View
2
Download
0
Category
Preview:
Citation preview
W&M ScholarWorks W&M ScholarWorks
Dissertations, Theses, and Masters Projects Theses, Dissertations, & Master Projects
2005
Vibrational lifetimes of hydrogen and oxygen defects in Vibrational lifetimes of hydrogen and oxygen defects in
semiconductors semiconductors
Baozhou Sun College of William & Mary - Arts & Sciences
Follow this and additional works at: https://scholarworks.wm.edu/etd
Part of the Condensed Matter Physics Commons, Materials Science and Engineering Commons, and
the Optics Commons
Recommended Citation Recommended Citation Sun, Baozhou, "Vibrational lifetimes of hydrogen and oxygen defects in semiconductors" (2005). Dissertations, Theses, and Masters Projects. Paper 1539623477. https://dx.doi.org/doi:10.21220/s2-27ah-f695
This Dissertation is brought to you for free and open access by the Theses, Dissertations, & Master Projects at W&M ScholarWorks. It has been accepted for inclusion in Dissertations, Theses, and Masters Projects by an authorized administrator of W&M ScholarWorks. For more information, please contact scholarworks@wm.edu.
VIBRATIONAL LIFETIMES OF HYDROGEN AND
OXYGEN DEFECTS IN SEMICONDUCTORS
A D issertation
Presented to
T he Faculty of the Department, of Applied Science
T he College of W illiam & M ary in Virginia
In P artia l Fulfillment
Of the Requirem ents for the Degree of
Doctor of Philosophy
by
Baozhou Sun
2005
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A P P R O V A L S H E E T
This dissertation is submitted in partial fulfillment o f
the requirements for the degree of
Doctor of Philosophy
(Baozhou Sun)
Approved, August 2005
Gunter Liipke (Advisor)
Robert L. Void
Anne C. Reilly (Physics)'
r M/lhtZJH fGwyn P. Williams (Jefferson Lab)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
To my parents and my wife.
iii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table o f C ontents
A cknow ledgem ents v iii
List o f Tables x
List o f F igures xvii
A bstract xviii
1 Introduction 2
1.1 Light, Im purities in S em ic o n d u c to rs ................................................................. 2
1.1.1 H y d r o g e n ................................................................................................... 4
1.1.2 Oxygen ....................................................................................................... 8
1.2 Local V ibrational M o d e s .................................................................................... 10
1.2.1 Local V ibrational Mode S p ec tro sco p y .................................................. 11
1.2.2 Dynamics of Local V ibrational M o d es ............................................... 14
1.3 O u tlin e ....................................................................................................................... 16
2 T heoretical C onsiderations 18
iv
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.1 V ibrational Relaxation and L in ew id th ............................................................ 18
2.2 M ultiplionon Relaxation and Energy G ap L a w ........................................... 21
2.3 Anharrnonicity ...................................................................................................... 25
2.4 Sym m etry of Accepting M o d es .......................................................................... 27
3 E xperim ental D eta ils 33
3.1 Fourier Transform Infrared S p e c tro sc o p y ..................................................... 33
3.2 Generation of Picosecond Mid-Infrared P u l s e s ............................................ 37
3.2.1 In tro d u c tio n ............................................................................................... 37
3.2.2 O ptical Param etric Amplifier ............................................................ 38
3.2.3 Free Electron L a s e r ................................................................................. 46
3.3 V ibrational Pum p-probe Spectroscopy ......................................................... 50
3.3.1 Experim ental S e t - u p .............................................................................. 50
3.3.2 Alignment of the Infrared B e a m s ..................................................... 51
3.3.3 Experim ental P ro c e d u re s ....................................................................... 52
3.4 Sample P r e p a r a t io n s ............................................................................................ 53
3.4.1 Hydrogen in S ilic o n ................................................................................. 53
3.4.2 Oxygen in Si and G e .............................................................................. 57
4 H ydrogen B end ing M odes in Sem iconductors 58
4.1 In tro d u c tio n ............................................................................................................. 58
4.2 E x p e r im e n ta l .......................................................................................................... 61
4.3 Results and Discussions ..................................................................................... 63
v
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.3.1 Lifetime M easurem ents of Bend Mode of Hj in Si and Ge . . . 63
4.3.2 Decay Channel of the Ht, Bend M o d e ........................................ 66
4.3.3 Lifetime; of O ther H-related Bend Modes and Energy G ap Law 70
4.4 C onclusion ................................................................................................................. 73
5 H ydrogen Stretch M odes in Silicon 74
5.1 In tro d u c tio n ............................................................................................................. 74
5.2 E x p e r im e n ta l .......................................................................................................... 78
5.3 R e s u l ts ........................................................................................................................ 80
5.3.1 V ibrational Lifetime of H j and H V -V H (no).............................. 80
5.3.2 Tem perature Dependence of Lifetime and Decay Channels . . 81
5.3.3 S tructural Dependence of Lifetime of Hydrogen Defects . . . . 85
5.4 C onclusion ................................................................................................................. 87
6 Interstitia l O xygen in Si and G e 88
6.1 In tro d u c tio n ............................................................................................................. 88
6.1.1 V ibrational Spectroscopy of In terstitia l Oxygen in Si and Ge . 90
6.1.2 Isotope Effect of Oxygen in S i ...................................................... 95
6.2 E x p e r im e n ta l .......................................................................................................... 96
6.3 Results and Discussions ..................................................................................... 97
6.3.1 Lifetime M easu rem en ts.................................................................... 97
6.3.2 M ultiphonon Density of S tates and Decay C h a n n e ls ............ 102
6.3.3 Sym m etry A n a ly s is ........................................................................... 106
vi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6.4 C o n clusion ................................................................................................................ 107
7 Sum m ary and Future W ork 109
7.1 Sum m ary ................................................................................................................ 109
7.2 Future W o rk ............................................................................................................ I l l
B ibliography 114
V ita 124
vii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ACKNOWLEDGEMENTS
It is my pleasure to acknowledge dozens of people who have given me generous help through my Ph.D . study. F irst and foremost I would like to express my deepest thanks to Professor G unter Liipke, my dissertation advisor, who has guided me w ith wisdom and patience throughout my graduate; career a t W illiam and Mary. His overwhelming knowledge and physical insight contributed invaluably to this work. I will always be grateful to him for his advice, support and guidance in th is project.
I m ust also thank all of our cedlaborators. I gratefully acknowledge the aelvice and stim ulating eiiscussions w ith Dr. Feldman, Dr. Tolk and Dr. Sunkaranam at Vanderbilt University. Not only eliel t,he;y prepare the proton-im planted sam ples forme, they also invited me to Vanelerbilt 011 two occasions for wonderful discussions. I also thank Dr. Stavola’s group at Lehigh University for providing us silicon samples with low-concentration im purities, as well as the d a ta of the linewidth m easurem ents of loe;al vibrational modes. Dr. Staveda has always been available anel helpful to me when I needed help. I have beimfited from m any insightful eiiscussions with Dr. Pajot. in Paris and Dr. Newman in Im perial Cedlege; Lonelon. P articu lar thanks go to Dr. P ajo t for provieliug silicon sam ples w ith erxygen isotopes. I have been fortunate in having so m any intelligent scientists to dismiss anel work w ith in various projects. W ithout these collaborators, it is impossible for me to finish this work.
I wish to thank Dr. Xinliui Zhang, whei taugh t me so much about aligning optics in the first year. I woulel alsei thank Dr. Qiguang Yang for his help in the nonlinear optics anel assistance in the lab. I have learned things of value; from the either group members and lab m ates inclueling Dr. Yuhang Ren, Haibin Zhao, Shuyang Zhang, Dr. D iyar Talbayev and Dr. A ndrea Luearelli. It is always a pleasure to work with them.
I would like to erxpress my appreciation to the; rest of my com m ittee meunbers: Dr. R obert L. Void of Applied Sciene;e, Dr. Anne Reilly of Physieis and Dr. Gwyn W illiams of Jefferson Lab. I especially appreciate their encouragement, during my study at W illiam and M ary anel their instructive; euimments on my thesis.
I also extend my gratitude to the; staff a t the: T JN A F FEL who have given their beam tim e anel equipm ents for our e;xpe;riments, anel in particular, Joe Gubeli and Dr. Michelle Shinn were always available when I neeeled their help. I want to thank Prof. Michael Kelley who let me use his FTIR. spectrom eter many times, anel th an k him for his encouragement, in my career. T hanks also go to Amy W ilkerson at Applieei Research Center who always offers me her support 011 any favor I asked.
I m ust thank my parents in C hina for their dedicate*! love anel affectionate encenir-
viii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
agement throughout my stay in the USA. I also want to express my sincere gratitude to the brothers and sisters in Peninsula Chinese', B ap tist Church for their love, support and prayers in .Jesus Christ.
No words can express how grateful I am to my wife, Min Huang for her sacrifices, understanding, and patience. I would never have finished this Ph.D .w ork w ithout your undying love and endless care throughout our stay a t W illiam and Mary. This dissertation is dedicated to her.
Lastly b u t most im portantly, I thank God for giving me strength and courage to endure and for leading me on this pa th and illum inating the way even when it is difficult to go through sometimes.
ix
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
List o f Tables
2.1 One example showing the sym m etry of the com bination modes by using
the direct product and sum ................................................................................. 29
2.2 High sym m etry points, phonons and irreducible representation in dia
mond s tru c tu re 31
3.1 Specification of the FEL IR-Denio a t the .Jefferson L ab ............................ 49
5.1 Lifetimes of H -rclated stretch modes in Si measured a t 10 K. To obtain
the lifetimes shown in column 3, the linewidths were corrected for
instrum ental broadening and then entered into equation 1\ = \/{2-kcY q). 86
6.1 Three-phonon processes in Si and Ge close to the frequency of the
asym m etric stretch m ode of in terstitia l oxygen. The; infrared activity
M is also listed ......................................................................................................... 107
x
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
List o f Figures
1.1 Point defects in crystalline solids........................................................................ 3
1.2 Hydrogen defects in sem iconductor.................................................................... 6
1.3 (a). Representative model for H2 in Si proposed by Zhou and Stavola
[23]. (b). S tructure of H2 defect in Si proposed by Holbech [22]. . . . 8
1.4 S tructure of in tersititial oxygen in Si and G e................................................ 9
1.5 Scheinatical description of infrared activity of local v ibrational modes. 12
2.1 Schematic description of the two common relaxation processes: Energy
relaxation and pure dephasing............................................................................ 19
3.1 Sketch of a FTIR. spectrom eter. It consists of a light source, a fixed
mirror, a movable mirror, a beam sp litter and a detec to r......................... 34
3.2 An illustration of obtaining a sample absorbance spectrum from an
interferogram: (a) reference spectrum , (b) sample spectrum , (e) ab
sorbance spec trum .................................................................................................. 36
3.3 Sketch of optical param etric generation........................................................... 39
xi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.4 The amplified T i:sapphire laser system consists of an oscillator M ira
900-F, diode-pum ped laser Verdi 5V, frequency-doubled Nd:YLF laser
DQ527 and T O !'A S.................................................................................................... 41
3.5 A utocorrelation trace of the compressed amplifier ou tpu t, recorded
from a BBO crystal. The solid curve is a sech2 w ith a FW HM of
2.5 ps............................................................................................................................. 42
3.6 The traveling-wave optical param etric amplifier TOPAS, the 800 mil
pum p beam is shown in grey, the signal+idler beam in black. All
m irrors are 800 mil high-reflective dielectric mirrors, except M3, M7
and M8 which are metallic mirrors. NC: nonlinear crystal (BBO); TD:
birefringent plate to generate delay between signal and idler; GP: glass
plate; B Sl and BS2: beam splitters................................................................... 43.
3.7 Tuning curves of the TOPAS at the College of W illiam and Mary.
S: Signal; I: Idler; DF1: Difference C rystal AgGaS,: DF2: Difference
Crystal G aSe................................................................................................................ 45
3.8 Schematic diagram of a free electron laser [75]................................................. 46
3.9 Typical set-up for pum p-probe experim ents...................................................... 51
3.10 Pum p-probe signal at room tem peratu re from a Ge sam ple.......................... 54
3.11 Spectrum of a 1.7 ps laser pulse passing through a monochrom ator.
The dip at 860 cm ” 1 corresponds to the absorption of in terstitia l oxy
gen in Ge doped by Co — ~ 1018 cm ”3.............................................................. 54
3.12 Hydrogen im plantation profile................................................................................ 55
xii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.1 Schem atic potential energy surface for a Si-H bending mode, showing
vibrational energy levels in the first energy well. W hen H is vibra-
tionally excited to the top of the well, it may be em itted to transporting
s tates and dissociate the bond ............................................................................ 59
4.2 Absorbance spectrum of proton-im planted Si showing the 817 cm -1
bending mode of HJj, shown in the inset.......................................................... 62
4.3 Decay of the transient bleaching signal Si, from the 817-cm” 1 mode of
H -2 m easured a t 10 K with the TOPAS. S*, decays exponentially w ith
a tim e constant of 7 \ = 12 ± 0.8 ps.................................................................. 63
4.4 Spectrum of a 1.7 ps laser pulse passing through a monochromator,
showing the' absorption of the HI, bending mode at 813 cm -1 at room
tem peratu re ............................................................................................................... 64
4.5 IR. absorbance1 spectrum of e-irradiated, H-doped Si, showing the lines
of Hj and D J............................................................................................................. 66
4.6 Tem perature dependence of 7 \ of the 817-cm” 1 mode of H*,. The
dashed and solid lines are the theoretical predictions for decay into
three phonons {517, 150, 150} and two phonons {460, 357} cm -1 , re
spectively.................................................................................................................... 67
4.7 Two-phonon density of s ta tes in Si. The? Ill, braiding mode at 817 cm -1
coincides with L A +T O phonons, whereas the Dlj bending m ode a t 588
cm ” 1 lies in the; T A +T O plionon ban d ............................................................ 69
xiii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.8 V ibrational lifetimes of H -rclated bending modes vs decay order in
different hosts. T in1 line is a fit to Eq. (4.3) w ith two parameter's,
5.1 S tructures of the in terstitia l-type defects (top), and the vacancy-type
complexes (bottom ). Large: spheres are Si, whereas small spheres are H. 75
5.2 Absorption lines of hydrogen stretch modes in proton-im planted Si: Hf;
gives rise> to two absorption lines a t 1838 (BC site) and 2062 cm _1(AB
site); IH2 is a self-interstitial defect and has two absorptions lines at
1987 and 1990 cm -1 ; V2H2 has two Si-H stretch modes, but only one
Si-H m ode is IR active at 2072 cm -1 ; VH consists of one hydrogen
atom satu ra ting a elangling bond of the vacaney, which was assigned
to be 2038 c u r 1; VH2 gives rise to two absorption lines at 2267 and
2316 cm "1................................................................................................................... 76
5.3 Absorbance spectrum of proton-im planted Si showing the 2062.1-, 2068-
5.4 Transient bleaching signal from the 2062-cm 1 (left side:) and the 2072-
A = 0.15 ps anel B = 2.01 71
, and 2072.5-cm"1 lines of H*, IIV-V. and HV-VH(n0) 78
cm 1 (right side) moele me'asmcel at 10 K 80
xiv
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.5 Tem perature dependence of T i of the 2062-crn"1 inode of H2 (left side)
and the 2072-cm"1 mode of HV-VHpio) (right, side). The solid lines
are fits from Eq. (2.9) in chapter 2 for decays into the sets of accepting
modes (165, 165, 165, 522, 522, 522} c u f 1 and {343, 343, 343, 521,
521} cm "1 of the 2062- and 2072-cm"1 modes. T he dashed lines are
fits of lowest-order decay channels..................................................................... 82
5.6 Phonon spectrum of Si adapted from Reference [113]...................... 83
6.1 S tructure of in terstitia l oxygen in Si and G e ................................... 90
6.2 Fundam ental vibrational modes of (a) the puckered Ge20 and (b) the*
linear Si20 quasimolecule. The trad itional molecular nom enclature has
been adopted for G e20 . For Si2() the nom enclature corresponds to the
D u point, group, appropriate for the linear s tru c tu re ................................. 91
6.3 Absorption spectrum of the asym m etric m ode of 160,; in Si measured
at 60 K showing the side bands on the lower energy side......................... 92
6.4 Absorption spectrum of th e asym m etric mode i/3 of O; in natu ra l Ge
at 6 K. T he spectrum shows the complex fine structu re arising from
different Ge isotopes and pairs of lines due to coupling between v3 mock1
and oxygen ro tational motion. Each peak is identified by the average
Ge isotopic mass. The average linewidth is ~ 0.04 cm "1.......................... 93
xv
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6.5 Delocalization of oxygon in Si:Oj: and G e:0 , in the plane perpendicular
to the Si-Si (Ge-Ge) axis. T he 0 ,; m otion is radial in Si and ro tational
in G e............................................................................................................................ 93
6.6 Absorption spectrum of the asym m etric stretch mode of 0,- for different
G isotopes in Si samples enriched w ith 17() and 18(). T he linewidth of
1'0 mode is 1.2 cm -1 , while it is 0.6 cm -1 for 16() and 18( ) ....................... 95
6.7 (a) T he transient bleaching signal St, decays exponentially with a time'
constant Ti = 11.5 ± 1 ps for the asym m etric stretch mode of 160.; in
Si a t 10 K; (b) while Ti = 4.5 ± 0.4 ps for 170.,............................................. 98
6.8 The transient, bleaching signal St, measured a t a center frequency of
1123 cm -1 , which lies between the 160 ,; and 170 , mode. The signal
does not show any decay. It is fitted by a Gaussian function, giving
the pulse w idth of ~ 1.7 ps.................................................................................. 99
6.9 T he laser pulse spectrum has a band w idth of 16 cm -1 at ~ 9 /urn
measured w ith a m onochrom ator....................................................................... 99
6.10 A bsorption spectrum of the asym m etric stretch mode of 16().; measured
in a fioat-zone Si sam ple with oxygen concentration of ~ 0.2 ppm. The
linewidth is 0.55 cm "1............................................................................................ 100
6.11 The transien t bleaching signal St, decays exponentially with a tim e
constant Ti = 125 ± 10 ps for the asym m etric stretch mock' of 16Oi in
Ge at 10 K ................................................................................................................. 101
xvi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6.12 Pliuiioii dispersion curves and density of states for Si (top) and Ge
(bottom ). Si figure is courtesy of Wei [138], Ge figure is courtesy of
Giannozzi et al [137].............................................................................................. 103
6.13 Normalized three-phonon density of state's in Si at 0 K. T he asym m etric
stretch m ode of 170 ,; coincides with the highest three-phonon density
in Si, whereas the l(’0 , and 1R0 , modes are on both sides of the 2TO
+ 1TA phonon peak ............................................................................................... 104
6.14 Normalized three-phonon density of sta tes in Ge a t 0 K. T he 160.;
and irO, modes lie in the 3TO phonon band, whereas the 18Oj mode
coincides w ith a 2TO + 1LO phonon process............................................... 105
7.1 S tructure of O-H bond in ZnO and absorption spectrum a t 3326 cm -1
[151].............................................................................................................................. 112
xvii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ABSTRACT
C haracterization of defect and im purity reactions, dissociation, and m igration in sem iconductors requires a detailed understanding of the rates and pathways of vibrational energy flow and of the coupling mechanisms between local modes and the phonon bath of the host m aterial. Inform ation 011 the inelastic microscopic interaction can be obtained by m easuring the lifetime of local v ibrational modes (LVMs). This dissertation presents vibrational lifetime m easurem ents of hydrogen and oxygen defects in sem iconductors by m eans of time-resolved infrared (IR) pum p-probe spectroscopy. In these experim ents, a picosecond IR pulse excites a v ibration of a significant fraction of point defects in a sample. This excitation causes a transient decrease; in the absorbance; of the sample, as there are fewer defects in the v ibrational ground s ta te th a t can absorb. T he transm ittance change; is measureel w ith a weaker probe pulse th a t passes through the sam ple after a variable; delay, see th a t the evolution of the excited vibration can be followed in time. This technique enables the direct time-elomain m easurenmnt of the; lifetime of LVMs.
First, we measureel the vibrational lifetime of H- anel D-relat.ed bending modes in Si and other semiconelue:tors. Time-resolve;d pump-probe; anel linewidth m easurem ents reveal th a t the lifetime of bcmeling moeles can be; explained by an energy gap law, i.e., th e decay tim e increases exponentially w ith increasing decay order.
Second, we present the v ibrational lifetime m easurenmnts of a selection of Si-H stretch modes in crystalline Si. The lifetime's of in terstitia l-type elefeets are found to be; a few picoseconds, whereas vacancy-type de;fe;e;ts have; lifetimes* up to 300 picoseconds. T he strong dependence of lifetime on the deflect s tructu re suggests th a t pseudolocalized modes are; involveel in the vibratiemal relaxation of the hydrogen stretch moeles in Si. It is found th a t the energy relaxation of Si-H stretch modes does not decay by lowest orelcr, i.e., low frequency mexles are involveel in the decay process.
Furtherm ore, we perform ed lifetime; m easurenmnts of in terstitia l oxygen in Si anel Ge. T he lifetime of 17()i in Si is half of 16( ) j and 180 j . A e;alculation of the three- phonon elensity of s ta tes shows th a t 170,; lies in the highest phonon density resulting in the shortest lifetime. T he lifetime; of the 160.,; mode in Ge is m easured to be 10 times longer than in Si. The; interaction be;tween the leecal moeles and the lattice vibrations is discussed accoreling to the; activity of phonon combination.
These studies elucidate the elynamies of energy dissipation anel vibrational decay channels of point defects in sem iconductors. The;y provide; a bet,te;r understanding of the dissociation of Si-H anel S i-0 bonds and the strong hydrogem and deuterium isotope effect found in H-passivated semiconeiuctor devices. The experim ental results provide an indispensable beneihmark for fu ture theoretk 'al investigations.
xviii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
VIBRATIONAL LIFETIMES OF HYDROGEN AND
OXYGEN DEFECTS IN SEMICONDUCTORS
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 1
Introduction
1.1 Light Impurities in Semiconductors
G roup IV element sem iconductors such as silicon and germanium wore identified as
the most promising m aterials in sem iconductor industry. Most electronic and optical
properties of these sem iconductors are determ ined by the type and concentration of
defects they contain. Many defects are present in as-grown sem iconductors. They
include intrinsic defects (self-interstitial and vacancies) and extrinsic defects or im
purities which come from the source' m aterials, the; ambient, the crucible, th e heating
element, etc. Defects m ay also be introduced during processing of the devices: E tch
ing injects vacancies as well as hydrogen, the growth of layers cause's self-interstitial
and vacancies in the bulk, ion plantation creates dam age [1]. Figure 1.1 shows a
simple structu re of intrinsic and extrinsic elefeeits in crystalline; solids.
T he adelition of these im purities to a sem iconductor will alt,e;r its v ibrational anel
electronic properties. For example, dopants can aeld charge carriers (electrons e~ to
the conduction band or holes h + to the valence' band), which will form an e~- h +
2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 1. INTRODUCTION 3
Substitutionalimpurity
lnterstiti<impurity
V a ca n cy
F igu re 1.1: Point, defects in crystalline solids.
recom bination center and reduce carrier lifetimes. In general, defects or im purities
break the periodicity of the crystal, in troduce local strain, and result in new energy
levels which are often in the gap. Depending on the Fermi level anti the position of
defect level, some defects may exist in several charge states. Furtherm ore, defects can
be mobile, leading to a wide: range of reactions between them . Such reactions create
new, more complicated, defects with different electronic and optical properties than
the original o i k :. Defect diffusion and reaction is enhanced by therm al treatm ents.
Since defects are very im portan t in semiconductors, it is crucial to understand
their fundam ental properties, including how they diffuse, in teract with the crystal,
w ith each other, and with other defects. T he most common im purities in semi
conductors are hydrogen and oxygen, which have been of interest for m any years
because of their simple atom ic structures. Experimentally, m any techniques are used
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 1. INTRODUCTION 4
to probe their microscopic properties. Fourier transform infra-red (FTIR.) absorption
and R am an spectroscopy are; used to study local vibrational modes (LVMs). Elec
tron param agnetic resonance (EPR ) is a technique to find unpaired electron spins.
