Measurement of the Energy-Band Relations of Stabilized Si Photoanodes Using
Operando Ambient Pressure X-ray Photoelectron Spectroscopy
M. H. Richter
a
*, M. F. Lichterman
a,b
*, S. Hu
a,b
*, E. J. Crumlin
c
*, T. Mayer
d
, S. Axnanda
c
,
M. Favaro
a,c
, W. Drisdell
a,c
, Z. Hussain
c
, B. S. Brunschwig
b,e
, N. S. Lewis
a,b,e,f
, Z. Liu
c,g,h
,
and H.-J. Lewerenz
a‡
a
Joint Center for Artificial Photosynthesis, California Institute of Technology, Pasadena,
CA 91125, USA.
b
Division of Chemistry and Chemical Engineering, California Institute of Technology,
Pasadena, CA 91125, USA.
c
Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA.
d
Surface Science Division, Materials Science Department, Darmstadt University of
Technology, 64287 Darmstadt, Germany.
e
Beckman Institute, California Institute of Technology, Pasadena, CA 91125, USA
f
Kavli Nanoscience Institute, California Institute of Technology, Pasadena, CA 91125,
USA.
g
State Key Laboratory of Functional Materials for Informatics, Shanghai Institute of
Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai
200050, People’s Republic of China.
h
School of Physical Science and Technology, Shanghai Tech University, Shanghai
200031, China.
‡
Corresponding author E-mail: lewerenz@caltech.edu
The energy-band relations and electronic properties for the light-
absorber/protection-layer stack of TiO
2
-stabilized Si photoanodes
have been determined by ambient pressure x-ray synchrotron
radiation photoelectron spectroscopy under an applied potential
(
operando
), from single core-level emission lines. The experiments
have
also
been
complemented
with
laboratory-based
monochromatic XPS data. Electrochemical parameters are
additionally derived directly from x-ray photoemission data, and a
method is presented to derive interface-state densities from such
operando data.
Introduction
The electrochemical potential at semiconductor/liquid, semiconductor/metal, and
semiconductor/semiconductor junctions equilibrates by charge transfer across the
interface between the two contacting phases. This charge transfer consequently produces
band bending, (partial) Fermi level (E
F
) pinning at interface states, and the formation of
interface dipoles (1,2). The energy-band alignment also affects the electronic properties
and performance of the resulting photoelectrochemical cell. Typically, the energy-band
alignment is determined experimentally by a combination of x-ray photoelectron
spectroscopy (XPS) and ultraviolet photoemission spectroscopy (UPS) during the step-
wise growth of a contacting phase on top of the substrate of interest (3,4).
10.1149/06606.0105ecst ©The Electrochemical Society
ECS Transactions, 66 (6) 105-113 (2015)
105
) unless CC License in place (see abstract).
ecsdl.org/site/terms_use
address. Redistribution subject to ECS terms of use (see
131.215.70.231
Downloaded on 2016-11-21 to IP
This approach does not, however, consider the integral nature of photoemission
spectroscopy, and is not feasible for
operando
investigation of solid-liquid interfaces.
We describe herein the use of ambient pressure photoelectron spectroscopy (AP-PES)
(5,6) and standard x-ray photoelectron spectroscopy (XPS) to analyze the energetics of
light-absorber/protection-layer stacks (7-9). Specifically, the integral nature of PES and
the large inelastic mean-free path (IMFP,
λ
) of tender X-Rays for
i
n-situ
PES has been
used to determine the electrostatic (Galvani) potential of semiconductor junctions by
x-ray photoelectron spectroscopy. All of the required parameters have been determined
experimentally from single core-level emission-line profiles at each applied potential,
U
.
For the
ex-situ
XPS experiment, this method has been applied to the growth by atomic-
layer deposition (ALD) of a TiO
2
protection layer on a Si surface.
Experimentation
Deposition of TiO
2
was performed by ALD on degenerate, p-type boron
(4 × 10
19
cm
-3
) and n-type arsenic doped (3 × 10
19
cm
-3
) Si(100) substrates (7). Silicon
wafers were cleaned with an RCA SC-1 procedure by immersion in a 3:1 (by volume)
“piranha” solution of ~ 18.4 M H
2
SO
4
and ~ 11 M H
2
O
2
for 10 min, followed by a 10 s
etch in 10 % by volume of hydrofluoric (HF) acid, and finally, an RCA SC-2 etch of
5:1:1 (by volume) solution of H
2
O, 11.6 M hydrochloric acid, and ~11 M H
2
O
2
for
10 min
at
75 °C.
