Interface broadening due to ion mixing during thin film growth at the radio-frequency-

biased electrode in a plasma-enhanced chemical vapor deposition environment

A. Amassian, M. Svec, P. Desjardins, and L. Martinu

Citation: Journal of Vacuum Science & Technology A 24, 2061 (2006); doi: 10.1116/1.2348642
View online: http://dx.doi.org/10.1116/1.2348642
View Table of Contents: http://scitation.aip.org/content/avs/journal/jvsta/24/6?ver=pdfcov
Published by the AVS: Science & Technology of Materials, Interfaces, and Processing

Articles you may be interested in

Plasma-enhanced chemical vapor deposition synthesis of silica-silicone nanolaminates using a single
precursora)
J. Vac. Sci. Technol. A 29, 021012 (2011); 10.1116/1.3553149

Ion-surface interactions on c - Si ( 001 ) at the radiofrequency-powered electrode in low-pressure plasmas: Ex

situ spectroscopic ellipsometry and Monte Carlo simulation study
J. Vac. Sci. Technol. A 24, 45 (2006); 10.1116/1.2134709

Annealing and oxidation of silicon oxide films prepared by plasma-enhanced chemical vapor deposition
J. Appl. Phys. 97, 014913 (2005); 10.1063/1.1829789

Interface formation during the yttrium oxide deposition on Si by pulsed liquid-injection plasma enhanced metal-
organic chemical vapor deposition
J. Vac. Sci. Technol. A 22, 2490 (2004); 10.1116/1.1810163

Surface relaxation during plasma-enhanced chemical vapor deposition of hydrocarbon films, investigated by in
situ ellipsometry
J. Appl. Phys. 81, 1531 (1997); 10.1063/1.364184

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 128.248.155.225 On: Sun, 23 Nov 2014 08:37:25
 Interface broadening due to ion mixing during thin film growth
at the radio-frequency-biased electrode in a plasma-enhanced
chemical vapor deposition environment
A. Amassian,a兲 M. Svec,b兲 P. Desjardins, and L. Martinuc兲
Regroupement québécois sur les matériaux de pointe (RQMP) and Department of Engineering Physics,
École Polytechnique de Montréal, P.O. Box 6079, Station Centre-ville, Montréal, Québec H3C
3A7, Canada
共Received 18 November 2005; accepted 9 August 2006; published 11 October 2006兲
The authors show that ion bombardment in the range of tens to a few hundreds of eV, used in ion-
and plasma-assisted deposition processes, can lead to thin film growth dominated by subsurface
deposition due to subplantation 共shallow implantation兲. This can cause significant interface
broadening during the initial stages of film deposition as a result of ion mixing. First, by studying
the modifications of a c-Si共100兲 target exposed to an O2 plasma at the radio-frequency 共rf兲-biased
electrode using in situ real-time spectroscopic ellipsometry 共RTSE兲, the authors detect implantation,
damage, and oxidation to a depth of up to ⬃10 nm. They validate these results using high resolution
transmission electron microscopy and simulate the effects of ion-surface interactions at the rf-biased
electrode by using Monte Carlo TRIDYN simulations. The simulation code, which was modified
specifically to consider a broadband ion energy source, enabled the authors to reproduce depth and
time relevant experimental results with good agreement. In situ RTSE was then used to monitor
TiO2 deposition on SiO2 under similar ion bombardment conditions. The authors observed the
formation of a 2- to 4-nm-thick interfacial layer, depending on the ion-to-neutral flux ratio 共␾i / ␾n兲,
which was controlled by varying the deposition rate. TRIDYN simulations revealed that oxygen
subplantation causes interfacial broadening during the growth through ballistic mixing of Ti and Si
atoms at the interface; the interface width scales as ⬃共␾i / ␾n兲1/2. Intensive ion mixing at ␾i / ␾n
⬎ 1 is also shown to be responsible for the ballistic displacement of the majority of
surface-deposited Ti atoms into the bulk, so that the growth appears to be dominated by subsurface
deposition under conditions of intense ion bombardment. © 2006 American Vacuum
Society. 关DOI: 10.1116/1.2348642兴

I. INTRODUCTION The IB characteristics mentioned above depend indirectly

Ion bombardment 共IB兲 assisted deposition has been used
for the fabrication of high performance optical films and
coatings—with dense and amorphous microstructures—that
are essential for photonics and telecommunications
applications.1 This includes IB assisted evaporation, dual ion
beam sputtering, unbalanced magnetron sputtering, and other
techniques described in numerous reviews.2,3 More recently,
plasma enhanced chemical vapor deposition 共PECVD兲 has
emerged as a way to fabricate transparent thin films4 with
good optical and mechanical properties. This highly versatile
process allows one to sensitively control the film
microstructure5 as well as optical6 and interfacial
properties.7,8 In particular, a selective control of the ion-to-
atom flux ratio 共␾i / ␾n兲 and the mean ion energy 共Em兲, such
as in dual-mode microwave/radio frequency 共rf兲
discharges,9–11 has enabled us to finely tune film porosity and
density.5,7,12
a兲
Present address: Materials Science and Engineering, Cornell University,
Ithaca, New York 14850.
b兲
Present address: Department of Physics, Jan E. Purkyne University, Horeni
13, 400 96 Usti Nad Labem, Czech Republic.
c兲
Electronic mail: lmartinu@polymtl.ca
on external process parameters, such as plasma power, sub-
strate bias 共VB兲, and working pressure 共p兲. Significant efforts
have been devoted to study plasma chemistry and IB char-
acteristics in terms of the external process parameters, in
order to bridge the gap between the bulk properties of the
plasma, ion-surface interactions, and their effects in terms of
surface, interface, and thin film modifications. The effects of
IB on thin film formation,13,14 have been studied using many
conventional techniques, including Rutherford backscatter-
ing spectroscopy and transmission electron microscopy
共TEM兲, which are highly invasive, destructive, and cannot be
operated in situ. In addition, the low-pressure PECVD envi-
ronment is not permissible toward the use of ion and electron
beams during processing, owing to the chemically and elec-
trically active environment. Consequently, it appears that
nondestructive and noninvasive in situ optical diagnostics
can serve to investigate plasma-surface interactions during
processing in the PECVD environment.
Recently, we have used ex situ variable angle spectro-
scopic ellipsometry15 to analyze c-Si共001兲 samples pre-
treated for 10 min at the rf-biased electrode in an O2 rf dis-
charge. For typical deposition conditions of oxide films,
where VB can vary anywhere between −60 and −600 V,
modifications of c-Si were found to be quite dramatic, in-

