Initial stages of ITO/Si interface formation: In situ x-ray photoelectron spectroscopy
measurements upon magnetron sputtering and atomistic modelling using density
functional theory
O. M. Løvvik, S. Diplas, A. Romanyuk, and A. Ulyashin
Citation: Journal of Applied Physics 115, 083705 (2014); doi: 10.1063/1.4866991
View online: http://dx.doi.org/10.1063/1.4866991
View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/115/8?ver=pdfcov
Published by the AIP Publishing
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
129.240.152.50 On: Fri, 28 Feb 2014 15:08:32
JOURNAL OF APPLIED PHYSICS 115, 083705 (2014)
Initial stages of ITO/Si interface formation: In situ x-ray photoelectron
spectroscopy measurements upon magnetron sputtering and atomistic
modelling using density functional theory
O. M. Løvvik,1 S. Diplas,1 A. Romanyuk,2 and A. Ulyashin1
1
SINTEF Materials and Chemistry, Forskningsveien 1, NO-0314 Oslo, Norway
University of Basel, Kingelbergstr. 82, CH-4056 Basel, Switzerland
2
(Received 7 January 2014; accepted 13 February 2014; published online 28 February 2014)
Initial stages of indium tin oxide (ITO) growth on a polished Si substrate upon magnetron sputtering
were studied experimentally using in-situ x-ray photoelectron spectroscopy measurements. The
presence of pure indium and tin, as well as Si bonded to oxygen at the ITO/Si interface were
observed. The experimental observations were compared with several atomistic models of ITO/Si
interfaces. A periodic model of the ITO/Si interface was constructed, giving detailed information
about the local environment at the interface. Molecular dynamics based on density functional theory
was performed, showing how metal-oxygen bonds are broken on behalf of silicon-oxygen bonds.
These theoretical results support and provide an explanation for the present as well as previous
ex-situ and in-situ experimental observations pointing to the creation of metallic In and Sn along
C 2014 AIP Publishing LLC.
with the growth of SiOx at the ITO/Si interface. V
[http://dx.doi.org/10.1063/1.4866991]
I. INTRODUCTION
A. TCO-Si hetero-junctions
It is well established that indium tin oxide (ITO), deposited under appropriate conditions, is a degenerate n-type
semiconductor with low resistivity and high transmission in
the visible range of the solar spectrum. In general, transparent conducting oxides (TCO) such as ITO and ZnO:Al are
widely used for different types of solar cells, both as antireflection coatings and transparent conducting electrodes due
to their attractive combination of electrical and optical properties.1 TCO thin films deposited on any p-type semiconductor material like Si, copper indium gallium selenide, etc.,
form heterojunctions, which can be used for fabricating efficient solar cells. In most of these applications, the quality of
interfaces limits the performance of solar cells and other
electronic devices in which transport of electrical current
through TCO/Si interfaces is required. As was noticed in
Ref. 2, the use of two chemically different semiconductors,
which is the characteristic feature of heterojunction solar
cells, introduces a new set of problems not encountered in
homojunction cells, such as: (i) chemical compatibility and
stability, (ii) reproducibility of the chemical and physical
interface, and, in the case of crystalline and polycrystalline
materials, (iii) lattice compatibility at the metallurgical junction. Thus, interfaces play an important role in photovoltaic
systems based on heterojunctions containing ITO, since a
significant and increasing part of charge carrier losses is due
to defects at the interfaces. It is interesting to note that already long time ago it was reported that efficient ITO/Si solar cells can be made when ITO is deposited on both n- and
p-type Si substrates.3–5
Previous research on TCO (ITO, SnO2)/Si solar cells
indicated that the junction between two semiconductors can
be considered as a metal-insulator-semiconductor (MIS)
0021-8979/2014/115(8)/083705/7/$30.00
(Schottky-barrier) device, since it is expected that a thin
SiOx layer formed at the TCO/Si interface during processing
of such type of solar cell structures can determine their properties.3 Indeed, TCO/SiOx/n-Si solar cell structures in which
thin SiOx layers were purposely formed by heating the Si
