 nm. Substituting the appropriate numbers into Eq. (8.2-2), and assuming that op < 107!8/cm?
(see Ch. 7.2), one obtains

Pp < 0.05ep.

Pp is small compared to cp, and we therefore neglect the second term in Eq. (8.2-1) and have
Cp Yp.- (8.2 — 3)

The extracted defect concentration in the CoSi, films is plotted as a function of the implantation .
dose in Fig. 8.2-2 (e). It is undetectable (< 1%) after the implantation to 2x 10'4/cm? and increases

monotonically with the dose.

134

In comparison, we also computed the Frenkel pair concentration in a 150 keV 78Si implanted
CoSi2(50 nm)/Si amorphous matrix by TRIM88 simulation (dashed line in Fig. 8.2-2). A displace-
ment threshold energy of 15 eV and a binding energy of 1 eV were assumed. It is evident that
the measured damage is only a small fraction of that directly produced by collision cascade, which
suggests that the majority of the defects produced anneal out at room temperature during and
after implantation. This observation leads us to hypothesize that the higher radiation resistance of
CoSi2 than that of Si at room temperature is due to its higher defect recombination rate, probably
resulting from the higher mobility of defects in CoSig.

A heterostructure differs from a bulk crystal in that there exists intrinsic strain in an as-grown
heterostructures. Implantation produces defects, which induce additional strain in the heteroepi-
taxial film. We use x-ray double crystal diffraction to monitor the strain change in the CoSig film
as a result of the implantation. Figure 8.2-3 shows a set of x-ray rocking curves diffracted from the
(111) symmetrical plane of the CoSi2/Si(111) samples implanted to a dose of (a) 0x, (b) 0.5x, (c)
1x, (d) 2x, (e) 5 x 10!4/cm?.

On the basis of previous studies on implantation in Si(100) (see Ch. 7), we expect that the
damage in the Si substrate induces positive strain, and that additional diffraction peaks will appear
on the lower angle side of the bulk Si peak at 6g = 18° in Fig 8.2-3. Implantation to < 2x 10!4/cm?
produces low damage in the Si substrate and induces small perpendicular strain of < 0.1% (see Ch.
7.2), The corresponding shift in the position for the (111) diffraction peak of the implanted layer is
< 0.02’, which lies within the main peak and hence is undetectable (see Fig. 8.2-3). Implantation
to > 5 x 10'4/cm? produces a heavily damaged or continuous amorphous layer (see Ch. 7.2); the
corresponding diffraction peak then becomes weak and remains buried in the background intensity
of ~ 0.003% (Fig. 8.2-1).

The peak intensity of x-ray diffracted from the CoSig (111) planes decreases with increasing
dose (Fig. 8.2-3). After implantation to a dose of 2 x 10'4/cm?, the intensity drops to ~ 1/4 of that
of the as-grown film, unlike the channeling yield, which is the same as that of the as-grown sample
(Fig. 8.2-1&2). This fact demonstrates that x-ray diffraction is much more sensitive to low defect
concentrations than channeling. The x-ray intensity from the CoSi, layer of the sample implanted
to a dose of > 10'°/cm? drops below the background (~ 0.003%) and hence becomes unmeasurable.
The defect concentration extracted from the channeling yield of that same sample is only about
27%. A comparison of the channeling spectra (Fig. 8.2-1) and x-ray rocking curves (Fig. 8.2-3) of

the implanted CoSig films clearly illustrates that channeling is insensitive to very low damage

(111)CoSi,

rH)

¢ (10!4/cm?)

(a) 0
(b) 0.5
(c) 1
(d) 2 4
e) 5

Intensity ( % )

FIG. 8.2-3. Fe Kg, x-ray rocking curves diffracted from the symmetrical (111) planes of (a) as-grown
CoSiz/Si(111); and the samples implanted to doses of (b) 0.5x, (c) 1x, (d) 2x, (e) 5 x 10!4/cm?.
Bragg peak from the bulk Si substrate is 0g = 18°.

150 keV *8si into CoSi,(50 nm)/Si(111)

0.15
r as-implanted
_~ 0.10 L
g i 250°C/60 min
& r 400°C/60 min
30.05
F >250°C/60 min | >600°C/60 min
0 1 2 3 4 5

¢ ( 10'*/cm® )

FIG. 8.2-4 The static atomic displacement induced by the defects in the CoSio films vs the ?8Si
dose, for the as-implanted sample (¢), and for those annealed in vacuum for 60 min at 250°C (v)

and 400°C (oc). The shaded area represents the error in estimating the displacement of a perfect

CoSig film.

136

(cp < 1%), while the x-ray diffraction from the highly damaged films (ep > 27%) becomes unmea-
surable. These two techniques are therefore complementary and together provide a good picture of
damage production by ion implantation in CoSig.

In the following, we will analyze quantitatively the x-ray rocking curves diffracted from the
CoSig films implanted to < 5 x 10'4/cm?. The angular position of the (111) diffraction peak does
not change with the dose (~ 0.33°, Fig. 8.2-3). This means that the perpendicular strain e+ in
the films is a roughly constant, ~ —1.76%, within the experimental sensitivity (~ 0.08%). In other
words, the upper limit in the difference of the magnitude of the strain between the implanted and
as-grown films, Ae*, is ~ 0.08%. The defect concentration cp in the corresponding films ranges
from < 1% to ~ 4% (¢ Fig. 8.2-2). The above results imply that in the implanted CoSig films, one
has

|Aet| < 0.02ep. (8.2 — 4)

We were unable to measure the parallel strain in the CoSig film, which is B-type (Ch. 2.2). Given
the knowledge that the interfacial misfit dislocations do not shear at room temperature (see Ch.
2.5-6), it is reasonable to assume that the parallel strain of the film remains unchanged upon the
implantation; i.e., Aell = 0. From the above results and the relationship between the strain and

lattice mismatch (see Ch. 7.2), we obtain the lattice dilatation of the CoSiz induced by the defects,

ew = (Fa )1det| <0.0lep, (8.2—5)
where a is the lattice constant of the as-grown CoSi, film, and v is Poisson’s ratio of CoSiz (=1/3,
see Ch. 2.6). The coefficient relating |Aa/a| to cp for CoSiz is < 0.01, consistent with that for Si
(~ 0.007, see Ch. 7.2).

The full-width at half-maximum of all diffraction peaks that are strong enough to be detected
(films implanted to < 5 x 10'4/cm?) is an invariant, independent of the damage induced by ion
implantation (Fig. 8.2-3). The peak broadening of the as-grown film is due to the finite film
thickness. This result implies that the x-ray diffraction from the implanted film is highly coherent,
and suggests that the damage in the implanted film consists mainly of randomly distributed point-
like defects. The x-ray diffraction can therefore be modeled by a static Debye-Waller factor, which

does not change the peak width but decreases only the intensity.?8 The ratio of the peak intensity

from the implanted film, Ip, to that from the as-grown one, Jy, equals,?8

Ip 167? , sin?6p
a exp(——3— Up) (8.2 — 6)

137

where up is the mean-square root of the atomic displacement caused by the point-like defects in the

implanted film. For Fe Ky, x-ray diffracted from CoSi(111) planes, Eq. (8.2-6) results in

up = 0.085, /In z (nm). (8.2 —7)

The displacement up in the implanted CoSi, film thus extracted from the measured intensities in
Fig. 8.2-3 rapidly rises to ~ 0.06 nm after implantation to 5 x 10!3/cm?, and then increases linearly
with dose (¢ in Fig. 8.2-4).

We also measured the resistivity in the implanted film to monitor the defects build-up. Four-
point-probe methods were used to measure the sheet resistance of the CoSiy films. The resistivity
of the selected samples was also measured by the van der Pauw method, and agrees with that from
four-point-probe measurements. Current-voltage measurements show that the vertical resistance
across the silicide-silicon interface is always much greater than the sheet resistance of the silicide
film for all samples, implying that the film is practically insulated from the substrate. The measured
resistivity of the CoSis films increases from ~ 16 ~Qcem for the as-grown sample to ~ 300 wQcem for
the sample implanted to 5 x 10'4/cm?. It flattens off up to 3 x 1015/cm?.

Resistivity, p, in metals can be decomposed into two terms according to Matthiessen’s rule,)3

P=PpLt pp, (8.2 — 8)

where pz is the lattice (Bloch-Gruneisen) contribution, and pp results from carrier scattering by
defects. Hensel et al.!3 measured the resistivity of 2 MeV He bombarded CoSiz films at 4-300K
and found good agreement with Eq. (8.2-8). These authors also established that the resistivity of
the samples implanted to different doses has a similar temperature dependence, meaning that the
lattice contribution py is the same regardless of the damage (so long as the film is not amorphized
to cause localization of carriers). The difference of p between the bombarded and the as-grown
films, Ap, hence equals App, the resistivity contribution from radiation-induced defects. Hensel et
al.'3 discovered that App rises approximately linearly with dose till ~ 100 pOQcm and then flattens
off with further dose increase. The resistivity jumps abruptly to ~ 1000 wQcm once the CoSig is
amorphized, and remains constant thereafter.

We used the van der Pauw method to measure the resistivity of the CoSiy film for selected
samples at 130-300K. The temperature dependence of the resistivity agrees with the prediction of
Eq. (8.2-8), and that reported by Hensel.'? We assume in the following that the resistivity difference

Ap between the implanted and the as-grown CoSiy films equals that induced by the implantation

138

App for all our samples. App increases with dose to ~ 280 wQcm at 5 x 10!4/cm? and flattens off
up to 3 x 10!5/cm? (¢ in Fig. 8.2-5).