2061 J. Vac. Sci. Technol. A 24„6…, Nov/Dec 2006 0734-2101/2006/24„6…/2061/9/$23.00 ©2006 American Vacuum Society 2061

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 128.248.155.225 On: Sun, 23 Nov 2014 08:37:25
 2062 Amassian et al.: Interface broadening due to ion mixing
cluding damage layer formation and oxide growth down to a
VB-dependent depth of ⬃3 – 10 nm. Monte Carlo 共MC兲 TRIM
simulations16 showed that the depth-dependent transport of
implanted oxygen is responsible for damage formation at
high depth and for near-surface oxidation.
These results raise important questions about the possible
effects of ion-surface interactions during the early stages of
thin film deposition. We need to determine, for instance,
where—at or below the surface—the film grows, and
whether ion bombardment conditions in the PECVD envi-
ronment can lead to significant interface broadening by mix-
ing atoms at substrate/film or film/film interfaces. Lifshitz
et al.17 were the first to propose the subplantation 共shallow
implantation兲 model, which demonstrated that in the case of
diamondlike carbon coatings, dense films with sp3 bonding
grow as a result of subsurface deposition of hyperthermal
carbon species, namely, film-forming ions with energies in
tens to hundreds of eV. More recently, it has been demon-
strated that thin films prepared by ion beam assisted
deposition,18 sputtering,19 and pulsed cathodic arc
deposition20 processes are subject to interface broadening,
which increases with ion energy 共Ei兲 and ion mass, and de-
creases with deposition rate 共R兲. Interface broadening in op-
tical interference filters 共OIFs兲 is highly undesirable because
it can reduce reflections at individual interfaces in the
multilayer stack, and therefore modify the spectral perfor-
mance of the OIF,21 especially in cases in which ultrathin
films are used in the OIF design. In such cases, the problem
may be addressed either by minimizing interface broadening,
which likely requires tweaking of the ion bombardment and
growth conditions, or by accounting for the interfacial broad-
ening in the starting OIF design. Either way, to address this
problem, we must understand how interface broadening oc-
curs, and then develop tools to accurately predict it in typical
process conditions.
The research on interface broadening has greatly ben-
efited from MC TRIDYN simulations,18,20,22 which enable cal-
culations of the depth-dependent transport of ions and atoms
on time and fluence scales of interest for thin film deposition
processes 共⬃10−2 ⬍ time⬍ 103 s兲.23,24 While TRIDYN is de-
signed to simulate monoenergetic ion beam interactions, in-
cluding ion beam oxidation,25–27 we have recently
demonstrated28 that it can be modified to simulate the inter-
action of complex ion sources possessing broad ion energy
distribution functions 共IEDFs兲, such as what is typically ob-
served at the rf-biased electrode in a rf discharge.10,11
In this work, we demonstrate that the effects of ion-
surface interactions manifest themselves early on during the
plasma exposure 共10−1 – 101 s兲 in a PECVD environment,
first by impinging damage, and then by causing significant
composition changes in the c-Si共001兲 substrate by subplan-
tation of hyperthermal ions 共100 – 103 eV兲. Furthermore, we
investigate how ion bombardment leads to interface broad-
ening during the deposition of TiO2 on SiO2 at the rf-biased
electrode. We use in situ real-time spectroscopic ellipsometry
共RTSE兲 to monitor both the O2 plasma pretreatment 共plasma
oxidation兲 of c-Si at the rf-biased electrode and the subse-
J. Vac. Sci. Technol. A, Vol. 24, No. 6, Nov/Dec 2006
Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 128.248.155.225 On: Sun, 23 Nov 2014 08:37:25
2062
quent TiO2 deposition on SiO2 formed during the pretreat-
ment study. The modified TRIDYN simulation code is used to
simulate the O2 plasma pretreatment step on c-Si and then
the interface broadening during thin film deposition.
II. EXPERIMENTAL METHODOLOGY
The experiments were performed in a parallel plate rf
PECVD system 共30 liter volume兲 共Ref. 29兲 with a base pres-
sure ⬍10−5 torr. The c-Si共001兲 substrates were etched in hy-
drofluoric acid, rinsed in de-ionized water, and N2 dried be-
fore being placed on a 15-cm-diameter rf-powered electrode
inside the vacuum chamber.
The experiments involve two distinct steps: 共1兲 plasma
pretreatment in an O2 rf discharge of c-Si placed at the rf-
biased electrode and 共2兲 subsequent PECVD deposition of
TiO2 on the oxide layer formed on c-Si in step 共1兲. The
pretreatment and deposition steps were performed at p
= 20 mtorr. The substrates were first exposed for 10 min at
VB = −600 V 关step 共1兲兴, then VB was reduced, without extin-
guishing the plasma, to −450 V in preparation for step 共2兲.
This bias condition is required to fabricate dense TiO2 films
of optical grade.30 After ⬃1 min at VB = −450 V, TiCl4 was
introduced into the reactor, thereby initiating TiO2 growth,
without interrupting the plasma between steps 共1兲 and 共2兲.
The partial pressure of TiCl4 was kept low at around
0.1 mTorr in order to obtain R ⬍ 0.1 nm/ s and was subse-
quently used to control the deposition rate.
The pretreatment 关step 共1兲兴 and the deposition 关step 共2兲兴
were monitored in situ by RTSE 共M-2000, J. A. Woollam
Co.兲. The ellipsometer, equipped with a rotating compensa-
tor, was mounted at an angle of incidence near 65° from the
substrate normal. Spectral range for data acquisition and
analysis was limited to 477 photon energies 共E兲 from
1.24 to 5.05 eV 共245– 1000 nm兲. RTSE spectra were ac-
quired at 0.4 s intervals, ensuring that multiple spectra are
recorded for each monolayer of TiO2 共assuming 1 ML
⬃0.24 nm兲.
III. MONTE CARLO TRIDYN SIMULATIONS WITH
BROADBAND ION SOURCE CAPABILITY
The TRIDYN simulations are based on the binary collision
approximation 共BCA兲; they take dynamic target modifica-
tions into consideration, enabling them to compute the ef-
fects of high fluence implantations. The TRIDYN code has
been designed for single-energy ion sources, and has to be
modified in order to simulate broadband IEDF sources like at
the rf-biased electrode.9–11
Ignoring the resonance peaks, the IEDF at the rf-biased
electrode 关Fig. 1共a兲兴 may be approximated by a rectangular
function with Ei ranging from a few eV up to Emax
⬃ 1.15e兩VB兩, while the O+ distribution is a saddlelike struc-
ture, centered on Ei ⬃ e兩VB兩, with energy ranging from
⬃0.85e兩VB兩 to ⬃1.15e兩VB兩. Working with the rectangular dis-
tribution shown in Fig. 1共a兲, we first dissociate all molecular
ions 共艌95% of all ions11兲 into atomic species 共the dissocia-
tion energy of O+2 molecules is much smaller than Ei兲, then
we split the IEDF into ten discrete energy channels 关see Fig.
 2063 Amassian et al.: Interface broadening due to ion mixing
FIG. 1. 共a兲 Ion energy distribution function 共IEDF兲 of an O2 rf discharge at