TiO
2
was
then
deposited
by
ALD
from
a
tetrakis(dimethylamido)titanium (TDMAT) precursor. A 0.1 s pulse of TDMAT was
followed by a 15 s purge of N
2
at 20 sccm, followed by a 0.015 s pulse of water and
another 15 s N
2
purge. The layer growth rate was determined to be 0.04 nm per cycle (7).
Operando (in-situ)
ambient pressure x-ray photoelectron spectroscopy data were
collected on a Scienta R4000 HiPP-2 system, with the photoelectron collection cone
aligned to the beamline x-ray spot at a distance of ~ 300 μ m. The system used differential
pumping supplied by four turbo pumps, backed by rough pumps that were protected by
liquid-nitrogen cold traps, to maintain a pressure of ~ 5 x 10
-7
mbar at the detector while
allowing a stable pressure of 27 mbar at the sampling position. Beamline 9.3.1 at the
Advanced Light Source was used to provide “tender” x-rays with an energy of
h
ν
=
4000 eV. The potential,
U
, was applied to the substrate (working electrode) in a
three-electrode potentiostatic configuration (6).
Ex-situ
x-ray photoelectron spectroscopy measurements were performed using a
Kratos Axis Ultra system with a base pressure of < 5 x 10
-10
mbar. X-rays were produced
by a monochromatic Al K
α
(h
ν
=
1489.6 eV) source with a power of 150 W.
ECS Transactions, 66 (6) 105-113 (2015)
106
) unless CC License in place (see abstract).
ecsdl.org/site/terms_use
address. Redistribution subject to ECS terms of use (see
131.215.70.231
Downloaded on 2016-11-21 to IP
Figure 1. Model of the light-absorber/protection-layer/catalyst stack used in this
investigation, consisting of a silicon substrate (e.g. n-doped Si bulk), the Si space-charge
region (Si-SCR), the SiO
2
interface region, a TiO
2
protection layer, and an electrolyte
layer for
in-situ
investigation. The expected potential distribution,
U
(x), is depicted for
each layer. An applied potential will drop partially in the silicon space-charge region, in
the SiO
2
, in the TiO
2
at the interface with the electrolyte, and in the electrolyte double
layer. For a highly doped silicon substrate, the potential drop will appear almost
exclusively at the TiO
2
/electrolyte interface. The thickness of the electrolyte, TiO
2
, and
SiO
2
layer is specified for both
ex-situ
(h
ν
= 1489.6 eV) and
in-situ
(h
ν
= 4000 eV)
experiments.
Results and Discussion
A rigorous description and evaluation (data fitting) of the experimental data requires a
precise calculation of the potential distribution in the semiconductor. In the Schottky
junction model (10), the charge density in the depletion region is considered to be equal
to the (spatially constant) acceptor density,
N
A
(or donor density,
N
D
), with the charge
density assumed to abruptly become zero at the edge of the depletion layer (i.e. the
abrupt-junction approximation). In the abrupt-junction approximation, upon
approximating the Fermi-Dirac distribution function with a Boltzmann distribution
function, the width of the depletion layer,
d
SCR,
is given by the doping concentration,
N
,
of the material in conjunction with a potential drop,
U
0
, in the space-charge region
(equation 1):
=
∙
[1]
Taking the boundary conditions as
U
(x
= 0) = 0 (surface) and
U
(x
= d) =
U
0
(bulk),
the electrostatic potential,
E
pot
(x)
= -e
U
(x), across the space-charge region is given by
equation 2:
|
=
∙
∙−
;
|
=
[2]
ECS Transactions, 66 (6) 105-113 (2015)
107
) unless CC License in place (see abstract).
ecsdl.org/site/terms_use
address. Redistribution subject to ECS terms of use (see
131.215.70.231
Downloaded on 2016-11-21 to IP
In the space-charge region with 0 <
x
<
d
scr
,
U
(x) depends quadratically on
x
. For
d
>
d
scr
, the potential is constant and is given by
U
0
, i.e.
U
(
x
>
d
scr
)
=
U
0
.