Deep-level transien t spectroscopy (DLTS) is used to probe the gap for defect levels.
More macroscopic m ethods such as secondary-ion m ass spectroscopy (SIMS) also pro
vide crucial inform ation. Recently, time-resolved pum p-probe experim ents have been
perform ed to provide new inform ation abou t the dynam ics of these defects.
1.1.1 H ydrogen
Hydrogen is the sim plest element in the periodic tab le considering the atom ic struc
ture. However, hydrogen has very rich chemical properties and forms more m olecular
com pounds th an any other element [2],
Hydrogen is a very common im purity in semiconductors. It can be incorporated
during crystal growth or processing steps such as wafer polishing, wet chemical etch
ing, or reactive ion etching [3]. It is present a t virtually every stage of the processing
of devices. This makes it an unavoidable im purity even in devices fabricated w ith
the most perfect single crystals. T he solubility of hydrogen in sem iconductors is low.
For example, the solubility of hydrogen in intrinsic silicon is ~ 6 x 1015 cm ~3 ( ~ n r 2
ppm) at 1200°C [4] and ~ 106 c u r 3 (~ 2 x H R 12 ppm) at 300°C [4, 5], However,
the effective solubility a t m oderate tem peratures is determ ined by the concentration
of hydrogen traps, e.g., dopants, dangling bonds, etc., and exceeds the intrinsic value
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 1. INTRODUCTION 5
by several orders of m agnitude. Early h igh-tem perature perm eation d a ta showed th a t
hydrogen is a rapid diffuser in silicon w ith the diffusivity
D lt = (9 ± 4) x l ( r 3e x p [ - °-4 8 ^ (i ;Q6e1i/]cm2/ s (1.1)Kt,l
in tem peratu re range of 970-1200°C [4]. This high diffusivity is observed only at
high tem peratu re or w ith low im purity concentration, where trapping of hydrogen is
negligible. T he effective' diffusivity is typically several orders of m agnitude smaller
than the value given by Eq. (1.1). T he effect of hydrogen on a sem iconductor de
vice can be harmful, or in o ther cases, beneficially. Hydrogen may also be delib
erately introduced. For example, hydrogen annealing is performed as the last step
of polycrystalline-silieon solar-cell processing. This boosts the perform ance of these
cells by a substantial am ount [6]. Because hydrogen is so im portan t in semiconduc
to r industry, a large am ount of work has been done during the last three decades.
T he properties of hydrogen in sem iconductors have been discussed in m any recent
reviews [3, 6, 7, 8, 9].
W hen hydrogen is used to passivate deep and shallow defects in semiconductors,
it behaves as an am photeric im purity; th a t is, it acts as a donor (H+ ) in p-type
m aterial and an acceptor (H- ) in n-type m aterial. T he charge sta te determ ines the
most favorable location of hydrogen in the lattice. In the positive charge s ta te (H+)
the im purity is essentially a proton. Since the maximum negative? charge density
is found at the bond-center (BC) site in most covalent sem iconductors, H+ locates
a t this site [10]. In the negative charge s ta te (H “ ) hydrogen prefers regions of low
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 1. INTRODUCTION 6
electronic charge density, in which its distance! to the host atom s is maximized. So
H “ was found a t the te trahedral in terstitia l site. Based on the calculations by Van
De Walle [10], neutra l hydrogen was also found at the BC site, bu t it is less stable
than H+ a t this site.
Figure 1.2: Hydrogen defects in semiconductor.
Hydrogen can form strong covalent bonds with m any other elem ents in the periodic
table which makes it a very diverse impurity. Hydrogen in teracts w ith virtually any
im perfection in the crystal such as im purities, intrinsic delects, interfaces and surfaces.
Figure 1.2 shows a two-dimensional illustration of hydrogen defects in sem iconductor.
T he m ajor role of hydrogen is the passivation of both shallow and deep im purity
centers and other defects through the form ation of complexes. On the surface, H
forms covalent bonds w ith silicon dangling bonds [11]. In the bulk, H forms all
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 1. INTRODUCTION 7
kinds of interstitial, self-interstitial, vacancy-type defects or complexes w ith o ther
im purities [11]. At the interface, H passivation can reduce the density of Pb centers,
which is crucial in sem iconductor devices [12]. For example, hydrogen satu ra tes
dangling bonds at surfaces [13] and a t S i/S i()2 interfaces [14, 15], which generally
removes the electronic levels associated with the dangling bond from the band gap
and thus neutralizes its electrical activity. These H-related defects have been well
characterized [3, 16].
Hydrogen can also in teract with each other and form two types of hydrogen dim er
called H2 molecule and H |. T he H-2 molecule locates at a te trahedral site and its
existence in silicon was first predicted by C orbett at al [17] in 1983 and confirmed by
observation of an H-H vibrational mode with Ram an spectroscopy [18] and infrared
absorption spectroscopy [19]. Recently, it was proposed th a t the H2 molecule behaves
as a nearly free rotor [20]. A nother hydrogen dim er structu re has C3„ sym m etry
w ith one hydrogen close to the bond-center site and the o ther at an antibonding site.
T he existence of H?] has been predicted by C hang [21] and confirmed by Holbech [22].
Figure; 1.3 shows the structu re of these two dimers. The; H2 molecule in Si was
suggested to be oriented approxim ately along a (100) direction and has a sym m etry
of Ci point group [23, 24]. HJ] defect has C$v sym m etry with one hydrogen close to
the bond-center site and the o ther at an antibonding site.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 1. INTRODUCTION 8
r-̂ tij£l*
F igu re 1.3: (a). Representative model for H2 in Si proposed by Zhou and Stavola [23]. (b). Structure of H2 defect in Si proposed by Holbech [22].
1.1.2 O xygen
At present about 80 % of all the silicon wafers used for microelectronic circuit fab
rication are prepared by the C/,ochralski (CZ) m ethod or its modification. These
silicon wafers contain oxygen 011 the order of 1018 atom s/cm 3. Oxygen im purities
critically affect the properties and yield of electronic: devices because of the effects on
the mechanical and electrical properties of the wafers as well as on the lattice defects
incorporated. During the last four decades both academy and industry have devoted
a great deal of atten tion to the investigation of oxygen in silicon [25].
Oxygen is mainly introduced during crystal growth by the CZ pulling technique
from a silica crucible. By means of infrared absorption spectroscopy, Kaiser, Keek,
and Lange first confirmed th a t CZ silicon crystals contain oxygen as an im purity in
concentration on the orefer of m agnitude higher than usual eloping im purities [26],
They found th a t e«ygcn atom s incorporated into silicem elemiinantly occupy the in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 1. INTRODUCTION 9
terstitia l site, which lies midway between two neighboring silicon atom s along the
four equivalent (111) bond directions. T he structu re is shown in Figure 1.4. Oxygen
is located into an off-axis bond-centered position, giving rise; to a strong absorption
band at 1136 cm -1 . Early X-ray diffraction m easurem ents showed th a t the lattice
param eter a 0 is increased when oxygen is present [27]. Due to the high concentration
of oxygen in Si, annealing silicon can produce tin; precipitation of oxygen into vari
ous forms of silica (S i0 2) precipitates. This process is reversible and oxygen can be
dispersed again by dissolution of the precipitates at very high tem peratu re (~1300-
1350°C). A discussion of the relationship between the shape of the precipitates and
their infrared absorption has been given in Reference [28].
[Ill]
F igu re 1.4: Structure of intersititial oxygen in Si and Ge.
Group VI doping elem ents O, S, Se and Te are double donors in silicon [29]. This
electrical activity is explained by a substitu tional location of these elem ents in the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 1. INTRODUCTION 10
same way as the single donor electrical activ ity of P, As, or Sb. As-grown CZ silicon
displays n-type electrical activity related to oxygen [30]. This n-type behavior comes
indeed from a series of donors, which are called therm al donors (TDs). Oxygen could
act as a double donor on a substitu tional site. This location is unstable because of
the small te trahedral radius of the oxygen atom (0.68 A), bu t it has been argued
th a t TD s could be due to substitu tional oxygen surrounded by in terstitia l oxygen
to com pensate an inward distortion by an outw ard distortion [31]. T he to ta l oxygen
concentration of these TD s is typical less th an 1% of the to ta l oxygen concentration
in as-grown silicon. Tin; study of therm al donors is a very active area of oxygen in
semiconductors. Detailed inform ation can be found in soiik; reviews [25]. This thesis
will include the studies of v ibrational dynam ics and oxygen isotope' effects in Si and
Ge by time-resolved infrareel spectroscopy.
1.2 Local Vibrational M odes
In a crystalline semiconductor, atom s collectively oscillate about their equilibrium
positions, resulting in quantizeel vibrational modes, or phonons [32]. Phonons in
crystalline m aterials have a band of phonon frequencies anel can propagate through
the lattice'. In a perfect lattice, phonons have-, a well-defined frequency w and wave
vector q. The w vs q elispersion relation can be experim entally determ ined by neu
tron scattering [33]. Most of the phonon modes are Ram an active. W hen an im purity
is introeluceel, the translation sym m etry is broken and one eir more' new vibrational
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER !. INTRO D U CTIO N 11
modes may appear. If an im purity replaces a heavier host atom , its v ibrational fre
quency will be higher than the phonon frequency. Unlike the; phonon, the v ibrational
modes of point defects cannot, propagate in the crystal and remain localized near
the defects. Hence they are called local v ibrational modes or LVMs. T he physical
description of local v ibrational modes closely resembles th a t of molecular vibrations,
except th a t point defects do not have ro tational degrees of freedom. These defects
can also induce; a dipole mom ent which makers the vibrational moele infrareel active.
Much informatiem has been gaineel from infrared absorption stuelies of loe:al vibra
tional modes because they can generally be; obse;rved as sharp liirns on a monotonous
background.
1.2.1 Local V ibrational M ode Spectroscopy
Local v ibrational moele spectroscopy has become; an im portan t probe of defects in
solids and often provides inform ation abou t defect, strim tures anel properties th a t
cannot be obtained by other methoels. This techniepie has been applieel extensively
to study isolated im purities anel im purity complexes in crystalline sediels, including
ionic crystals [34] and sem iconductors [16].
A local vibrational moele; in teracts w ith light by absorption anei/or R am an scat
tering. At lew tem peratures, the: population of the: e:xoit,eel sta tes of leeeal v ibrational
modes is negligible. The local moele is infrareel active if the light-induced transition is
alloweel between vibrational states. T he transitiem probability is proportional to the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 1. INTRO D U CTIO N 12
square m atrix element of the electric dipole mom ent |e x /r // |2. Here e is a un it ventol
in the direction of the polarization of the light and m j is the m atrix element of the
electric dipole moment of the center between the initial and final v ibrational states.
For harmonic vibrations, a mode is infrared active if the motions of th e defect atom or
atom s give rise to a change in the dipole moment of the neuter. Consider the simple
example of a linear defect of the form X-Y-X. Their sym m etric vibrational m ode in
which the Y atom is stationary and the X atom s vibrate 180° out of phase is infrared
inactive because there is no net oscillating dipole moment (shown in Figure 1.5(a)).
T he antisym m etric mode in which the X atom s vibrate in phase with each other and
opposite to the direction of Y has an allowed elentrie dipole transition (shown in
Figure 1.5(b)). Furtherm ore, the vibrational mode will only be excited by incident
light th a t has a com ponent of its electric vector polarized along the direction of the
oscillating dipole moment. For the simple example of the antisym m etric stretching
mode of a linear defect molecule, tin; direction of the oscillating dipole m om ent is
along the axis of the defect molecule.
X Y X X Y X
Q - • O © — •< — — ► — ► < — — ►
(a) (b)
Figure 1.5: Schernatical description of infrared activity of local vibrational modes.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 1. INTRO DU CTION 13
T he integrated absorption coefficient for a defect is well known to be proportional
to the concentration of the defect. T he commonly used expression for the integrated
absorption strength for the local modes of an oscillating defect with mass m and an
effective charge q is given by
where N is the concentration of defects in cm -3 , and rj is the refractive index of the
host. This equation has been w ritten in CGS units to be consistent, with the absorp
tion coefficient, which is conventionally determ ined in units cm - 1 and the frequency
a given in wavcnumber units, cum 1. An infrared absorption experim ent typically
allows one to determ ine the intensity of the absorption line, i.e., the area under the
absorption peak. T he reduced mass p is given as [35]
where fximp and are the atom ic masses of im purity and host atom s, and x is
a param eter which represents the coupling of the v ibrating entity to the crystalline
host. T he x param eter depends on the structu re of the defect, bu t is presum able close
an infrared absorption line associated w ith a local v ibrational m ode depend on the
sam ple tem perature a t which the m easurem ent is carried out. The frequency shifts
to higher wavcnumber and the line becomes sharper with decreasing tem perature.
( 1 .2 )
P 1 “ Pimp + iXPho.u) \ (1.3)
to unity for hydrogen-related local modes in Si and Ge. The' frequency and w idth of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 1. INTRO D U CTIO N 14
R am an spectroscopy is com plem entary to infrared absorption spectroscopy and
can provide inform ation about v ibrational modes th a t are infrared inactive. For
R am an scattering m easurem ents of local v ibrational modes, a m onochrom atic laser
beam is incident on a sample. T he light is inelastically scattered by the excitation
of the crystal. M easurem ents are usually m ade a t low tem peratures and only Stokes
scattering processes th a t excite im purity vibrations have1 appreciable intensity [16].
In this case, the frequency of light lj„ scattered from local vibrational modes satisfies
the relationship
w s = u l — u i v m , (1-4)
where and u w m are the laser and local modes frequencies, respectively. Thus the
local modes can be detected from their Ram an shifts. Since Ram an scattering is a
second-order process in pertu rbation theory, it is subject to different selection rules
than electric-dipole allowed optical transitions. T he Ram an intensity is proportional
to \es x R x e /J2, where R is the second-order R am an tensor and eg and are unit
vectors in the polarization directions of the scattered and incident light.
1.2.2 D ynam ics o f Local V ibrational M odes
Most spectroscopy studies of LVMs in sem iconductors have; been carried out in the fre
quency domain, which probes the time-averaged optical response of the modes. The
interaction between local modes and phonons has been studied by tem peratu re depen
dence of linewidth and frequency shift. As tem peratu re increases, phonon modes are
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 1. INTRO D U CTIO N 15
therm ally populated. Two models have been used to understand the coupling between
the local modes and phonon bath [32]. Those studies are very limited to understand
the relaxation process of defects because inhomogeneous broadening may contribute
significantly to the observed absorption linewidth. T he energy relaxation tim e cannot
be derived from the line w idth of the linear absorption spectrum [30, 37]. Until now
very little was known about the dynam ics of local v ibrational inodes in crystalline
semiconductors. T he tim e scales and mechanisms for population and phase relaxation
upon excitation were still unknown. The dynam ics of local modes is responsible for
the flow of energy into and ou t of im purities and defect complexes. U nderstanding
the dynam ics of defects is im portan t for defect m igration and reaction. W hen defects
act as recom bination or scattering center in semiconductors, the electronic excitation
th a t is deposited a t the defects is then available to prom ote defect m igration and
reaction. Such inform ation is crucial since excited v ibrational s tates may be involved
in the dissociation of the bond between defects and the lattice [38]. T his investigation
is m ade possible only recently by the advances in tunable, u ltrafast, high-intensity
infrared light sources, including optical param etric amplifiers (OPAs) [39, 40] and
free-electron lasers (FELs) [41, 42]. T he tim e scale of energy relaxation and dynam
ics of the elastic and inelastic local interaction can be obtained directly by pum p-
probe transien t bleaching spectroscopy. This dissertation focuses on time-resolved
dynam ical studies of hydrogen and oxygen defects in silicon and germanium.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 1. INTRO D U CTIO N 16
1.3 Outline
In C hapter 2, we discuss the im portan t theories for vibrational relaxation dynam ics
of local vibrational modes. We discuss the relationship between vibrational lifetime
and linewidth, multiphonon relaxation, energy gap law, and symmetry of accepting
modes.
In C hapter 3, we describe the1 experimental techniques including FTIR and pum p-
probe spectroscopy, together w ith the method for generating short m id-infrared pulses.
In C hapter 3, we also give a brief introduction for doing experiments w ith the Free
Electron Laser a t Jefferson Lab.
In C hapter 4, we present the first study of the vibrational lifetime of hydrogen
bending modes in semiconductors. Time-resolved pum p-probe and linewidth m ea
surem ents reveal th a t the lifetime of hydrogen bend modes can be explained by an
energy gap law, i.e., the decay tim e increases exponentially w ith increasing decay
order. This study provides a b e tte r understanding of the dissociation of Si-H bonds
and the strong hydrogen and deuterium isotope effect found in H-passivated devices.
In C hapter 5, we present measurements of the vibrational lifetime of H-related
stretch modes in crystalline silicon. T in 1 vibrational lifetimes are found to be strongly
structu re dependent. T he interstitial-type defect. HJ; has a lifetime of 4.2 ps a t 1 0 K,
whereas the lifetime of the vacancy-type complex HV-VH(no) is 2 orders of m agnitude
longer (~ 270 ps). T he decay channel and relaxation dynam ics arc1 also discussed in
C hapter 5.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 1. INTRO D U CTIO N 17
In C hapter 6 , we focus on the study of the lifetime of in terstitia l oxygen and its
isotope effects in silicon and germanium. The linewidth of the 1 7 Oi m ode in Si is two
times broader than tin; corresponding 16(),; and 18Ot- mode. This difference is also
observed by time-resolved pum p-probe spectroscopy. T he lifetimes of in terstitia l 160 j
and 170 i isotopes in Si are measured a t 10 K to be 11.5 and 4.5 ps, respectively. We
also calculate the three-phonon density of s ta tes and found th a t the 170 m ode lies
in the highest phonon density resulting in the shortest lifetime, which explains the
isotope effect. T he lifetime of the 16(),; m ode in Ge was m easured to be 125 ps, i.e.,
10 tim es longer than in Si. T he interaction between the local inodes and the lattice
vibrations is discussed according to the activity of phonon combinations.
In C hapter 7, the last chapter of this dissertation, we give1 a sum m ary and outlook
for fu ture experim ents.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 2
T heoretical C onsiderations
In this chapter, wc present the theoretical description of the v ibrational dynam ics of
local v ibrational inodes. F irst, the relationship between vibrational relaxation and
linewidth is described. A short introduction to the existing theory of m ultiphonon
relaxation process is given. We will also discuss the im portant factors which deter
mine the v ibrational lifetime of LVMs, including energy gap law, anharm onicity, and
sym m etry of accepting modes.
2.1 Vibrational Relaxation and Linewidth
R elaxation processes are usually divided into two classes. F igure 2.1 schematically
describes the two relaxation processes. The first class is called population relaxation,
which changes the quantum num ber of the LVM, thus causing the states to have
finite lifetimes. Tf characterizes th e tim e scale of energy relaxation of the first excited
state. In the low tem perature lim it the homogeneous absorption linewidth of a LVM
is determ ined by the uncertainty relation. Given the excited s ta te lifetime 7 \, the
18
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 2. TH EO RETICAL CONSIDERATIONS 19
natura l linewidth is
r = 1
27TcTi( 2 .1 )
Population relaxation
-|2)
-1 0
1r ,JO)
Characteristic time: 7",
P h ase relaxation
tftV af t t T t
/fVf t
Characteristic time: 7V
F igure 2.1: Schematic description of the two common relaxation processes: Energy relaxation and pure dcphasing.
The second class of relaxation processes are caused by interactions th a t conserve
the quantum num ber of the LVM, bu t cause a loss of phase coherence. These are
elastic scattering processes which do not contribute to population or energy relax
ation. This process is also called pure dephasing. The lifetime of pure dephasing is
characterized by T2*. The to tal dephasing tim e T2 is determ ined by both the pure
dephasing processes and the energy decay time. T he homogeneous linewidth is given
bv
r = 1 1(2 .2 )
2ncTi 7rcT |
Since these processes involve exponential decay, the corresponding absorption line
shapes are Lorentzian, which is the Fourier transform of an exponential function. In
the low tem perature limit, the linewidth caused by pure dephasing can be neglected.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 2. TH EO RETIC AL CONSIDERATIONS 20
At higher tem peratures, elastic, phonon scattering becomes therm ally activated which
results in homogeneous dephasing broadening of the absorption line.
In many cases, the line shape is dom inated by inhomogenous broadening. In an
am orphous host or a crystal solid with high concentration of im purities, each im purity
may experience a slightly different local environm ent. This can affect the v ibrational
frequency of the LVM through slightly different interactions with the surrounding host
atoms. This leads to additional broadening of a LVM. This broadening is referred
to as inhomogeneous broadening, which is always occurring in surface? vibrations and
am orphous host. For example, the lifetimes of Si-H stretch modes were' believed to be
in the nanosecemel range based on direct m easurem ents of the lifetime of Si-H bonds
on H -term inated surfaces [43]. Sue:h lifetimes correspond to natura l linewidths on the
order of 0.005 cru"1, inueih less than the? wielths usually eebserved feu LVM’s in seuni-
coneluctors. Knowledge of the linewidth is not sufficient to understanel the elynamies
of LVMs in semicemductors. Howeveu, by decreasing the e'emcentratiem of H im puri
ties in a crystalline Si sample, Budele e.t al founel th a t the? inhomejgemoous linewidth
can be elim inated or neglected when the concentration of im purities is rethmed down
to 1 ppm [44]. Based on Bueleic’s observation, a t low tem perature and low defeed,
concentration the linewidth of a LVM m easured in a crystalline sem iconductor is a
good estim ate of its v ibrational lifetime'.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 2. TH EO RETIC AL CONSIDERATIONS 21
2.2 M ultiphonon Relaxation and Energy Gap Law
The theoretical description of v ibrational energy relaxation can be very complicated.
The vibrationally or electronically excited im purities in a crystal can, in principle,
decay into photons, electronic degrees of freedom, or other vibrational modes. The
radiative lifetime can be estim ated to be of the order of milliseconds [45], which rules
out radiative decay as the dom inating mechanism. Electronic decay can also be ruled
out in undoped Si and Ge, because the H- or O -related defects studied here cannot in
troduce shallow electronic levels in the bandgap. Consequently, electronic transitions
require more energy th an th a t of local v ibrational modes. It is most likely th a t the
LVM decays into v ibrational modes of the system consisting of phonon b a th or o ther
low frequency modes. If the vibrational relaxation process involves transferring energy
directly to the phonons in a single step, and the am ount of energy transferred exceeds
by m any tim es th e m aximum energy of a single phonon, then m any phonons will have
to be created simultaneously. This process is know as multiphonon relaxation [46].
One m ight reasonably imagine th a t the ra te of such a high-order process would be
quite small. Therefore, it is rem arkable th a t even if the num ber of phonons em itted
is as high as ten or more, m ultiphonon processes typically com pete successfully w ith
radiative decay, and are in fact often the dom inant relaxation mechanism.