the rf-powered electrode, including distributions of O+2 and O+ ions, 共b兲 the
resulting atomic oxygen energy distribution function used in TRIDYN simu-
lations, and 共c兲 the optical model used for the analysis of in situ RTSE data
acquired during plasma pretreatment 共oxidation兲 of c-Si共001兲 and subse-
quent TiO2 deposition 共see text for more details兲.
1共b兲兴. Simulations are performed at each energy channel us-
ing ten pseudoparticles 共or packets兲 of 1011 O cm−2 fluence,
for a total O fluence of 1012 cm−2/channel. The overall im-
pact of the IEDF on the target after exposure to 1012 O cm−2
is calculated by weighted averaging simulation results 共such
as elemental depth profiles, surface densities, sputtering
yield, surface recession, etc.兲 from all energy channels. The
results of the weighted averaging are taken as the starting
point for the next incremental exposure to 1012 O cm−2 at
each channel. This procedure is repeated until a total fluence
of 1018 O cm−2 is obtained, equivalent to a plasma exposure
time of a few minutes.
TRIDYN simulations were also performed to evaluate the
effects of IB during deposition of up to 15-nm-thick TiO2
films on SiO2. TiO2 deposition at the rf-biased electrode was
simulated by alternately adding a 0.1-nm-thick slice of TiO2
film 共3.4 g / cm3 共Ref. 6兲 for amorphous TiO2 by PECVD兲
and irradiating it with an O fluence determined by ␾i / ␾n.
Simulations were performed for R = 0.04– 4.00 nm/ s, corre-
sponding to ␾i / ␾n ⬃ 6.00– 0.06.
IV. RESULTS AND DISCUSSION
A. Plasma pretreatment of c-Si at the rf-biased
electrode in O2 rf discharge †step „1…‡
1. In situ RTSE monitoring of the pretreatment
We have previously shown that modification of the
c-Si共001兲 substrate induced by an O2 plasma pretreatment at
the rf-biased electrode can be optically modeled by a two-
layer structure:15,28,31 a SiO2 overlayer on top of a mixed-
phase interfacial layer 共IL兲 关modeled as a Bruggeman effec-
tive medium approximation 共BEMA兲 mixture of c-Si, a-Si,
and SiO2兴 on the c-Si substrate 关see Fig. 1共c兲, step 共1兲兴.
JVST A - Vacuum, Surfaces, and Films
Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 128.248.155.225 On: Sun, 23 Nov 2014 08:37:25
2063
FIG. 2. In situ RTSE analysis results of O2 plasma pretreatment of
c-Si共001兲 关step 共1兲兴 and TiO2 deposition 关step 共2兲兴 at the rf-powered elec-
trode. All surface and interface positions are presented in reference to the
initial surface position of the substrate.
The optical model was fitted to the RTSE spectra and the
results are plotted in Fig. 2 关step 共1兲兴. The dynamic evolu-
tions of oxide and IL thickness 共tIL兲 are reported relative to
the initial surface position, so as to include the effects of
oxidation-induced swelling and rf sputtering. Swelling was
calculated from the SiO2 thickness and the SiO2 fraction in
the IL. The sputtering rate was obtained from SRIM
simulations16,32 and estimated to be ⬃0.014 nm/ s for VB
= −600 V. The ion fluence scale shown at the top of step 共1兲
in Fig. 2 was obtained from ␾i measurements performed
earlier.9 We estimate ␾i ⬃ 1.0⫻ 1015 and ⬃1.3
⫻ 1015 cm−2 s−1 for VB = −450 and −600 V, respectively.
Since more than 95% of ions are O+2 molecules, we consider
that the total oxygen flux leading to ion bombardment is
⬃1.95 times larger than ␾i. In the remainder of this article,
we report results as a function of O fluence: 1.95⫻ ␾i
⫻ time.
IL formation is observed after an exposure of
⬃5 ⫻ 1014 O cm−2, corresponding to ⬃0.2 s of plasma expo-
sure. This result is in agreement with the amorphization
threshold of bare c-Si, which was found to be
⬃4 ⫻ 1014 cm−2 for a 500 eV Ar+ beam.33
Oxidation is detected a few seconds later
关⬃2 ⫻ 1015 O cm−2兴, and saturates following 40 s of plasma
exposure 关⬃1017 O cm−2 in Fig. 2, step 共1兲兴. Saturation of
the oxide thickness is expected to occur as oxygen implan-
tation decays rapidly with depth. As oxygen accumulation
saturates near the surface, swelling slows down and is sub-
sequently replaced by surface recession due to sputtering
关see step 共1兲 of Fig. 2 for the transition from swelling to
recession兴. At high fluence, a self-limiting steady-state re-
gime is observed which is characterized by continuous SiO2
 2064 Amassian et al.: Interface broadening due to ion mixing
FIG. 3. TRIDYN simulation of the O2 plasma pretreatment process on a Si
target. 共a兲 Oxide composition and displacements/at. after exposure to 5
⫻ 1014 and 5 ⫻ 1016 O cm−2, and 共b兲 evolution of the absolute surface posi-
tion of the target obtained by TRIDYN simulations 共for a broad ion energy
source and Emax ions兲 and calculated from a combination of RTSE and TRIM
results.
removal from the surface and by damage and oxide forma-
tion in the IL as the modifications’ front moves into the
target.
2. Monte Carlo TRIDYN simulations of plasma
pretreatment
TRIDYN calculations of the compositional depth profile in
the Si target are shown in Fig. 3共a兲 for O2 plasma irradiation
at the rf-biased electrode, as calculated by the IEDF method
共see Sec. III兲 at two fluences 共5 ⫻ 1014 and 5 ⫻ 1016 O cm−2兲.
The target’s composition is expressed in terms of the stoichi-
ometry of SiOx. At low fluence 共5 ⫻ 1014 cm−2兲 oxygen is
subplanted as deep as ⬃8 nm, but it is present in very low
quantity 共not likely to be detectable by optical means兲. At the
same time, we observe significant atomic relocations, with
values greater than 0.1 displacement/ at., at a depth of up to
4 nm. This dpa value is obviously sufficient to cause damage
formation in room temperature Si, as evidenced by the early
detection 关by RTSE, see Fig. 2, step 共1兲兴 of damage in c-Si
共IL formation兲.
At higher fluence 共5 ⫻ 1016 cm−2兲 oxygen accumulation
proceeds inward from the surface and damage accumulation
moves deeper. The fact that atomic relocations are observed
beyond the reach of O ions is likely due to the predominance
J. Vac. Sci. Technol. A, Vol. 24, No. 6, Nov/Dec 2006
Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 128.248.155.225 On: Sun, 23 Nov 2014 08:37:25
2064
FIG. 4. HRTEM micrograph of the O2 plasma pretreated 关step 共1兲兴 and
TiO2-coated 关step 共2兲兴 c-Si共001兲 along the 关011兴 zone axis. Inset shows
selected area electron diffraction 共SAED兲 pattern on c-Si along the 关011兴
zone axis.
of Si atoms in the bulk and to the preferential displacement
of Si in comparison to O; the bulk displacement energy 共Ed兲
of Si is usually taken as Ed ⬃ 12 eV and Ed ⬃ 28 eV for O in
SiO2.32 Consequently, Si recoils are more likely to be formed
than O recoils by end-of-range ions and energetic recoils.
Surface motion was obtained from TRIDYN simulations by
using the IEDF method and in the single-energy Emax ap-
proximation. For comparison, we also calculated surface mo-
tion by combining the oxide thickness 共RTSE results兲 with
the sputtering rate 共SRIM simulations兲. The results are plotted
in Fig. 3共b兲; they show that TRIDYN simulations with the
IEDF method agree very well with surface motion data cal-