The resistivity from carrier scattering by point-like defects is proportional to the density of
scatterers. The initial rise of App (¢ < 5 x 101*) suggests that the implantation produces point-like
defects, which build up with increasing dose.! This conclusion is very similar to that drawn from
x-ray diffraction results (see previous discussion). Figure 8.2-6 plots the resistivity caused by defects
as a function of the atomic displacement induced by defects in the CoSiz film implanted to a dose
of < 5 x 10'4/cm?. The plot shows a good correlation between these two indicators of the defect
concentration in the CoSiz films. From 5 x 10'*/cm? to 3 x 10!5/cm?, App is roughly a constant
(~ 280 ~Qcem, Fig. 8.2-5), while the defect concentration cp extracted from the channeling yields
increases from ~ 4% to ~ 54% (¢ in Fig. 8.2-2). This suggests that the flat App is probably
indicative of some agglomeration of defects.!3

(ii) damage annealing

2 MeV *He axial channeling was also used to monitor the change of the channeling yield as a
function of the annealing temperature. The annealing characteristics of the damaged Si substrate
is the same as that of implanted bulk Si.?” At low damage level (¢ < 2 x 10!4/cm?, Fig. 8.2-1), the
channeling yield decreases as the annealing temperature rises. The dominant process is probably
the recombination of point-like defects. At high damage levels near the amorphization threshold
(¢ = 5 x 10'4/cm?, Fig. 8.2-1), the channeling spectrum does not change after 60 min isochronal
annealing at 250°C. The channeling yield decreases after 400°C annealing, and becomes the same
as that of the as-grown sample after 600°C annealing. The channeling spectrum of the amorphized
samples (¢ > 101°/cm?, Fig. 8.2-1) does not change after 400°C annealing. In that case, appreciable
solid-phase epitaxial growth occurs after annealing at 600°C. However, the channeling yield is very
high (~ 50%) and remains that high after 800°C annealing. A high density of extended defects (e.g.,
dislocation loops, microtwins) is probably present.

The qualitative annealing features of the CoSig films are simpler than those of the Si substrate,
since the film is not amorphized after implantation to the highest dose of 3 x 101° /cm? (see Fig. 8.2-
1). The channeling yields of the implanted films decrease with increasing annealing temperature.
The defect concentration in the film after annealing at various temperatures, extracted from the
channeling yield according to Eq. (8.2-3), is shown in Fig. 8.2-2. The films implanted to a dose of
< 5 x 10'4/cm? completely recover, while those implanted to doses > 10!5/cm? do not. We believe

that the non-zero cp (~ 10%, Fig. 8.2-2) after annealing is more indicative of the presence of

139

150 keV *8Si into CoSi2(50 nm)/Si(111)
300

—*

e¢ as~implanted

~ 200 Vv 250°C

E o 400°C

& 60 min
cy A 600°C

~ a 800°C

3 100

—8
—4
30

0 10 20
¢ ( 10'*/cm? )

FIG. 8.2-5 The resistivity difference between the implanted and as-grown CoSiz films vs the dose,

for the as-implanted sample (e), and for those annealed in vacuum for 60 min at 250°C (v), 400°C

(0c), 600°C (A), 800°C (square).

150 keV *8Si into CoSig(50 nm)/Si(111)

300

as—implanted Bx
“-200- |
oO
or L A
3 100+ 4

OF FA a
0 0.05 0.10 0.15
Up ( nm )

FIG. 8.2-6 The resistivity difference vs the static displacement of the lightly damaged CoSiy films
(cp < 4% or 6 < 5 x 1014/cm?). The approximately linear relationship indicates a good correlation

between the concentration of the carrier scatterers and the structural defects in such films.

140

extended defects than of a measure of the point-like defect concentration. The imcomplete recovery
of the film coincides with that of the Si substrate.

Upon thermal annealing, both the angular position and width of the x-ray diffraction peak
from the lightly damaged CoSiz films (cp < 4%) remanin unchanged. This means that the strain of
the CoSiz films does not change, and indicates that the presence of a small defect concentration in
the film does not enhance the strain relaxation upon thermal annealing. The x-ray peak intensity
rises with increasing temperature. For the films implanted up to 2 x 10!4/cm?, the intensity after
250°C is the same as that of the as-grown film within the experimental uncertainty (~ 20%). The
uncertainty in the intensity measurements causes a corrsponding uncentainty in the estimation of

the atomic displacement extracted according to Eq. (8.2-7),

(Sup/up) _ 0.5
V(lp/Ip)? + (Iv /Iv)2 Inv /Ip) (8.2 — 9)

This relationship shows that up of the film with the intensity close to that of Iy has a large
percentage error. For a percentage error of the x-ray intensity of ~ 20%, the error of up is about
0.04 nm, which means that measured displacements of 0-0.04 nm all correspond to a perfect CoSis
film. The displacement extracted for the films implanted to < 2 x 10!4/cm? after all annealings is
< 0.04 nm (Fig. 8.2-4). This suggests that these films recover completely after all annealings. The
displacement of the film implanted to 5 x 10'*/cm? decreases with increasing annealing temperature,
and becomes indistinguishable from that of the as-grown film after 600°C annealing (Fig. 8.2-4).

The heavily damaged CoSiz films (¢ > 10'°/cm?) have very different annealing characterisitics.
Upon thermal annealing, the x-ray peak intensity rises above the background (~ 0.003%) and
becomes measurable. The peak position is about the same as that of the as-grown film, meaning
that the strain of the heavily damaged films does not change after annealing. However, the peak
is much broader and weaker than that of the as-grown film, even after 800°C. This result signifies
that the strain in the annealed films is very inhomogeneous, and indicates the presence of extended
defects. The above results correlate well with those obtained from channeling measurements.

The resistivity of the CoSiz films of all samples decreases drastically after 250°C (V7 in Fig. 8.2-
5), and becomes about the same as that of the as-grown sample after 600°C (A in Fig. 8.2-5). The
decrease of the resistivity of the low damaged films (¢ < 5 x 10'4/cm?) after annealing correlates well
with the recovery of structure defects probed by channeling and x-ray diffraction measurements. For
the highly damaged films (¢ > 10'5/cm?), the structural recovery after annealing is incomplete, with
a channeling yield of ~ 10% (greater than that of the as-implanted films to a dose < 5 x 10!4/cm?),

and x-ray diffraction suggests that there exist extended defects in these annealed films. Yet the

141

resistivity of such films is about the same as those of the structurally perfect ones, much less than
that of the as-implanted films of any dose. Apparently the extended defects that still exist in the
highly damaged films after annealing are ineffective scattering centers for the carrier transport.

In summary, we discovered that MeV ion channeling is well suited to characterize heavily
damaged CoSiz films (cp > 1%), while x-ray diffraction is for lighly damaged ones. Resistivity
is a valid indicator of the damage over the entire range. A linear relationship exists between the
structural and electrical defects in lightly damaged films. Such a relationship is absent in heavily
damaged ones, where the resistivity flattens but the defect concentration further increases with
dose. The lightly damaged films recover completely upon thermal annealing, whereas the heavily
damaged ones do not. The residual defects after annealing are of the type of extended defects such

as dislocation loops and microtwins, and are ineffective for carrier scattering.

8.3 Radiation damage in ReSi, by an MeV ‘He beam.

Backscattering and channeling spectrometry are extensively used to probe compositional and
structural properties of materials within submicron depths below the surface.?9° They have become
routine analytical tools to characterize electronic materials. The effect of an analysis beam (of MeV
“He or 'H ions) on the materials studied is therefore important both from a practical point of view
for the correct interpretation of experimental data, and from a fundamental point of view for the
understanding of MeV ion-solid interactions.

In the past two decades, there have been about a dozen studies explicitly concerned with the
radiation damage in solids by MeV *He and !H beams.3!~4! Alkali halides are the most extensively
studied class of materials*!~3*, where the ionization of target atoms by incident ions is the main
mechanism for damage production. In Si and Ge, the damage is mainly produced by elastic nuclear
collisions.3°-3” In GaP, both inelastic electronic ionization and elastic nuclear collisions contribute
to the damage.** Extensive radiation damage has also been observed in some oxides (BaTiO3*9,
NbO*°, Al2O3*!). Little work has been done on the radiation damage in transition-metal silicides
by MeV *He and !H ions, in spite of the widespread utilization of the backscattering and channeling
technique in the characterization of these thin films. J.C. Hensel et al.}3 studied the effect of 2 MeV
4He irradiation on the resistivity of CoSig and NiSig thin films. H. Ishiwara et al.” used MeV *He
ion beam to analyze the radiation damage produced by 100 keV ?°Ar ions in epitaxial Pd2Si and
NiSig thin films grown on Si substrates.

We present here some experimental results on the damage induced by MeV ‘He ion irradiation

142

in epitaxial ReSi2 thin films grown on Si(100) substrates. Both the minimum yield of [100] axial
channeling and the half angle were measured as a function of sample exposure to the *He analysis
beam. The damages induced at room temperature by beams incident along a random direction and
an axial {100] direction are compared. We will show that the damage is produced by elastic nuclear
collisions. The measured amount of damage produced by a random incident beam agrees with that
computed from TRIM”°. This agreement indicates that the total amount of damage produced by
elastic nuclear collisions is preserved at room temperature.

The epitaxial ReSi film of ~ 150 nm thick was grown on a hot Si(100) substrate (~ 650°C)
by “teactive deposition epitaxy” in ultrahigh vacuum (~ 10~° Torr) at Colorado State University.4?
Details of the growth procedure and the characterization of the epitaxial ReSi2/Si(100) structure
were described in Ch. 3.2. The fundamental parameters of channeling, the minimum yield, ymin,
and the critical angle, ~/2, of as-grown ReSi2/Si(100) sample were discussed in Ch. 3.3.43 Radiation
damage produced by the analysis beam is the focus here.