Given the wide use; and the general im portance of m ultiphonon relaxation, a theory
of relaxation rates based on a microscopic Hamiltoniarn is clearly desirable. Exist
ing theories involve one of two approaches: T he adiabatic (Born-Oppenheim er) or
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 2. TH EO RETIC AL CONSIDERATIONS 22
’’Static-coupling” (Crude B oni-O ppenheim er) m ethods. The first m ethod is ususallv
invoked when discussing relaxation between electronic states, but can be applied
to high-frequency vibrations as well. T he coordinates are divided into fast (elec
tronic or high-frequency vibration) and slow (phonons) components. One neglects
for the moment the kinetic energy of the phonons, and finds the eigenstates of
the H am iltonian for fixed phonon coordinates. T he eigenvalues of th is procedure
generate the usual adiabatic potential surfaces, and transitions between these! sur
faces are due to the ”non-adiabatie” coupling (phonon kinetic energy) term in the
full Ham iltonian. Even when this 11011-adiabatie coupling is taken to lowest order
in perturbation theory, m ultiphonon transitions emerges. This was the approach
pioneered by K ubo anel Toyozawa [47], Perlin [48], Miyakawa anel D exter [49],anel
others [50, 51, 52, 53, 54, 55],
The second approach goes by several different names but its essence is as follows:
C ue assumes th a t the H am iltonian can be1 w ritten as
H = H s + H n + H , . (2.3)
H s is the H am iltonian for the LVM (twee-level system ) of the impurity,
H s = huxAa, (2.4)
where uj is the v ibrational frexpmimy, ad anel a are the e:reatiou and annihilation
operator, respectively. Hu is the H am iltonian of the bath (phonons), which is giveai
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 2. TH EO RETIC AL CONSIDERATIONS 23
H ,t = J 2 hujh<fh- (2.5)k
where W and b are the phonon creation and annihilation operator, respectively. In
the most common approach, Hj is strictly off-diagonal in the two-level basis of the
local vibrational m ode [46, 56, 57], T he off-diagonal m atrix elements of Hj are ex
panded to high order in th e phonon coordinates, and are treated by lowest-order
tim e-dependent pertu rbation theory. T he interaction H am iltonian is then given by
independent coupling strength of the channel {u}. {u} is characterized by the set
The contribution of the decay ra te due to H j is given by the standard result of first-
order tim e-depended perturbation theory [58]. T he to tal decay ra te (inverse lifetime)
is given as the sum of the rates of all of the decay channels [46, 56, 57]:
( 2 .6 )
M
where B{v) = n ^ Z ^ p ,} , and Gq„} denotes the tem perature-
W2</\ . . . , of accepting mode frequencies. Energy is conserved in the
decay process and the frequency sum is
(2.7)
( 2 .8)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 2. TH EO RETIC AL CONSIDERATIONS 24
T he function n{„} describes the tem perature-dependent population of the receiving
modes:
exp(jMp) - 1
n M = T w • (2-9)1 1 ?=!i e x p i i f r ) ~ 1]
T he com pound spectral density of accepting states, pp,}, can be expressed in term s
of a convolution of single spectral densities of states:
P i O = J duJi ] ■■■ J • • • P [n M n ] ) - (2 . iO)
In the low-tem perature limit (K/,T < , the decay ra te reflects spontaneous
decay into N„ accepting modes and n{„} = 1. At higher tem peratures the decay ra te
increases due to stim ulated emission in a fashion determ ined by the frequencies of the
accepting modes [46, 56, 57].
The general expression for the v ibrational relaxation ra te Eqs. (2.8) and (2.9) con
sists in general of a large num ber of term s corresponding to higher order m ultiphonon
processes. T he relaxation ra te is determined by the m agnitude of the coupling term s
between local v ibrational modes and accepting modes. It is reasonable to assume th a t
the |( j { „ } |2 term s decrease fast w ith increasing order of the m ultiphonon process AT^j.
As G{v] is expected to exhibit a very strong dependence on iVp,}, the relaxation ra te
can be approxim ately expressed by the dom inant contribution. T he largest term in
Eq. (2.8) is determ ined by the collection of phonon states {i/} = 1, 2, • • • , N{„), where
./V{„} is the smallest number of phonons th a t can result in a vibrational relaxation
process subjected to the energy conservation [57]. The approximate expression for
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 2. TH EO RETICAL CONSIDERATIONS 25
the vibrational relaxation ra te is now
7 - 7t|G >}|2PM n M- ( 2 .1 1 )
In very good approxim ation [57], the coupling constant can be expressed as G{v) ~
CSN , where C is a constant, 0 < d <C 1 , and N is the decay order which is determ ined
by the energy gap between local v ibrational modes and phonon modes. Such a relation
together w ith Eq. (2.11) implies th a t 7 oc S2N (a t constant T). A phenomenological
approach to the energy-gap dependence of relaxation ra te 7 has been developed in
Reference [46, 59, 60]. It is generally believed th a t the relaxation ra te decreases
exponentially as the energy gap is increased. The; energy gap law can be expressed
by
where A and B are two param eters which depend on the coupling constants between
LVM and accepting modes.
2.3 Anharm onicity
T he decay of LVMs into o ther modes is an anharm onic process. So the anharm onicity
of the defect po ten tial will play an im portan t role in the vibrational lifetime of LVMs.
T he im purities do not reside in a perfectly parabolic potential, ra ther, the potential
becomes weaker for larger displacements. The: interaction between the defect and the
Ti = A e x p (B ■ N ) ( 2 .1 2 )
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 2. TH EO RETICAL CONSIDERATIONS
lattice can be described by a Morse potential [61], winch is given by
26
V ( z - zc) = D t.\exp[—2a(z - zc)] - 2exp[a(z — zK)]}, (2-13)
where ze is the equilibrium value of the v ibrational coordinate z, and D r is the binding
energy. T he Schrddinger equation can be solved exactly for the Morse potential to
give the vibrational frequencies
uj{n) = ojc(n + i ) - uj,Xc(n + ^ ) 2, (2.14)
where u>(, = a (h D (,/ircn)1/2, u>cXe = hoc2 /(Akc/ji) and /x is the reduced mass. The
observed fundam ental frequency is given by uj — luc — 2a>*x«- The second harmonic
frequency is given by u 2nd = 2 iot. — 6 u)v.Xe- So the anharom inicity param eter can be
given by
2 uq — i0 2nd / o i r \X, : “ 2 ( 3 ( ' ]
T he anharm onicity can be estim ated from the overtone and fundam ental v ibrational
frequencies. This simple model has been used to explain the difference in the lifetime
of VH4 and VD 4 complexes qualitatively [62]. The D-related stretch mode in VD4 has
sm aller v ibrational am plitude, which gives rise; to smaller anharm onicity and longer
v ibrational lifetime th an VH4 s tretch mode in Si.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 2. TH EO RETICAL CONSIDERATIONS 27
2.4 Sym m etry of Accepting M odes
In the discussion of the decay process of LVMs, there can be no coupling between
v ibrational inodes belonging to different irreducible representations [63], T h a t is, to
ensure nonzero anhannonic coupling coefficients, the sym m etry representation of the
high-frequency local mode is required to contain the same irreducible representation
of the exchange m ode or com bination of low-frequency modes. More generally, the
Kroneckcr product of the two irreducible representations lnus to contain the to tally
sym m etric representation of the space group [64], So a full-scale space group analysis
of the sym m etry of local modes and the accepting modes is necessary to understand
the coupling between them . G roup theory is often employed as a m ethod to deter
mine the sym m etry of the norm al modes and their optical activity. T he first, step is
to establish the sym m etry of the modes such as rotations, reflections and inversions,
which leave the system invariant. For a perfect crystal system which has transla
tion symmetry, space groups m ust be considered [65]. The sym m etry of the norm al
modes can be classified into different groups, which is characterized by character ta
bles. Space group theory allows classification of the local modes and the delocalized
states in crystalline solids (electrons, phonon, etc.) w ith respect to irreducible repre
sentations. T he in troduction to group theory is given in the books by C ardona [6 6 ]
and Kuzm any [67]. This section focuses on the selection rules of the decay of LVMs.
Q uantum -m echanical selection rules determ ine the transition probability for a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 2. TH EO RETICAL CONSIDERATIONS 28
system changing from a state; i to / . T he relevant quantity is the m atrix element
where r<./ are the sym m etry representations of the initial and final state, respectively,
i/q,/ are the wave; functions for the initial and final state;, respectively. Pp is the
sym m etry representation of the operator driving the transition, p is the m om entum
operator. T he selection rules can be; generalized by expressing the; direct product
T; <S> r p <8 > Ty as a direct sum of irreducible represemtatious [6 6 ]
To obtain the irreducible representations of P /s , we; first m ultiply the; characters of r*,
Y'p and Tf column by column. Then, we find a set, of represemtatious which add up to
the products. If one; of the; representations is T i, the totally sym m etric representation,
the integral may be nonzero anel the transition is allowed.
In the relaxation process of LVMs, sta tes e;an interact via a m atrix element,
( r k \SH\ri), where SH is a pertu rbation due to the surrounding atom s and has sym
m etry P i. T he direct product, is given by
States only interact if they have the same sym m etry component. For example, Hsu
et al [6 8 ] discussed the interaction between th e antisym m etric stretch inode anel the
(2.16)
r ; © r ?, 0 P / — Pa ® Pti © r<: (2.17)
r f c ® r i « > r / = rfc<8>r/. (2.18)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 2. TH EO RETICAL CONSIDERATIONS 29
resonant, inode of oxygen in Si. T he in tersititial oxygen in Si belongs to D u point,
group. Table 2.1 lists the character table for this defect. The stretch m ode and the
resonant inode have A%u and E g symmetries, respectively. They cannot in teract w ith
each other because they do not have th e same symmetry. From Table 2.1, we know
the second harm onics (n = 2) of the resonant mode (E g <g> E g) have A \ g, A 2g and E g
sym m etries, none of which can interact with the stretch modi' (A*,). However, the
com bination of an E g and E u mock; is represented by E g ® E u = A lu © A 2u ® E u,
which contains an A 2u com ponent. So the stre tch m ode (A 2u) can interact, w ith the
com bination modi; of E g and E u modes.
D u E 2 C 3 3C2 i 2 S6
A i fl 1 1 1 1 1 1
A 2g 1 1 -1 1 1 -1
E a 2 -1 0 2 -1 0 )Alu 1 1 1 -1 -1 -1
A'2u 1 1 -1 -1 -1 1
E u 2 -1 0 - 2 1 0
Ey <8 > Eg 4 1 0 4 1 0Alg © A 2y © Ey 4 1 0 4 1 0
Eg 0 Ey 4 1 0 -4 -1 0Aiu ® A 2u © E u 4 1 0 -4 -1 0
Table 2.1: One example showing the symmetry of the combination inodes by using the direct product and sum.
If the accepting modes are propagating phonons in a perfect, crystal, they have
translation symmetry. Space groups m ust be considered [65]. C alculating the sym
m etry of phonon com binations is com plicated because the phonons can occur a t dif
ferent sym m etry points, and thus have different sym m etry point groups. For example,
the in terstitia l oxygen defect has the same sym m etry as the L-point, which has D:id
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 2. TH EO RETICAL CONSIDERATIONS 30
symmetry. One of the allowed phonon decay channels is T O (L ) + T A ( L ) + T O ( X ) .
Two of the phonons are at the L-point, bu t one is a t the X -point, and therefore has
different symmetry. In general, the sym m etry of a three-phonon com bination mode
consisting of phonons with wave vectors *ki, *k -2 and *k3 is the direct: p roduct of their
irreducible representations:
*ki 0 *k -2 0 */u3 - (2.19)
T he reduction of the direct, product of different, irreducible representations of the
space group into a direct sum of space group irreducible representations is thus the
central problem. The coefficients of the irreducible representations in the direct sum
have been calculated and tabu lated by Birm an [69]. Oik: uses the point, group of
the highest sym m etry point since all of the o ther point groups will be its subgroups.
T he Oh group is the point group for the F point in the diam ond crystal structure.
For tlie decay of a LVM w ith infrared activity, the direct sum of the com bibation of
accepting modes must, contain the p (15~) irreducible representation in the 0 /t point
group. r<15-) is the representation of the sym m etry of the infrared dipole operator
in diam ond structure. It, is irreducible in O/, point group, but, can be reduced to
A 2u ® E fj in Dzd point group since D 3 4 at, L-point, is the subgroup of 0 /( group a t F
point, in the diam ond structure. F^15-) and A2u are expressed by different notations.
A 2u and Eg notations are often used in molecular physics. F plus a superscript, i to
lable these irreducible representations is common in solid s ta te physics.
Two- and three-phonon decay channels and their activities w en1 also calculated by
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 2. TH EO RETIC AL CONSIDERATIONS 31
C ritical Point Phonon Symm etry Critical Point Phonon Symm etry
r O(P) p(25+) L T O( L ) L(3-)
X T O ( X ) X W LO (L) L(i+)L ( X ) X d) LA{L) L (2©
T A { X ) X(3) T A ( L ) L( 3+)
Table 2.2: High symmetry points, phonons and irreducible representation in diamond structure.
B irm an [70], bu t an example is included here to apply this principle to the analysis
of the sym m etry of m ultiphonon decay channels. T he critical sym m etry points, the
phonons a t those points and their irreducible representations are listed in Table 2.2.
Let us consider the sym m etry and activity of the phonon combination T O ( L ) +
T O ( X ) + T A ( L ) , which is one of the decay channels of the in terstitia l oxygen stretch
m ode in Si. T he calculation of the direct sum is as follows:
T O ( L ) + T A ( L ) + T O ( X ) -> L(3“ } 0 L (3+) 0 X (4>
= (r<2-> © r (12" } © 2 ( r (25“ ) © x (1) © x (2) © x (3) © x (4))) 0 x (4)
= i 2 ( x (1) © x (2) © x (3)) © n x (4) © 2 ( r (1+) © r (1” } © r (2+) © r ^ )
©4(p(12+) © r (12-)) © 6( r (15+) © r (15_) © r (25+) © r (25-) (2.20)
The above three-phonon com bination contains r^15-) with a m ultiplicity or activity
of 6 . So it is an allowed decay channel for in terstitia l oxygen in Si. The detailed
theory is given in Ref. [70].
The level of theoretical investigation of the lifetime of local vibrational inodes is
less advanced. To our knowledge, first-principle calculations of v ibrational lifetimes of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 2. TH EO RETIC AL CONSIDERATIONS 32
point defects in crystalline sem iconductors are not available until now. T he theoretical
considerations presented in this chapter are used as a framework for discussing the
v ibrational lifetimes of local vibrational modes in semiconductors.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 3
E xperim ental D etails
In this chapter, the techniques involved in perform ing these experim ents are described
in detail. We begin w ith a discussion of Fourier transform infrared (FTIR.) spec
troscopy, followed by a description of the use of the Jefferson Lab Free Electron
Laser and optical param etric amplifier. We also discuss the vibrational pum p-probe
spectroscopy in detail in this chapter. A description of sam ple preparation for each
experim ent concludes this chapter.
3.1 Fourier Transform Infrared Spectroscopy
Local v ibrational modes of an im purity can be excited by infrared light of the proper
frequency. In infrared absorption spectroscopy (IRAS) o ik ; measures the attenuation
of the beam intensity as a function of frequency caused by the penetration of in
frared light through a particular sample:. T he use of FTIR. spectrom eter allows high
resolution IRAS m easurements.
Figure 3.1 shows the schematic diagram of a FTIR. spectrom eter. T he main eom-
33
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 34
Movable Mirror
SampleDetector
Fixed Mirror
oBroadband Source
F igu re 3.1: Sketch of a FTIR spectrometer. It consists of a light sourer', a fixed mirror, a movable mirror, a beam splitter and a detector.
portent, of a F T IR spectrom eter is a Miehelson interferom eter. T he interferom eter
consists of a beam sp litter, a fixed m irror and a movable' mirror. Collim ated light
is directed to a sem itransparent beam splitter, which is designed to transm it half of
the light to the movable m irror and half of the light to the fixed mirror. T he two
beam s will reflect off their respective m irrors and recombine at the beam splitter. One
beam will travel through the sample', where light is absorbed because of electronic or
vibrational excitations. Finally they enter a detector, where the intensity integrated
over all frequencies is measured.
The beam sp litter splits the beam into two parts of electrom agnetic waves moving
towards the detector, which have equal electric field am plitudes E 0 (a) for a given
wave num ber a. T he electric field a t the detector is the sum of the electric fields of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 35
the two waves:
E( o) = E 0 (a ){eilwt-'27rrTZl] + } (3.1)
where z \ and z -2 are the optical pa th lengths from the source to the detector for the
waves reflected from tin ' movable and fixed m irror, respectively. T he contribution to
the light intensity at the detector from waves w ith a particular wave num ber o (in
CGS unit)is given by [71]
where z = z\ — Z2 is the difference in optical path length of the two waves. W hen
the movable m irror and fixed m irror are at the same distance from the beam splitter,
this condition is called zero path difference. T he detector measures the maximum
intensity at this position of the movable1 m irror. A plot of light intensity versus z is
called interferogram. W hen a broadband infrared source w ith many wavelengths is
used, the to tal interferogram m easured by the detector is proportional to the intensity
given in Eq. (3.2) integrated over all wave numbers:
where C is a constant, which depends on the design of the detector. T he first term in
Eq. (3.3) is a constant representing the to ta l energy in the spectrum . This constant
is sub tracted from the interferogram before the Fourier transform is performed. The
= ^ - |£ ( c r ) | 2 = E (a j 2 { 1 + co*[2 n a z}} , (3.2)
I ( x ) — f I(o . x )dx = ̂ ̂+ C f — E 0 2 (o) cos[2i\oz]do, (3-3)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 36
spectrum , B( < r ) can he obtained by calculating the inverse cosine Fourier transform
of the interferogram:
r 2 I 00B( a ) = j ^ o V ) = c j Q “ 7 ( ° ) /2] ™&$™z]dz. (3.4)
The limits of this integral should extend from 0 to plus infinity. In reality, the
movable m irror has finite moving range (0 to L). The1 interferogram is therefore
truncated at some maximum path difference zmax. T his truncation function is often
called a boxcar truncation function.
E3
O®CL
if)OOc4>OO 0C
E3
OoQ_
CO©Q.E(0
CO
©oc<0Aom<
0 500 1000 1500 2000 2500 3000 3500 4000 4500
Frequency (c m 1)
F igu re 3.2: An illustration of obtaining a sample absorbance spectrum from an interferogram: (a) reference spectrum, (b) sample spectrum, (c) absorbance spectrum.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 37
In a F T IR experim ent a background spectrum m ust be m easured as a reference.
As shown in Figure 3.2 (a), the background spectrum contains contribution from
the instrum ent and the environment. Next, an interferogram is m easured with a
sam ple containing defects and a raw spectrum is produced as shown in Figure 3.2
(b). It looks sim ilar to the background spectrum except th a t additional dips are
superim posed upon the background spectrum . The line observed around 1136 cm -1
corresponds to in terstitia l oxygen in Si, which will be discussed in detail in chapter
6. To reduce the atm ospheric contributions to the spectrum , the raw spectrum must,
be divided by the background spectrum (See Figure 3.2 (c)). This will produce an
absorbance spectrum . We use FTIR. spectroscopy to check the absorption line before
the lifetime m easurem ent by the pum p-probe experim ent. This ensures th a t enough
defects are present in the sample. F T IR spectroscopy is also used to m easure the
natu ral linewidth of local v ibrational modes.
3.2 Generation of Picosecond Mid-Infrared Pulses
3.2.1 In troduction
Time-resolved spectroscopy of v ibrational transitions requires u ltrashort pulses th a t
are tunable in the m id-infrared region of the electrom agnetic spectrum . T he pulses
have to be intense since: local vibrational mode transitions are hard to sa tu ra te due to
their small absorption cross sections. T he dynam ics of vibrational transitions occurs
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 38
oil a picosecond tim e scale, which requires u ltrafast laser pulses.
Continuously tunable pulses in the m id-infrared can be generated by different
lasers and techniques, such as free-eleetron lasers [41, 42], color-center lasers, differenee-
frequeney mixing [72, 73], and optical param etric generation and amplification. The
light sources used in this dissertation are the; Free Electron Laser a t Thom as Jefferson
N ational Accelerator Facility (T.JNAF) and an optical parametric: amplifier (OPA).
FELs do not use atom s or molecules as an active medium, but ra ther ’’free” electrons
in a specially designed magnetic: field. A brief description of the use of the FEL at
Jeffeson Lab will be presented in the: Section 2.2.3.
3.2.2 O ptical Param etric Am plifier
Due to their availabilities, optical parametric: devices comprise the vast m ajority of
IR sources. OPA is based on a nonlinear interaction in which two light waves of
frequencies and cu2 are amplified in a medium which is irradiated w ith an intense
pum p wave of frequency o>3 = u>x + u>2. The: higher of the: two frequencies uj2 is
referred to as the signal, the lower as the idler. For the parametric: amplification to
occur efficiently, the phase-m atching condition k$ = ki + k 2 m ust be satisfied, where
A*3 , k x and k 2 are wave vectors of pum p beam, signal and idler, respectively. This is
generally achieved by using birefringent nonlinear crystals such as /?-BaB20 4 (BBO)
and K T iO P 0 4 (K T P). By changing the angle of the optical axis of the crystal with
respect to the: polarization of one or two of the three interacting fields, a particu lar
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL D ETAILS 39
set of frequencies lo\, uj2 is selected for which bo th the conditions of phase m atching
(nzu>3 = niUi + and energy conservation (W3 = uj\ + m2) are fulfilled {n.j is the
index of refraction of the medium at frequency cjj). Figure; 3.3 is an illustration of
this process.
Nonlinear Crystal
Pump «
Signal or Idler Seed
F igu re 3.3: Sketch of optical parametric generation.
OPA is a second-order nonlinear optical process. I11 general, the dielectric po
larization P(t ) of a medium w ith nonlinear susceptibility x can be w ritten as as
expansion in powers of the applied field
P(t ) = e0( x W m + X {2)E \ t ) + Xi3)E 3(t) + (3-5)
where x ^ is the Arth-order susceptibility tensor of rank Ah-1. T he electrom agnetic
wave
E{t) = Ei 0 0 3 (0 ;! — ki z) + E -2 cos(cu2 — k-iz), (3-6)
consists of two com ponents incident on the nonlinear medium. At a fixed position
(z=0) the quadratic term x ^ E ( t ) in Eq. (3.5) gives the second-order polarization
Signal
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 40
p ( 2) = £oXi2)E 2(z = 0)
— £ 0X^' \E ' i COS* UJ\t 4~ E ‘2 COS UJ‘2̂ 4~ ‘2 E \E i2 COS Cd \t ' OOSLL^t)
= + E% + -E 'l cos2wi^
— ~ E 2 cos 2 cC'2 ̂4“ A'i E 2 [<‘os(^i 4“ (^2 ) ̂4~ ('os(u.'i — cco)^]}, (3.7)
which includes a dc polarization, ac com ponents at the second harmonics 2 u>i, 2 u>2
and com ponents at the sum or difference frequencies u \ ± u>2.
The phase condition can be fulfilled in uniaxial birefringent crystals th a t have two
different indices n 0 and n,; for the ordinary and the extraordinary waves. T he BBO
crystal is a good choice due to its high dam age threshold [74] and large bandw idth.
BBC) crystals are therefore often used in commercial OPA systems. If the OPA is
pum ped with the fundam ental T bsapphire wavelength at 800 inn. signal and idler
cover a wavelength range of 1-2.6 gin. Longer wavelengths cannot be generated
because of the onset of infrared absorption in BBO at ~ 2.6pm [74]. T he tuning
range can be extended to 2 0 pm by difference frequency mixing of the signal and
idler in a different crystal e.g., AgGaS2 or GaSe, bu t the resulting intensities are
ra ther low.
In our OPA system depicted in Figure 3.4, a tunable mode-locked T bsapphire
oscillator (Coherent M ira Model 900-F) is used to generate pulses of 200 fs at a
repetition ra te of 76 MHz, with a center wavelength of 800 11111. The oscillator is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 41
200 fs, 800 nm-1 0 nJ at 76 MHz
2.5 ps, 800 nm2.5 mJ at 1 kHz
Coherent Mira 900-F
Coherent Verdi 5 V pump laser
Light Conversion TOP AS
Quantronic DQ-527 flash lamp pumped
Quantronic Titan RGA and MPA amplifiers
F igure 3.4: The amplified Tbsapphire laser system consists of an oscillator Mira 900-F, diode-pumped laser Verdi 5V, frequency-doubled Nd:YLF laser DQ527 and TOPAS.
pum ped by a Coherent Verdi V-5 D iode-Pum ped laser, which has an o u tp u t power
of 5W. The oscillator o u tpu t pulses are first stretched to a length of several hundreds
of picoseconds (ps) by using a single grating. T he pulses from the stretcher art!
used to seed a Q uantronix T itan regenerative T bsapphire amplifier (RGA), which is
pum ped by 15% of the beam from a Q uantronix DQ527 Nd:YLF laser. T he DQ527
Q-switehed laser ustss a flash lam p to pum p the Nd-doped LiYF4 crystal to generate!