culated from RTSE and SRIM results. The large discrepancy
between the IEDF method and the Emax approximation origi-
nates from the fact that Emax ions are much more energetic
than Em 共mean ion energy of the IEDF兲; hence Emax ions
form a thicker oxide and sputter SiO2 at a much higher rate
than IEDF ions.
From these simulations, we conclude that the dynamics of
surface modifications, such as damage formation, oxidation,
and swelling are modeled with sufficient accuracy by BCA-
based MC codes, without performing any parameter fitting.
3. High resolution TEM analysis of plasma pretreated
c-Si
Figure 4 shows a typical bright-field high resolution trans-
mission electron microscopy 共HRTEM兲 image of the c-Si
sample along the 关011兴 zone axis after a 10-min-long plasma
pretreatment at VB = −600 V, followed by TiO2 deposition at
VB = −450 V and ␾i / ␾n ⬇ 6. The electron diffraction pattern
taken from an area selected exclusively in the c-Si substrate
 2065 Amassian et al.: Interface broadening due to ion mixing
is shown in the inset. The image clearly reveals the forma-
tion of a 9.4± 0.2-nm-thick light gray amorphous layer sand-
wiched between c-Si and the TiO2 film. More notably, the
thickness of this layer closely corresponds to the combined
thickness of SiO2 and IL layers reported by RTSE before
共⬃9.9± 0.2 nm兲 and after 共⬃8.9± 0.2 nm兲 TiO2 deposition
共see Fig. 2兲. Given that the top interface is significantly
graded as a result of interface broadening during TiO2 depo-
sition 共see next section兲, it is not clear whether the layer
thickness reported by TEM is equivalent to RTSE data be-
fore or after TiO2 deposition. It may be more appropriate to
compare the TEM-determined thickness to the average
thickness of the combined SiO2 and IL layers before and
after TiO2 deposition, which yields ⬃9.4± 0.4 nm in excel-
lent agreement with TEM.
The light gray layer is found to be stoichiometric SiO2 at
its center and it is sandwiched between stoichiometric TiO2
共top兲 and pure Si 共bottom兲 according to energy electron loss
spectroscopy 共EELS兲 analysis 共nominal spot size of 0.2 nm;
results not shown兲. EELS analyses of the top and bottom
25% of the SiO2 layer show it to be Ti rich near the
TiO2 / SiO2 interface and significantly oxygen deficient when
probing closer to the interface with c-Si. The TEM image
also reveals the existence of a rough amorphous/crystalline
interface between SiO2 and c-Si, which is also consistent
with the RTSE finding that an oxygen-deficient and partially
amorphous interfacial layer lies between SiO2 and c-Si. The
TEM results therefore confirm that a two-layer model con-
sisting of a SiO2 layer and a damaged, O-deficient interfacial
layer is an accurate optical representation of plasma-induced
modifications at the rf-biased electrode, while also revealing
that the IL detection is partly due to interfacial roughening
between SiO2 and c-Si. The analysis also reveals a graded
TiO2 / SiO2 interface formation, which is the subject of de-
tailed studies in the next section.
B. Deposition of TiO2 on SiO2 †step „2…‡
1. In situ RTSE during TiO2 deposition at the rf-
biased electrode
TiO2 deposition was initiated by introducing TiCl4 in the
reactor during O2 plasma pretreatment. TiO2 growth was
monitored by in situ RTSE; the optical model used to fit the
RTSE spectra is shown in Fig. 1共c兲 关step 共2兲兴. Parameters
fitted in step 共1兲 were fixed during the fitting process for step
共2兲, except for SiO2 thickness, which was allowed to change
in order to account for the formation of a mixed oxide by
relocation of Ti into SiO2 and Si displacement into TiO2.
TiO2 thickness 共tTiO2兲 and interface thickness 共tinterface兲 were
allowed to vary, while the TiO2 optical constants were fixed
during the fitting procedure. TiO2 optical constants 共n550
= 2.4兲 were obtained from Tauc-Lorentz31,34 parametrization
of a 200-nm-thick TiO2 film, which was deposited on glass
under identical conditions. The interfacial layer was modeled
as a fixed 1:1 BEMA mixture of TiO2 and SiO2.
The results of the fitting procedure are presented in step
共2兲 of Fig. 2. We initially detect formation of the TiO2 / SiO2
mixed-oxide layer 共660⬍ t ⬍ 730 s兲 at the expense of the
JVST A - Vacuum, Surfaces, and Films
Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 128.248.155.225 On: Sun, 23 Nov 2014 08:37:25
2065
FIG. 5. HRTEM micrographs of TiO2 deposited on SiO2 at 共a兲 ␾i / ␾n ⬇ 6 and
共b兲 ␾i / ␾n ⬇ 0.1 along the 关001兴 zone axis of c-Si共001兲 substrate 共not
shown兲. Inset show SAED patterns taken from the TiO2 films. A TiO2 crys-
talline nanoparticle is encircled in 共a兲.
SiO2 film, which is seen to thin slightly. The bulk-like TiO2
layer starts to increase only after t ⬃ 730 s. TiO2 deposition
rate 共R兲 is found to be R ⬃ 0.037 nm/ s, and corresponds to
␾i / ␾n ⬃ 6.7 共␾i ⬃ 1 ⫻ 1015 cm−2 s−1 at VB = −450 V;9 ␳
⬃ 3.4 g cm−3兲.
It is quite revealing that tinterface remains between ⬃3 and
4 nm throughout the remainder of the deposition; and does
not shrink with time, as would be expected in case of a
nucleation layer undergoing coalescence. It is well known,
for instance, in the case of TiO2 thermal evaporation pro-
cesses, that the growth begins with the formation of a low-n
nucleation layer, which gradually disappears as coalescence
sets in.35 The results in Fig. 2 suggest that the growth of
TiO2 proceeds differently in the presence of intense ion bom-
bardment. Ti atoms initially deposited near the surface are
knocked into the substrate and ballistically mixed with the Si
and O atoms to form a mixed-oxide layer. The growing film
does not show evidence of nucleation, as islands are likely
dissolved as a result of intense IB 共Ref. 36兲 and a continuous
film is formed, instead. As the film grows beyond ⬃3 or
⬃4 nm, it buries the interface and reduces ion mixing of
interfacial atoms. The IB continues to ballistically mix the
newly deposited Ti and O atoms into the bulk during subse-
quent growth, ensuring densification of the top ⬃4 – 6 nm of
the growing film and a continuous supply of oxygen. In the
next section, we present simulation results in support of this
growth model, and explore ways to control and minimize
interface broadening.
2. High resolution TEM analysis of TiO2 – SiO2
interfaces
Figure 5 shows bright-field HRTEM images of the inter-
facial region between TiO2 and SiO2 for two different
100-nm-thick TiO2 films grown at ␾i / ␾n ⬇ 6 关Fig. 5共a兲兴 and
␾i / ␾n ⬇ 0.1 关Fig. 5共b兲兴 in a TiO2 / SiO2 multilayer stack de-
posited by PECVD on O2 plasma pretreated c-Si. The im-
ages reveal that TiO2 deposition at high ␾i / ␾n is accompa-
nied by significant interfacial broadening, as evidenced by
the gradual variation of the gray tone near the interfacial
region 关Fig. 5共a兲兴. The TiO2 film itself contains some nano-
crystals, which are mainly located in the bulk of the film,
 2066 Amassian et al.: Interface broadening due to ion mixing 2066