Experiments were performed at room temperature, using an MeV *He beam as both an irradi-
ation source and analysis probe, with the ReSi2/Si(100) sample mounted on a goniometer with x-y
translations and with two axes of rotation. To eliminate the effect of irradiation during the process
of aligning the [100] channel with the incident beam, the channeling spectra were taken according
to the following procedure: we first used the two rotation axes to find the [100] axial channel at
one corner of the sample (beam size ~ 0.2 x 0.2 cm?, sample size ~ 1 x 1 cm?) and then translated
the sample so that a virgin region of the sample was exposed to the irradiation beam for analysis.
Figure 8.3-1 shows the backscattering spectra of the sample for a beam with random incidence (solid
line) and for a beam incident along the [100] axial channel at three different damage stages, (a) as-
grown (the dose during the channeling measurement of the as-grown sample is less than ~ 1014 /cm?
and the damage induced is negligible), after irradiation by a 1.4 MeV ~ 10!7/cm? 4He ion beam
incident (b) along the [100] axial direction or (c) along a random direction. Three facts are evident
from the spectra: (1) the as-grown ReSig sample is highly epitaxial with a Re minimum yield ymin
of ~ 2% (the fraction of counts below the surface peak of the aligned spectrum normalized with
respect to that of a random spectrum); (2) substantial doses of the analysis beam produce damage
in the ReSig film, which results in noticeable increases of the minimum yield; and (3) the amount
of damage produced by irradiation with an aligned beam is much smaller than that with a beam
of random incidence. This last fact suggests that the damage is produced predominantly by elastic

collisions among nuclei.

143

1000

| 1.4 MeV 4He
goo- 150 nm ReSiz on Si(100)

5 random spectrum

n 600F [100] aligned spectra:
= a (a) 9° as—grown
a after irradiation by
Oo 400- (b) « aligned beam
- (c) x random beam
200

| 1 i
0.6 0.8 1.0 1.2
Energy ( MeV )

FIG. 8.3-1 1.4 MeV *He backscattering and channeling spectra of a 150 nm thick epitaxial ReSig
layer grown on a Si(100) substrate. All four spectra were taken at room temperature and are
plotted by normalizing incident doses to a common value. The solid line is the spectrum for random
incidence. The three [100] channeling spectra are for samples irradiated at room temperature with

doses of (a) ~ 10*4/cm?, ~ 10'7/cm? (b) in a [100] aligned direction and (c) in a random direction.

1.0
~ 0.8
fs
Pt 0.6b ; { (a) ° as—grown
oO q ReSig i after irradiation by
30.4, [100] channel (b) * aligned beam
bo i mn. (c) x random beam
0.2
re) 1 l L ! n el 1 ! nl J 1
-6 —4 ~2 0 2 4 6
Tilt Angle ( ° )

FIG. 8.3-2 The normalized backscattering yield of the Re signal versus angle of tilt between the

incident beam and the [100] direction of the sample for the three damage stages of Fig. 8.3-1.

144

To further probe the damage structure of ReSig by an MeV 4He beam, we also measured the
critical angles for the [100] axial channel before and after irradiation. In this channeling orientation,
the atomic columns of the ReSig lattice consist of only Si or only Re atoms (see Fig. 3.3-3). There
are therefore two critical angles, one for Si columns and one for Re columns (see Fig. 3.3-2).*? The
angular scan and the critical angle of Re at the three damage stages discussed above (Fig. 8.3-1)
are shown in Fig. 8.3-2 For the virgin sample, the critical angle for Re (as well as Si) agrees with
Lindhard’s prediction.*° The critical angle decreases as the minimum yield (or damage) increases.
It is known that disorder in the form of amorphous regions** or a mosaic structure*®:** increases
both the minimum yield and the critical angle. On the other hand, a spatially correlated disorder
similar to that produced by lattice vibrations increases the minimum yield and decreases the critical
angle.9© The angular scan measurements on irradiated ReSia (Fig. 8.3-2) therefore suggest a defect
structure of correlated displacements for MeV *He irradiated ReSiz. The same conclusion about the
defect structure was obtained from the angular scans of the Si signal.

The damage induced by 1.4 MeV ‘He irradiation was quantified by monitoring the minimum
yields of the [100] axial channeling spectra as a function of the total dose of exposure. To measure
damage by a beam incident along a random direction, the sample was repetitively irradiated and
then oriented in the [100] direction to take a channeling spectrum at increasing dose levels. Since a
channeled beam generates little damage compared to that generated by the random beam (see Fig.
8.3-1), the channeling yield may be measured without significantly increasing the state of damage.
Figure 8.3-3 shows the dependence of the minimum yield ymin versus dose of random irradiation.
The minimum yield ymin of Si (and of Re as well, A and o in Fig. 8.3-3) initially increases rapidly
(starting value: 14% and 2% respectively) up to a dose of ~ 2 x 1015 /cm? and then with a slower
rate (~ 2.5%/10'°/cm? for Si and ~ 1.9%/10'6/cm? for Re).

The fact that the minimum yield for Si is always larger than that for Re is peculiar to chan-
neling of MeV ions in a diatomic crystal,** not an indication of higher initial or subsequent defect
concentration for the Si sublattice (see Ch. 3.3). The reason is that the minimum yield of the
element with low atomic number (Si) is enhanced and dominated by the ions deflected from the
columns with the element of high atomic number (Re), while the minimum yield of Re is affected
little by the deflection of He from Si columns.*3 We therefore use the minimum yield ymin of Re as
a measure of irradiation damage in ReSip.

To obtain the dose dependence of irradiation damage by an aligned beam, we first oriented the

sample in the [100] direction and then monitored the damage build-up by recording channeling

145

60+ 1.4 MeV *He > ReSig A

_-7 random incidence

40h Ay i
| a& ° [100] incidence ___-----~-~ a

a-"

Xmin ( % )

ee eee
“ja
20

random incidence ‘

_[100] incidence

4. 4 { 1 1 i st | L 41

0 5 10
¢ ( 10% /cm? )

FIG. 8.3-3 The minimum channeling yields of the Si and Re signals for an epitaxial ReSiy film as

a function of the 1.4 MeV ‘He irradiation dose for both a random and a [100] aligned incidence.

1.4 MeV *He radiation damage in ReSig
20 TRIM88 -
&§ 10

$ ( 10!8/em? )

FIG. 8.3-4 Comparison of the measured defect concentration versus irradiation dose in ReSi2
produced by 1.4 MeV ‘He ion beams of random (0) and [100] aligned (e) incidence at 300K, and
one calculated (dashed line) by a TRIM88 computer simulation of a beam of random incidence at

OK.

146

spectra at increasing dose levels during irradiation. The minimum yield Xmin(¢) of the sample after

irradiation of dose, ¢, is defined as

xmin(#) = aa(e) Pn), (83-1)

where Na(¢) and Np(¢) are the total backscattering counts resulting from a dose ¢ of a [100] aligned
and randomly incident beam, respectively. aot was obtained from channeling measurements by
numerical differentiation. avat@) is a dose-independent normalization constant, obtained from a
random backscattering spectrum. Figure 8.3-3 shows the dependence of the minimum yield ymin
versus irradiation dose for a [100] aligned beam. Again, the minimum yield ymin of Si (and of Re
as well, filled triangle and e in Fig. 8.3-3) initially increases rapidly up to a dose of ~ 2 x 1015 /cm?
and then with a slower rate ( ~ 0.9%/10!°/cm? for Si and ~ 0.3%/10!®/cm? for Re). The rapid
initial rise, although clearly present in all four plots of Fig. 8.3-3, is difficult to grasp experimentally
because of the poor statistics involved and will not be discussed further.

Energetic ions lose energy to target atoms in two processes: electronic and nuclear stopping.
According to TRIM88 simulation, almost all the energy of an MeV ‘He ion generates electronic
excitation in ReSig, and only a very small fraction of the energy (~ 0.3%) goes to the target nuclei
to produce displacements. In the regime of high energy and light incident ions, the persistent damage
resulting from these displacements consists mainly of isolated interstitial-vacancy pairs,” although
some small defect clusters may also be produced. We used the TRIM88 program to simulate damage
production. The simulation assumes an amorphous 150 nm thick ReSiz film on an amorphous Si
substrate and computes the concentration (defect density/atomic density of ReSiz) of displaced
atoms as a function of depth in the linear cascade approximation.”® A typical value for displacement
threshold energy of 15 eV and binding energy of 1 eV was chosen. The maximum defect concentration
in the simulation locates at a depth of ~ 4 ym, well into the Si substrate where the majority of
stopped *He ions also reside. In the ReSig film, the defect concentration is very small and is roughly
uniform in depth through the entire film. This defect concentration from the TRIM simulation is
plotted versus dose in Fig. 8.3-4 (dashed line).

Experimentally, the defect concentration in a damaged crystal can be estimated from channeling
measurements. When the defect concentration at depth z is cp(z) and the probability that an aligned
incident beam is dechanneled by the defects over the region from the surface to the depth z is Pp (2),

we have27

Xp(z) — xv(z) (8.3 — 2)

ep(z) + (1 — ep(z))Pp(z) = l-xv@)”

147

where xv(z) and yp(z) are the normalized channeling yields at depth z for virgin and damaged
crystals, respectively. In the near-surface region, the dechanneling probability Pp is small compared

to the defect concentration cp. Equation (8.3-2) therefore becomes

_ Xmin,p — Xminv (8.3 — 3)

cD
1— Xmin v

where Ymin,y and Xmin,p are the minimum yields. When the minimum yield of Re is used in
Eq. (8.3-3), one obtains the defect concentration in ReSiz shown in Fig. 8.3-4. An aligned beam
produces only about 1/7 the number of defects produced by a random beam. This is in accord with
the observation that the close encounter probability between the incident ion and the target nuclei
for an aligned beam is about one order of magnitude smaller than that for a random beam?° and
our assertion above that the defects are produced by elastic collisions among nuclei.