20-niJ pulses at a wavelength of 527 nm. T he seed pulse is coupled into and out of
the RGA a t a frequency of 1 KHz using a Pockel’s coll. A photodiode! m onitoring the
ou tpu t pulses from the RGA is used together with a frequency divider to trigger the
Pockel’s cell and the Q-switched DQ527, ensuring th a t they are synchronized with
the seed pulses. T he ou tpu t from the RGA are amplified by a two-pass amplifier.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 42
which is pum ped by 85% of the beam from the DQ527. T he amplified pulses will be
compressed by a pair of gratings. Pulses of 2.5 ps and 2.5 m.J at a repetition ra te of
1 KHz are obtained. A typical autocorrelation trace of amplified pulses are shown in
Figure 3.5. T he pulses from T itan will be used as pum p source for the OPA.
Slow-Scan Autocorrelation Trace1.1
SoBd Hue: A U T O n tA ry ;n « am l Mm: H ypatoM c n e b Fit. FW HM-2-Sp«
1.0
a?
0.4e4)
0.1
0.0- 15.0 - 5.0 15.0
Time Delay, ps
F igure 3.5: Autocorrelation trace of the compressed amplifier output, recorded from a BBO crystal. The solid curve is a sech2 with a FWHM of 2.5 ps.
The OPA used for lifetime m easurem ents is a commercial traveling-wave optical
param etric amplifier of superfluroreseence (TOPAS), which can generate signal and
idler pulses th a t are continually tunable from 1 to 2.6 fun. It is based on the one
stage for generation of superfluroreseence (SFL) and four light amplification stages
arranged in the same 5 mm BBO crystal.
The TOPAS is depicted schematically in Figure 3.6. The 800 nm pum p beam
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 43
OUTPUT
F igure 3.6: The traveling-wave optical parametric: amplifier TOPAS, the 800 nm pump beam is shown in grey, the signal+idler beam in black. All mirrors are 800 nm high-reflective dielectric mirrors, except M3, M7 and M8 which are metallic mirrors. NC: nonlinear crystal (BBO); TD: birefringent plate to generate delay between signal and idler; GP: glass plate; BSl and BS2: beamsplitters.
is split into three parts by the beam splitters B Sl and BS2. T he first two are used
for the fourth and fifth amplification stage. The th ird one (only 4%) passes through
a telescope and is directed to the BBO crystal by the m irrors M l and M2. There,
param etric generation takes place, resulting in a broadband seed pulse. The seed is
backward reflected from CM1, through the BBO crystal (NC), to CM2. This is the
second pass, which amplifies the eollinear com ponent of SFL produced in the first
pass. T he th ird pass is the reflection from CM2 through NC to the diffraction grating
(DG). In the th ird pass the beam of amplified SFL goes ju s t above C M 1 and hits
the center of DG. T he beam from DG is relatively weak and broad-band param etric
radiation w ith significant divergence in vertical plane. T he beam will be amplified
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 44
by the first part of the pum p beam, which is directed to the BBO by m irrors M3-5.
It passes through a glass p late G P which compensates the parallel displacem ent of
the seed when the phase angle of the BBO is changed. T he pre-amplified beam goes
through M 2 and is reflected from M7 and passes the time delay crystal th a t separates
signal and idler pulses in time. Then it will be directed to NC by m irrors M8 . There
it is strongly amplified by the th ird and most intense part of the pum p beam, directed
to the crystal by m irrors M9-11.
The TOPAS performance1 is critically dependent on the pum p beam param eters,
such as energy, pulse1 duration, contrast ratio and beam quality to achieve the specified
values. T he beam energy from T itan should be over 2 .2 m.T. The power from T itan
can be optimized by aligning the pump m irrors of the RGA and m ultipass amplifier.
T he contrast ratio measured by a fast photodiode; should be > 100. If the contrast,
ratio is less th an 100, it can be optimized by turn ing the screws the of PoekeTs cell
slightly. If all these parameters are dose; to the specification of T itan , the daily
operation of the TOPAS is very easy.
The first step to align the TOPAS is to ensure th a t the beam goes through the
center of the first two irises. Then block the 4th and fifth pa th by placing a beam
blocker before M3 and M9. Second, place a card before C M l and cheek th a t the
beam s after the first and th ird passes are in one vertical plane. By scanning in height
of the card, one can see the first and th ird pass alternatively. By using a small card
before grating (DG) and aligning the pick-up m irror outside of TOPAS, the th ird
path can be aligned through the center of CL2. Third, one can now unblock the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 45
4th and 5th pass and put a power m eter after the TOPAS to observe the ou tput.
If the power is too low, then the SFL generated in the third pass is most likely too
weak. One reason is th a t the pulse duration is slightly different from previous setting.
T he angle of stretcher grating can be tweaked slightly to optimize the dispersion of
stretcher in T itan according to the reading of the power meter.
TOPAS4/80^sTunin^cu^^j|
W&M2004.04.09
Pump: Quantronix Titan E=2.6 mJ, t = 2 ps, A=795 nm
100
o DF1 ♦ DF2
....:
•■o
1000 10000Wavelength, nm
F igure 3.7: Tuning curves of the TOPAS at the College of William and Mary. S: Signal; I: Idler; DF1: Difference Crystal AgGaS2 ; DF2: Difference Crystal GaSe.
Very high conversion efficiencies can be reached with the TOPAS. Signal+idler
energies of 650 //.I can be obtained with pum p energy of 2.3 m J. To extend the tuning
range to infrared wavelengths, a difference frequency generator is used to mix the
TOPAS signal and idler pulses. Nonlinear crystals for frequency mixing are installed
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL D ETAILS 46
in com puter controlled ro tation stage th a t can be controlled by TOPAS module.
Two different nonlinear crystals are used for the difference frequency generation. The
wave-length range can be obtained from 2.6-13 gin for AgGaS -2 and 4.5-22 gm for
GaSe crystal. The tuning curves are shown in Figure 3.7. More details are given in
the TOPAS m anual (TOPAS m anual, Light Conversion).
3.2.3 Free E lectron Laser
O verview
Vrtgglci magnet array
Ffectnat
Total ^ reflector 1 Lorentz transform
x I Doppler shift
F igu re 3.8: Schematic diagram of a free electron laser [75],
A free-eleetron Laser (FEL) provides intense, powerful beam s of laser light th a t
can be tuned to a. precise wavelength. FELs differ from most o ther types of lasers by
not having atom s or molecules as an active medium, but ra ther ’’free” electrons in a
specially designed wiggler. They can absorb and release energy a t any wavelength.
This key feature enables the FEL to produce intense powerful light, over a large
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL D ETAILS 47
wavelength range. The; lack of a lasing medium in the; cavity allows the laser to
operate at very high power levels w ithout the usual cavity heating problems.
T he key com ponents of an FEL oscillator include an electron beam of given en
ergy and intensity and the associated accelerator, a wiggler (also called an undu-
lator), and the electrom agnetic (EM) wave and the associated optical components.
Figure 3.8 [75] is a schematic representation of an FEL. The principles behind free
electron lasers are not discussed here. A detailed description of FEL physics is given
in Ref. [76]. Briefly, the electrons are accelerated to the velocity close to light veloc
ity by a superconducting R F linac. The: relativistic electron beam oscillates w ithin
a cavity w ith a m agnetic field, which is controlled by a linear arrangem ent of m ag
nets of alternating polarity. K inetic energy from the electron beam is tran sfe ra l to
EM waves w ithin the cavity, allowing for regenerative energy build-up in the cavity.
W hen the electron beam is accelerated and confined to bunches, the FEL cavity can
be synchronously pum ped, generating sub-picosecond IR. pulses.
The radiation produced by an FEL has a wavelength given by
A = ^ L ( l + a' ;̂), (3.8)
where A,„ is the wiggler period, and 7 is the relativistic! Lorentz factor, and aw is
a characteristic of the wiggler. Unlike; most conventional solid s ta te or g;rs lasers,
which produce radiation at specific frequencies corresponding to transitions in a gain
medium, the FEL wavelength can be quickly and continually tuned merely by vary
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 48
ing the electron beam energy. This wavelength tuuability is one of the most useful
features of the FEL. A review com paring and contrasting the uses of FEL and optical
param etric sources for the study of v ibrational dynam ics has been presented in Ref.
[77].
C haracteristics o f th e IR -D em o FEL at Jefferson Lab
T he Jefferson Lab FEL fatality is designed to delivery the optical beam to sev
eral user stations. The optical beam is transported to each experim ental room via
an evacuated beam line. A small fraction of the beam is delivered to a diagnostics
table where the beam is continuously characterized. These diagnostics consist of
m acropulse power m easurem ents, spectral wavelength m easurem ents of FEL beam
using a m onochrom ator, and the tem poral pulse length m easurem ents using an au
tocorrelator. Table 3.1 summarizes some of the; im portan t characteristics of the FEL
IR-Demo at Jefferson Lab.
T he average power used in our experim ents described in C hap ter 4 is around 2 0
W , with a repetition ra te of 18.7 MHz. The beam delivers pulses w ith a duration of ~
1 ps a t 5 n in. T he beam size is ~ 2.5 cm in d iam eter at 5 pm when it is transported
to the lab. A telescope has to be used to reduce the beam size.
In order to work in a safe environm ent, the labs a t Jefferson Lab FEL have three;
different modes of operation. Before running the full power beam, the beam is aligned
in a so-called ’’alignm ent m ode” . T he laser provides less th an 400 m.I of light to the
user lab in 2 Hz pulse operation. Users have to wear goggles all tim e while working in
the lab. A fter the laser is aligned, the laser can be operated in the full power mode.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 49
Basic C haracteristicsW avelength range 3 - 6.2 /mi
B andw idth (FW HM ) 0.3-2%Pulse length (FW HM ) 0.5-1.7 ps
Transverse mode <2x diffraction limitBeam diam eter a t lab 1.5-3.5 cm
Pulse Structure, Energy, and PowerRepetition R ate 74.85, 37.425, 18.7 MHz
Micropulse energy up to 70 i l lAverage power 1720 w atts
Electron Beam C haracteristicsElectron Energy 42 MevAverage current 5 niA
Peak C urrent 50 A
StabilityWavelength Stability < 2 cm - 1
A m plitude jitte r (RMS) < 1 0 % p-pEnergy stability (RMS) 4 x u r 4
Beam position ji tte r (RMS) 1 0 0 pmBeam Angle ji tte r (RMS) 250 prad
T able 3.1: Specification of the FEL IR-Demo at the Jefferson Lab.
An interlocked hutch ensures th a t users are not exposed to the FEL beam while it
is being delivered to the room. This is called ’’hutch mode:” . W hen the interlocked
hutch is open and the full power beam is delivered to the experim ental table, users
are not allowed in the lab during beam delivery. This is called exclusionary mode.
All the d a ta acquisition is accomplished by a rem ote com puter in the control room.
If the user needs to go to the lab to adjust some optics, the FEL beam has to be shut
off by the operator.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 50
3.3 Vibrational Pum p-probe Spectroscopy
3.3.1 E xperim ental Set-up
The vibrational lifetimes in this thesis were m easured by pum p-probe spectroscopy,
also called transient bleaching spectroscopy. T he experim ental setup is a standard
one-color pum p-probe set-up which is shown in Figure 3.9. T he laser beam of the
TOPAS or FEL must be collim ated at different wavelengths by a pair of lenses or
curved mirrors. The collim ated laser beam is split into two beam s by a ZnSe beam
splitter. One is pum p beam, which carries about 90% of the to ta l energy. T he other
is probe beam, which is much weaker th an the pum p beam (about 10%). T he probe
beam will pass through a com puter-controlled translation stage. Both pum p and
probe beam s are focused by a pair of parabolic mirrors. T he pum p and probe are
spatially overlapped in the sample. T he strong pum p beam will excite a significant
fraction of the defects to the first excited sta te , which will cause a transient increase
in the transm ission coefficient of tin; sample. By varying the delay between the
pum p and probe pulses, the evolution of the excitation can be followed in time. The
transm itted probe; beam is m easured by a detector placed after a m onochrom ator. In
order to m easure the decay tim e of the first excited vibrational s ta te of th e defect,
the laser pulse m ust be significantly shorter th an the decay time. T he pum p beam is
chopped a t 300 Hz for the OPA system and 3 KHz for the FEL beam. T he chopped
beam causes a m odulation of the pum p-probe signal, which is recorded by a lock-in
amplifier. In the experim ents we use M C T /InSb ’’sandwich” detectors from Judson
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 51
Technologies, LLC. T he InSb and M CT detectors can be used a t wavelengths at 2 to 6
/rrn and 6 to 16 /mi, respectively. T he detected signal is first amplified by a low-noise
voltage preamplifier. A SR250 gated in tegrator and boxcar averager, and a lock-in
amplifier are used to average and ex tract the transient change of the transm itted
signal of the probe beam.
I Laser
Iris
*1 if* 1 l«V ' I
I
Sample/cryostat
IHeNe Laser
Chopper Pump detector
VOptical delay
t Probe detector
^% <4=*C t O r V y .
Monochromator
F igu re 3.9: Typical set-up for pump-probe experiments.
3.3.2 A lignm ent o f the Infrared B eam s
T he alignm ent of the invisible beam is the most im m ediate challenge to perform ing
any experim ent w ith the IR. beam. Basic alignment, tasks, trivially perform ed with
a visible laser, become difficult and time-consuming with the infrared beam. It is
therefore im portan t to develop alignment, m ethods which are quick and accurate so
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 52
th a t beam alignm ent can be accomplished efficiently. The; most generally useful
technique is to coalign a visible; beam with the infrared beam to provide a visual
reference to guide the alignm ent of the IR. beam. R.ed Helium-Neon (HeNe) lasers are
generally the visible laser of choice, because of their simple operation and excellent
transverse mode quality.
In the experim ental set-up shown in Figure; 3.9, two irises are; placed before the
beam splitter. During the alignm ent procedure, the IR. lase;r beam is blocked and
the HeNe laser be;am is directeel into the center of the two irises by a pick-up mirror.
T he pick-up m irror is m ounted on a m agnetic base; and can be; easily removed and
placed back. All the optics after the second iris are aligned accoreling to the HeNe
laser beam. A fter all the aligmneaits are finished, one; needs to block the HeNe beam,
remove the pick-up m irror anel let the; IR. beam go through the twee irises. A detector
is placed after the irises and the power is optim ized when the; beam goes through the
irises. This ensures th a t the IR. be;am passes through the; ceaiter of bo th irises.
3.3.3 E xperim ental Procedures
The vibrational lifetime; mesasureanents of local vibrational modes are very challenging
due to the small absorption cross section of the elefects, especially when they are
measured by our OPA system. The; meausurement, is very critically dependent on the
alignm ent anel laser stability. Before running the experimeait, we m ust make sure th a t
the lasers run stable and make sure th a t the laser beam goes through the eemter of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 53
the first two irises. Generally there arc' three steps in the experim ent. F irst, we use a
reference sample, which gives rise to a strong signal, to chock the alignm ent and the
spatial overlap of the pum p and probe beam. A pin-hole w ith a diam eter of 50 fim
is used to help align the spatial overlap of the two beams. The sample is m ounted
on a three-dim ensional translation stage1, which can be moved to optim ize the signal.
Once the set-up is optimized, a big spike shown in Figure 3.10 can be observed. This
spike is independent of wavelength. Next, the sam ple is m ounted in the cryostat and
one needs to check w hether the spike still exists. Then, the sam ple is cooled down
to 80 K and the OPA is tuned to the absorption resonance of the defect. W hen the
spectrum of the- probe beam is m easured w ith the monochrom ator, an absorption dip
will be observed, which is shown in Figure 3.11. This spectrum is m easured in a Ge
sam ple w ith a very high oxygen concentration. T he dip corresponds to the absorption
of the oxygen stretch mode in Gc. Most defects do not have; a strong dip due to their
weak absorption. The; last step is to lower the tem perature to 10 K and m easure the
v ibrational lifetime.
3.4 Sample Preparations
3.4.1 H ydrogen in Silicon
Hydrogentated Si prepared by proton-im plantation
The proton-im planted Si samples used in v ibrational lifetime m easurem ents are
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 54
0.8
0.6
CO0.4
0.2
J * * tV0.0
■5 0 5■10 10 15 20
Time Delay (ps)
F igure 3.10: Pump-probe signal at room temperature: from a Ge sample.
0.04
"Ge:0 sample from K. Itoh
80K0.03
0.02
0.01
860 cm
0.00
X X820 830 840 850 860 870 880 890 900
wavenumber (cm'1)
F igure 3.11: Spectrum of a 1.7 ps laser pulse passing through a monochromator. The clip at 860 cm-1 corresponds to the absorption of interstitial oxygen in Ge doped by C o = ~ 1018 cm-3.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL D ETAILS 55
m ade by M. Budde [78], It is a 2-mm thick, high resistivity, disk-shaped Si single
crystal, coated w ith a m id-IR antireflection coating on one side. T he coating pre
vents spurious signals due to internally reflected light. T he Si sam ple is im planted
w ith protons at different, energies in the range l - 1 .8 MeV. A to ta l dose of 2 x 1017
p ro tons/cm 2 is im planted into the uncoated side of the sample at 80 K yielding a 47
gm deep, uniform H concentration profile. Figure 3.12 shows the im plantation profile
obtained by the software package SRIM [79]. T he im plantation dose at each energy is
adjusted in order to obtain a uniform hydrogen concent ration of 0.02%. T he sample
tem perature is kept at 80 K during the im plantation, and the background pressure
is about 10- 6 torr. A uniform H concentration of 1.7 x 102 ppm is formed from the
surface to a depth of 47 pm by variable-energy im plantation of the uncoated side of
the sam ple a t 80 K.
0.020
.2 0.015
£ 0.010
o 0.005
10 20 30 40Depth (pm)
Figure 3.12: Hydrogen implantation profile.
H ydrogentated Si prepared by electron-irradation
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL D ETAILS 56
T he samples which are used in high resolution IRAS m easurem ents are prepared
by electron irradiation at Lehigh University. An advantage' of producing these defects
by electron irradiation, ra th e r th an by proton im plantation, is th a t the defects can
be d istributed uniformly w ith lower concentrations throughout bulk samples and,
therefore, have reduced inhomogeneous broadening and have narrower vibrational
linewidths. T he samples consist of n-type Si ([As] = 3 x 1 0 14 c n r ,!) grown by the
floating zone (FZ) m ethod. T he samples have dimensions 7 x 7 x 15 m m 3 w ith the
optical viewing direction along the long axis. T he viewing surfaces of the samples
are polished w ith SiC and A120 3 grits and 0.5 gm diam ond paste. T he samples
doped with H (Si:H) or D (Si:D) are annealed a t 1250° for 30 min in quartz ampules
containing H2 or D2 gas at 0 .6 6 atm . T he anneals were term inated by a rapid quench
in w ater to room tem perature. The concentration of H or D is determ ined to be 1016
cm - 3 by m easuring the intensity of the 3618.4 cm - 1 line assigned to IT- T he H and D
defects are created by irradiating the samples with 2.5 MeV electrons at RT to a to ta l
dose of 1017 cm ” 2. The samples are irradiated w ith half the to ta l dose from opposite
sides to make the defect concentration nearly uniform. T he defect concentration in
the samples is estim ated to be ~ 1016 cm”3, which is ~50 tim es smaller th an in the
proton im planted samples.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 3. EXPERIM EN TAL DETAILS 57
3.4.2 O xygen in Si and G e
T he oxygen isotopes 170 and 180 are diffused into FZ-Si doped w ith boron concen
tra tion of 5.7 x 1015 cm - 3 a t the University of Paderborn [80]. In order to avoid
contam ination with transition m etals such as Fe or Ti, the diffusion was done under
extrem ely clean conditions using an IR. heat chamber. The sam ple is placed into an
()2-gas atm osphere enriched w ith 17() and 180 in a cooled quartz am poule onto a short
quartz rod. T he sample is held for about 14 days at an oxygen pressure of 3 bars
at 1400°C. The tem peratu re gradient is w ithin 5% over 50 mm of the sam ple length.
The; ho ttest point in the am poule and support system eould be held below 900°C by
forced air cooling, thus over 500°C less than the sample tem perature. The sam ple is
cooled slowly from 1400°C to 800°C at a ra te of 200°C /hour and them quenched at
room tem perature.
The "G e:0 sample is an as-grown sample taken from a CZ Ge:() n-type (n ~
1 .6 x 1014 cm -1 ) crystal grown by M etallurgie Hoboken Overlept in Belgium about
2 0 years ago.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 4
H ydrogen B ending M odes in
Sem iconductors
4.1 Introduction
Hydrogen passivation of defects is a standard processing step in t he production of
m etal-oxide-sem iconduetor (MOS) electronit: devices. This process reduces the den
sity of Pfc centers, the most abundant electrically active defect a t the S i/S i0 2 interface,
by 3 orders of m agnitude [1 2 , 81], Recently, an isotope effect of Si-H and Si-D bonds
has been observed during this passivation process. By using deuterium instead of
hydrogen to passivate Si dangling bonds of Pf, centers at the S i/S i0 2 interface, Lyd-
ing, Hess, and Kizilyalli found th a t hot-electron degradation of MOS transistors was
reduced by factors of 10-50 [82]. This reduction increased the operational lifetime of
a silicon chip. A similar isotope effect has been observed in a. study of the dissociation
kinetics of Si-H and Si-D complexes in GaAs. Chcvallier ei al. observed th a t, for a
given incident ultraviolet (UV) photon density, the concentration of dissociated Si-D
58
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C H A P T E R 4. H Y D R O G E N B E N D IN G M O D ES IN SE M IC O N D U C TO R S 59
complexes is 10-20 times below the concentration of dissociated Si-H complexes [83].
Energy
E=0
HEmission
TransportStates
Si-HVibrational
Levels
Figure 4.1: Schematic potential energy surface for a Si-H bending mode, showing vibrational energy levels in the first energy well. When H is vibrationally excited to the top of the well, it may be emitted to transporting states and dissociate' the bond.
T he isotope effect of H and D implies th a t Si-D bond-breaking is slower than
for Si-H. The origin of this isotope effect is not well established and m any possible
mechanisms have been proposed [82, 84, 85, 8 6 ]. T he key feature' is th a t the elec
tronic properties of Si-H and Si-D are identical, and only their vibrational properties
are different. It is believed th a t the dissociation of Si-H and Si-D bonds is caused
by inelastic electron scattering, which excites the defect to either a dissociative elec
tronic s ta te or to excited vibrational states. T he vibrational excitation mechanism
of hydrogen-related modes has been described by the truncated harmonic oscillator
model [84, 87, 8 8 , 89], which is shown in Figure 4.1. This model describes the Si-H
and Si-D bonds as harm onic oscillators. Si-H bonds can be excited to a more ener-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 4. H YD RO G EN BENDING M ODES IN SEM ICONDUCTORS 60
getic v ibrational s ta te by non-radiative electron-hole (e-h) recom bination or by hot
electrons at Si/SiC^ interfaces. If the excited Si-H oscillation can be sustained, hydro
gen can be em itted over the barrier to a mobile transport s ta te (Figure 4.1), causing
the Si-H bond to dissociate. W ithin th is model, the desorption ra te is approxim ately
given by
where T x is the lifetime of the v ibrational s ta te | 1 ) and F i s the excitation ra te of
the transition |0) —> |1). From Eq. (4.1) the vibrational lifetime T i is an im portant
param eter for the dissociation ra te of Si-H or Si-D bonds.
nism is most likely controlled by the Si-H bending mode [90]. Based on their calcu
lations by density functional theory, the bending mode at anti-bonding sites provides
bulk TO phonon states at the X point (463 can-1 ). So one may expect th a t the cou
pling of the Si-D bending mode; to the Si bulk phonons results in an efficient channel
for de-excitation. Evidence for the coupling between the Si-D bending mode and Si-Si
lattice vibrations is found a t the S i/S i0 2 interface [92] and in deuterated am orphous
silicon [93]. To our knowledge the v ibrational lifetimes of Si-H and Si-D bending
modes in Si have never been measured and vibrational dynam ics of bending modes is
(4.1)
Chris G. Van de Wallo et al. proposed th a t carrier-enhanced dissociation mecha-
a more favorable dissociation pa th than the Si-H stretch mode [91]. T he v ibrational
frequency of bending mode for Si-H was calculated to be around 650 cun J, and the;
estim ated frequency for Si-D is around 460 cm x, which is close: to the frequency of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 4. H YDRO GEN BENDING M ODES IN SEM ICONDUCTORS 61
essentially unknown. These experim ents are difficult to perform because the bending
modes lie close to the phonon bath and the laser pulses are weak in th a t wavelength
range.