away from the interface 关see circle in Fig. 5共a兲兴. Selected

area electron diffraction 共SAED兲 results taken for the TiO2
films are shown in the insets of Figs. 5共a兲 and 5共b兲; they
confirm the assertion that the film grown at ␾i / ␾n ⬇ 6 is
polycrystalline. The absence of crystallinity in the interfacial
region of the polycrystalline film is expected because Si in-
corporation into TiO2 is known to inhibit crystal growth.6,37
In contrast, the film grown at low ␾i / ␾n is amorphous, and it
forms a much more abrupt interface with SiO2, in agreement
with TRIDYN simulation results presented in the next section.

3. TRIDYN simulations of TiO2 deposition

at the rf-biased electrode
TRIDYN simulations of ion bombardment at the rf-biased
electrode during TiO2 deposition at VB = −450 V were per-
formed for ␾i / ␾n ranging from ⬃6 to ⬃0.06, which is
equivalent to R ranging between 0.04 and 4 nm/ s. We ne-
glect the effects of TiCl4-based molecular ions in the IEDF
because their overall impact is negligible compared to O+2
and O+ ions: 共1兲 a very small fraction of Ti and Cl atoms and
molecules are ionized compared to O2, as suggested by the
ratio of partial pressures; pTiCl4 / pO2 ⬃ 0.005; 共2兲 Ti and Cl
ions have smaller kinetic energies and larger radii and
weights than O atoms following molecular dissociation;
hence their implantation is shallower too. For example, as-
suming a TiCl+3 ion incident at Emax ⬃ 500 eV, we estimate
the maximum kinetic energies of Ti and Cl ions after disso-
ciation to be Ei共Ti兲 ⬃ 190 eV and Ei共Cl兲 ⬃ 100 eV; hence
their penetration depths in TiO2 and SiO2 are predicted by
FIG. 6. TRIDYN simulation results of interface formation during TiO2 depo-
SRIM to be less than half the value for the most energetic O+
sition on SiO2. 共a兲 Refractive index 共n550兲 depth profiles of 8-nm-thick TiO2
ions. films 共␾i / ␾n = 0.06, 0.6, and 6兲 calculated from Ti, Si, and O molar fractions
TRIDYN calculations of Ti, Si, and O molar fraction depth shown in the inset. n550 depth profiles of a TiO2 film grown at 共b兲 ␾i / ␾n
profiles are plotted in the inset of Fig. 6共a兲 for ␾i / ␾n = 6 and 共c兲 ␾i / ␾n = 0.6, after tTiO2 = 0.5, 4, and 8 nm.
⬃ 0.06 and 6. The oxygen molar fraction is stable at 67%
共maximum molar fraction allowed in the simulations兲, with
We observe the following general trends: 共1兲 n550 depth pro-
the remainder 共Si and Ti molar fractions兲 going from 0% to
files 共and molar fraction profiles兲 resemble an error function
33%. We can assume that a binary oxide of the form
distribution 共consistent with measured ion mixing profiles38兲;
TixSi1−xO2 is created, as previously observed during PECVD
共2兲 interface broadening increases with ␾i / ␾n; 共3兲 a slight
fabrication of mixed TiO2 / SiO2 films.6 The equivalent re-
profile asymmetry is formed as a result of preferential dis-
fractive index at 550 nm 共n550兲 of this mixed-phase oxide
placement of Si into TiO2 than Ti into SiO2 关ED共Si兲
can be expressed according to the following expression:
⬇ 12 eV and ED共Ti兲 ⬇ 25 eV兴.
n550 = nTiO2x + nSiO2共1 − x兲, 共1兲 The significant difference between interface broadening at
␾i / ␾n = 0.06 and 6 is purely due to ballistic transport pro-
where nTiO2 = 2.4 and nSiO2 = 1.45 at 550 nm. We can use Eq. cesses and can be explained by the hundredfold increase of
共1兲 to express the O, Ti, and Si molar fraction depth profiles ion-atom collisions per unit of deposited film thickness.
in terms of a single parameter 共n550兲, thus simplifying the More collisions lead to more atomic relocations as well as
graphical representation and facilitating the comparison of increased mixing of Ti and Si atoms about the interface per
TRIDYN and RTSE results. unit thickness of the deposit.
The presentation of n550 depth profiles is also appealing We now take a closer look at the effect of ␾i / ␾n on thin
from the perspective of OIF applications, where the interface film growth. The simulated n550 depth profiles are plotted in
width is usually expressed in terms of an n depth profile. The Figs. 6共b兲 and 6共c兲 for various stages of TiO2 growth at
information obtained from TRIDYN can be incorporated into ␾i / ␾n ⬃ 6 共low R兲 and 0.06 共high R兲. They show that the
the design of OIFs, allowing one to compensate for the effect subnanometer growth of TiO2 共tTiO2 ⬍ 1 nm兲 at ␾i / ␾n ⬃ 6 is
of interface broadening on the overall spectral performance dominated by subplantation of O, as most Ti atoms end up
of the OIF by further refining its design. being transported below the SiO2 surface, and a large num-
The simulated n550 depth profiles of 8-nm-thick TiO2 ber of Si atoms are also mixed into the growing film. Sub-
films with ␾i / ␾n = 0.06, 0.6, and 6 are plotted in Fig. 6共a兲. plantation and ion mixing reduce n of the growing oxide as a