However, an exact solution for
U
(x) is required to cover the full potential range that
extends beyond the thermodynamic potentials for the oxidation (
U
OER
) and reduction
(
U
HER
) of water, as is needed to describe
operando
AP-PES measurements, and/or to
describe the situation for highly or degenerately doped semiconductors. This relationship
can be obtained by solving the full Poisson equation with Femi-Dirac statistics,
f
(
E
)
(equation 3):
!
"#$
!
=
%
&
'
(
)
−(
*
)
]∙,
)
-)+/
0
1
−/
2
3
4
[3]
I
n equation 3,
(
E
and
(
*
E
are the effective density of states in the conducti
on band
and valence band, respectively, and
/
0
1
and
/
2
3
are the density of ionized donors and
acceptors, respectively.
Figure 2 depicts the potential distribution for both of the potential descriptions
(equation 2 and 3) at the outer layer of bulk Si. A potential shift of
U
= +0.1 V vs
U
fb
(depletion) was assumed in Figure 2a. The potential distribution was calculated for a
moderate doping concentration of
N
D
= 1 x 10
17
cm
-3
. For n-type Si and a positive bias
(+
U
), downward band bending is produced, in which the difference between E
F
and the
conduction band edge, E
C
, increases towards the surface.
In the bulk, i.e. for
x
>
d
SCR
,
U
reaches its bias value, and thus the difference between
E
F
and either E
CB
or E
VB
is constant and depends solely on the doping concentration.
Hence, for a neutral sample, zero difference in binding energy,
E
B
, will be produced
between the measured core-level emission lines relative to their bulk values. Only when
0 <
d
<
d
SCR
, where
U
approaches zero, will the Fermi energy shift because of band
bending, with
E
B
≠
0.
F
or x-ray photoemission spectroscopy, 95 % of the contribution to the measured core-
level emission-line profile arises from an information depth of 3
⋅λ
(three times the
i
nelastic electron mean free path). Thus, a change in the electric potential,
U
(x), i.e. the
Galvani Potential of the semiconductor, across 0 <
d
< 3
⋅λ
will have a strong influence on
t
he position of the core-level emission maximum, as well as on the line shape, because
the photoelectrons emitted at different distances from the surface will be exposed to a
different potential
U
, i.e. will have different binding energies. XPS integrates over the
different core-level peak signals produced by emission from different depths in the solid,
which consequently produces a broadened core-level emission that results in a poorly
resolved electronic shift and an apparent band-energy alignment signal. In particular, the
maximum core-level emission shift at the interface is, in most cases, underestimated. This
behavior occurs because in the presence of a space-charge layer, the observed peak
maximum position does not represent the core-level position of the topmost layer, and the
emission lines are convoluted in the observed signal due to superposition of the lines
from different depths with different weighing factors.
ECS Transactions, 66 (6) 105-113 (2015)
108
) unless CC License in place (see abstract).
ecsdl.org/site/terms_use
address. Redistribution subject to ECS terms of use (see
131.215.70.231
Downloaded on 2016-11-21 to IP
Figure 2. Electrostatic potential for n-Si (a) as calculated by the abrupt junction
approximation (Boltzmann, dashed line), and the full Poisson equation (Fermi-Dirac,
solid line) for a potential of
U
= +0.1 V (depletion). The change of the core-level binding
energy
E
B
(x) is
E
B
(x) =-e[
U
(d)-
U
(x)]. The inset in (a) depicts the movement of the
conduction-band minimum (CBM) and valence-band maximum (VBM) upon application
of a positive potential,
U,
to the electrode. Due to the assumption of an abrupt junction,
the quadratic form of
U
(x) obtained from Boltzmann statistics shows a kink at
d
SCR
,
whereas
U
(x) obtained assuming Fermi-Dirac statistics is smooth over the complete
range of
x
. In (b), the simulated core-level emission line profile for the neutral sample
(
U
= 0.0 V, solid line) and for
U
= +0.1 V (dashed line) are shown, calculated relative to
the binding energy signals for the bulk semiconductor. The solid-dashed line shows the
additional broadening of the spectra at
U
= +0.1 V with respect to
U
0 +0.1 V, obtained
as the difference spectrum between the data at the applied potential of interest and the
data at the flat-band potential. Due to the broadening, the apparent peak (maximum) shift
of the Si 2p
3/2
and Si 2p
1/2
signals for the
U
= +0.1 V spectrum with respect to the
U
= 0.0 V spectrum is
E
B
= -0.07 eV instead of the expected value of
E
B
= -0.1 eV.