This chapter presents the first m easurem ent of H -related bending modes in Si. The
lifetime of the bending m ode in Si a t 817 cm - 1 is measured to be = 12.2 ± 0.8
ps by transient bleaching spectroscopy. T he temperature; dependence of the lifetime
and calculations of phonon-density of s ta te indicate th a t the bending m ode of H*
decays into two phonons (LA + TO phonons a t the X point). T he vibrational life
times of o ther H -related bending modes are estim ated from their infrared absorption
linewidths. T he different, lifetimes of H and D -related bending modes can be described
by an energy gap law [93].
4.2 Experim ental
The sample consists of a 2-min thick disk cut from single; crystalline; high-resistivity
Si. T he sam ple was polished em both sides, and coated with a miel-IR antireflection
coating on one side to prevent spurious signals arising from m ultiple internal reflec
tions. A uniform H concentration of 1.7 x 102 ppm was formed from the surface to a
dep th of 47 pm by variable-energy im plantation of the uncoateel side; of the sam ple at
80 K. After the im plantation, infrared absorption spectroscopy was perform ed with
a commercial Nicolet Nexus 670 F T IR spectrom eter at tin; Applied Research Center
of Jefferson Lab. F igure 4.2 shows the spectrum of the bending mode at 817 cm -1 .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 4. H YD RO G EN BENDING MODES IN SEM ICONDUCTORS 62
The inset shows the defect structure: one H close- to the bond-center side (BC) and
the o ther near the antibonding (AB) site. T he Hg in Si gives rise to four absorption
lines a t 817, 1599, 1838, 2062 cm -1 . T he 1838- and 2062 -cm - 1 lines were assigned
to the stretch inodes at the AB side and the BC site, respectively. The 817-em-1 line
corresponds to the bending mode of hydrogen at the- AB site. 1599-cm - 1 line is the
overtone of the 817-em "1 modi!.
0.25
817 cm 80 K0.20
CDOcCD.aotn<
0.05
0.00 -
805 810 815 820 825 830
Wave Number (cm'1)
Figure 4.2: Absorbance spectrum of proton-implanted Si showing the 817 cm"1 bending mode of H?i, shown in the inset.
T he IR, pum p-probe setup is described in C hapter 3. The laser source is the
TO PAS which delivers pulses at 1-kHz repetition ra te with a tim e duration of ~ 2
ps, spectral w idth of 10 cm "1, and a pulse energy of 1 2 p.J a t 817 cm "1. T he tran
sient bleaching signal Su is detected using an amplified liquid nitrogen-cooled M CT
detector. To allow for low tem perature m easurem ents, all experim ents are performed
w ith the sample m ounted in an optical cryostat. The Si sam ples used for linewidth
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 4. H YDRO GEN BENDING MODES IN SEM ICONDUCTORS 63
m easurem ents are 15 mm thick, grown by the float-zone m ethod, and prepared by
electron irradiation in Prof. Stavola’s group at Lehigh University. T he details about
the electron irradiation are described in C hapter 2. The' IR. absorbance spectra are
m easured w ith a resolution of 0.1 cm - 1 using a Bonien DA 3.16 Fourier transform
infrared spectrom eter.
4.3 R esults and Discussions
4.3.1 L ifetim e M easurem ents o f B end M ode o f Hjj in Si and
Ge
0.010
0.008
0.006
-CC/5
0.004
0.002
0.000
-10 -5 0 5 10 15 20 25 30 35
Time Delay (ps)
Figure 4.3: Decay of the transient bleaching signal Si, from the 817-cm-1 mode of H| measured at 10 K with the TOPAS. St, decays exponentially with a time constant of T\ = 12 ± 0.8 ps.
Bend Mode of H2 1 0KOn resonance 817 cm T = 12 + 0.8 ps
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 4. HYD RO G EN BENDING MODES IN SEM ICONDUCTORS 64
Figure 4.3 shows a semi-log plot of the transient bleaching signal versus tim e
delay m easured at 10 K w ith the laser frequency centered on the 817-em-1 mode.
The signal decreases w ith a tim e constant 7 i = 12 ± 0.8 ps. Figure 4.4 shows the
spectrum of our laser pulses. The dip at 813-cm-1 corresponds to th e absorption of
the bending mode of H j a t room tem perature. F igure 4.4 also shows the band w idth
w ith a FW HM of 10 cm - 1 a t 12.2 /am. W hen the wavelength is tuned off 817 cm - 1
by 10 cm -1 , the exponential decay S*, disappears. M easurem ents of Si, vs wavelength
show th a t the transien t bleaching signal originates from th e 817 cm - 1 mode.
813 cm' at RT
0.06
~ 0.04 HWFM: -10 cm
0.02
0.00
790 850780 800 810 820 830 840
Wave Number (cm'1)
Figure 4.4: Spectrum of a 1.7 ps laser pulse passing through monochromator, showing the absorption of the H2 bending mode at 813 cm-1 at room temperature.
As discussed in C hapter 2, the vibrational lifetime of the excited sta te of a. local
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C H A P T E R 4. H Y D R O G E N B E N D IN G M O D E S IN SE M IC O N D U C TO R S 65
v ibrational mode can be estim ated from
where F 0 corresponds to the natu ra l linewidth, which has to be m easured a t the
highest resolution possible using samples with very low defect concentrations. The
sample m ust be cooled to below 10 K for the absorption m easurem ent to insure th a t
the line broadening due to pure dephasing is negligible.
Figure 4.5(a) shows the absorption spectrum of HJ measured a t 5 K using samples
with H concentration of ~ 1 x 1 0 16 cm -3 . T he line shape of H j is fitted by a L o ren taan
w ith a FW HM of 0.42 cm -1 , which corresponds to a lifetime of 12.2 ps calculated
by Eq. (4.2). This result shows th a t the lifetime measured in the tim e dom ain is
exactly consistent w ith the one obtained in the frequency domain. T he absorption
linewidth is dom inated by the homogeneous natu ra l linewidth. T he lifetime estim ated
from this low concentration sam ple is reliable. Since the v ibrational lifetime of Dg is
hard to measure due to the weak laser pulses a t ~ 17 /zm, we obtained the lifetime
of Dj a t 588 cm - 1 by measuring its linewidth. T he silicon sam ple containing D.J is
prepared in the same way as II ) , and the line broadening a t 5 K is dom inated by its
natu ra l linewidth. From Figure 4.5(b) we obtain the absorption line w idth of 0.46
cm -1 , which corresponds to a lifetime 7 \ = 1 1 .6 ps. For this special case, it is very
interesting to find th a t the lifetimes of HJ; and bending modes are almost the same.
In the following, we will show th a t this result can be explained by the energy gap
law, i.e., both modes fall into the two-phonon band of Si. To gain further insight, the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 4. H YDRO GEN BENDING MODES IN SEM ICONDUCTORS 66
relaxation channel and decay mechanism need to be understood.
0.05
0.040§ 0.03a
■£ 0.02
1 0 .0 . 0.00
r = 0.42 ± 0.01 cm'
815 816 817 818 819 820
0.02
4 )OJ 0.01S-lo</>x><*
^ 0.00
r„ = 0.46 ± 0.02 cm
586 587 588 589 590
W avenumber(cm"')
F ig u re 4.5: IR absorbance spectrum of e-irradiated, H-doped Si, showing the lines of Hlj and Dj.
4.3 .2 D ecay C hannel o f th e H 2 B end M ode
T he na tu re of the accepting modes of H -related stretch modes is unknown [44, 94],
They can be Si-H bending modes, pseudolocalized modes, a n d /o r phonon modes. In
contrast, the decay channel of the bending modes may be less com plicated because
they art: low-frequency modes and m ay decay directly into the phonon bath . To iden
tify the relaxation channel we measured th e tem perature' dependence of the lifetime.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 4. H YDRO GEN BENDING MODES IN SEM ICONDUCTORS 67
14
12
_ 10wQ.CD.i 8
- 1 6
4
2
0 20 40 60 80 100 120 140
Temperature (K)
Figure 4.6: Temperature dependence: of T \ of the 817-ein-1 mode of HJ. The dashed and solid lines are the theoretical predictions for decay into three phonons {517, 150, 150} and two phonons {460, 357} cm-1, respectively.
817 cm'1 = 0(517cm'1) +2 TA(150 cm'1) - 817 cm'1= TO(460cm'1) +LA(357 cm'1)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 4. H YD RO G EN BENDING MODES IN SEM ICONDUCTORS 68
Figure 4.6 shows T i versus tem peratu re for the bending mode of H.J a t 817 cm - 1 in
silicon. T x is nearly constant up to 130 K. T he lifetime measured above 130 K is not
reliable due to the low signal-to-noise ratio. T he solid line in Figure1 4.6 shows a fit,
using Eq. (2.9) in C hapter 2 w ith the v ibrational relaxation channel of the 817-cm "1
mode represented by two accepting modes {460 and 357} cm -1 . T he dashed line is a
fit using three accepting mode's {517, 150 and 150} cm "1, which corresponds to the
phonon com bination of one1 optical phonon at F point anel two TA phonons. It is
clear th a t th e beneling moele at 817 cm - 1 decays via two-order proe'e'ss. T he phonon
decay channels are not uniquely determ ined by the tem peratu re dependence of the
lifetime. They can also be fitted well by o ther two-order processes. For example, the
tem perature dependence of the 817 cm - 1 is equally well represented by 409 and 408
cm " 1 accepting modes. However, any three-order process will not fit the d a ta points.
Since D£ and fall into the two-phonon band of Si, we would expect th a t D | also
decays via the two-order proems. Figure; 4.7 shows the calculation of the two-phonon
density of states by straightforw ard convolution of the one-phonon density of s ta tes
according to Eq. (2.10) in chapter 3. The bending mode at 817 cm " 1 falls on the
peak of the T O + L A phonon combination, whereas the DJ bending mode a t 588 cm " 1
coincides w ith T A + T O phonon combination. According to the sym m etry selection
rules described in C hap ter 2 , the decay channels of H j and D | are infrared active and
the two-phonon decay channels are allowed.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 4. HYDROGEN BENDING MODES IN SEMICONDUCTORS 69
20
0)S 15 coo >,
S 10Q c o c of 5o g I-
0
0 200 400 600 800 1000
W ave Number (cm '1)
F igure 4.7: Two-phonon density of states in Si. The H?; bending mode at 817 c a r 1 coincides with LA+TO phonons, whereas the bending mode at 588 cm-1 lies in the TA+TO phonon band.
2 TOTA+TO
LA+TOSi-D
2 TA
Si-H
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 4. HYDRO GEN BENDING MODES IN SEMICONDUCTORS 70
4.3.3 L ifetim e o f O ther H -related B end M odes and Energy
G ap Law
Nitssan at al. developed a theory of v ibrational energy relaxation in solids, which is de
scribed in C hapter 2 . T he vibrational relaxation ra te is expressed in Eqs. (2.8), (2.9),
and (2.10). T he relaxation ra te is determ ined by the m agnitude of the coupling term s
between local vibrational modes and accepting modes. Since the hydrogen bending
mode decays by lowest order, it is reasonable to assume th a t the |Gq,q | 2 term s decrease
fast w ith increasing order of the m ultiphonon process N ^ . According to Eq. (2.12),
the relaxation ra te decreases exponentially as the energy gap between local modes
and phonon modes or decay order is increased. T he energy gap law can be expressed
by [95, 96]
71 = A e llN, (4.3)
where N is the decay order according to the lowest-order decay, and A and B are
two param eters which depend on the coupling constants between tin ' local vibrational
modes and the accepting modes.
F igure 4.8 shows a plot of tin- v ibrational lifetimes of H (D )-related bending modes
vs decay order for various im purities in different semiconductors. The linewidth of Si
ll bending mode; in Si-doped GaAs at 896 cm - 1 is ~0.02 cm - 1 [97], which corresponds
to a lifetime of ~260 ps. Since the m aximum phonon frequency in GaAs is a t 291
cm -1 , there must, be a t least 4 phonons involved in the decay process. T he linewidth of
Si-H bending mode at 890 cm - 1 in Si-doped AlAs is 0.07 cm - 1 [97], which corresponds
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 4. HYDROGEN BENDING MODES IN SEMICONDUCTORS 71
1000 r o Si-D in GaAs: o Se(Te)-H in GaAs
* Si-H in AlAs .X o
100A S-H in GaAs A
: 0 Si-H in GaAs• □
• ★ X ^10 r • x ▼ Si-D Calc.X A ▲ Si-H Calc.
• Se(Te)-D in GaAs
1♦ Si-D of H2* in Si
▼ ■ Si-H of H2*Si★ S-D in GaAs
0.1 1 1 . l i
1 2 3 4 5
Decay Order
F igure 4.8: Vibrational lifetimes of H-related bending modes vs decay order in different, hosts. The line is a fit to Eq. (4.8) with two parameters, A = 0.15 ps and B = 2.01
to a lifetime of ~75 ps. The maximum phonon frequency of AlAs is 402 cum 1, which
leads to a decay order of 3. The linewidth of Si-D in GaAs at 641 cum 1 is around 0.13
cm corresponding to T i ~ 40 ps [98]. S (Se,Te)-H modes have' almost the same
frequencies w ith a linewidth of 0.045-0.09 cun corresponding to T x = 55-110 ps [99],
and a three-order process. In contrast, the linewidth of the corresponding D-modes
are 0.3-0 .6 cm " 1 (T i = 8-16 ps) due to their two-order decay process. T ight-binding
molecular dynam ics sim ulations predict a lifetime of 0.3~0.5 ps for Si-D at 460 cm " 1
and 5-6 ps for Si-H at 640 cm "1. The solid line in Figure 4.8 is a fit using Eq. (4.3).
This fit, reveals the general trend th a t the v ibrational lifetime of H (D )-related bending
modes does follow closely an energy gap law. The d a ta points are fitted with the values
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 4. HYDRO GEN BENDING MODES IN SEMICONDUCTORS 72
A = 0.15 ps and B = 2.01. A is determ ined by the tim e scale of the first-order decay
and B depends only logarithm ically on the coupling constant and on N [95]. The
coupling constant B = 2 .0 1 is close1 to the value of the CN “ stretch mode in silver
halides, which also follows the energy gap law w ith the fitting param eter B = 1.94
[96]. T he variation of B from host to host is therefore small enough to justify the
d a ta analysis.
T he overall fitting of the; d a ta point is satisfactory, bu t by no m eans perfect.
There are a few possible errors in the: lifetime of H (D )-related bending modes plotted
in Figure 4.8: (1 ) Most of the lifetimes arc estim ated from their linewidths measured
a t low tem peratures, which may have some inhomogenous broadening and not their
real natu ra l linewidth. (2) T he linewidth may contain the instrum ental broadening,
especially those sharp lines. (3) Isotope broadening may contribute: to the linewidth
of Sc (Te)-H or D mode's in GaAs [100]. (4) The: coupling constant e:an be slightly
different in different hosts even though the calculations by Biswas [89] indicate th a t
lifetimes of bending modes are: insensitive to the: environment. (5) T he calculated
Si-H lifetime is about half of our measure*! lifetime of H[>. We m easured the Si-H
lifetime to be ~ 12 ps, while the calculate*! lifetime of Si-H bemeling m ode is only 5-6
ps. So the lifetime for Si-D moele may be larger as well, i.e., ~ p s according to the
experim ental elata.
We already kimw th a t the Si-H stretch modes do not, follow the: energy gap law.
For example, the HI, and HV-VH(n0) complex in Si have: imarly the same freepieimy
bu t very different lifetimes by two-orders of m agnitude [101, 94]. Likely, the coupling
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 4. HYDRO GEN BENDING MODES IN SEMICONDUCTORS 73
constant, accepting modes and decay mechanism of stretch modes are different and
show a strong structu re dependence as discussed in the next chapter. O ur lifetime
m easurem ent of bending modes and analysis of energy gap law clearly shows th a t
the lifetime of bending m ode critically depends cm the decay order, i.e., the energy
gap between the local v ibrational inodes and accepting modes. T he energy relaxation
tim e can be different by 3 orders of m agnitude when the; bending m ode decays into
different num ber of phonons. Since; the energy of the bending mode; is sm aller for
deuterium th an for hydrogen the former m ay decay via lower-order process which
results in shorter lifetime. This result can explain the giant H (D) isotope effect and
supports the finding th a t the dissociation of Si-H and Si-D bonds is controlled by the
dynam ics of hydrogen bending modes.
4.4 Conclusion
In conclusion, we have measured the vibrational lifetime's of Si-H and Si-D bending
m odes of H j in Si and extended this study to m any other bending modes in semicon
ductors. This study is key to understanding the physics of the giant isotope effect.
We show th a t the v ibrational lifetime; of bending modes can be; desci'ibed by an energy
gap law. T he lifetime increases dram atically w ith increasing de;cay oreler.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 5
H ydrogen Stretch M odes in Silicon
5.1 Introduction
For a com plete understanding of the v ibrational dynam ics of hydrogen in semiconduc
tors, it is important, to study also the lifetime of hydrogen stretch modes. In contrast,
to the bending modes, the stretch modes are high-frequency modes and can decay
into phonons, bending modes, an d /o r pseudolocalized modes related w ith hydrogen
defects. In 1975, Stein reported a pioneering infrared absorption study of crystalline
silicon im planted with protons at room tem peratu re (RT) [102]. The spectrum ob
served by Stein after im plantation a t ~ 80 K is dom inated by an intense absorption
line a t 1990 cm - 1 [103], which was identified as bond-center hydrogen (H ^ |) shown in
Figure 5.1(a). In proton-im planted Si, more than ten absorption lines are observed in
the range 1800~2250 cm - 1 a t R.T (Figure 5.2). The lines shift down in frequency by a
factor of 1 / \ /2 when deuterons are im planted instead of protons. These Si-H or Si-D
absorption lines can also be observed in Si grown in a. H2 am bient and subsequently
irradiated w ith electrons [104, 105] or neutrons [106], Local vibrational m ode spec-
74
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 5. HYDRO GEN ST R E T C H MODES IN SILICON 75
troseopy and uniaxial-stress techniques have been used to assign these Si-H stretch
modes to specific hydrogen-related defects.
(e). VH2 (f). v h 4
Figure 5.1: Structures of the interstitial-type defects (top), and the vacancy-type complexes (bottom). Large spheres are Si, whereas small spheres are H.
A num ber of' defects have been identified, part of which are shown in Figure 5.1.
Figure 5.1(a) shows the bond-center hydrogen defect in Si. It gives rise to an intense
absorption line at 1998 cm * 1 due to the excitation of the stretch mode, where the
hydrogen vibrates parallel to the Si-H-Si bond axis. Figure 5.1(b) depicts the 11*2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 5. HYDRO GEN ST R E T C H MODES IN SILICON 76
Hz’
i_________ i__________i__________i__________ i__________i__________i__________ i
1800 1850 1900 1950 2000 2050 2100 2150Wave Number (cm 1)
F igure 5.2: Absorption lines of hydrogen stretch inodes in proton-implanted Si: Hj gives rise to two absorption lines at 1838 (BC site) and 2002 cm-1(AB site'.); IH-2 is a self- interstitial defect and has two absorptions lines at 1987 and 1990 cm-1; V2H2 has two Si-H stretch modes, but only one Si-H mode is IR active at 2072 cm-1; VH consists of one hydrogen atom saturating a dangling bond of the vacancy, which was assigned to be 2038 cm” 1; VH2 gives rise to two absorption lines at 2267 and 2316 can” 1.
defect, which is described in C hap ter 1 and 4. F igure 5.1(c) shows the self-interstitial
hydrogen defects in Si. This defect contains two nearly perpendicular and equivalent
Si-H bonds th a t are almost, aligned w ith the [011] and [011] directions and has a C 2
axis parallel to the [001] direction [107].
Vacancy type defects are also fundam ental, intrinsic point defects in crystalline Si.
Theoretical calculations [91, 108, 109, 110] predicted th a t a dangling Si bond repre
sents the strongest trap for hydrogen in bulk Si and th a t a vacancy may trap up to
four H atoms. T he vibrational stretch frequencies were also calculated and exhibited
a steady increase w ith increasing n for the V H n complex. The atom ic configuration
of VmH.„ complexes are shown in Figure 5.1(d), (e) and (f). VjIIz was identified
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 5. HYDROGEN ST R E T C H MODES IN SILICON 77
by Suezawa in a hydrogenated Si sample by electron-irradiation [111). This complex
with a divacariey binding two H atom s is nam ed H V ■ H V (n0) to emphasize th a t the
Si-H bonds art' located in the sam e (110) plant!, in separate halves of the divacariey.
V H -2 has C-2v sym m etry and consists of two hydrogen atom s each satu ra ting a dan
gling bond of the vacancy. V H -2 has two Si-H stretch modes a t 2122 and 2145 cm -1 .
V H 4 has all four dangling bonds of the vacancy satu ra ted by hydrogen, which renders
the complex electrically inactive. This complex has T(i sym m etry and gives rise to a
absorption line at 2223 cm "1.
In order to understand the dynam ics of these defects, th e vibrational lifetimes of
these local modes need to be measured. Initial vibrational lifetime m easurem ents of
H in crystalline Si were done by Buckle [44] et al. for the bond-center hydrogen at
1998 cm -1 . The vibrational lifetime of the stretch mode was m easured to be 7.8
ps. T he tem peratu re dependence of v ibrational lifetime showed th a t the H # q s tretch
mode does not decay via lowest order. The; lifetimes of o ther H -related stretch modes
were also estim ated from their spectral w idths. A structure' dependence of the lifetime
was found from the linewidth analysis in samples with very low defect concentrations
[44]. T he lifetimes of in terstitia l-type defects are much shorter than vacancy-type
defects. This disparity in m easured lifetimes is unexpected based on simple theories
and is as yet unexplained [56, 57]. Experimentally, the most straightforw ard way to
characterize a m ultiphonon relaxation ra te from a given vibrational level is from the
tem peratu re dependence of its lifetime. In this chapter, we will discuss a detailed com
parison of the tem perature dependence of the vibrational lifetime of two structu rally
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 5. HYDRO GEN ST R E T C H MODES IN SILICON 78
distinct point defects in crystalline Si, H%, an in terstitia l defect, and HV-VH(110), a
vacancy complex. T he very different, temperature! dependences show th a t these vi-
brationally excited defects decay into m arkedly different accepting vibrational modes,
giving new insights into the origin of the observed largo difference' in their lifetimes.
5.2 Experimental
H V V
20802060 20702050Wave Numbers (cm-1)
F igure 5.3: Absorbance spectrum of proton-implanted Si showing the 2062.1-, 2068-, anel 2072.5-cm-1 lines of IF>, HVF, and HV-VH(110).
The Si sample useel for the lifetime measurement, of H2 boneling me)de is the same
one used for stretch mode m easurem ents as described in C hapter 4. After warming to
room tem perature, IR absorbance m easurem ents from the im planted sample revealed
a series of strong absorption lines in the Si-H stretch region, including lines a t 2062.1,
2068, and 2072.5 cm -1 . Figure 5.3 shows an absorbance spectrum taken w ith 0.15-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 5. HYDRO GEN STR E TC H MODES IN SILICON 79
cm - 1 resolution at 10 K from a Si-H sam ple w ith a low H concentration (~ 7 ppm) to
clearly resolve the different absorption lines. T he 2068-cm-1 line has been assigned
to the divacancy binding one H atom HV-V [112], and was ~ 3 tim es smaller th an
the 2072.5-cm“ 1 line of V2 H 2 in the sam ple w ith the high H concentration. Since the
linewidth is inversely related to the lifetime, it is im m ediately apparent th a t HJ] and
the vacancy complexes have very different excited sta te lifetimes. However, a direct
determ ination of the lifetime from the w idth may be uncertain due to inhomoge-
neous broadening, instrum ental resolution, and dephasing at elevated tem peratures.