J. Vac. Sci. Technol. A, Vol. 24, No. 6, Nov/Dec 2006

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 128.248.155.225 On: Sun, 23 Nov 2014 08:37:25
 2067 Amassian et al.: Interface broadening due to ion mixing 2067

result of Si incorporation; for instance, we see that n550

⬃ 1.75 near the growth surface of a 0.5-nm-thick TiO2 film
deposited at ␾i / ␾n ⬃ 6 共low R兲, corresponding to a TiO2
fraction of only ⬃30% 关Eq. 共1兲兴. After 4 nm of deposition,
n550 has increased substantially both near the growth surface
and in the bulk of the film, but we still have n550 ⬍ nTiO2. The
deposition of TiO2 leads to a natural increase of 关Ti兴 and n550
located near the surface, while at the same time deep-lying Si
atoms get out of range of O ions and become more difficult
to mix into the growing film. 关Ti兴 and n550 also appear to
increase in the bulk of the film 共well below the immediate
growth surface兲 as a result of the ballistic transport of newly
deposited Ti atoms many nanometers into the bulk. This be-
havior is consistent with RTSE results shown in step 共2兲 of
Fig. 2, where the initial growth appears to take place inside
the SiO2, forming a mixed-oxide layer during the initial
⬃3 nm of TiO2 deposition 关Fig. 2, step 共2兲兴. The existence of
significant subsurface transport of surface deposited Ti atoms
is one of the main conclusions of this week. It suggests that
existing thin film growth models should be revised to incor-
porate the subsurface effects of intense ion bombardment, as
will be discussed shortly.
In contrast, subsurface transport and ion mixing are much
less important at high R, as suggested by the abrupt variation
of n550 at the interface 关Fig. 6共c兲兴. We conclude that high
␾i / ␾n 共low R兲 conditions lead to broader interface formation
than at low ␾i / ␾n simply because IB can displace and mix
more atoms per unit time or thickness in the former case.
The simulated interface broadening can be quantified in
terms of an interfacial thickness 共tinterface兲, which is calcu-
lated by taking the distance between the 90% and 10% po-
sitions of the Ti molar fraction depth profiles 共Fig. 6兲. We
have plotted tinterface in Fig. 7共a兲 as a function of tTiO2 for
different ␾i / ␾n; a reference line corresponding to tinterface
= tTiO2 is included in the plot to judge whether surface depo-
sition 共tTiO2 ⬎ tinterface兲 or subsurface growth by subplantation
共tTiO2 ⬍ tinterface兲 is the dominant growth mechanism of TiO2 FIG. 7. 共a兲 Evolution of tinterface vs tTiO2 for different R and ␾i / ␾n values
during deposition of TiO2 on SiO2; 共b兲 tinterface as a function of ␾i / ␾n for
at the rf-biased electrode in PECVD. For thin films deposited different tTiO2 values 共tTiO2 = 0.5, 1, 4, 8, and 15 nm兲 plotted on logarithmic
at ␾i / ␾n 艌 0.6, we observe that tTiO2 ⬍ tinterface during the ini- scale; the inset shows the same data 共tTiO2 = 1 and 15 nm, solid and dashed
tial ⬃2 nm of deposition, indicating that the majority of Ti curves, respectively兲 plotted on a linear scale.
atoms are displaced below the growth surface and that TiO2
growth is dominated by subplantation. By increasing
R共␾i / ␾n ⬍ 0.6兲 we observe a saturation of tinterface at smaller ␾i / ␾n = 0.82 are situated between simulations at ␾i / ␾n = 0.6
thickness, so that tTiO2 ⬎ tinterface is always true, indicating and 1.2, as expected, while the data points for ␾i / ␾n = 6.7 are
that fewer surface atoms are displaced into the bulk; in this ⬃1 – 1.5 nm below the TRIDYN predictions for ␾i / ␾n = 6.
case, surface deposition and diffusion are likely the dominant This discrepancy may be linked to the restrictive nature of
growth processes. the two-layer optical model, where both layers’ optical prop-
Experimental results of the evolution of tinterface at ␾i / ␾n erties were kept constant.
= 0.82 and 6.7, obtained by in situ RTSE analysis, are shown The lack of fully quantitative agreement between the re-
in Fig. 7共a兲 共solid symbols兲. Initially, interface broadening sults presented above can also be due in part to inherent
evolves linearly with tTiO2; saturation is reached with limitations of the TRIDYN code, which does not account for
tinterface ⬃ 3.5± 0.5 and ⬃2.3± 0.3 nm for ␾i / ␾n ⬃ 6.7 and thermal or transient diffusion processes, leading to either un-
0.82, respectively. The linear increase of tinterface is likely an derestimation or overestimation of the overall interfacial
artifact of the data analysis and is believed to be caused by broadening. When diffusion is activated, chemical and ther-
using a two-layer model with fixed optical constants. TRIDYN modynamic driving forces determine the transport direction
predictions agree reasonably well with RTSE results and of atoms, either giving rise to additional broadening of the
trends. Quantitatively, the experimental data points at interface or to its contraction. The latter is possible, if both