Figure 8.3-4 shows that the measured defect concentration produced by a random beam ap-
proximately equals that computed from TRIM88. If a value for the threshold energy Eq other than
15 eV were chosen, the slope of the TRIM88 result would change inversely with Eg. However, within
the range of acceptable threshold values, the qualitative agreement between the simulation and the
experiment would still hold. We therefore conclude that the defects are stable at room temperature.
This result is in contrast with that obtained for other silicides such as CoSi» (Ch. 8.2) and other
semiconductors such as Si,?’ where the measured damage produced by light energetic ions (no dense
cascade) at room temperature is much less than that predicted by TRIM88. The stability of defects
may be explained by the semiconductor character and the relatively large cohesive energy of ReSig.
Semiconductors are more sensitive to irradiation than metals because the strong chemical bond-
ing in semiconductors gives a higher activation energy for vacancy-interstitial pairs to recombine.
Defects are therefore more stable in semiconducting ReSig than in metallic silicides such as CoSiz.
In addition, ReSig has a cohesive energy of 8.0 eV/atom (obtained from the cohesive energy*® of
elemental Re and Si and the heat of formation? of ReSiz) compared with 4.6 eV/atom*® for Si.
This gives rise to a larger energy barrier for the migration of point defects in ReSig than in Si. De-
fects in ReSi2 are therefore more stable than those in Si. The sublinear rise of the measured defect
concentration in Fig. 8.3-4 suggests that defects formed late in the irradiation are increasingly likely
to be annihilated,*’ a process that results in a gradual saturation of the defect concentration as the

damage increases.

148

In summary, epitaxial ReSi2 thin films grown on Si(100) substrates were analyzed at room
temperature by MeV *He backscattering and channeling spectrometry. The minimum yield of [100]
axial channeling increases with increasing exposure of the ReSiz sample to the analyzing He beam.
This means that ReSi2 suffers irradiation damage induced by an MeV *He beam at room temper-
ature. The damage in the film induced by a beam incident along a random direction is about one
order of magnitude larger than that induced by a beam with an aligned incidence, indicating that
the damage is mainly generated by elastic collisions of nuclei. The experimentally measured defect
concentration produced at 300 K by a beam of random incidence is compared with the theoretically
estimated one produced at 0 K in an amorphous target. The agreement is fairly good, suggesting

that the defects produced by elastic nuclear collisions are stable at room temperature.

8.4 Amorphization and recrystallization of epitaxial ReSi, films

Transition-metal silicides have attracted much attention in the last decade because of their
importance in Si-based microelectronics.*”:*° Some silicides, such as ReSig and CrSig, are semicon-
ductors with narrow bandgaps.***! In the previous section, we showed that epitaxial ReSi» films
suffer irradiation damage by an MeV *He beam.!® We report here the results of the effect of 300
keV ?8Si and 380 keV 4°Ar ion implantation on some structural properties of epitaxial ReSi. films
and of the damage recovery by thermal annealing.!®

Epitaxial ReSig films 150 nm thick were grown on phosphorous-doped (p ~ 50Q-cm) n-type
3-in $i(100) wafers in ultrahigh vacuum (~ 10-° Torr) at ~ 650°C by reactive deposition epitaxy
(see Ch. 3.2).4? The thickness and composition of the as-grown films were confirmed by 2 MeV
*He backscattering spectrometry. Channeling measurements give a minimum yield of ~ 2% for Re,
showing that the films are highly epitaxial. Furthermore, 6-20 x-ray diffractometry shows only the
ReSi2 (200) peak in the 20 angular scan from 20° to 170° (the relevant section is shown in Fig. 8.4-1
a), indicating that the films are epitaxially oriented with, respect to the Si (100) substrate.

Film-substrate current-voltage (I-V) measurements on ReSip films grown on both p- and n-type
Si substrates were performed. While the film on the p-type Si substrate shows ohmic behavior, the
sample on the n-type Si substrate has the characteristics of a rectifying junction (junction resistance
~2kQ at ~ +10 V and ~ 1 MQ at ~ —-10 V, sample size ~ lcm x lcm). This means that the ReSig
film is electrically isolated to a degree from the n-type Si substrate. The van der Pauw method was
used for Hall and resistivity measurements on the ReSi, film grown on the n-type Si substrate and

showed that the ReSig film is p-type. The resistivity of the ReSiy film was also measured from 90 K

149

300 keV *8si implanted ReSiz/Si(100)

a: as—grown

— ReSiz (200)

ray)
T | T | UJ i T

(400)

a?)

b: 1x1013/cm?

Intensity ( 10° counts/s )
ao
T J T | Ll | Ls

0.47 c: 1x10!*/cem?
0.2L
0 i ~
4} 14 2
i d: 5x10°*/em
2 a Si
P (400)

Kg \
0 4 ah en eee ee ;
62 64 66 68 70
26(°)

ol
o>)

FIG. 8.4-1 X-ray diffraction spectra of epitaxial ReSi2/Si(100) samples: (a) as-grown; and implanted
by 300 keV **Si to doses of (b) 10'3/cm?, (c) 10'*/cm?, and (d) 5 x 10!4/em?. A Cu X-ray source

and 0-26 geometry were used.

| 300 keV “8sj implanted ReSi, film :
— -1 L_ a
g 10 'E E
(s) E 4
fo F :
by ' 4
5 _
5 107?- 3
2 C J
‘DD 5 -~
(3) - 4
me - 4
J an ]
10-3 ] l l Ls

1/T ( 103/K )

FIG. 8.4-2 Resistivity of the samples shown in Fig. 8.4-1, measured at temperatures ranging from

90K to 330K in the dark.

30 min isochronal annealing
100k « 300 keV *8si implanted ReSi2/Si(100) |
- d
° a: as-grown
x b: 1x10!3/em?
ce: 1x10!4/em?
q 50- d: 5x10!*/cm? 4
oF po Lo an
0 200 400 600

Temperature ( °C )

FIG. 8.4-3 Changes of the minimum yield for Re upon thermal annealing in vacuum at 500°C,

600°C and 700°C for 30 min for the samples shown in Fig 8.4-1.

151

to 330 K (Fig. 8.4-2 curve a). It equals ~ 23 mQcm at room temperature, nearly the same as those
reported for polycrystalline films (~ 20 mQcm).5:°3 All electrical measurements were carried out in
the dark.

To investigate the effect of implantation damage and of chemical species on the structural and
electrical properties of the ReSiz films, either 300 keV ?8Si or 380 keV 4°Ar ions were implanted
into selected samples at room temperature with doses ranging from 103/cm? to 10!°/cm?. At the

implantation energies chosen, the damage peaks before the film-substrate interface.

2 MeV *He backscattering and channeling spectrometry were used to monitor the damage build-
up in the ReSig film. Because the minimum yield from only the element with the largest atomic
number indicates the epitaxial quality in a polyatomic crystal,*3 we use the minimum yield for Re as
an indicator of damage in the ReSiz film. The minimum yield for Re increases monotonically with
increasing implantation dose from ~ 2% for the as-grown sample to 100% for the 5 x 10'4 28Si/em?
implanted sample. This suggests that the ReSig is amorphized at room temperature by the 300 keV
5 x 1014 ?8Si/cm? implantation. The channeling spectrum of the sample implanted to 5 x 10!4/cem?
also shows that the Si substrate is not amorphized. Similar results on damage build-up were found
for the *°Ar implanted samples. The epitaxial ReSig film is amorphized at room temperature by

implantation of 380 keV 10!4 #°Ar/cm?.

The damage induced in semiconducting ReSi2 by room temperature implantation is significantly
larger than that found in metallic silicides. NiSiz is amorphized at room temperature only after
implantation of 100 keV 3 x 10'° #°Ar/cm?,’ which is about 30 times the dose for amorphization of
semiconducting ReSiz. CoSig is not amorphized by room-temperature implantation of 150 keV ?8Si
to a dose of 3 x 10'®/cm??8 and Pd2Si by 100 keV *#°Ar to 10!7/cm?.” On the other hand, the critical
dose for amorphization of ReSiz is about the same as that for bare Si.?” The radiation sensitivity
of ReSig is typical of the covalent bonding of semiconductors rather than of metallic bonding. This
result is also consistent with the high susceptibility of ReSiz to irradiation damage by an MeV *He

beam.}5

The crystalline quality of the epitaxial ReSi film was also evaluated from the intensity and
full-width at half-maximum (FWHM) of the x-ray diffraction peak from the film. Fig. 8.4-1 shows
the evolution of the x-ray diffraction spectrum as the ?8Si ion dose increases from 0 to 5 x 10!4/cem?
(Fig. 8.4-1 a tod). A strong ReSiz (200) peak is detected from the as-grown sample. That the Si
(400) peak is weak and observable only in a magnified plot is probably due to the strong diffraction

of the incident x-ray by the epitaxial ReSig film. As the implantation dose rises, the FWHM of

152

the ReSiy (200) diffraction peak increases and the intensity decreases. For the 5 x 10!4 8Si/cm?
implanted sample, the ReSiz (200) peak disappears while the Si (400) peak simultaneously becomes
intense, indicating that the film becomes relatively transparent to the incident x-rays. An x-Tay
diffraction pattern was taken with a Read camera for the 5 x 10'4 ?8Si/cm? implanted sample.
Diffuse rings were detected that are characteristic of amorphous material. We conclude that the
ReSiz film becomes x-ray-amorphous upon 5 x 1014 ?8Si/cm? implantation. Again, similar results

were obtained for the 40 Ar implanted samples.