The calculations for obtaining the data of the core-level emission line profile are
described in detail in the text.
For a combination of Gaussian and Lorentzian line profiles (also known as pseudo
Voigt profile), the core-level emission intensity is a function of the binding energy E
B
,
electric potential
U
(x), half-width at half-maximum
Γ
, a scaling parameter
C
1
that is
proportional to the peak height, and a shape parameter
C
2
(fraction of Lorentz to Gauss
peak shape). The observed signal intensity,
I
(
E
), can be expressed as an integration over
the contribution from each atomic layer at a depth
x
(equation 4):
6
)
=
7
8
∙
'
9
3
:
;
∙<
1−7
∙9
3>?∙
@A@
B
CDEF
G
+
81
@A@
B
CDEF
G
H-
[4]
Figure 2b depicts the simulated core-level emission-line profile (equation 4) as
obtained for the given potential
U
(x) calculated from the full Poisson equation, using
tender X-Ray radiation of h
ν
=
4000 eV (
λ
Si
= 7.2 nm (11,12)). As expected from the
(a)
(b)
E
F
E
CB
E
VB
+
U
ECS Transactions, 66 (6) 105-113 (2015)
109
) unless CC License in place (see abstract).
ecsdl.org/site/terms_use
address. Redistribution subject to ECS terms of use (see
131.215.70.231
Downloaded on 2016-11-21 to IP
underlying approximation of the electrostatic potential, the resulting core-level emission-
line profiles are different for different potentials,
U
. Asymmetric line shapes should be
observed for high doping concentrations as well as for large positive applied potentials.
Figure 2b includes the additional broadening induced by the potential, depicted as the
solid-dashed line in the figure. The peak profile asymmetry and broadening will increase
with doping and with applied potential. In this case, the core-level maximum shift is
smaller than the applied potential. For the case shown in figure 2, the shift is
E
B
=
-0.07 eV, in contrast to the expected value of
E
B
= -0.1 eV at
U
= +0.1 V. This
behavior is due to integral nature of XPS and
U
(x)
≠
constant. The maximum shift of
E
B
= -e
U
(
d
) is observed only at the surface (
x
= 0), but core-level emissions with
E
B
= -eU(
x
<
d
) will also contribute to the XPS signal, resulting in peak broadening and
a reduced experimentally observed core-level shift at the emission maximum.
For the potential range used in
operando
AP-PES, a quantitative description over the
complete range of
U
requires use of the full Poisson equation. As the range of applied
potential exceeds the band-gap of the semiconductor, the Fermi energy will approach the
conduction band, or the valence band, and thus Boltzmann statistics will not be an
appropriate description. A quantitative description of the core-level emission-peak profile
can be obtained from the actual function
U
(x) by fitting equation 4 to the experimental
PES data. With known
U
(x) profiles, the energy-band relations can then be obtained for a
single bias, as well as over the complete bias range of interest.
Several additional important parameters, such as the flat-band potential (
U
fb
), the
valence-band (
U
VB
) and conduction-band (
U
CB
) offsets, and the density of in-gap
interface states (N
IS
), can also be obtained from the experimental core-level emission-line
profile. Specifically, at the flat-band potential, a minimum in the FWHM of the
corresponding core-level signals is expected. Explicitly considering the dependence of
the FWHM on
U,
at
U
=
U
fb
the first derivative is zero (
!IJKL
!
=0
) and the second
d
erivative is positive (
!
IJKL
!
>0
). Also, at
U
fb
, the binding energy of the core-levels is
not a function of depth
x
, because no space-charge region (band bending) is present. In
contrast, at potentials positive or negative of
U
fb
, the actual core-level binding energy
depends on
x,
and consequently an asymmetric peak broadening will be observed in the
PES data. Hence the smallest value of FWHM (sharpest core-level emission-line profile)
should indicate the position of
U
=
U
fb
.
For potentials energetically close to, and beyond, the values of
U
CB
and
U
VB
the
change of the core-level binding energy will be constant, i.e.