JVST A - Vacuum, Surfaces, and Films

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 128.248.155.225 On: Sun, 23 Nov 2014 08:37:25
 2068 Amassian et al.: Interface broadening due to ion mixing
film materials are immiscible with each other. However, evi-
dence in the literature shows that ion mixing of materials
with an average atomic number 共Z兲 below 18 is of pure
ballistic character. Furthermore, as the substrate temperature
does not exceed ⬃100 ° C at VB = −450 V,31 and remains be-
low 20% of the melting point of either TiO2 or SiO2, it is
reasonable to neglect bulk diffusion processes as well. Bulk
diffusion is generally a dominant process when substrate
temperature approaches 75% of the melting temperature.39
The simulated tinterface data is plotted on a logarithmic
scale in Fig. 7共b兲 as a function of ␾i / ␾n for tTiO2 = 0.5, 1, 4, 8,
and 15 nm. We observe that at steady state, tinterface scales
with 共␾i / ␾n兲1/2 for all ␾i / ␾n. This scaling law is similar to
what has been observed during ion beam mixing at the inter-
face between two static thin films.38 However, the situation is
slightly more complicated here because ion mixing occurs
during thin film deposition; two important consequences
should be noted: 共1兲 the growing film continually pushes the
ion penetration front away from the interface, and 共2兲 the
film is a finite source of atoms, with very little Ti compared
to Si during the early stages of growth. The first point is
illustrated by the saturation behavior of tinterface with tTiO2
关Fig. 7共a兲兴. Given that the penetration depth of hyperthermal
ions is less than ⬃10 nm in SiO2 and ⬃6 nm in TiO2, and
that the ion implantation depth profile decays with depth, it is
clear that ion-interface interactions will decrease substan-
tially during thin film growth, resulting in the saturation be-
havior observed in Fig. 7共a兲.
The second point suggests that in some conditions 共at
high ␾i / ␾n兲, the ion fluence per unit film deposit can be
sufficiently high to displace most, if not all, deposited atoms,
as well as many bulk atoms within range. This happens when
there is not a sufficiently large supply of any one of the
atomic species to continue the broadening process at the in-
terface. From Fig. 6共b兲, it appears that at least half of the Ti
atoms deposited at tTiO2 = 0.5 nm are displaced into the SiO2
and are replaced by Si atoms, thereby reducing the mixing
potential. This is illustrated by the “premature” saturation of
tinterface at ␾i / ␾n ⬎ 2 for tTiO2 = 0.5, 1, 4, and 8 nm 关Fig. 7共b兲兴,
which is also visible in the inset of Fig. 7共b兲 for tTiO2
= 1 nm. Not surprising, either, that the mixing law is gradu-
ally restored for all ␾i / ␾n, as enough Ti is supplied to the
target for the mixing to take place 关see inset of Fig. 7共b兲 for
tTiO2 = 15 nm兴.
We propose a simple rule of thumb to estimate interface
broadening as a function of deposition conditions, based on
the results of Fig. 7共b兲. Combining the notions that the num-
ber of incident ions for each incident atom is given by ␾i / ␾n,
and that the number of recoils 共NIV兲 created by each ion can
be estimated from the Kinchin-Pease relation, given by NIV
= Em / 2Ed, where Em ⬃ 0.3e兩VB兩 共Ref. 15兲 is the mean ion
energy of the IEDF, we find that in light of Fig. 7共a兲 thin film
growth mechanism can be limited to surface deposition and
the interface broadening 共by ballistic processes兲 can be kept
below ⬃1 nm, if the condition 共␾i / ␾n兲共Em / 2Ed兲 艋 1 is met.
For instance, when VB = −450 V 共Em ⬃ 135 eV兲, NIV is in the
vicinity of ⬃5 to 6 for Si atoms in SiO2 共ED = 12 eV兲 and ⬃2
J. Vac. Sci. Technol. A, Vol. 24, No. 6, Nov/Dec 2006
Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 128.248.155.225 On: Sun, 23 Nov 2014 08:37:25
2068
to 3 for Ti atoms in TiO2 共ED = 25 eV兲. Taking the average,
i.e., NIV ⬃ 4, we calculate that tinterface 艋 1 nm only if ␾i / ␾n
艋 0.25, or R ⬎ 1 nm/ s for fixed ␾i. This rule can point out
the extent of interface broadening-due to ballistic ion
mixing-for given deposition conditions; and it can be very
useful for optimizing the fabrication of OIFs and other ap-
plications that require abrupt interfaces, or to dealing with
the interface broadening by incorporating the interfacial
width in the OIF design.
V. CONCLUSION
We have demonstrated, using a combination of in situ
RTSE, TRIDYN simulations, and HRTEM, that ion-surface
interactions at the rf-biased electrode in a low-pressure
PECVD environment can lead to significant subsurface
modifications as a result of subplantation, similar to ion
beam processes, operating under conditions of intense ion
bombardment, and that this will cause interface broadening
during the early stages of thin film growth. In situ RTSE
monitoring of O2 plasma pretreatment of c-Si has revealed
that oxygen subplantation is capable of causing damage for-
mation, and near-surface oxidation on time scales of Ⰶ1 s
and ⬃2 − 3 s, respectively. While monitoring the early stages
of TiO2 deposition on SiO2, in situ RTSE revealed that oxy-
gen subplantation also causes the formation of a mixed-oxide
interfacial layer with thicknesses of ⬃3.5± 0.5 and
⬃2.3± 0.3 nm for ␾i / ␾n ⬃ 6.7 共R = 0.037 nm/ s兲 and 0.82
共R = 0.3 nm/ s兲, respectively. TRIDYN simulations confirmed
that interface broadening is physically viable and that it is
mainly caused by ballistic ion mixing of interfacial atoms.
TRIDYN simulations aslo enable us to identify growth re-
gimes which are dominated by subsurface deposition and
revealed that any material deposited on the surface can be
ballistically transported below the growth surface. The rela-
tive importance of subsurface transport with respect to sur-
face deposition was found to depend mainly on ␾i / ␾n and on
the mean ion energy. In addition, simulations also predicted
that the interfacial thickness scales as 共␾i / ␾n兲1/2, in agree-
ment with ion mixing laws.
The methodology combining in situ real-time monitoring
by RTSE with TRIDYN simulations has given us the ability to
quantify and predict some of the effects of ion-surface inter-
actions at the rf-biased electrode in a PECVD environment.
This approach can be easily extended to study other IB-
intensive processes, such as ion beam assisted deposition and
ion plating, where ion energy typically reaches a few hun-
dreds of eV or higher. This approach is also promising in that
it can be used to diagnose and quantify ion-surface interac-
tions in the case of more novel processes, such as high power
impulse magnetron sputtering40 or arc deposition,41 where all
the sputtered particles and gas phase species become ionized,
leading to films entirely made up of so-called “film-forming”
ions with complex energy distribution functions. The devel-
opment of in situ diagnostics strategies and of predictive
capabilities will become essential to the understanding of the
subsurface transport of ions and their effects on interfacial
 2069 Amassian et al.: Interface broadening due to ion mixing 2069