Figure 8.4-2 shows the resistivity from 90 to 330 K of the as-grown and 28Si implanted ReSi,
films. The resistivity decreases with increasing temperature for all the samples, suggesting that
ReSig remains a semiconductor upon implantation. The resistivity decreases monotonically with
increasing implantation dose at all measurement temperatures. The same behavior was observed for
the 4°Ar implanted samples. I-V measurements confirmed that the films are still electrically isolated
from the substrates up to the highest dose, with about the same forward and reverse resistances as

those measured before implantation.

The resistivity decrease of ReSiz we observed is associated with structural changes produced by
implantation, not with the chemical identity of the added impurity. A decrease in resisitivity upon
ion implantation was also observed in semiconducting CrSi2.9!” This is opposite to the response of
metallic transition-metal silicides to implantation damage.!*!4.23 There, the carrier concentration
is roughly constant. Implantation produces static disorder, which reduces the carrrier scattering
length. The resistivity hence increases. The amount of increase is typically proportional to the
density of disorder (see Ch. 8.2).?8 However, the resistivity of the amorphized ReSiy is about the
same as that of the amorphized CoSiz, ~ 1 mQem.'* This indicates that the electrical conduction
in amorphous materials is governed by the hopping of charge carriers among the localized states!4

and is roughly material-independent.

Upon thermal annealing, the channeling yields through the entire film decrease approximately
uniformly, indicative of none layer-by-layer growth. Figure 8.4-3 shows how the minimum yield of
the Re signal changes as a function of the temperature for 30 min isochronal annealing in vacuum
(~ 5x 10-7 Torr). The lightly damaged ReSip recovers after 500°C annealing (Fig. 8.4-3, curve
b). On the other hand, the minimum yield of fully amorphized ReSiy remains unchanged upon
annealing at 600°C, and drops to ~ 2% after 700°C annealing (Fig. 8.4-3, curve d). Clearly,
amorphous ReSig recrystallizes completely after 700°C annealing and recreates a highly epitaxial

film. The heavily damaged but non-amorphized ReSig film (Fig. 8.4-3, curve c) recovers incompletely

153

after 700°C annealing, as the minimum channeling yield of ~ 10% indicates. Heavily damaged but
non-amorphized Si also regrows poorly upon thermal annealing.®4

The recovery was also monitored by the ReSi2 (200) x-ray diffraction peak intensity and FWHM.
The annealing characteristics are similar to those obtained from channeling analysis (Fig. 8.4-3).
The diffraction peak of the lightly damaged sample (Fig. 8.4-1, curve b) intensifies and sharpens with
increasing annealing temperature, while that of the amorphous sample (Fig. 8.4-1, curve d) changes
only after 700°C annealing. The diffraction spectrum becomes the same as that of the as-grown
sample (Fig. 8.4-1, curve a). This result suggests that solid phase crystallization of amorphous

ReSiz films could be an alternative technique to grow epitaxial ReSi» films on Si(100) substrates.

The resistivity of all implanted ReSiz films also increases with increasing annealing temperature
and returns to about the value of the as-grown sample after 700°C annealing.

By both x-ray diffractometry and channeling spectrometry, the 4°Ar implanted ReSi films that
are not amorphized recover similarly to the ?°Si implanted films upon thermal annealing. However,
unlike the ?°Si implanted amorphous film, the *°Ar implanted amorphous one fails to recrystallize
epitaxially up to 1000°C annealing for 30 min. For all annealing temperatures, the #°Ar amorphized
ReSiz films retain a channeling yield of 100% and produce no detectable (200) x-ray diffraction
peak. The Read camera x-ray diffraction pattern shows that the film after 1000°C annealing is
polycrystalline. One possible explanation, based on an assumed analogy with crystallization of
4°Ar implanted amorphous Si,** is that the implanted *°Ar atoms form bubbles in the matrix that

interfere with the epitaxial regrowth of the amorphized ReSig.

The present results and the analogy with damage production and annealing in Si,°4 suggest
the following processes in implanted ReSig: at low doses, ion implantation at room temperature
produces damaged zones dispersed in a crystalline matrix. As the dose increases, the zones overlap
and interact until the entire matrix is amorphized. The individual zones, which probably consist of
defect clusters or small amorphous regions, reorder at a relatively low temperature (< 600°C). The
amorphous film recrystalizes abruptly at a higher temperature (~ 700°C), in contrast to the layer-
by-layer growth of amorphous Si on crystalline Si. In that case, a seed exists and recrystallization
is limited by the growth. In the present case, no seed is present and recrystallization is presumably
nucleation-limited, and is therefore abrupt. The recrystallization temperature is roughly 1/3 of the

melting point, similar to that for semiconducting CrSig films.9:5®

In summary, we used 2 MeV ‘He backscattering spectrometry, x-ray diffractometry, and the van

der Pauw technique to study how 150 nm epitaxial ReSi» films on $i(100) change structurally and

154

electrically upon room-temperature implantation of 300 keV 28Si or 380 keV #°Ar. The as-grown
film has a minimum channeling yield of ~ 2% for Re, and a resistivity of ~ 23 mQcm at room
temperature. Ion implantation produces damage in the film, which increases monotonically with
dose. At a dose of either 5 x 10'* ?8Si/cm? or 1 x 10'4 *°Ar/cm?, the entire ReSig film becomes
both x-ray- and channeling-amorphous. The resistivity of the film decreases monotonically with
dose. The amorphous film has a resisitivity of ~ 1.2 mQcm at room temperature. Upon annealing
in vacuum at 700°C for 30 min, the damage anneals out and the amorphous ReSig film recrystallizes
epitaxially, once again exhibiting a minimum channeling yield of ~ 2% for Re and a resistivity of ~

23 mQcm.

8.5 Strain induced by ion implantation in GeSi films

Recent interest in GeSi-base heterojunction bipolar transistors has spurred extensive studies
on various properties of GeSi/ Si structures, including the effect of ion implantation. The difference
in atomic numbers of Ge and Si results in selective damage of GeSi and Si layers, which can in-
duce amorphous-crystalline superlattices.>:?° Amorphized GeSi layers on crystalline Si recrystallize
epitaxially upon thermal processing.'®!° Defects produced in a metastable strained layer by ion
implantation could enhance strain relaxation.”! In this section, we report some results on the strain
and damage induced by 320 keV **Si implantation into pseudomorphic GeSi layers on 5i(100), and

the effects of thermal annealing.®’ The stability of ion-implanted strained layers is also discussed.

A. Experimental Procedures

Pseudomorphic metastable GerSij_, (x ~ 0.04, 0.09, 0.13) layers ~ 170 nm thick were grown
on Si(100) wafers at ~ 600°C by ultrahigh vacuum chemical vapor deposition at IBM. The samples
were sent to Caltech, degreased, and loaded into the implanter. 320 keV 28Si ions were implanted
into the samples at room temperature in vacuum (~ 1077 Torr). The damage peaks at ~ 200 nm
beneath the interface inside the Si substrate, according to TRIM88 simulation.?> Doses range from
10'*/cm? to 2 x 10'5/cm?, and the flux was kept below 10!2/cm?/s. Double crystal diffractometry
and MeV ion channeling were used to analyze the strain and damage induced by implantation.

Samples were also annealed for 30 min at 300-700°C in vacuum (~ 5 x 10-7 Torr). The recovery

of damage and the change of strain in the GeSi layers were monitored.

B. Results and Discussion

(i) damage and induced strain

155

Perpendicular and parallel strain, «+ and ell, in GeSi/Si(100) heterostructures were extracted
from x-ray rocking curves diffracted from both symmetrical (400) and asymmetrical (311) planes (see
Ch. 2.2). All the as-grown samples are pseudomorphic within experimental sensitivity (ell < 0.01%),
in agreement with the experimental results of Ch. 4.2. These samples are metastable, meaning that
they are all thicker than the equilibrium critical thickness predicted by Matthews-Blakeslee’s model
(see Fig. 4.2-2). Figure 8.5-1 shows the (400) rocking curves of the as-grown Geo .o9Sio.91 /Si(100)
(solid line) and of the samples implanted to doses of 2x (e) and 5 x 10!4/cm? (0). The rocking
curve of the as-grown sample shows clearly visible small-amplitude oscillations, indicating a high
crystalline perfection of the GeSi layer. The angular width of the diffraction peak from the layer is
due entirely to the finite thickness of the layer. The perpendicular strain of the as-grown sample
is e+ = 0.69%. After implantation to 2 x 10!4/cm?, the diffraction peak from the GeSi layer shifts
farther away from the substrate Si peak, meaning that the GeSi layer develops an additional positive
strain (see ein Fig. 8.5-1). The peak intensity decreases, while the angular width remains about the
same. These facts suggest that defects are produced in the layer, that the layer still diffracts x-ray
coherently (see discussion in Ch. 8.2), and that the defects are uniformly distributed in the layer and
induce uniform lattice expansion. The perpendicular strain of the implanted sample is e+ = 0.97%,
while the parallel strain remains zero. The sample is thus still pseudomorphic and becomes more
metastable than it was before. The additional perpendicular strain induced by the implantation
damage, Act, is 0.28%. One can also see the diffraction peak (at ~ —0.05° near the Si reference
peak) from the damaged substrate Si ( in Fig. 8.5-1). The strain induced (~ 0.06%) is about the
same as that in the implanted bulk Si samples, suggesting that the GeSi overlayer has no influence
on the implantation damage in the Si substrate. As the dose rises to 5 x 10'4/cm?, the diffraction
peak from the GeSi layer is buried in the background and becomes undetectable, while that from
the damaged Si substrate is still measurable (0 in Fig. 8.5-1). These findings indicate that the GeSi
layer is more severely damaged than the Si substrate. In other words, GeSi alloy is more susceptible
to radiation damage than Si is, in agreement with others’ results.*:?° As the dose increases further
to 2 x 10'5/cm?, the rocking curve from the sample becomes featureless (not shown) because both
the layer and the substrate are amorphized.27