Therefore, a transien t bleaching (pum p-probe) technique was used for tim e dom ain
m easurem ents of the lifetimes.
T he transient bleaching signal S<> was m easured as a function of tim e delay be
tween the pum p and probe pulses, as described in C hapter 2. T he experim ents were
performed using the high-power, IR. free-electron laser (FEL) at the Thom as Jeffer
son N ational Accelerator Facility (T.JNAF). T he FEL delivered pulses w ith a time
duration of ~ 1 ps, spectral w idth of 1 2 cm -1 , and pulse energy of 1 p J a t a pulse
repetition ra te of 18.6 MHz.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 5. HYDROGEN ST R E T C H MODES IN SILICON 80
1 .0 -
HV*VH,(110)
0.8 -
0.6A
0) T = 291 + 4 ps0 .4 -
0.2
0.00 2 0 0 4 0 0 6 0 0 8 0 0
0 .8 -
0 .6 -
co0 .4 -
0 .2 -
0.00 20 4 0
Time Delay (ps) Time Delay (ps)
F igu re 5.4: Transient bleaching signal from the 20(>2-em-1 (left side) and the 2072-cm-1 (right side) mode measured at 10 K.
5.3 Results
5.3.1 V ibrational L ifetim e o f and H V V H (n0)
Figure 5.4 shows the transient bleaching signal versus tim e delay measured a t a tem
perature of 10 K w ith the laser frequency centered either on the 2062-enm1 (left) or
the 2072-cni-1 line (right). The signal of the former decreases w ith a tim e constant
Ti = 4.2 ± 0.2 ps (left side of Figure 5.4). The lifetime of Ti = 4.2 ± 0.2 ps ob
tained directly from the transient bleaching experim ent is longer than the Ti = 1.9
ps estim ated from the w idth of the 2062.1-cm-1 line measured in a low hydrogen
concentration sample [101]. This indicates th a t the low -tem perature linewidth is still
broadened by intrinsic (homogeneous) dephasing processes. In addition, a long-lived
tail is clearly present in the decay of S*. Since the laser spectrum has a FW HM
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 5. HYDROGEN ST R E T C H MODES IN SILICON 81
of ~ 12 cm -1 , wo expect th a t Si, also exhibits a response; from the; ‘2068- and 2072-
cm _ 1 modes. Indeed, by centering the laser frequency on the 207'2-cm-1 mode, the
long-lived com ponent of St, is enhanced; i.e., it corresponds to the stretch m ode of
H V ■ V i7(no), possibly with a small contribution from H V V (right side; of F igure 5.4).
T he lifetime of the H V ■ V H ^ W) s tretch mode; is constant w ithin experim ental error
up to 60 K, with a mean value of 7) = 291 ± 4 ps. The; contribution of the; 2068-
crn- 1 line to Si, is expee;ted to be very small since the; transient bleae:hing signal is
a nonlinear function of the; laser intensity. In addition, the HV-VH(n0) and HV-V
defects arc; structurally very similar, and we1 therefore; expect them to have similar
lifetimes. Both signals in Figure' 5.4 exhibit a fast transient, which e;oineidcs w ith the
pum p pulse. This instantaneous response results from the nonresonant excitation of
the bulk crystal.
5.3.2 T em perature D ependence o f L ifetim e and D ecay Chan
nels
Figure 5.5 shows T i versus tem peratu re for HJ (left side) and HV-VH(110) (right
side). In the case of the 2062-cm_1 moele, T x is nearly constant up to ~ 60 K, where
it s ta rts to decrease, reaching half of its low-tem perature value a t 130 K. T he solid
lines in Figure 5.5 are fits using Eq. (2.9) in chapter 3 with the v ibrational relaxation
channel of the 2062-cm-1 m ode represented by a set of six accepting modes {165,
165, 165, 522, 522, 522} cm -1 . This set of accepting modes is very similar to th a t
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 5. HYDRO GEN STR E TC H MODES IN SILICON 82
350
4.5 HV-VH( 110)300
4.0
250— 3.5
2> 2003.0
2 X 521 cm '1 + 3 x 344 cm'1
3 x 522 cm'1 + 3 x 165 cm'1
== 2.5 □ 150
2.0100
i1.5
150 100
Temperature (K)
200 3000 50 100
Temperature (K)
F igu re 5.5: Temperature dependence of Ti of the 2062-cm-1 mode of FT, (left side) and the 2072-cm-1 mode of HV-VH(110) (right side). The solid lines are fits from Eq. (2.9) in chapter 2 for decays into the sets of accepting modes {165, 1(55, 165, 522, 522, 522} cm-1 and {343, 343, 343, 521, 521} cm-1 of the 2062- and 2072-cm-1 modes. The dashed lines are fits of lowest-order decay channels.
of' the 1998-cm-1 mode of bond-center H in Si, which is well described by decays
into six accepting modes, {150, 150, 150, 516, 516, 516 } cm - 1 [44]. In contrast,
the lifetime of the 2072-cm-1 mode sta rts to decrease at 120 K reaching half of its
low-tcm perature value a t 250 K. The decay channel of the 2072-cm-1 mode is well
represented by five accepting modes with frequencies {343, 343, 343, 521, 521} cm -1 .
We note th a t the v ibrational relaxation channels are not uniquely determ ined by the
tem perature dependence of the lifetimes. In particular, the high-frequency phonon
modes of the decay channels do not have a strong influence on the tem perature
dependence of the lifetimes. For example, the tem perature dependence of the 2072-
cin - 1 mode is equally well represented by a set of accepting modes of {343, 343, 343,
343, 700} cm -1 . However, the wave num ber of the accepting mode w ith the lowest
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 5. HYDRO GEN ST R E T C H MODES IN SILICON 83
frequency is quite' well determ ined by the onset of the tem peratu re dependence of
T i, for example, 60 K for IT2 and 120 K for HV-VH(n0). F igure 5.5 also shows th a t
in bo th eases the calculations using Eq. (2.9) deviate from the m easured lifetimes at
elevated tem peratures, indicating th a t additional relaxation channels may increase
the decay ra te at higher tem peratures. We note th a t the low-frequency accepting
m odes of 165 cm - 1 for (and 150 cm ” 1 for bond-center hydrogen) nearly coincide
with transverse acoustic phonons of the undistorted Si crystal (Figure 5.6), which
have been calculated in Reference [113]. Likewise, the 343-crn- 1 inodes involved in
the decay of the stretch mode of HV-VH(110) nearly coincide w ith the peak in the
longitudinal acoustic phonon density (Figure 5.6). T he natu re of these accepting
modes is still unknown. However, likely candidates are pertu rbed optical, acoustic
an d /o r pseudolocalized modes (PLM s) of these' defects.
TO
TA
LA
100 300 500
Wave Numbers (cm 1)
F igure 5.6: Phonon spectrum of Si adapted from Reference [113].
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 5. HYDROGEN ST R E T C H MODES IN SILICON 84
It has been accepted th a t high-frequency vibrational modes decay by lowest order
according to N itzan’s theory [56, 57]. T he coupling strengths are proportional to
the derivative of the potential energy surface1 w ith respect to the norm al coordinates
of the Si-H stretch mode and the N< accepting modes, corresponding to enharm onic
term s of order N, + 1 . Since; the m agnitude of the anharmemic teams is believed to fall
off rapidly w ith increasing order, channels w ith small N< are; likely to dom inate the
decay. T he decay channel of bo th and HV-VH(no) with lowest order is the decay
into four modes with frequeneaes at about 515 cm -1 , e:orresponding to e>ptie;al phouons
or peessibly Si-related LVMs split off from the phonon bane Is by the strain fielels in
the vicinity of the defects. Figure 5.5 also shows th a t the; tem peratu re dependence
of T i is incompatible1 with the Si-H stretch me)de;s elec:aying predom inantly into four
modes a t abou t 515 cm -1 . This is consistent w ith the fact th a t a com bination of four
optical phouons are not infrared active [69, 70] and therefore are not allowed as a
decay channel. A low-ffequency m ode m ust be involved in the decay process of Si-H
stretch modes in Si. This is different, with the; Si-H moelees in ameuphous Si, which
decays into bending modes via lowesst order [114] since phouons do not show high
sym m etry in am orphous hosts.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 5. HYDROGEN ST R E T C H MODES IN SILICON 85
5.3.3 Structural D ependence o f Lifetim e o f H ydrogen D e
fects
The Hj arid HV-VH(110) complexes show 2 -onlers-of-ma.gnitude difference in their vi
brational lifetimes. This is surprising a t first glance considering th a t bo th Si-H stretch
modes have nearly the same frequency and symmetry, and the com pound density of
accepting phonon states of the unperturbed lattice is almost identical. T he vibra
tional lifetimes of other Si-H modes m easured by infrared absorption spectroscopy
a t the highest resolution possible using sam ples w ith very low defect concentrations
also show a strong structu ra l dependence [1 0 1 ]. Table 5.1 shows th a t the lifetimes of
Si-H stretch modes depend strongly on the bonding configuration of the defects. For
instance, th e in terstitia l-type defects , IH 2 , and H ^ . have lifetimes of 4.2 — 12 ps,
whereas th e lifetimes of vacancy-hydrogen complexes are a t least 42 — 295 ps. The
strong structu ra l dependence m ight be explained by either pseudolocalized modes or
LVMs being involved in the decay process. Being localized around the defect, such
modes have much larger vibrational am plitudes on the atom s close to the Si-H bond
than , e.g., lattice phonons, which m ay cause a strongly enhanced anharm onic cou
pling to the Si-H stretch mode. However, significant distortions of the Si-Si bonds
in th e vicinity of the defects are required for pseudolocalized modes or LVMs to
form [34] which is consistent w ith highly distorted in terstitia l-type defects having
shorter lifetimes th an vacancy-type defects. Furtherm ore, to explain the large differ
ence in lifetimes, one should also consider the anharm onic coupling strength between
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 5. HYDRO GEN ST R E T C H MODES IN SILICON 86
the stretch m ode and the low-frequency modes. In the HV-VH(no) center, it certainly
seems plausible th a t the relatively large open volume of the vacancies would lead to
a small interaction between the hydrogen and the surrounding silicon atom s. This
small interaction would lead to a longer v ibrational lifetime.
DefectSi-H
u> (cm-1 )Si-H
T i (ps)f a 2062.1 1.9 (4.2)
I f a 1987.1 12
I f a 1990.0 11
V fa 2 2122.3 60V f a 2145.1 42V f a 2223.0 56
H V • V H aw) 2072,5 262 (295±6)
Table 5.1: Lifetimes of H-related stretch modes in Si measured at 10 K. To obtain the lifetimes shown in column 3, the linewidths were corrected for instrumental broadening and then entered into equation Ti = l/(27rcro).
T he coupling term of H *2 and HV-VH(110) has been sim ulated by Buckle [78, 115]
using a simple Morse potential, which includes anharm onic effects and is known to
represent the bonding of diatom ic molecules quite accurately. T he model ham iltonian
of the two coupled oscillators is given by
H = - C m ' + “ l p + ^ ' l f + ( w )
where hi and /r2 are the effective masses, and and r 2 are the displacement coordi
nates corresponding to bond stretching, f a r H z is the coupling term which determines
the coupling strength of the two oscillators. In this simulation, the coupling constant
f a of H -2 is about 50 times larger than of HV-VH(110). This calculation shows th a t the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 5. HYDROGEN ST R E T C H MODES IN SILICON 87
two Si-H stretch modes strongly couple to each other in Hj and the energy dissipation
is much faster th an in HV-VH(110).
5.4 Conclusion
Vibrational lifetime m easurem ents of Si-H stretch modes have great significance in
excitation processes which involve the dynam ics of energy dissipation in solids. A
specific example is the desorption of H at the S i/S iO -2 interface [116, 8 8 , 90]. T he ra te
of H loss is a sensitive function of the excitation to vibrational s ta tes of the Si-H bond
and the lifetimes of these states. The results presented here indicate th a t the relevant
lifetime is a strong function of the local defect configuration. Thus quantita tive models
of these processes require detailed knowledge of the local structu re and the associated
local phouons. In conclusion, we have m easured the vibrational lifetime of the Si-H
stretch mode of two structurally distinct, point defects in Si, H£, and HV-VH(110),
as a function of tem perature. We; have shown th a t the dom inating decay channels
of these defects are significantly different, involving low-frequency accepting modes
(165 and 343 cm -1 ) of very different symmetry. It is suggested th a t this highly
selective; behavior is due to the nature; of the local distortions and the associateel
pseudoloealized phonon modes. A more; detailed unelerstaneling of the nature of these
PLM s is required to fully elucielate the v ibrational energy relaxation and transfer
channels in solids.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 6
Interstitial O xygen in Si and Ge
6.1 Introduction
Oxygen is one of the most common and technologically im portan t im purities in
Czochralski (CZ)-grown silicon and germanium, which are the dom inant m aterials
used for the fabrication of electronic devices. Si grown by CZ has a high concentra
tion of oxygen (~ 1 0 18 cm "3), which is incorporated as a result of the dissolution of
the S i0 2 crucible at the growth tem perature. T he concentration of oxygen is less than
1016 cm " 3 in crystals prepared by the; floating zone method. T he oxygen im purity
critically affects the properties and yield of the devices. At high tem peratures it forms
aggregates or clusters w ithin silicon, as well as in terstitia l dislocation loops. These all
d isrupt the la ttice and lead to defect s ta tes in silicon. At lower tem peratures oxygen
can still be a problem, for example through the formation of therm al donors which
occur when silicon is annealed a t tem peratures of around 450°C. These defects are
electrically active, and act as electron-hole recom bination centers, lowering th e mo
bility, etc. For this reason, it is essential to understand exactly how oxygen behaves
88
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 6. IN TE R STITIA L O X YG EN IN SI AND GE 89
in silicon. A full understanding of the bonding and structure, as well as the diffusion
and kinetics of oxygen, will yield to improved m anufacturing processes. This in tu rn
should lead to faster, cheaper, and more efficient microchip technology [117].
Because oxygen is lighter than Si and Ge atoms, it forms local vibrational modes
in Si and Ge crystals. T he most well studied oxygen LVM absorption is known as
the 9-pm line appearing a t 1107 cm - 1 a t room tem perature [118, 119]. Gxvgen
in germanium, first investigated at about the same time [1 2 0 , 1 2 1 ], has a similar
s tructu re as oxygen in Si and gives rise to an absorption line; a t 11.7 pm . Due to
the lower m elting point and lower oxygen affinity of Ge, the intensity of the 11.7-pm
band is two orders of m agnitude sm aller th an th a t of the corresponding 9-pm band
in crucible-grown Si. Recently, oxygen in Ge has been reinvestigated more actively,
due to progress in spectroscopy and to the availability of quasi-monoisotopic (qmi)
germanium samples [122, 123, 124, 125, 126, 127].
In the early work by Kaiser et al. [118], a model was proposed in which the oxygen
atom occupies an in terstitia l position, bonded to two Si neighbors, sim ilar to the Si-
O-Si unit in quartz. T he uniaxial stress experim ents doin' by C orbett [1 2 1 ] confirmed
th a t Si-O-Si complexes are aligned along (111) direction. T he structu re is shown in
Figure 6 .1 . From ab initio calculations [128], it was found th a t the S i-0 bond length
is 1.59 A and the Si-O-Si angle is 172°. W hile in equilibrium geom etry the Ge- 0
bond length is 1.70 A and the Ge-O-Ge angle is 140° [127].
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 6. INTER STITIAL O X YG EN IN SI AND GE 90
[111]
F igu re 6.1: Structure of interstitial oxygen in Si and Ge
6.1.1 V ibrational Spectroscopy o f In terstitia l O xygen in Si
and Ge
Infrared v ibrational spectroscopy is one of th e most powerful techniques to charac
terize oxygen defects in Si and Ge. The in terstitia l oxygon in Si and Ge has been
assigned to the v irtual molecule w ith Czv sym m etry [129, 130, 131]. Early studies
on in terstitia l oxygen in silicon (Si:0.i;) showed two main absorption features in the
infrared ( 519 cm - 1 and 1136 cm -1 ) and another structu re in the far infrared (29
c n f 1). T he three m ain features were associated w ith the th ree fundam ental v ibra
tional modes of a puckered Si20 molecule, v\ (the 519 cm - 1 peak), u2 (the 29 cm - 1
band), and ^3 (the 1136 cm - 1 or 9 pm stretching band), as displayed in Figure 6.2.
T he analysis of the fine s tructu re of the high- and low-frequency features [129] of
fered insight into the natu re of the microscopic: s tructu re of the oxygen center and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 6. IN TE R STITIA L O X YG EN IN SI AND GE 91
its dynam ics. In spite of the correctness and m erit of this analysis, it was only re
cently shown th a t the puckered image was misleading, a be tte r one being th a t of
a pseudo-molecule Si3 = S i-0 -S i= Si3 w ith dynam ic Dzd sym m etry [132], T he as
signments were consequently revised: the 519 cm " 1 peak is not a vibration of the
Si2() pseudomolecule (not shown in Figure 6 .2 ), b u t a backbond Si-Si vibration of E,t
sym m etry of the corresponding D^d point group. T he stretching m ode (u3) remains
essentially unchanged, being of A2,, symmetry. T he equivalent to the vx m ode is an
A ig sym m etric stretching m ode invisible to the infrared, located at around 600 c u r 1.
T he observed com bination mode a t 1750 cm - 1 is a com bination of the A1(y and A2it
modes, or (vi+vz). Its isotope shift strongly supports this new model [133].
Ge \
V \ v<i 1/3
Si
E u A \g A^U
F igure 6.2: Fundamental vibrational modes of (a) the puckered Ge2() and (b) the linear Si20 quasimolecule. The traditional molecular' nomenclature has been adopted for Ge20. For Si20 the nomenclature corresponds to the D-̂ d point group, appropriate for the linear structure.
In the case of in terstitia l oxygen in germ anium (Ge:0,,), infrared-absorption bands
have been reported near 860 and 1270 cm -1 , which are equivalent to the 1136 and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 6. IN TE R STITIA L O X YG EN IN SI AND GE 92
1750 c u r 1 bands in Si:()j. Low-energy absorption lines have also been detected by
phonon spectroscopy, a t a much lower energy scale than the far-infrared absorption
in Si [123], but there is no report on any feature in the interm ediate frequency range.
The excitation dynam ics of oxygen in Si and Ge has been well studied. It was
observed th a t there are a few side bands on the low-energy side of 1136 cm - 1 line
when the tem peratu re was raised above 20 K [129] (see Figure 6.3). Y am ada-K aneta
[132, 134] e.t al. have quantified the coupling interaction by form ulating the Hamil
tonian for the problem as tin; sum of the low-energy mode' w ith a phenomenological
po tential a r '2 + j3r4, the one-dimensional simple harmonic i/3 mode, and an anhar
monic coupling between them . This model proposes a two dimensional low energy
anharm onic excitation of the oxygen complex due to the; coupled A%u and A 1(/ modes.
This coupling causes the fine structu re of 9-p band.
30
25
20
15
10
1050 1100 1150 1200 1250
Wavenumber (cm'1)
Figure 6.3: Absorption spectrum of the asymmetric mode of 160.; in Si measured at 60 K showing the side bands on the lower energy side.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 6. IN TE R STITIA L O X YG E N IN SI AND GE 93
3.0
Ge:U0 Paderbcm @ fcK2 .5 -
2 .0 -
§
0 .5 -
0 .0 -
863 864860 861 862
Wavenumber (cm1)
F ig u re 6.4: Absorption spectrum of the asymmetric mode of ()v in natural Ge at 6
K. The spectrum shows the complex fine structure arising from different Ge isotopes and pairs of lines due to coupling between ^3 mode*, and oxygen rotational motion. Each peak is identified by the average Ge isotopic mass. The average linewidth is ~ 0.04 e n f *.
G e .
#
o
is.
■ A ;-Y »• G e
F ig u re 6.5: Delocalization of oxygen in Si:0» and Ge:0,; in the plane perpendicular to the Si-Si (Ge-Ge) axis. The O, motion is radial in Si and rotational in Ge.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 6. IN TE R STIT IA L O X YG EN IN SI AND GE 94
The absorption band of oxygen stretch mock; in Ge (1/3 ) has a different structu re
than in Si. Unlike Si, where 28Si is much more abundant th an 29Si and 30Si, the Ge
isotopes occur with com parable abundances. As a result, a t very low tem peratures,
the 1/3 m ode of oxygen in Ge exhibits a complex spectrum as shown in Figure 6.4,
w ith as m any as 44 sharp lines within a 5-em - 1 range associated with the 11 distinct
isotopic com binations of the two Ge atom s [125], On the o ther hand, oxygen in Ge
does not show a two-dimensional motion, bu t a ro tation (see1 F igure 6.5). T he coupling
of th is ro tational m otion and asymmetric' stretch mode 1 /3 gives rise to additional lines.
T he microscopic! s tructu re of the defect center and the excitation dynam ics of
in terstitia l oxygen in Si and Ge are well understood. This is an ideal system for
studying vibrational relaxation mechanisms in semiconductors, which are im portant
to elucidate diffusion and defect reactions involving oxygen im purities [135], Re
cently, McCluskey and co-workers used hydrostatic pressure; to bring the asym m etric
stretch inode of 18(); in Si into resonance w ith a spatially extended mode and studied
the resonant interaction between these two modes [6 8 ]. The; au thors sta ted th a t the
transition from a localized to an extended mock' m ight be the; first step toward decay
into lattice phouons. However, little is known about the; actual decay dynam ics of
the local vibrational m ode into lattice' phouons.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 6. IN TE R STITIA L O X YG EN IN SI AND GE 95
5 - Iso to p e e ffe c t a t 6 K of th e a sy m m etric 4 )um o d e o f (3 in
silicon d o p e d with l70 - a n d
' “O -e n ric h e d o xygenEo
co>o<DOOcoQ.O(/>n<
1080 1100 1120 1140
Wavenumber (cm'1)
Figure 6.6: Absorption spectrum of the asymmetric stretch mode of 0 ? for different O isotopes in Si samples enriched with 170 and 180 . The linewidth of 170 mode is 1.2 cm-1 , while it is 0.6 c n f 1 for 160 and 18().
6.1.2 Isotope Effect o f O xygen in Si
In the past, isotope substitu tion of the oxygen center with 16(), 170 , and 180 has
been useful for experim ental understanding of oxygen-related LVMs in Si and Ge.
However, the1 full w idth at half maxim um (FW HM ) of the asym m etric stre tch inode
of 170,t in Si (1109 cm -1 ) is a factor of 2 larger than for the 16Oi (1136 cm -1 ) and
18()i mode (1085 cm -1 ) as shown in Figure 6 .6 , a puzzle th a t has not been solved
since 1995 [133]. This broadening does not depend on the; 17G concentration and
it is not observed for the low-frequency mode a t 29 cm ” 1. Thus, the ground s ta te
is not broadened. T he broadening m ust be from the excited state. In principle, a
broadening could result from the splitting of the 5 /2 spin level of 17() by quadrupole
interaction with an electric field gradient in the v ibrational excited state . However,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 6. IN TE R STIT IA L O X YG EN IN SI AND GE 96
in order to explain the observed broadening for i/3 (170 ), one would need a very large
field gradient, about a hundred times larger than typically found for ground states
in sem iconductors [133]. T he effect was not observed in Ge which rules out th a t it
is intrinsic to 170 . T he effect, may be related to the vibrational lifetime and to the
coupling between the local modes and the lattice phonons. It is therefore crucial to
m easure the vibrational lifetimes of oxygen isotopes in Si.
T he other interesting question is th a t the linewidth of 16() in Ge is much sharper
th an in Si. Infrared absorption m easurem ents show th a t the average line w idth of
160 m ode in Ge is 0.04 cm -1 , i.e., a factor of 10 narrower th an in Si [122]. In terstitial
oxygen in Si and Ge have; similar defect structures and excitation dynamics. The
different linewidths may indicate; different relaxation dynam ics of these two modes in
Si and Ge;. F igure 6.4 shows the; absorption spectrum measured a t 7 K a t a resolution
of 0.02 cm - 1 in Ge sample.