Δ)
P
≅7
. Hence, in this
p
otential range, the band edges will move as the applied potential is changed. At
potentials more positive than the VBM (
U
>
U
VB
)
or more negative than the CBM
(
U
<
U
CB
)
, the slope of the core-level emission-maximum shift,
E
B
vs
U,
will become
approximately zero, due to a change from a band-bending regime to a band-edge shifting
regime. Hence, in this regime, the difference between E
F
and the VBM or CBM will stay
approximately constant as E
F
approaches either the VBM or CBM. A further shift of the
Fermi energy into the conduction band or valence band will not result in a further
increase of
E
B
,
i.e.
!RS
B
!
≅ 0
.
F
or potentials between
U
CB
and
U
VB
, the difference between the applied potential
change and the sum of the binding energy change
E
B
of each layer can be attributed to a
charging or discharging of interface states as the Fermi energy is moving across these
states. Specifically, for
U
VB
>
U
>
U
CB
, the difference between overall band bending
ECS Transactions, 66 (6) 105-113 (2015)
110
) unless CC License in place (see abstract).
ecsdl.org/site/terms_use
address. Redistribution subject to ECS terms of use (see
131.215.70.231
Downloaded on 2016-11-21 to IP
(
E
B
)
and the applied potential
U
(derivative of
E
B
by
U
) is related to the density of
interface states N
IS
(
U)
at that potential.
The varied information contained in the line shape of the core-level emission signal
allows for the identification of the flat-band potential, usually extracted from
Mott-Schottky analysis, and also allows for the determination of the band gap from the
core-level AP-PES signals.
Figure 3. Experimental values of the FWHM for the respective core-level emissions of
ex
-
situ
Si/SiO
2
/TiO
2
(a) and
in
-
situ
Si/SiO
2
/TiO
2
/electrolyte (b) stacks. Panel (a) presents
the FWHM of the Si 2p
3/2
and Ti 2p
3/2
signals on n
+
-Si and p
+
-Si substrates. Panel (b)
presents the FWHM of the O 1s-TiO
2
and O 1s-H
2
O core-level emission. The FWHM of
the O 1s-TiO
2
signals showed a minimum at
U
= -0.9 V, whereas the FWHM of the O 1s
signal due to the electrolyte stayed nearly constant. The black lines are linear fits to the
data.
Figure 3 shows the FWHM values of the (a)
ex
-
situ
and (b)
in
-
situ
core-level
emission-line profiles of (a) the Si/SiO
2
/TiO
2
and (b) the Si/SiO
2
/TiO
2
/electrolyte stack.
For the growth of TiO
2
on p
+
-Si or n
+
-Si, a change was observed in the FWHM of the
Si 2p
3/2
and Ti 2p
3/2
core-level emissions. As described above, a minimum in the value of
the FWHM is expected at the flat-band condition, and an increase in FWHM should
occur for depletion or accumulation conditions. Hence, a decrease of the FWHM with
increased cycles of ALD indicated a trend toward flat-band conditions as the number of
ALD cycles was increased. Thus, for the growth of TiO
2
by ALD (figure 3a), the FWHM
trend of the Ti 2p
3/2
signal indicated that the TiO
2
layer approached the flat-band
condition, for both, n
+
-Si and p
+
-Si. The trend of the Si 2p
3/2
FWHM indicated an
approach to flat-band for p
+
-Si (decrease of the FWHM) but the formation of a depletion
layer for n
+
-Si (increase of the FWHM), as confirmed by the observed shift of the core-
level emission maximum towards lower binding energies. In addition, the ongoing
change of the FWHM as the number of ALD cycles increased demonstrated the dynamic
formation of a complex Si/TiO
2
interface.
For the TiO
2
/electrolyte interface (figure 3b) on a highly doped p
+
-Si substrate, the
FWHM of the O 1s core-level emission arising from the electrolyte was constant over the
complete potential range. In contrast, the FWHM of the O 1s core-level emission of TiO
2
displayed a minimum at
U
= -0.9 V. This behavior indicates that the applied potential
dropped essentially completely in the semiconductor, which exhibited a flat-band
(a)
(b)
ECS Transactions, 66 (6) 105-113 (2015)
111
) unless CC License in place (see abstract).
ecsdl.org/site/terms_use
address. Redistribution subject to ECS terms of use (see
131.215.70.231
Downloaded on 2016-11-21 to IP
potential of
U
fb
= -0.9 V vs. Ag/AgCl. In addition, for depletion and accumulation, a
difference in the slope (back line, figure 3b) is expected because the potential drop is
limited to a smaller depth
d
for accumulation, i.e. sharper core-level emission, as
compared to a depletion condition, for the same magnitude of |
U
|.