We also used 2 MeV *He [100] axial channeling to characterize the damage in the GeSi layer.
The defect concentration, cp, was extracted from-the measured channeling yields (see Ch. 8.2 and
8.3). The average defect concentration in the layer is ~ 14% after 2 x 1014/cm?, rises to ~ 80% after

5 x 10'4/cm?, and becomes 100% after 2 x 10!5/cm? when the layer is amorphized. Combining the

156

10 : (400) diffraction rocking curves 3
Be 490 | - as~grown 1
~~ E © 2x 10!4/cm? q
> EF o 5x 10!4/cem? ;
n a 4
§ 10°'L
~~ Cc
& E

107-2

-0.6 —0.4 ~0.2 0
Aa (°)

FIG. 8.5-1 X-ray rocking curves diffracted from the (400) symmetrical planes of the as-grown

Geo .o9Sio.91/Si(100) sample (solid line), and of those implanted at room temperature with 320 keV
28Si to 2x (©), 5 x 10'4/cm? (0).

320 keV *8si into Ge,Si;_,/Si(100)

0.4
Fo 1.2x 10!4/em? 4
0.3L 4 2x 10!4/em? oa a
e 5 4
~ 0.24 4
4 --"
0.1 2-7" 4
O i po L 1 ! rif oi 1 | 4
0 0.05 0.10 0.15

FIG. 8.5-2 The strain induced by 320 keV 78Si implantation into pseudomorphic GeSi layers vs

the Ge composition. The dashed lines are the strains predicted from a linear interpolation model of

Eq. (8.5-2).

157

x-ray and channeling results for the 2 x 10'*/cm? implanted sample, we obtain the following rela-

tionship between Ae+ and cp,

Act = 0.02cp. (8.5 —1)
We tentatively postulate that this relation applies for this sample implanted to any dose. To verify
this hypothesis, additional experiments with fine dose intervals are needed. The coefficient (0.02) is
of the same order of magnitude as that for implanted bulk Si (0.012; see Ch. 7.2). We are currently
in the process of measuring the coefficient for implanted bulk Ge, and of studying the dependence
of the coefficient on the Ge composition.

X-ray rocking curve measurements of other implanted samples with different Ge composition
give similar results as those described above. Firstly, the strain in the damaged substrate Si is
the same as that in the implanted bulk Si (Ch. 7.2). Secondly, the implantation damage induces
additional strain in the GeSi layer besides the intrinsic strain of the heterostructure. The layers
remain pseudomorphic. At low damage level (¢ < 2 x 10'*/cm?) where the diffraction peak from
the GeSi layers is measurable, the additional strain induced by damage, Ae+, increases with the
implantation dose (Fig. 8.5-2). We know that the strain increases linearly with the Si dose at
low damage levels in implanted bulk Si (Ch. 7.2) and Ge (Ref. 58) crystals. The limited data in
Fig. 8.5-2 and the analogy with Si and Ge lead us to propose that for any Ge composition, Aet
is proportional to the dose at low damage levels. For a given dose, the induced strain in the GeSi
alloy increases linearly with the Ge composition (Fig. 8.5-2). The dashed line is that obtained by
interpolation between the corresponding strain of implanted Si (Ch. 7.2) and Ge (Ref. 58). For that
interpolation, the strain in implanted Si can be found directly from the measurements (Ch. 7.2).
The strain for implanted Ge was computed by multiplying the slope of strain versus dose by the
corresponding dose (1.2x or 2 x 10'4/cm?), because this strain cannot be realized physically since
the Ge is amorphized beyond ~ 7 x 101° *8Si/cm?. Figure 8.5-2 shows that the interpolation fits
the data reasonably well. This result suggests that the slope, Sy(GezSi;_.), of the implantation-
induced strain Ae+ vs the dose ¢ for Ge,Si,_, alloy, may, be predicted from that of Ge and Si,

S¢(Ge) and Sg(St) according to

Further experiments are needed to prove or to invalidate this generalization.
(ii) damage annealing and strain change
To investigate the stability of these implanted layers upon thermal processing, we annealed the

samples at 300-700°C in vacuum for 30 min and monitored the change of strain. Firstly, we noticed

158

that the annealing characteristics of the damaged Si substrate are similar to those of implanted bulk
Si samples (see Ch. 7.2). The annealing behavior of the GeSi layers can be categorized into three
types according to the initial damage levels. For the lightly damaged samples (¢ < 2 x 10!4/cm?),
the annealing shifts the diffraction peak from the GeSi layer towards the main peak, and the peak
intensity increases. The peak width remains the same. The rocking curve becomes the same as
that of the as-grown sample after 700°C. For the heavily damaged samples (¢ = 5 x 10!4/cm?, o
in Fig. 8.5-3) where the peak from the implanted GeSi layer is undetectable, it is still undetectable
after 300°C (filled inverse triangle in Fig. 8.5-3), but becomes measurable after 400°C annealing
(e), and the samples completely recover after 700°C (solid line in Fig. 8.5-3). Figure 8.5-4 plots
the perpendicular strain of three samples studied here as a function of 30 min isochronal annealing
temperature. The filled symbols represent the as-grown samples. Two conclusions are evident:
(a) the major annealing stage occurs at 23-300°C, and (b) the strain and damage induced by the
implantation can be completely healed by a 700°C annealing. Furthermore, no relaxation of the
intrinsic strain was observed for any sample after 700°C annealing, meaning that the presence of
defects does not significantly enhance the relaxation of the metastable strain. Hull et al.?4 did observe
some enhancement of strain relaxation in implanted GeSi/Si structures by transmission electron
microscopy. These authors suggested that the defects promote the nucleation of dislocations, but
impede their propagation. The x-ray rocking curve technique is not sensitive enough to explore that
regime of initial dislocation nucleations because the dislocation density is below the x-ray detection
limit. Our results therefore do not contradict those of Hull et al.??

We also used x-ray diffraction and MeV ion channeling to study the solid-phase epitaxial re-
growth of the amorphized GeSi layers (@ = 2x 1915 /cm?). The regrown layers have larger channeling
yields, much weaker and broader x-ray diffraction peaks, than the as-grown ones, indicating the pres-
ence of extended defects in the layers. To maintain the crystalline perfection of epitaxial GeSi layers,
amorphization transformation should be avoided.

In summary, we found that ion implantation in GeSi layers on $i(100) produces damage. This
damage induces additional positive strain in the layers. At a low damage level, the induced strain
increases linearly with the dose. The slope of the induced strain vs the dose rises linearly with the
Ge composition, and can be predicted by interpolation of the slopes for bulk Ge and Si crystals.
Thermal annealing removes damage and eliminates the induced stram. Damaged, but not amor-
phized, samples fully recover after 700°C annealing for 30 min; amorphized samples recover only

partially.

107!

Intensity ( % )

159

320 keV 5x10!4 *8si/cem?
30 min isochronal annealing

T UTTTTHy
i Lagi

300°C 700°C
RT 400°C}

TM
Ll bosiinl

it tail

TTT TTT

FIG. 8.5-3 (400) x-ray

temperature (RT) by 320 keV 5 x 10'* ?8Si/cm? and annealed for 30 min at various temperatures.

The spectrum of the 700°C annealed sample (solid line) is indistinguishable from that of the as-grown

sample (Fig. 8.5-1).

rocking curves of a pseudomorphic Geo o9Sio.91/Si(100) implanted at room

320 keV *8si into Ge,Si,_,/Si(100)

On 200 400 600

T (°C )

FIG. 8.5-4 The strain

and circle are for the samples with z = 0.04, 0.09, and 0.13, respectively. The filled symbols represent

the unimplanted samples. The small (big) open symbols represent the samples implanted by 320

in the Ge,Si,_z layer vs the annealing temperature. The square, triangle,

keV 1.2x (2x) 1014 28Si/cm?.

160

References

10.
11.

12.

13.

14.
15.
16.

17.
18.

19.

J.J. Coleman, P.D. Dapkus, C.G. Kirkpatrick, M.D. Camras, and N. Holonyale, Jr., Appl. Phys.
Lett. 40, 904 (1982). |

A.H. Hamdi, J.L. Tandon, and M-A. Nicolet, Nucl. Instr. and Meth. B10/11, 588 (1985).
D.R. Myers, S.T. Picraux, B.L. Doyle, G.W. Arnold, and R.M. Biefeld, J. Appl. Phys. 60, 3631
(1986).

A.G. Cullis, P.W. Smith, D.C. Jacobson, and J.M. Poate, J. Appl. Phys. 69, 1279 (1991).
D.J. Eaglesham, J.M. Poate, D.C. Jacobson, M. Cerullo, L.N. Pfeiffer, and K. West, Appl.
Phys. Lett. 58, 523 (1991).

. S. Mantl, D.B. Poker, and K. Reichelt, Nucl. Instr. and Meth. B19/20, 677 (1987).