6.2 Experim ental
N atural Si is m ade up from 28Si (92.2%), 29Si (4.7%) and 30Si (3.1%), and its isotope
shift of v ibrational lines can be observed by infrared absorption spectroscopy. The
composition of oxygen in Si consists of 160 (99.76%),17() (0.04%) and 180 (0.2%).
A float-zone grown silicon sam ple enriched w ith oxygen isotopes by diffusion is used
for the lifetime experim ent. T he concentrations of 16(), 1' 0 and l8() isotopes are 20,
25 and 2 .4x lO 16 cm -3 , respectively. The germanium sample' contains isotopes 70Ge,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 6. IN TE R STIT IA L O X YG EN IN SI AND GE 97
72Ge, 73Ge, 74Ge, 77Gc, w ith relative (natural) abundances of 0.205, 0.274, 0.078,
0.365 and 0.079, respectively. The oxygen concentration of the Ge sam ple is 4.9 x 1016
cm -3 . The lifetime of the asym m etric stretch mode of in terstitia l oxygen isotopes is
measured directly by transien t bleaching spectroscopy. We use an IR pum p-probe
setup, as described in C hapter 3. The; TOPAS used in this experim ent delivers pulses
at 1-kHz repetition ra te w ith a tim e duration of ~ 1 .7 ps, spectral w idth of 16 cm -1 ,
and a pulse energy of 18 p.J a t 1136 cum 1 and 1109 cm -1 , and 9 //..I a t 861 cm -1 .
T he transient bleaching signal S& is detected using an amplified liquid nitrogen-cooled
HgCdTe detector. To allow for low tem peratu re m easurem ents, all experim ents are
performed with the sam ple m ounted in an optical cryostat.
6.3 R esults and Discussions
6.3.1 L ifetim e M easurem ents
Figure 6.7 shows a semi-log plot of So vs tim e delay from the asym m etric stretch mode
of 16Oj and 170 ,; in Si a t 10 K. Both decay profiles fit well to a single exponential
function, yielding a lifetime T\ = 11.5 ± 1 ps for 160 ,;, and T\ = 4.5 ± 0.4 ps for irO,j;.
We could not measure1 the lifetime of th e 180.; mode because of its low concentration.
T he d a ta shown in Figure 6 .8 are m easured at a center frequency of 1123 a n -1 , i.e.,
between the two modes. The curve does not exhibit any decay process, bu t only
the laser pulse with 1.7-ps width. T he TOPAS spectrum with a FW HM of 16 cm - 1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 6. IN TE R STITIA L O X YG EN IN SI AND GE 98
o1136 cm
“ -2
-4■6 0 6 12 18 2 4
01109 cmI'! = 4.5 + 0 .4 ps
•2
•4- 4 - 2 0 2 6 8 10 124
Time Delay (ps) Time Delay (ps)
Figure 6.7: (a) The transient, bleaching signal Si, decays exponentially with a time constant Ti = 11.5±1 ps for the asymmetric stretch mode of 16Oj in Si tit 10 K; (b) while T\ = 4.5±0.4 ps for 17 Oi.
(Figure 6.9) is therefore narrow enough to distinguish between tin ' two vibrational
modes of the oxygen isotopes in Si.
T he narrow band w idth enables us to tune the wavelength into resonance with
the 16Oi and 1 7 Ot m ode in Si. T he 1.7-ps pulse w idth and 16 cm - 1 band w idth are
ideal param eters to measure' the vibrational lifetime of these two modes. If the band
w idth is too narrow, the pulse will be1 too long to m easure the lifetime of 17Oi mode
since the lifetime of 1 7 Ot is only 4.5 ps. If the band width is too wide, it will not
distinguish the two modes.
Figure 6 .6 shows the infrared absorption spectrum of the asym m etric stretch mode
of different oxygen isotopes m easured from the Si sam ple by B. P a jo t in Paris. This
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 6. IN TE R STIT IA L O X YG EN IN SI AND GE 99
0.004
Gauss fit of laser pulseF W H M -1.67 psLaser wavelengh~1123 cm'1
0.003
0.002
_oCO
0.001
0.000
0-5 5 10
Time Delay (ps)
Figure 6.8: The transient bleaching signal S/, measured at a center frequency of 1123 cm-1 , which lies between the 160,; and 17O, mode. The signal does not show any decay. It is fitted by a Gaussian function, giving the pulse width of ~ 1.7 ps.
0.25
0.20Pulse Band Width - 16 cm1
0.15
toca>
0.05
0.00
1000 1050 1100 1150 1200 1250
Wave Number (cm'1)
Figure 6.9: The laser pulse spectrum has a band width of 16 cm” 1 at /nn measured with a monochromator.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 6. IN TE R STIT IA L O X YG EN IN SI AND GE 100
m easurem ent was perform ed at liquid helium tem peratu re using a BOM EN DA3
Fourier transform spectrom eter w ith spectral resolution of 0.1 c m -1. T he shape of an
absorption line is generally given by the convolution of its homogeneous line shape
w ith a function describing the inhomogeneous broadening [136]. T he FW HM of the
1 7 Oi line is ~ 1 .2 cm -1 , while for 16Oj and 180.; it is ~0 .6 (an - 1 a t 7.5 K. We also
m easured the linewidth of 160,; mode in a Si sam ple with lower oxygen concentration
1016 cin-3 ). T he linewidth of the 16O, m ode at 10 K was found to be F = 0.55
cm -1 , which is shown in Figure 6.10.
0.26
0.244>O C (0f 0.22<o S3 <
0.20
0.181132 1134 1136 1138 1140
W avenum ber (cm '1)
F igure 6.10: Absorption spectrum of the asymmetric stretch mode of 160.; measured in a fioat-zone Si sample: with oxygen concentration of ~ 0.2 ppm. The linewidth is 0.55 cm-1 .
The natu ra l linewidth deduced from the lifetime is 1.12 cm - 1 for 17Oj, and 0.44
cm - 1 for 1 6 Oi. O ur results in the tim e dom ain are therefore consistent w ith those in
the frequency domain, which shows th a t the absorption lines of in terstitia l oxygen are
homogeneously broadened in crystalline: silicon a t low tem perature. This consistency
0.55 cm
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 6. IN TE R STIT IA L O X YG EN IN SI AND GE 101
has been shown for hydrogen-related modeds in crystalline silicon in the tim e and
frequency dom ain [44, 101]. We can therefore deduce the lifetime of the lsO, mode
from the linewidth to be ~ 10 ps, close to th a t for 1 6 0{. The difference (factor of two)
between the linewidths of 17Oi com pared w ith ieO,; and 17O t is also observed in their
lifetimes.
Ge: H,0 862 cm"*T, = 125 ± 10 ps
0
cn •2
•3
-4o 50 100 150 200 250
Time Delay (ps)
Figure 6.11: Tlie transient bleaching signal S*, decays exponentially with a time constant Ti = 125 ± 10 ps for the asymmetric stretch mode of 16Oi in Ge at 10 K.
In order to elucidate the difference of the linewidth of ieO ; in Si and Ge, we
measured the v ibrational lifetime of 16Oj; in Ge. Figure G .ll shows the lifetime of
the asym m etric stretch mode, which was m easured to be Ti = 125 ± 10 ps at 10 K.
Since the absorption lines of the 16Oj m ode for different Ge isotopes are very close,
our laser pulses cam m easure only an average lifetime. This corresponds to an average
linewidth of 0.04 cm -1 , which is consistent with the absorption spectrum shown in
Figure 6.4. This indicates th a t a t low tem peratu re and low defect concentration the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 6. IN TE R STIT IA L O X YG EN IN SI AND GE
160 i m ode in Ge is dom inated by the natural linewidth.
102
6.3.2 M ultiphonon D en sity o f S tates and D ecay C hannels
T he asym m etric stretch mode of Oi lies in the three-phonon continuum of Si and is
expected to decay by a three-phonon process. It is therefore necessary to calculate
p{l/] = dE 2 ■■■ d E N-ip i{fiw - Ei)p-2(E i - E 2) . . . P n{E N- i). (6 .1 )
T he one-phonon density of states was obtained from ab initio calculations, which
are available in the literature! [137, 138]. T he dispersion curves in F igure 6.12 are
for Si and Ge. T in 1 graphs on the far right side show the density of states, which
are obtained by integrating over the dispersion curves, giving the density of phonon
identical crystal structure, so their dispersion and density of state's curves are very
similar, with Ge! having lower frequencies because its mass is higher th an Si. Plionon-
dispersion relations can be determ ined experim entally by inelastic neutron scattering.
The m arkers along the curves in Figure 6.12 are! experim ental d a ta points. The
theoretical calculations are very reliable.
Figure 6.13 shows the three-phonon density of s tates for Si calculated by stra igh t
forward convolution of the one-phonon density of sta tes aee:ording to Eq. (6 .1 ). The
the com pound three-phonon density of states. T he com pound three-phonon density
is given by
inodes as a function of frequency (in unit of waveuumber, cm r ). Si and Ge have
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 6. IN TE R STITIA L O X YG EN IN SI AND GE 103
Si15 TO
4i'A DOS
o r rK X LGe
TO15
1.012
Kfc. 8
4
TA DOS0 r rk x L
F ig u re 6.12: Phonon dispersion curves and density of states for Si (top) and Ge (bottom). Si figure is courtesy of Wei [138], Ge figure is courtesy of Giannoz'/i et. al [137].
1 1 Oi mode happens to fall right on the peak of the 2TO + 1TA phonon com bination,
whereas the 160 ,; and 18Ch mode are on both sides with sm aller combined phonon
densities. High phonon density corresponds to a short lifetime. The1 difference of the
phonon density provides a satisfactory explanation for the difference of the lifetimes.
T he ratio of the phonon densities at the frequencies of the 170 , and Ui0.; mode is ~1.3,
which is smaller th an the ratio of the lifetimes (~2.5). The overall agreem ent between
the calculations and the experim ent is satisfactory, bu t is by no m eans perfect. In
a full trea tm ent of v ibrational relaxation by anharm onic coupling between LVM and
lattice phonon modes, one should consider modifications introduced by the distortion
of the lattice because of the im purities, which may affect the phonon density of states
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 6. IN TE R STIT IA L O X YG EN IN SI AND GE 104
locally [139].
2 TO + TA
</)c<13Qcoco 2 TA + TOaCl
q>CD
LA+ 2 TO0.6
3 TO. cH0.4
3 TAccEoz
0.2
0.0
200 400 600 800 1000 1200 1400 1600
Frequency (cm'1)
Figure 6.13: Normalized tliroe-plioiion density of states in Si at 0 K. The asymmetric: stretch mode of 17 0 \ coincides with the highest three-phonon density in Si, whereas the ieO; and 18Oi modes are on both sides of the 2TO + 1TA phonon peak.
Recently, Kat.o and coworkers reported a surprisingly large effect of Si isotopic
mass disorder on the linewidth of the asym m etric stretch mode of 160 .;, which cannot
be explained well by their calculations [140]. O ur lifetime' m easurem ents indicate'
th a t the line broadening of the 160 , mode is dom inated by the natu ral linewidth at
low tem peratures. The linewidth of this mode in quasi-mono-isotopic (qmi) 29Si and
30Si samples are narrower th an those in natu ral Si and 28Si. This can be explained
by the isotopic shift, of the phonon frequency ra ther than tin' isotopic mass disorder.
Passing from 28Si to 30Si, the phonon frequency will shift, to a. lower wave number,
which results in lower phonon densities for 29Si and 30Si than for natu ral Si a t the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 6. IN TE R STIT IA L O X YG EN IN SI AND GE
frequency of the 160 ,; mode.
105
Ge
3 TO0.6
£ 0.4
=5 0.2
0.0
200 400 600 800
Frequency (cm 1)
F ig u re 6.14: Normalized three-phonon density of states in Ge at 0 K. The 160,; and 17Oi inodes lie in the 3TO phonon band, whereas the lsO; mode coincides with a 2TO + 1LO phonon process.
To elucidate the long lifetime of the asym m etric stretch mode of 160 ; in Ge requires
knowledge of the decay channels and phonon density of states. The 862-cm-1 mode
falls into the three-phonon continuum since the cut-off frequency of phouons in Ge
is 306 cm -1 . Figure 6.14 shows the calculated three-phonon density of states in Ge
similar to the one of Si. The 16Oj; and 170 , modes fall on the 3TG phonon peak in Ge
while they coincide w ith 2TO + TA phonon processes in Si. The oxygen isotope effect
has been investigated also in Ge [141]. A resonant broadening of the 17Oi absorption
line com pared w ith 16Oj and 180 ?; has not been observed in Ge, bu t the individual
lines of 18Cb are broader th an those for 1(,O t and 170 ,:. The 16O j and 17O i lines coincide
with the same peak (3TO phonons), bo th of them have the same decay channel and
their three-phonon density of states is close. So they show the same linewidth in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 6. IN TE R STIT IA L O X YG EN IN SI AND GE 106
Ge. T he 18Oi mode in Ge may have; different coupling constant since it decays into a
different channel (2TO + LO phouons).
6.3.3 Sym m etry A nalysis
T he difference in phonon density between the 2TO + 1TA and 3TO phonon combi
nations is only a factor of 2~3 , which is not enough to explain the large difference in
lifetimes of 1 6 Ot in Si and Ge. Since the decay channels are different, it is necessary
to account for the different coupling constants (?{„}, representing the strength of the
interaction between the local mode and the accepting modes. T he asym m etric stretch
m ode of in terstitia l oxygen in Si and Ge has A-2u symmetry, which is infrared active in
the D 3(i point group of the 0 ; defect. Tin; three-phonon com binations m ust contain
A'2u sym m etry [20], or otherw ise the decay channel is not allowed. T he infrared dipole
operator whose irreducible representation in the 0 /,. group is T15/ can be reduced to
A-2u + E u sym m etry in the D 3d group, since D3,; (group of L in the diam ond structure)
is a subgroup of ()/,. So we need to look into the infrared activity of three-phonon
processes in the diam ond structure. This has been published in Ref. [70]. Table 6.1
lists the com binations of 2TO + TA phonons in Si and 3TO phouons in Ge which
are infrared active [70]. Activity (M) is the coefficient of T15/ in the reduction of
the direct sum of sym m etry elements of three-phonon combinations. Larger M corre
sponds to a more infrared active three-phonon process [70]. Table 6 .1 shows th a t the
com binations of 2TO + TA phonons are more IR-aetive than 3TG phonon processes.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 6. IN TE R STIT IA L O X YG E N IN SI AND GE 107
The interaction will be stronger between the A 2u local mode and 2TO + TA phonons
th an w ith 3TO phonons. In addition, there an; more channels for the A 2u mode to
decay into 2TO + TA phonons, which results in a shorter lifetime of the oxygen mode
in Si.
Activity (M) Phonon com binations Frequency (cm x)Silicon
3 T 0 { X ) + 0 { F ) + T A ( X ) 11296 T O ( L ) + O ( r ) + T A ( L ) 1 1 2 2
3 2T O { L ) + T A ( X ) 11346 T O ( L ) + T O ( X ) + T A ( L ) 10712 2 T O { X ) + T A ( X ) 1078
Germ anium1 2 T O { X ) + O ( P ) 8563 2 T O { L ) + 0 (F ) 8 8 8
T able 6.1: Threo-phonon processes in Si and Ge close to the frequency of the asymmetric stretch mode of interstitial oxygen. The infrared activity M is also listed.
6.4 Conclusion
In conclusion, we have measured the v ibrational lifetimes of the asym m etric stretch
m ode of oxygen isotopes in silicon and germanium. In silicon, the 1 1 0.-, m ode lies in
the highest density of three-phonon states, which gives rise to a shorter lifetime (7 \
= 4.5 ps) than for the 160* and 180 ,; modes (7 \ ~ 10 ps). We also m easured the
lifetime of 160 ,; mode? ? in Ge, which show a much longer lifetime, T t — 125 ps, th an in
Si. We argue th a t the oxygen modes in Si decay into 2TO + TA phonons, while they
decay into 3TO phonons in Ge. These phonon com binations have different infrared
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 6. IN TE R STIT IA L O X YG EN IN SI AND GE 108
activities, which result in different coupling strengths and very different lifetimes. A
more quantita tive understanding of the v ibrational energy relaxation and transfer
channels in solids requires detailed knowledge' of anharrnonic coupling between LVM
and lattice phonon modes and phonon density of sta tes modified by im purities. The
m easured lifetimes presented here have great significance in elucidating the stability
and diffusivity of im purities in sem iconductors as well as complex formation and will
generate renewed interest in the dynam ics of energy dissipation from local modes into
solid-state phonon bath.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 7
Sum m ary and Future Work
7.1 Summary
The study of relaxation dynam ics of local v ibrational inodes down to the picosecond
timeseale is a now emerging field. In this dissertation, we have studied in detail the
v ibrational lifetime of various defects in sem iconductors by vibrational pum p-probe
spectroscopy. This study of vibrational lifetimes elucidates the decay dynam ics of
local v ibrational modes and provides unique inform ation about energy relaxation
processes to solid-state phonon bath . According to our studies in this dissertation,
there are five im portant factors which determine! the lifetime of local v ibrational
modes: ( 1 ) Order o f decay channel, (2)sym m etry o f the accepting modes, (3 )density o f
states o f accepting phonon inodes, (4 )anharm onicity , and (5) defect, structure.
V ibrational lifetimes of H- and D -related bending modes are determ ined by the
decay order, i.e., the decay tim e increases exponentially w ith increasing decay order.
This provides an estim ate of the lifetimes of m any other hydrogen bonding modes in
solids. For instance, when the local mode couples to only one phonon, the decay tim e
109
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 7. SU M M A R Y AND FUTURE W O RK 110
is around 1 ps or less. Two-order decay of hydrogen bending inodes corresponds to a
decay tim e of ~ 10 ps, three-order processes correspond to 40-100 ps, while the decay
tim e of four-order processes can be longer than 250 ps. This so-called energy-gap
law shows th a t the; lifetime of D -related bending modes are generally shorter than
H -related bending modes. This result can explain the large H /D isotope effect in
H-passivaster! sem iconductor devices.
In contrast, v ibrational lifetime's of H -related stretch modes critically depend on
defect structure. The in terstitia l-type defect HJj a t 2062 cm - 1 has a lifetime of 4.2
ps at 1 0 K, whereas the lifetime of the vacancy-type complex HV-VH(no) a t 2072
cm - 1 is 2 orders of m agnitude longer. The; decay mechanism of stretch modes is more
complicated than for bending modes. The accepting modes of hydrogen stretch modes
can lie phonons, bending modes, an d /o r pseuelolocalizoel modes. More theoretical
work is needed.
The study of in terstitia l oxygen and its isotope effects in Si and Ge shows th a t
both density of states and sym m etry of accepting modes are' im portant to eietermine
their v ibrational lifetime's. T he lifetimes of in terstitia l 160 and 18Q m ode is two timers
longer th an 170 m ode in Si. A calculation of the three'-phonon density of states
shows th a t 1,() lies in the highest phonon density resulting in the shortest lifetime.
The lifetime of 160 inoeles in Ge is measured to be ~ 10 times longer than in Si.
Sym m etry analysis reveals th a t the accepting moeles of in terstitia l oxygen in Si are
more infrared active and have more decay channels than in Ge .
Until now, no first-principle calculatiems have been done of the' vibrational life
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 7. SU M M ARY AND FUTURE W O RK 111
times of hydrogen or oxygen in crystalline semiconductors. T he level of theoretical
investigations into the dynam ics of these modes is lews advanced. We believe th a t the
m easured lifetimes presented in this dissertation provides an indispensable benchm ark
for fu ture theoretical calculations.
7.2 Future Work
In th is dissertation, the v ibrational lifetimes of H- and O- related defects are investi
gated in intrinsic: sem iconductors and the decay of LVMs are governed by m ultiphonon
relaxation processes. T he decay of LVMs via electronic processes represents an excit
ing new direction for this project. A lthough the energy of LVMs is small com pared to
the band-gap of' an intrinsic semiconductor, bo th p- or n- doped sem iconductors and
laser-induced free carriers provide energetically accessible; electronic channels. The
existence of electronic pathways will undoubtedly influence the population lifetimes
and linewidths.
One of the: m ost interesting systeaus is the O-H complex in ZnO, which is a wide
band gap sem iconductor th a t has a ttrac ted trem endous interest for optical, electronic,
and mechanical applications [142, 143]. It provides a possible alternative to GaN for
use in optoelectronic devices. Bulk ZnO single' crystals can be grown and they nearly
always exhibits n-type conductivity. Recently, Van Do Wallc> performed theorectical
calculations and proposed th a t H im purity in ZnO is a shallow donor th a t may be
introduced into the bulk during growth or processing [144]. This is unexpected be
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 7. SU M M A R Y AN D FUTURE W O RK 112
cause in all other sem iconductors, in terstitia l hydrogen has been found to counteract
the prevailing conductivity caused by extrinsic dopants. In ZnO, however, hydrogen
occurs exclusively in the positive; charge; state;, whieli can e;xplain the typical n-type
eionehmtivity of ZnO [145, 146].
By annealing ZnO in H 2 gas at ele;vateel te;mpe;rature, McCluskey et al. have;
observed a v ibrational line a t 3326.3 cm - 1 anel have asseKiiated this line; w ith O-H
stretch m ode eef a H -related donor [147]. F igure 7.1 slmws the s tructu re anel the
absorption line; of this stretch moele:
[0001]1.00
0.95
0.85o03XI<
0.75
0.70
3320 3322 3324 3326 3328 3330 3332Anti bondingFrequency (cm )
Figure 7.1: Structure of O-H bond in ZnO anel absorption spectrum at 3326 cm-1 [151].
We note th a t the absorption line of the O-H vibrational moele' a t 3326.3 cm - 1
is asym m etric. W ith elecre;asing concentration, the line be;e:oirie;s sharper anel more
sym m etric. In boron-eloped Si, this phenomenon was observed by M. Stovala [148].
Tlie acceptor-H vibrational banels are; asym m etric anel broadened by a Farm inter
action w ith the continuum absorption elue to reisidual fine; e:arriers. Fano interaction
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CH APTER 7. SU M M ARY AND FUTURE W O RK 113
effects are well known in Ram an studies of heavily doped m aterials. B oth the Si
optical phonon at 520 cm - 1 and the B local-mode features at 620 and 644 cm - 1
show pronounced Fano-interaetion effects in Si w ith high free-carriers concentration
[149, 150],
The O-H stretch mock' a t 3326 cm - 1 is associated with the H -related donors. W ith
decreasing the H concentration, the free-carrier concentration will also decrease. At
low hydrogen concentration, the Fano-shape is less pronounced [151]. To the best
of our knowledge, nobody has studied how this interaction between local m ode and
free carriers will affect the vibrational lifetime of local modes. It will be interesting
to study how the v ibrational lifetime will change with different H concentrations.
Through the tem perature1 dependence of lifetime m easurem ents, we learn about
the relaxation channel. As discussed above, relaxation through free carriers may
provide another channel for vibrational modes. The interaction between local mode
and free carriers may increase the relaxation rate. This exciting new direction will
lead to a be tte r general understanding of relaxation mechanisms.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Bibliography
[1] S. K. Estreieher. S tructure and dynanm ies of point defects in crystalline silicon. P H YSIC A ST A T U S SO L ID I D, 217:513, 2000.
[2] F. A. C otton, G. W ilkinson, and P. L. Gaus. Basic Inorganic Chemistry. Wiley, New York, 1987.
[3] S. J. P earton, J. W . C orbett, , and M. Stavola. Hydrogen in Crystalline Sem iconductors. Springer-Verlag, Berlin, 1992.
[4] A. Van W ieringer and N. W armholz. Physica, 22:849, 1956.
[5] T. Ichimiya and A. Furuichi. Int. J. Appl. Radial, h o t, 19:573, 1968.
[6 ] Stefan K. Estreieher. Hydrogen-related defects in crystalline sem iconductors: a theo rist’s perspective. Materials Science and Engineering, R 14:319, 1995.
[7] J. J. Pankove and N. M. Johnson. Hydrogen in Semiconductors I: semiconductors and semimetals, vol. 34- Academic Press, New York, 1991.