Conclusion
A single core-level emission-line profile of each respective layer contains all of the
parameters required to describe the complete energy-band relations in this layer, and
consequently of semiconductor/heterojunction and semiconductor/liquid junction
structures at an applied potential or during ad-layer growth. For
operando
AP-PES, this
method provides a powerful tool to describe the complete system at each potential, where
layer-by-layer growth of a contacting material and/or depth-profiling techniques are not
either feasible or possible. In particular, for
operando
AP-PES, in addition to accessing
the electrochemical parameters directly from core-level emissions, e.g.
U
fb
, the
conduction-band offset
U
CB
, valence-band offset
U
VB
, the energy gap (
U
BG
=
U
VB
-
U
CB
),
and the density of interface states, N
IS
, can be determined.
Single core-level energy-band relations have been used recently to describe a highly
stable photoanode, covering the complete potential range beyond the thermodynamic
limits of water splitting. Further use of tender AP-PES and the single core-level method
will enable detailed investigation of the electrolyte double-layer structure in a variety of
electrode systems of interest.
Acknowledgments
This work was supported through the Office of Science of the U.S. Department of
Energy (DOE) under award no. DE SC0004993 to the Joint Center for Artificial
Photosynthesis, a DOE Energy Innovation Hub. The Advanced Light Source is supported
by the Director, Office of Science, Office of Basic Energy Sciences, of the U.S.
Department of Energy under Contract No. DE AC02 05CH11231.We acknowledge Dr.
Philip Ross for contributions to the conceptual development of the AP-PES end station
and experimental design. We acknowledge Fadl Saadi, Beomgyun Jeong, and Sana Rani
for assistance during data collection at the beamline, Ravishankar Sundararaman for
helpful discussions and assistance with theory, and Joseph A. Beardslee for assistance in
XPS data collection.
References
1.
H. J. Lewerenz and H. Goslowsky,
J. Appl. Phys.
,
63
, 2420–2424 (1988).
2.
D. C. Gleason-Rohrer, B. S. Brunschwig, and N. S. Lewis,
J. Phys. Chem. C
,
117
,
18031–18042 (2013).
3.
E. Mankel, C. Hein, M. Kühn, and T. Mayer,
physica status solidi (a)
,
211
, 2040–
2048 (2014).
4.
A. Franciosi and C. G. Van de Walle,
Surf. Sci. Rep.
,
25
, 1–140 (1996).
5.
S. Axnanda et al.,
Meet. Abstr.
MA2013-02
, 921 (2013).
ECS Transactions, 66 (6) 105-113 (2015)
112
) unless CC License in place (see abstract).
ecsdl.org/site/terms_use
address. Redistribution subject to ECS terms of use (see
131.215.70.231
Downloaded on 2016-11-21 to IP
6.
S. Axnanda, E. J. Crumlin, B. Mao, S. Rani, R. Chang, W. C. Stolte, P. G.
Karlsson, M. O. M. Edwards, M. Lundqvist, R. Moberg, P. Ross, Z. Hussain, and
Z. Liu, Submitted (2015).
7.
S. Hu et al.,
Science
,
344
, 1005–1009 (2014).
8.
M. F. Lichterman et al.,
Energy Environ. Sci.
,
7
, 3334–3337 (2014).
9.
H. J. Lewerenz, in
Photoelectrochemical Materials and Energy Conversion
Processes
, R. C. Alkire, D. M. Kolb, J. Lipkowski, and P. N. Ross, Editors, vol.
12, p. 61–181, Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, Germany
(2010).
10.
S. M. Sze and K. K. Ng,
Physics of Semiconductor Devices
, John Wiley & Sons,
Inc., Hoboken, NJ, USA, (2006).
11.
S. Tanuma, C. J. Powell, and D. R. Penn,
Surf. Interface Anal.
,
43
, 689–713
(2011).
12.
W. Smekal, W. S. M. Werner, and C. J. Powell,
Surf. Interface Anal.
,
37
, 1059–
1067 (2005).
ECS Transactions, 66 (6) 105-113 (2015)
113
) unless CC License in place (see abstract).
ecsdl.org/site/terms_use
address. Redistribution subject to ECS terms of use (see
131.215.70.231
Downloaded on 2016-11-21 to IP