H. Ishiwara, K. Hikosaka, and S. Furukawa, Appl. Phys. Lett. 32, 23 (1978).

. M. Maenpaa, L.S. Hung, M-A. Nicolet, D.K. Sadana, and S.S. Lau, Thin Solid Films 87, 277

(1982).

C. A. Hewett, I. Suni, S. S. Lau, L. S. Hung, and D. M. Scott, Mat. Res. Soc. Symp. Proc.
27, 145 (1984).

C.A. Hewett, S.S. Lau, I. Suni, L.S. Hung, J. Appl. Phys. 57, 1089 (1985).

M. C. Ridgway, R. G. Elliman, R. P. Thornton, and J. S. Williams, Appl. Phys. Lett. 56, 1992
(1990).

K. Maex, A.E. White, K.T. Short, Y.F. Hsieh, R. Hull, J.W. Osenbach, and H.C. Praefcke, J.
Appl. Phys. 68, 5641 (1990).

J. C. Hensel, R. T. Tung, J. M. Poate, and F. C. Unterwald, Nucl. Instrum. and Meth. B7/8,
409 (1985).

B-Y. Tsaur and C. H. Anderson, Jr., J. Appl. Phys. 53, 94 (1982).

G. Bai, M-A. Nicolet, J.E. Mahan, and K.M. Geib, Appl. Phys. Lett. 57, 1657 (1990).

K.H. Kim, G. Bai, M-A. Nicolet, J.E. Mahan, and KM. Geib, Appl. Phys. Lett. April 29,
1991.

T. C. Banwell, X. A. Zhao, and M-A. Nicolet, J. Appl. Phys. 59, 3077 (1986).

B.T. Chilton, B.J. Robinson, D.A. Thompson, T.E. Jackman, and J.-M. Baribeau, Appl. Phys.
Lett. 54, 42 (1989).

S. Mantl, B. Hollander, W. Jager, B. Kabins, H.J. Jorke, and E. Kasper, Nucl. Instr. and
Meth. B39, 405 (1989).

21.

22.
23.
24.
25.
26.
27.
28.
29.

30.

31.
32,
33.

34.
35.

36.
37.

38.
39.
40.

41.
42.

161

- M. Vos, C. Wu, I.V. Mitchell, T.E. Jackman, J.-M. Baribeau, and J. McCaffrey, Appl. Phys.
Lett. 58, 951 (1991).

R. Hull, J.C. Bean, J.M. Bonar, G.S. Higashi, K.T. Short, H. Temkin, and A.E. White, Appl.
Phys. Lett. 56, 2445 (1990).

C.J. Tsai, H.A. Atwater, and T. Vreeland, Appl. Phys. Lett. 57, 2305 (1990).

G. Bai, M-A. Nicolet, unpublished.

J.P. Biersack and L.G. Haggmark, Nucl. Instr. and Meth. 174, 257 (1980).

G. Bai and M-A. Nicolet, unpublished.

G. Bai and M-A. Nicolet, J. Appl. Phys., July 15, 1991.

B.E. Warren, X-Ray Diffraction, Dover, New York, 1990.

W.K. Chu, J.W. Mayer, M-A. Nicolet, Backscattering Spectrometry, Academic Press, New
York, 1978.

L.C. Feldman, J.W. Mayer, T.C. Picraux, Materials Analysis by Ion Channeling, Academic
Press, New York, 1982.

S. Roth and R. Sizmann, R. Coutelle, and F. Bell, Phys. Lett. 32A, 119 (1970).

M.J. Hollis, Phys. Rev. B8, 931 (1973).

C.S. Newton, R.B. Alexander, G.J. Clark,, H.J. Hay, P.B. Treacy, Nucl. Instr. and Meth. 132,
213 (1976).

P.B. Price and J.C. Kelly, Phys. Rev. B17, 4237 (1978).

P. Baruch, F. Abel, C. Cohen, M. Bruneaux, D.W. Palmer, and H. Pabst, Rad. Eff. 9, 211
(1971).

S.U. Campisano, G. Foti, F. Grasso, and E. Rimini, Appl. Phys. Lett. 21, 425 (1972).

W.H. Kool, H.E. Roosendaal, L.W. Wiggers, and F.W. Saris, Nucl. Instr. and Meth. 132, 285
(1976).

G. Dearnaley and D.R. Jordan, Phys. Lett. 55A, 201 (1975).

D.S. Gemmel and R.C. Mikkelson, Phys. Rev. B6, 1613 (1972).

S. Yamaguchi, M. Koiwa, M. Hirabayashi, Y. Fujino, J. Takahashi, K. Ozawa, and K. Doi, Rad.
Eff. 40, 231 (1979).

T.F. Luera, J. Appl. Phys. 51, 5792 (1980).

J.E. Mahan, K.M. Geib, G.Y. Robinson, R.G. Long, X.H. Yan, G. Bai, M-A. Nicolet, and M.
Nathan, Appl. Phys. Lett. 56, 2439 (1990).

43.
44.
45.
46.

47.

48.
49.

50.
ol.
52.

53.
54.
50.
56.
57.
58.

162

G. Bai, M-A. Nicolet, J.M. Mahan, and K.M. Geib, Phys. Rev. B41, 8603 (1990).

E. Rimini, E. Lugujio, and J.W. Mayer, Phys. Rev. B6, 718 (1972).

H. Ishiwara and S. Furakawa, J. Appl. Pyhs. 47, 1686 (1976).

N. Cheung, M-A. Nicolet, and J.W. Mayer, Proceedings of the Symposium on thin film interfaces
and interactions, edited by J.E.E. Baglin and J.M. Poate (The Electrochemical Society Inc.,
Princeton, 1980), p. 176.

M.W. Thompson, Defects and Radiation Damage in Metals (Cambridge University Press, 1969),
p. 253.

C. Kittel, Introduction to Solid State Physics, 5th ed. (Wiley, New York, 1976), p. 74.

M-A. Nicolet and S. S. Lau, in VLSI Electronics: Microstructure Science Vol. 6, ed. by N.
Einspruch and G. Larrabee (Academic, New York, 1983), p. 391.

S.P. Murarka, in Silicides for VLSI Application, Academic, New York, 1983.

M. C. Bost and J. E. Mahan, J. Appl. Phys. 58, 2696 (1985).

C. Krontiras, L. Groenberg, I. Suni, F. M. d’Heurle, J. Tersoff, I. Engstroem, B. Karlsson, and
C. S. Petersson, Thin Solid Films, 161, 197 (1988).

R. G. Long, M.C. Bost, and John E. Mahan, Thin Solid Films, 162, 29 (1988).

J.F. Gibbons, Proc. IEEE, vol. 6, 1062 (1972).

P. Revesz, M. Wittmer, J. Roth, and J.W. Mayer, J. Appl. Phys. 49, 5199 (1978).

F. Nava, T. Tien, and K. N. Tu, J. Appl. Phys. 57, 2018 (1985).

G. Bai, M-A. Nicolet, unpublished.

V.S. Speriosu, B.M. Paine, M-A. Nicolet, and H.L. Glass, Appl. Phys. Lett. 40, 604 (1982).

163

Appendix I Some Studies of Compound Semiconductors

Summary

Al. Thermal Strain of Epitaxial In,Ga,_,As;_,P, Films on InP(100)

— in collaboration with Dr. S.J. Kim at AT&T Bell Laboratories

Single layers of ~ 0.5um thick InugGa;_,Asi_,P, (0.52 < u < 0.63 and 0.08 grown epitaxially on InP(100) substrates by liquid phase epitaxy at ~ 630°C. The compositions of
the films were chosen to yield a constant bandgap of ~ 0.8 eV (A = 1.55 pum) at room temperature.
The lattice mismatch at room temperature between the epitaxial film and the substrate varies from
—4 x 10-3 to +4 x 10-5. The strain in the films was characterized in air by x-ray double crystal
diffractometry with a controllable heating stage from 23°C to ~ 700°C. All the samples have an
almost coherent interface from 23°C to about. ~ 330°C with the lattice mismatch accommodated
mainly by the tetragonal distortion of the epitaxial films. In this temperature range, the x-ray strain
in the growth direction increases linearly with temperature at a rate of (2.0 + 0.4) x 10-°/°C, and
the strain state of the films is reversible. Once the samples are heated above ~ 330°C, a significant

irreversible deterioration of the epitaxial films sets in (publication No. 15).

A2. Damage in GaAs, Si, and Ge by MeV ‘He Irradiation at Room Temperature

MeV ‘*He backscattering spectrometry is routinely used to characterize the composition and
thickness of thin films (< 1 ym). It is therefore important to know what effects an He beam has
upon the materials analyzed. X-ray double crystal diffractometry analysis of bulk Si, Ge, and GaAs
crystals irradiated at room temperature by 2 MeV ~ 10!° *He/cm? (a typical dose needed to get a
channeling spectrum of good statistics) shows that while no measurable strain (< 0.01%) is induced
in Si and Ge, significant positive strain (~ 0.4%) is produced in GaAs. The strain in irradiated
GaAs decreases to zero upon annealing at 400°C for 15 min, indicating that the damage heals.
We conclude that backscattering spectrometry is practically a non-destructive analytical tool for Si
and Ge crystals, but not for GaAs. Special care is therefore needed in the analysis of GaAs and
AlGaAs/GaAs by MeV ion beams and in the interpretation of experimental results (publication No.