[8 ] S. M. Myers, M. I. Baskes, H. K. B irnbaum , G. G. Deleo J. W . C orbett, S. K. Estreieher, E. E. Haller, and S. J. Pearton. Hydrogen interactions w ith defects in crystalline solids. Physical Review B, 64:559, 1992.
[9] N orbert H. Nickel. Hydrogen in Semiconductors II: semiconductors and sem imetals. vol. 61. Academic Press, New York, 1999.
[10] Chris G. Van dc Wallc, P. J. H. Denteneer, Y. Bar-Yam, and S. T . Pantelides. Theory of hydrogen diffusion and reactions in crystalline silicon. Physical Review B , 39:10791, 1989.
[1 1 ] G.S. Higashi, Y .J. Chabal, G .W . Trucks, and K. Raghavachari. Ideal hydrogen term ination of the Si (111) surface. Applied Physics Letters, 56:656, 1990.
[12] A. Stem ans. Dissociation kinetics of hydrogen-passivated Pb defects a t the (11 l)S i/S i ( )2 interface. Physical Review B, 61:8393, 2 0 0 0 .
[13] Y. J. Chabal. Physica B, 170:447, 1991.
114
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
BIBLIO G RAPH Y 115
[14] K. L. Brower. Physical Review B , 38:9657, 1988.
[15] K. L. Brower. Physical Review B , 42:3444, 1990.
[16] M. Stavola. Hydrogen in Semiconductors II: semiconductors and semimetals, v o l.51. Academic Press, New York, 1999.
[17] J. W . C orbett, S. N. Sahu, T. Si. Shi, and L. C. Snyder. Physics Letters, 93A:303, 1983.
[18] A. W. R. Leiteli, V. Alex, and J. Weber. Ram an spectroscopy of hydrogen molecules in crystalline silicon. Physical Review Letters, 81:421, 1998.
[19] R. E. P ritchard , M. .J. Ashwin, J. H. Tucker, and R.. C. Newman. Isolatedin terstitia l hydrogen molecules in hydrogenated crystalline silicon. Physical Review B , 57:15048, 1998.
[20] E Elinor Chen, Michael Stavola, W. Beall Fowler, and J. A nna Zhou. R otation of molecular hydrogen in Si: Unambiguous identification of ortho-EB and para- D 2 . Physical Review Letters, 88:245503, 2002.
[21] K. J. C hang and D. J. Cliadi. Diatomic-hydrogen-coniplex diffusion and selftrapping in crystalline silicon. Physical Review Letters, 62:937, 1989.
[22] J. D. Holbeeh, B. Bech Nielsen, R. Jones, P. Sitcli, and S. Obcrg. W2 defect in crystalline silicon. Physical Review Letters, 71:875, 1993.
[23] J. A nna Zhou and Michael Stavola. Sym m etry of m olecular H2 in Si from a uniaxial stress study of the 3618.4 can- 1 vibrational line. Physical Review Letters, 83:1351, 1999.
[24] J. A nna Zhou, E. Chen, and Michael Stavola. Physica B . 273-274:200, 1999.
[25] Fumio Shimura. Oxygen in Silicon. Academic Press, London, 1994.
[26] W. Kaiser, H. L. Frisch, and H. Reiss. Mechanism of the form ation of donor s tates in hea t-treated silicon. Physical Review, 112:1546, 1958.
[27] Y. Takano and M. Maki. Semiconductor silicon. Electrochem ical Society, NJ,1973.
[28] P. Gaworzowshi, E. Hild, F. G. K irscht, and L. Veesornyes. P H Y SIC A ST A T U S SO LID I A, 85:133, 1984.
[29] J. W agner, K. Thonke, and R.. Sauer. Physical Review B , 29:7051, 1984.
[30] W. Kaiser and P. H. Keck. Journal o f Applied Physics, 28:882, 1957.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
BIBLIO G RAPH Y 116
[31] W. W . Keller. Journal o f Applied Physics, 55:3471, 1984.
[32] M .D. McCluskey. Local v ibrational modes of im purities in semiconductors. Journal o f Applied Physics, 87:3593, 2000.
[33] B. N. Brockhouse and P. K. Iyengar. Physical Review, 111:747, 1958.
[34] A. S. Barker Jr. and A. .J. Siever. Review o f Modern Physics, 47:Sl, 1975.
[35] R. S. Leigh and R. C. Newman. Semicond. Sci. Technol., 3:84, 1988.
[36] Y. J. Chabal. S u r f Set. Rep., 8:211, 1988.
[37] Y. J. Yan and S. Mukamel. Journal o f Chemical Physics, 94:179, 1991.
[38] G. Liipke, N. H. Tolk, and L. C. Feldman. Journal o f Applied Physics, 92:2317, 2003.
[39] C. L. Tang and L. K. Cheng. Fundamentals o f Optical Parametric. Processes and Oscillator's. Harwood Academic, A m asterdam , 1995.
[40] J. Y. Zhang, J. Y. Huang, and Y. R. Shen. Optical Parametric Generation and Amplification. Harwood Academic, Anm sterdam , 1995.
[41] W. B. Colson, E. D. .Johnson, M. J. Kelley, and H. A. Sehwettm an. P H Y S IC S TO DAY, 55:35, 2002.
[42] G. R. Neil, C. L. Bohn, S. V. Benson, G. Biallas, D. Douglas, H. F. Dylla, R. Evans, .J. Fugitt, A. Grippo, .J. Gubeli, R. Hill, K. Jordan, R. Li, L. Mer- minga, P. P iot, J. Preble, M. Shinn, T. Siggins, R. Walker, and B. Yunn. Sustained kilowatt lasing in a free-eleetron laser w ith same-cell energy recovery. Physical Review Letters, 84:662, 2000.
[43] P. Guyot-Sionnest, P. Dumas, Y. .J. Chabal, and G. S. Higashi. Lifetime of an adsorbate-substratc vibration: H on Si (111). Physical Review Letters, 64:2156, 1990.
[44] M. Budde, G. Liipke, C. Parks Cheney, N. H. Tolk, and L. C. Feldman. V ibrational lifetime of bond-center hydrogen in crystalline silicon. Physical Review Letters, 55:1452] 2000.
[45] B. H. Bransden and C. .J. .Joachain. Physics o f Atom s and M olecules. Prentice Hall, Harlow, 2003.
[46] S. A. Egorov and J. L. Skinner. On the theory of m ultiphonon relaxation rates in solids. Journal o f Chemical Physics, 103:1533, 1995.
[47] R. K ubo and Y. Toyozawa. Prog. Theorct. Phys., 13:160, 1955.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
BIBLIO G RAPH Y 117
[48] Y. E. Perlin. Sov. Phys. Usp., 6:542, 1964.
[49] T. Miyakawa and D. L. Dexter. Physical Review B , 1:2961, 1970.
[50] F. K. Fong, S. L. Naberhuis, and M. M. Miller. Journal o f Chemical Physics, 56:4020, 1972.
[51] S. H. Lin. Journal o f Chemical Physics, 65:1053, 1976.
[52] S. H. Lin. Radiationless Transitions. Academic Press, New York, 1980.
[53] R. Englnian. Non-radiative Decay o f Ions and Molecules in Solids. N orth Holland, A m sterdam , 1979.
[54] B.-Z. Zhang, Y.-X. Li, M.-R. Lin, and W.-.J. Chen. Chin. Phys., 10:876, 1990.
[55] F. K. Fong. Theory o f Molecular Relaxation. Wiley, New York, 1975.
[56] A. N itzan and J. Jortner. Mol. Phys., 25:713, 1973.
[57] A. N itzan, S. Mukamel, and J. Jortner. Journal o f Chemical Physics, 60:3929,1974.
[58] R. P. Feynm an and A. R. Hibbs. Quantum Mechanics and Path Integrals. McGraw-Hill, New York, 1965.
[59] L. A. Riseberg and H. W . Moos. Physical Review Letters, 19:1423, 1967.
[60] L. A. Riseberg and H. W. Moos. Physical Review, 174:429, 1968.
[61] P. M. Morse. Physical Review, 34:57, 1929.
[62] B. P ajo t, B. Clerjaud, and Z.-.J. Xu. Physical Review B, 59:7500, 1999.
[63] G. Herzberg. Molecular Spectra and Molecular Structure. II. Infrared and Ram an Spectra o f Polyatomic Molecules. Van N ostrand, New York, 1990.
[64] J.-K . Wang, C.-S. Tsai, C.-E. Lin, and .J.-C. Lin. V ibrational dephasing dynam ics at hydrogenated and deuterated sem iconductor surfaces: Sym m etry analysis. Journal o f Chemical Physics, 113:5041, 2000.
[65] T. Inui, Y. Tanabc, and Y. Onodera. Group Theory and Its Applications in Physics. Springer-Verlag, Berlin, 1990.
[6 6 ] P eter Y. Yu and Manuel Cardona. Fundamentals o f Semiconductors : Physics and Materials Properties. Springer-Verlag, Berlin, 1996.
[67] H. Kuzmany. Solid-State Spectroscopy: A n Introduction. Springer-Verlag, Berlin, 1998.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
BIBLIO G RAPH Y 118
[6 8 ] L. Hsu, M. D. McCluskoy, and J. L. L indstro. Resonant, interaction between localized and extended vibrational inodes in Si: 18() under pressure. Physical Review Letters, 90:095505, 2003.
[69] Joseph L. Birman. Space group selection rules: Diamond and zinc: blende. Physical Review, 127:1093, 1962.
[70] Joseph L. Birm an. Theory of infrared and Ram an processes in crystals: Selection rules in diam ond and zincblende. Physical Review, 131:1489, 1963.
[71] YR. J. Bell. Introductory Fourier Transform Spectroscopy. Academic Press, New York, 1972.
[72] M. R. X. de Barros, R. S. M iranda, T. M. Jedju, and P. C. Becker. High- repetitionrate femtosecond m id-infrared pulse generation. Optics Letters, 20:480, 1995.
[73] P. C. M. Planken, E. Snoeks, L. D. Noordam, H. G. Muller, and H. B. van den Linden van den Heuvell. G eneration of intense' suRpicosecond pulses in the midinfared. Optical Communications, 85:31, 1991.
[74] V. Petrov and F. Noack. Tunable femtosecond optical param etric amplifier in m id-infrared w ith narrow -band sending.
[75] George R. Neil and Lia Merminga. Technical approaches for high-average-power free-eleetron lasers. Review o f M odem Physics, 74:685, 2002.
[76] E. L. Saldin, E. A. Selmcnlmiller, anel M. V. Yurkov. The Physics o f Free Electron Lasers. Springea-Verlag, New York, 2000.
[77] A. E. Siegman. Lasers. University Science' Books, Mill Valley, CA, 1986.
[78] M. Buckle, B. Bech Nielsen, R.. Jones, J. Goss, anel S.Oberg. Local modes of the H -2 dimer in germanium. Physical Review B, 54:5485, 1996.
[79] ,J. F. Ziegler, J. P. Biersack, and U. L ittm ark . The stopping and Ranges o f Ions in Solids. Pergam on Press, New York, 1985.
[80] J. Michel, J. R. Niklas, and J. M. Spaeth. S tructure of therm al donors (NL8 ) in silicon: A study w ith electron-nuclear double resonance. Physical Review B, 40:1732, 1989.
[81] A. Stesmans. Interaction of pfc defects a t the (1 1 l)S i/S iO a interface w ith molecular hydrogen: Sim ultaneous action of passivation and dissociation. Journal o f Applied Physics, 88:489, 2000.
[82] J. W . Lyding, K. Hess, anel I. C. Kizilyalli. Applied Physics Letters, 68:2526, 1996.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
BIBLIOG RAPH Y 119
[83] J. Chevallier, M. Barbe, E. C onstant, D. Loridant-Bem ard, and M. C onstant. Strong isotope effects in the dissociation kinetics of Si-H and Si-D complexes in GaAs under ultraviolet illum ination. Applied Physics Letters, 75:112, 1999.
[84] T.-C. Sheri, C. Wang, G. C. Abeln, J. R. Tucker, J. W . Lyding, Ph. Avouris, and R,. E. W alkup. Science, 268:1590, 1995.
[85] Ph. Avouris, R. E. Walker, A. R. Rossi, H. C. Akpati, P. Nordlander, T. C. Shen, G. C. Abeln, and J. W . Lyding. Chemical Physics Letter , 254:148, 1996.
[8 6 ] J.-H. Wei, M.-S. Sun, and S.-C. Lee. A possible mechanism for improved light- induced degradation in deuterated amorphous-silicon alloy. Applied Physics Letters, 71:1498, 1997.
[87] K. Hess, L. F. Register, B. T uttle, J. Lyding, and I. C. Kizilyalli. Physica (Am sterdam ), 3E:1, 1998.
[8 8 ] E. T. Foley, A. F. Kam, .J. W . Lyding, and Ph. Avouris. Physical Review Letters, 80:1336, 1998.
[89] R. Biswas, Y.-P. Li, and B. C. Pan. Enhanced stability of deuterium in silicon. Applied Physics Letters, 72:3500, 1998.
[90] C. G. Van de Walle and W. B. Jackson. Applied Physics Letters, 68:2526, 1996.
[91] C. G. Van de Walle and R. A. Street. Physical Review B, 49:14766, 1994.
[92] Zhi Chen, Jun Guo, and Pangleen Ong. Evidence! for energy coupling from the Si-D vibration mode to the Si-Si and S i-0 vibration modes at the S i0 2/S i interface. Applied Physics Letters, 83:2151, 2003.
[93] An Shill, Si-Chen Lee, and Chili ta Cilia. Evidence for coupling of Si-Si lattice vibration and Si-D wagging vibration in deuterated am orphous silicon. Applied Physics Letters, 74:3347, 1999.
[94] G. Liipke, X. Zhang, B. Sun, A. Fraser, N. H. Tolk, and L. C. Feldman. S tructure-dependent v ibrational lifetimes of hydrogen in silicon. Physical Re- view Letters, 88:135501, 2002.
[95] A. N itzan, S. Mukamel, and J. Jortner. Journal o f Chemical Physics, 63:200,1975.
[96] U. Happek, C. E. M ungan, W. von der O sten, and A. .J. Sievers. Relaxation of the CN_ stretching vibration in silver halides: T he role of accepting modes. Physical Review Letters, 72:3903, 1994.
[97] E. Tuncel and H. Sigg. T em perature dependence of the vibrations of the si-h complex in GaAs and AlAs. Physical Review B, 58:5225, 1993.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
BIBLIO G RAPH Y 120
[98] B. P ajo t. private communication.
[99] J. Vetterhoffer and J. Weber. Hydrogen passivation of shallow donors S, Se, and Te in GaAs. Physical Review B, 53:12835, 1996.
[100] Note. T he natu ra l isotope of S is dom inated by 32S(95.02%), so the isotope broadening can be neglected. T he isotope broadening is much larger for Se(Te)-H(D) modes due to the com parable aboundanecs: 82Se(9.2%), 80Se(49.8%), 78Se(23.5%), 77Se(7.6%), 76Se(9.0%),74Se(0.9%) and 130Te(34.5% )’ 128Te(31.7%), 126Te(18.7%), 125Te(7%), 124Te(4.6%), 123Te(0.9%), 122Te(2.5%), 120Te(0.1%).
101] M. Budde, G. Liipke, E. Chen, X. Zhang, N. H. Tolk, L. C. Feldman, E. Tarhan,A. K. Ram das, and M. Stavola. Lifetimes of hydrogen and deuterium related vibrational modes in silicon. Physical Review Letters, 87:145501, 2001.
102] H. J. Stein. Bonding and therm al stability of im planted hydrogen in silicon. Journal o f Electronic Materials, 4:159, 1975.
1031 H. J. Stein. Vacancies and the chemical trapping of hydrogen in silicon. Physical Review Letters, 43:1030, 1979.
104] G. R. Bai, M. W . Qi, L. M. Xie, and T. S. Shi. Solid State Comnium., 56:277, 1985.
105] T. S. Shi, L. M. Xie, G. R. Bai, and M. W. Qi. P H YSIC A S T A T U S SO LID IB, 131:511, 1985.
1061 X. T. Meng, Y. C. Chang, and Y. D. Fan. P H YSIC A ST A T U S SO L ID I A, 101:619, 1987.
1071 M. Budde, B. Bcoh Nielsen, P. Leary, .J. Goss, R.. Jones, P. R. Briddon, and S.Oberg.
1081 H. Xu. Physical Review B, 46:1403, 1992.
109] C. G. Van de Walle. Physical Review B, 49:4579, 1994.
110] M. A. Roberson and S. K. Estreieher. Physical Review B, 49:17040, 1994.
l l l l M. Suezawa. Physical Review B, 63:035201, 2001.
112] P. Stallinga, P. Johannesen, S. Herstrom , B. Bech Nielsen, and J. R. Byberg. Physical Review B, 58:3842, 1998.
[113] C. F lensburg and R. F. Stewart. Physical Review B, 60:284, 1999.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
BIBLIOG RAPH Y 121
[114] M arjolein van der Voort, Chris W. Rella, Lex F. G. van der Meer, Audrey V. Akim ovl, and .Jaap I. Dijkhuis. Dynamics of Si-H vibrations in an am orphous environment. Physical Review Letters, 84:1236, 2 0 0 0 .
115] M. Budde. unpublished.
116] B. N. .J. Porsson and Ph. Avouris. Surf. Sci., 390:45, 1997.
117] R. Jones, h ttp ://n ew to n .ex .ac .u k /research /q sy stem s/jo n es/p ro jec ts/o x y g en /.
118] W. Kaiser, P. H. Keck, and C. F. Lange. Physical Review , 101:1264, 1956.
119] H. J. Hrostowshi and R.. H. Kaiser. Physical Review , 107:966, 1957.
120] W. Kaiser. J. Phys. Chem. Solids, 23:255, 1962.
121] J. W. C orbett, R. S. McDonald, and G. D. W atkins. J. Phys. Chem. Solids, 25:873, 1964.
122] B. Pajot, and P. Clauws. In O. Engstrom , editor. Proceedings o f the 18th In ternational Conference on the Physics o f Semiconductors, page 911, Singapore, 1987. World Scientific.
123] M. Gienger, M. Glaser, and K. La be ta maim. Solid State Commun., 86:285, 1993.
124] A. Lizon-Nordstrom and F. Y ndurain. Solid State Commun., 89:819, 1994.
125] A. J. Mayur, M. D. Sciacca, M. K. Udo, A. K. Ram das, K. It.oh, J. Wolk, and E. E. Haller. Physical Review B, 49:16293, 1994.
126] E. A rtacho and F. Y ndurain. Mater. Sci. Forum, 196-201:103, 1995.
127] A. J. Mayur, M. D. Sciacca, M. K. Udo, A. K. Ram das, K. It.oh, J. Wolk, and E. E. Haller. Physical Review B, 49:16293, 1994.
128] R. Jones, A. Umerski, and S. Oberg. Physical Review B, 45:11321, 1992.
129] D. R. Bosomworth, W . Hayes, A. R. L. Spray, and G. D. W atkins. Proc. R. Soc. London Ser. A, 317:133, 1970.
130] M. Stavola. Applied Physics Letters, 44:514, 1984.
131] B. P ajo t, H. J. Stein, B. Cales, and C. Naud. J. Elcctrochein. Soc., 132:3034, 1985.
[132] Hiroshi Y am ada-K aneta, Chioko K aneta, and Tsutom u Ogawa. Theory of loeal- phonon-coupled low-energy anharm onic excitation of the in terstitia l oxygen in silicon. Physical Review B, 42:9650, 1990.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
BIBLIO G RAPH Y 122
[133] B. P ajo t, E. Artacho, C. A. .J. Am m erlaan, and J. M. Spaeth. Theory of local- phonon-coupled low-energy anhannonie excitation of the in terstitia l oxygen in silicon. J. Phys., Condens. M atter ., 7:7077, 1995.
[134] Hiroshi Yarnada-Kaneta. Theory of local-phonon-coupled low-energy anhar- monic excitation of the in terstitia l oxygen in silicon. Physica B, 308-310:309,2001 .
[135] M. R am am oorthy and S. T. Pantelides. Conpled-barrier diffusion: T he case of oxygen in silicon. Physical Review Letters, 76:267, 1996.
[136] D. D. D lott. Laser Spectroscopy o f Solids II, edited by W. M. Yen, chapter 5. Springer-Vcu'lag, Berlin, 1989.
[137] Paolo Giannozzi, Stefano de Gironeoli, Pasquale Pavone, and Stefano Baroni. Ab initio calculation of phonon dispersions in semiconductors. Physical Review B, 43:7231, 1991.
[138] Siqing Wei, Changlin Li, and M. Y. Chou. Ab initio calculation of therm odynamic properties of silicon. Physical Review B, 50:14587, 1994.
[139] J. F. Angress, A. R. Goodwin, and S. D. Smith. Proc. R. Soc. London, A287:64, 1965.
[140] .Tiro Kato, Kohei M. Itoh, Hiroshi Yam ada-K aneta, and Hans-.Joachim Pohl. Host isotope effect on the localized vibrational modes of oxygen in isotopieally enriched 28Si, 29Si, and 30Si single crystals. Physical Review B, 68:035205, 2003.
[141] B. P ajo t, P. Clauws, J. L. Linelstrom, anel E. Artacho. O isotope effects and v ibration-rotation lines of in terstitia l oxygen in germanium. Physical Review B, 62:10165, 2000.
[142] D. C. Look. Reerent advances in ZnO m aterials and dewices. M aterials Science and Engineering B, 80:283, 2001.
[143] D. P. N orton, Y. W. Heo, M. P. Ivill, K. Ip, S. .J. Pearton, M. F. Chisholm, and T . Steiner. M aterials Today, 7:34, 2004.
[144] C. G. Van de Walle. Hydrogen as a cause of eloping in zinc oxide. Physical Review Letters, 85:1012, 2000.
[145] D.M. Hofmann, A. Hofstaetter, F. Leiter, H. Zhou, F. Henecker, B. Meyer, S.B. Orlinskii, .J. Schmidt, and P.G. Baranov. Hydrogen: A relevant shallow donor in zinc oxide. Physical Review Letters, 88:045504, 2002.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
BIBLIO G RAPH Y 123
[146] S. F. J. Cox, E. A. Davis, S. P. C ottrell, P. J. C. King, ,J. S. Lord, J. M. Gil, H. V. A lberto, R. C. Vil?o, J. P iro to D uart, N. Ayres de Campos, A. Weidinger, R. L. Licliti, and S. J. C. Irvine. Experim ental confirmation of the predicted shallow donor hydrogen s ta te in zinc oxide. Physical Review Letters, 86:2601, 2001 .
[147] M.D. McCluskey, S..J. Jokela, K.K. Zhuravlev, P.J. Simpson, and K.G. Lynn. Infrared spectroscopy of hydrogen in ZnO. Applied Physics Letters, 81:3807,2002 .
[148] M. Stavola, S. J. Pearton, J. Lopata, and W. C. D autrem ont-Sm ith. Physical Review B, 37:8313, 1988.
[149] F. Cerdeira, T. A. Fjeldly, and M. Cardona. Physical Review B, 8:4734, 1973.
[150] F. Cerdeira, T. A. Fjeldly, and M. Cardona. Physical Review B, 9:4344, 1974.
[151] G. Alvin Shi, M arjan Saboktakin, Michael Stavola, and S. J. Pearton. ’’Hidden hydrogen” in as-grown ZnO. Applied Physics Letters, 85:5601, 2004.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
VITA
Baozhou Sun
Baozhou Sun was born on November 13, 1977 in W uhan, Hubei province of China.
He entered Jilin University in Changchun in 1996 and obtained his Bachelor degree
in Electronic Engineering in 2000. In August of the same year, he entered graduate
school in the Department, of Applied Science a t the College? of W illiam and Mary,
where he received his M aster degree in 2002. He? m arried his wile, Min Huang on
A ugust 28, 2004. Sun completed his Ph.D degree under the supervision of Dr. G unter
Liipke in 2005. During his study at W illiam and Mary, Sun was elected to be a
m em ber of Sigma Xi in 2005 and he? has been a member of American Physical of
Society since? 2003.
124
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Recommended