15
properties, microstructure, adhesion, and mechanical proper- A. Amassian, P. Desjardins, and L. Martinu, J. Vac. Sci. Technol. A 24,
45 共2006兲.
ties, among others. 16
J. F. Ziegler, J. P. Biersack, and U. Littmark, The Stopping and Range of
Ions in Matter 共Pergamon, New York, 1985兲, Vol. 1.
ACKNOWLEDGMENTS 17
Y. Lifshitz, S. R. Kasi, and J. W. Rabalais, Phys. Rev. Lett. 62, 1290
共1989兲.
The authors wish to thank Stéphane Larouche for his as- 18
W. Moeller and S. Mukherjee, Curr. Sci. 83, 237 共2002兲.
sistance with PECVD experiments, and Etienne Bousser and 19
A. Barna, M. Menyhard, G. Zsolt, A. Koos, A. Zalar, and P. Panjan, J.
Jean-Philippe Masse for HRTEM measurements. The authors Vac. Sci. Technol. A 21, 553 共2003兲.
20
also acknowledge Matthias Posselt of Forschungszentrum S.-Y. Chun, J. Ceram. Proc. Res. 4, 115 共2003兲.
21
Rossendorf 共FZR兲 for providing a copy of the TRIDYN simu- Optical Interference Coatings, edited by N. Kaiser 共H. K. Pulker,
Springer, 2003兲.
lation code and Oleg Zabeida for helpful discussions. This 22
E. B. Svedberg, J. Girch, I. Ivanov, E. P. Münger, and J.-E. Sundgren, J.
work was supported by the Natural Sciences and Engineer- Vac. Sci. Technol. A 16, 633 共1998兲.
ing Research Council 共NSERC兲 of Canada and the Canadian 23
W. Möller and W. Eckstein, Nucl. Instrum. Methods Phys. Res. B 2, 814
Foundation for Innovation 共CFI兲 programs. One of the au- 共1984兲.
24
W. Möller, W. Eckstein, and J. P. Biersack, Comput. Phys. Commun. 51,
thors 共P.D.兲 acknowledges the support from the Canada Re- 355 共1988兲.
search Chair 共CRC兲 program. Another author 共A.A.兲 ac- 25
S. S. Todorov and E. R. Fossum, Appl. Phys. Lett. 52, 48 共1988兲.
26
knowledges the support from the NSERC Postgraduate O. Vancauwenberghe, N. Herbots, and O. C. Hellman, J. Vac. Sci. Tech-
Scholarship 共PGS兲 and Postdoctoral Fellowship 共PDF兲 nol. A 10, 713 共1992兲.
27
J. S. Williams, M. Petravić, B. G. Svensson, and M. Conway, Appl. Phys.
programs.
Lett. 76, 1840 共1994兲.
28
1
A. Amassian, M. Svec, P. Desjardins, and L. Martinu, J. Appl. Phys.
S. M. Reiss, Opt. Photonics News 8, 31 共1997兲. 共in press兲.
2
Handbook of Deposition Technologies for Films and Coatings, edited 29
J. E. Klemberg-Sapieha, O. M. Küttel, L. Martinu, and M. R. Wertheimer,
by R. F. Bunshah 共Noyes, Park Ridge, NJ, 1994兲. Thin Solid Films 193/194, 965 共1990兲.
3
Thin Film Processes, edited by J. L. Vossen and W. Kern 共Academic, 30
S. Larouche, A. Amassian, S. C. Gujrathi, J. E. Klemberg-Sapieha, and L.
New York, 1991兲, Vol. 2.
4 Martinu, Proceedings of the Society of Vacuum Coaters 44th Annual
L. Martinu and D. Poitras, J. Vac. Sci. Technol. A 18, 2619 共2000兲.
5 Technical Conference, Philadelphia, 2001, p. 277.
R. Vernhes, O. Zabeida, J. E. Klemberg-Sapieha, and L. Martinu, Appl. 31
A. Amassian, P. Desjardins, and L. Martinu, Thin Solid Films 447/448,
Opt. 45, 97 共2004兲.
6 40 共2004兲.
S. Larouche, H. Szymanowski, J. E. Klemberg-Sapieha, L. Martinu, and 32
S. C. Gujrathi, J. Vac. Sci. Technol. A 22, 1200 共2004兲. J. F. Ziegler, SRIM 2003, computer simulation software available for down-
7
A. Amsassian, R. Vernhes, J. E. Klemberg-Sapieha, P. Desjardins, and L. load via the internet at http://www.SRIM.org.
33
Martinu, Thin Solid Films 469/470, 47 共2004兲. W. Bock, H. Gnaser, and H. Oechsner, Surf. Sci. 282, 333 共1993兲.
34
8
R. Vernhes, A. Amassian, J. E. Klemberg-Sapieha, and L. Martinu, J. G. E. Jellison, Jr. and F. A. Modine, Appl. Phys. Lett. 69, 371 共1996兲.
35
Appl. Phys. 99, 114315 共2006兲. Y. Leprince-Wang and K. Yu-Zhang, Surf. Coat. Technol. 140, 155
9
O. M. Küttel, J. E. Klemberg-Sapieha, L. Martinu, and M. R. Wertheimer, 共2001兲.
36
Thin Solid Films 193/194, 155 共1990兲. L. Hanley and S. B. Sinnott, Surf. Sci. 500, 500 共2002兲.
10 37
A. Hallil, O. Zabeida, M. R. Wertheimer, and L. Martinu, J. Vac. Sci. P. Ahmet, M. Takakura, K. Nakajima, T. Koida, M. Yoshimoto, H.
Technol. A 18, 882 共2000兲. Koinuma, and T. Chikyow, Appl. Surf. Sci. 189, 307 共2002兲.
38
11
C. Wild and P. Koidl, J. Appl. Phys. 69, 2909 共1991兲. W. Bolse, Mater. Sci. Eng., A 253, 194 共1998兲.
39
12
L. Martinu, J. E. Klemberg-Sapieha, O. M. Küttel, A. Raveh, and M. R. J. A. Thornton, J. Vac. Sci. Technol. 11, 666 共1974兲.
40
Wertheimer, J. Vac. Sci. Technol. A 12, 1360 共1994兲. A. P. Ehiasarian, P. Eh. Hovsepian, L. Hultman, and U. Helmersson, Thin
13
C. R. Ottermann and K. Bange, Thin Solid Films 286, 32 共1996兲. Solid Films 457, 270 共2004兲.
14
M. Ishimaru, S. Harada, and T. Motooka, J. Appl. Phys. 81, 1126 共1997兲. 41
A. Anders, Vacuum 67, 673 共2002兲.

JVST A - Vacuum, Surfaces, and Films

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 128.248.155.225 On: Sun, 23 Nov 2014 08:37:25