164

A3. Defects Annealing Near Room Temperature of ?°Si implanted GaAs(100)

The annealing behavior near room temperature of the defects in 300keV 78Si implanted GaAs
was monitored by using x-ray double crystal diffractometry to measure the strain relaxation as a
function of time duration. The maximum strain of the samples stored in ambient air at 23°C and
100°C decreases with time, revealing that the defects are mobile near room temperature and can
recombine and reorder. Strain relaxation is exponential in time. At least two time constants of ~ 0.24
hr and ~ 24 hr are needed to fit the experimental data, indicating that two different kinds of defects
are responsible for the strain relaxation. Time constants are also obtained for different implantation
doses and temperatures (23°C or —194°C), and are insensitive to both these parameters, suggesting

that the time constants are intrinsic properties of the defects in GaAs (publication No. 4).

A4. Sequential Nature of Damage Annealing and Dopant Activation in Implanted GaAs

—in collaboration with Dr. J.L. Tandon at McDonnell Douglas Astronautics Company

Rapid thermal processing of implanted GaAs reveals a definitive sequence in damage annealing
and electrical activation. Removal of the damage occurs first. At that stage, the GaAs is n-type
with a very low carrier concentration (~ 10°/cm?) and high electron mobility (~ 5000 cm?/Vs),
regardless of the implanted species. Electrical activation is achieved next. The GaAs becomes n-
or p-type, or remains semi-insulating, depending on the chemical nature of the dopants. Similar
experiments were performed in implanted Ge. No such two-step sequence is observed, suggesting

that the sequential nature may be unique to compound semiconductors (publication Nos. 9, 12, 14).

10.

11.

165

Appendix IT List of Publications

Y.C. Kao, K.L. Wang, E. deFresart, R. Hull, G. Bai, D.N. Jamieson, and M-A. Nicolet, “Study

of CoSiz/Si strained layers grown by molecular beam epitaxy,” J. Vac. Sci. Technol. B5, 745

(1987).

D.N. Jamieson, G. Bai, Y.C. Kao, C.W. Nieh, M-A. Nicolet, and K.L. Wang, “On the critical

layer thickness of strained-layer heteroepitaxial CoSi, films on (111)Si,” Mat. Res. Soc. Symp.

Proc., Vol. 91, 479 (1987).

Y.C. Kao, D.N. Jamieson, G. Bai, C.W. Nieh, T.L. Lin, B.J. Wu, H.Y. Chen, and K.L. Wang,

“Epitaxial CoSi2/porous-Si strained layer structures grown by MBE,” Mat. Res. Soc. Symp.

Proc., Vol. 91, 473 (1987).

G. Bai, D.N. Jamieson, M-A. Nicolet, and T. Vreeland Jr., “Defects annealing of Si implanted

GaAs at room temperature and 100 C,” Mat. Res. Soc. Symp. Proc., Vol. 93, 67 (1987).

. Y.C, Kao, K.L. Wang, B.J. Wu, T.L. Lin, C.W. Nieh, D.N. Jamieson, and G. Bai, “Molecular
beam epitaxial growth of CoSig on porous Si,” Appl. Phys. Lett., 51, 1809 (1987).

. C.H. Chern, Y.C. Kao, C.W. Nieh, G. Bai, K.L. Wang, and M-A. Nicolet, “MBE growth of
SiGe on porous Si,” J. Electrochem Soc. 134, 543 (1987).

. G. Bai, D.N. Jamieson, M-A. Nicolet, T. Vreeland Jr., “Misoriented epitaxial growth of CoSig

on offset (111)Si substrates,” Mat. Res. Soc. Symp. Proc., Vol. 102, 259 (1988).

G. Bai, C.J. Tsai, A. Dommann, M-A. Nicolet, and T. Vreeland, Jr., “Characterization of semi-

conductors by MeV He backscattering spectrometry, channeling and double crystal diffraction,”

presented at TECHCON’S88, Oct. 12-14, 1988, Dallas, TX.

J. L. Tandon, J.H. Madok, 1.8. Leybovich, and G. Bai, “Damage removal and activation in

rapid-thermally-annealed Si implanted GaAs,” Mat. Res. Soc. Symp. Proc., Vol. 126, 207

(1988).

Y.J. Mii, R. Karunasiri, K.L. Wang, G. Bai, “Growth and characterizations of GaAs/AlGaAs

multiple quantum well structures on Si substrates for infrared detection,” J. Vac. Sci. Technol.

B7, 341 (1989).

G. Bai, M-A. Nicolet, T. Vreeland, Jr., Q. Ye, Y.C. Kao, K.L. Wang, “Thermal strain measure-

ments in epitaxial CoSi2/Si by double crystal x-ray diffraction,” Mat. Res. Soc. Symp. Proc.,

Vol. 130, 35 (1989).

. J.L. Tandon, J.H. Madok, I.S. Leybovich, G. Bai, and M-A. Nicolet, “Sequential nature of

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

20.

26.

27.

166

damage annealing and activation in implanted GaAs,” Appl. Phys. Lett., 54, 30 (1989).

G. Bai, M-A.Nicolet, T. Vreeland, Jr., Q.Ye, and K.L. Wang, “Strain in epitaxial CoSi, films on
Si(111) and inference for pseudomorphic growth,” Appl. Phys. Lett. 55, 1874 (1989).

J. L. Tandon, I.S. Leybovich, and G. Bai, “Activation analysis of rapid-thermally-annealed Si
and Mg implanted GaAs,” J. Vac. Sci. Technol. B7, 1090 (1989).

G. Bai, S.J. Kim, M-A. Nicolet, R.G. Sobers, J.W. Lee, M. Brelvi, P.M. Thomas, and D.P.
Wilt, “Thermal strain and photoluminnescence study of epitaxial InGaAsP films on InP,” Mat.
Res. Soc. Symp. Proc. Vol. 160, 171 (1990).

J.E. Mahan, K.M. Geib, G.Y. Robinson, R.G. Long, X.H. Yan, G. Bai, M-A. Nicolet, and
M. Nathan, “Epitaxial tendencies of ReSiz on (001)Si,” Appl. Phys. Lett. 56, 2439 (1990).

G. Bai, M-A. Nicolet, J.E. Mahan, K.M. Geib, “Backscattering and channeling study of epitax-

ial ReSi2/Si(111),” Phys. Rev. B41, 8603 (1990).

J.E. Mahan, K.M. Geib, G.Y. Robinson, R.G. Long, X.H. Yan, G. Bai, M-A. Nicolet, and
M. Nathan, “Large-area single-crystal films of semiconducting FeSi2,” Appl. Phys. Lett. 56,
2126 (1990).

Q. Ye, T.W. Kang, K.L. Wang, G. Bai, and M-A. Nicolet, “RHEED study of CoSi2/Si multilayer
structures,” Thin Solid Films, 184, 269 (1990).

G. Bai, M.A. Nicolet, J.E. Mahan, K.M. Geib, “Radiation damage in ReSig by a MeV *He
beam,” Appl. Phys. Lett. 57, 1657 (1990).

G. Bai, K.H. Kim, M-A. Nicolet, “Strain in porous Si layers formed on p*-Si(100) substrates,”
Appl. Phys. Lett. 57, 2247 (1990).

K.H. Kim, G. Bai, M-A. Nicolet, and A. Venezia, “Strain in porous Si with and without capping
layers,” J. Appl. Phys., Feb. 15, 1991.

G. Bai, M-A. Nicolet, “Defects production and annealing in self-implanted Si,” J. Appl. Phys.,
July 15, 1991.

K.H. Kim, G. Bai, M-A. Nicolet, “Amorphization and recrystallization of ion-implanted epitax-
ial ReSiz films on Si(100),” Appl. Phys. Lett., April 29, 1991.

J.E. Mahan, K.M. Geib, G.Y. Robinson, G. Bai, and M-A. Nicolet, “RHEED patterns of CrSio
films on Si(111),” J. Vac. Sci. Technol. B (in press).

G. Bai, M-A. Nicolet, T. Vreeland, Jr., “Elastic and thermal properties of CoSiz layers on
Si(100) substrates,” J. Appl. Phys. May 1, 1991.

K.M. Geib, J.E. Mahan, R.G. Long, G.Y. Robinson, M. Nathan, G. Bai, and M-A. Nicolet,

167

“Epitaxial orientation and morphology of 3-FeSiz on $i(100),” J. Appl. Phys. (in press).

28. J.E. Mahan, G. Bai, M-A. Nicolet, R.G. Long, and K.M. Geib, “Microstructure and morphology
of some epitaxial ReSiz films on $i(100),” submitted to Thin Solid Films.

29. G. Bai, M-A. Nicolet, “Defects production by /°F, ?8Si, #°Ar, and !5!Xe implantation into

Si(100) at room temperature,” submitted to J. Appl. Phys.

Papers in preparation

30. G. Bai. M-A. Nicolet, “Damage in implanted Si(100) at liquid nitrogen temperature.”

31. G. Bai, M-A. Nicolet, “Defects production and annealing in Si implanted CoSi2/Si(111).”

32. G. Bai, M-A. Nicolet, C.H. Chern, K.L. Wang, “Some properties of epitaxial GeSi alloys grown
on Si(100).”

33. G. Bai, M-A. Nicolet, “Misorientation in heterostructures.”

34. G. Bai, M-A. Nicolet, C.H. Chern, K.L. Wang, “Kinetics of strain relaxation of metastable GeSi
layers on Si(100).”

35. G. Bai, M-A. Nicolet, C.H. Chern, K.L. Wang, “Asymmetrical tilt boundary in epitaxial GeSi/Si
heterostructure.”

36. G. Bai, M-A. Nicolet, C.H. Chern, K.L. Wang, “Strain modification and relaxation of GeSi/Si

heterostructures by ion implantation and thermal annealing.”