wafers at temperatures below 600 C or by immersing them
in a chemical solution have been intensively investigated in
the past.6–8 Remarkable progress for TCO/n-Si solar cells
has been demonstrated by SANYO using an ultra thin (few
nm) pþ a-Si:H buffer layer between the TCO top electrode
and the n-type Si substrate.9,10 It has to be noticed that in the
latter case, formation of a thin oxide layer at the interface
between ITO and a-Si:H occurs. Thus, no single mechanism
can explain so far the performance of several ITO/Si based
solar cell structures. It is most probable that a number of
mechanisms (including heterojunctions and Schottky barriers) are responsible for the efficiency of all types of
TCO/Si based solar cells with different types of buffer
layers. It is necessary to note that the conversion efficiency
of TCO/Si based solar cells depends crucially on the properties of these buffer layers as well as interfaces, which can be
strongly influenced and regulated by chemical pre-treatment
of Si substrates, deposition conditions, as well as prior- and
post-deposition heat treatments of Si substrates and processed solar cell structures.11–19
B. ITO-Si interfaces
The properties of the interfaces depend strongly on the
properties of the TCO layers grown. It is known that ITO
forms upon substitution of In for Sn at the In sub-lattice in
the In2O3 structure up to a tin content of approximately 10%.
This implies the presence of Sn3þ and thus an excess of negative fixed charge. At higher tin content, a network of stoichiometric In4Sn3O12 starts to form; this time with Sn being
115, 083705-1
C 2014 AIP Publishing LLC
V
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
129.240.152.50 On: Fri, 28 Feb 2014 15:08:32
083705-2
Løvvik et al.
present as Sn4þ. The latter compound does not have the
desired properties, and its formation should thus be avoided.
Nevertheless, detailed knowledge about the electronic and
geometric structure of this compound is essential, since its
accidental formation may affect the properties and the subsequent performance of the TCO film.
Rhombohedral In4Sn3O12 can be formed by, e.g.,
electron-beam evaporation,20 magnetron sputtering,21 or isothermal annealing of In2O3 and SnO2 nanopowders.20–22 The
band gap of single-crystalline In2O3 was for a long time a
source of controversy,23,24 but has recently been determined
to be 2.93 6 0.15 and 3.02 6 0.15 eV for the cubic bixbyite
and rhombohedral polymorphs, respectively.25,26 The band
gap of ITO is about 3.5 eV27,28 and depends on the composition and deposition conditions.29
A (222) preferential orientation of ITO was seen in thin
films grown by thermal evaporation in vacuum.30 However,
the texture of ITO films depends on the deposition method
and processing conditions. For example, magnetron sputtered and electron beam deposited ITO films on amorphous
and crystalline Si substrates are usually amorphous in the as
deposited state with the substrate being at room temperature
(RT) during deposition, whilst they crystallize upon heat
treatment and/or upon deposition at higher substrate temperatures. The crystalline films have in this case no preferred
orientation.31,32 It is furthermore known that Sn segregates
to the surface of Sn-doped In2O3,33 and that metallic Sn-In
in ITO can be formed after heat treatment,34 as well as upon
electron beam evaporation.31,35,36 In the case of e-beam deposition, the amount of metallic In present in the amorphous
as-deposited film decreases with increasing the incident
angle of the ITO vapour to the Si substrate, with increasing
the deposition temperature and/or the post deposition annealing temperature, as well as with reducing the deposition
rate.35,36
In a number of studies, the Si-ITO interface has been
studied experimentally; see Refs. 31, 32, 37–42 and references therein. Early studies (e.g., Ref. 40) have attributed the
thermal degradation of ITO/p-Si solar cell structures, which
were fabricated by ion beam sputtering of an ITO target on
both polycrystalline and single crystal Si substrates, to the
formation of a SiO2 interfacial oxide. The formation of this
oxide was connected to inwards oxygen diffusion through
the ITO layer towards the Si substrate. However, no direct
interfacial characterisation methods were employed to support these claims. The presence of a non-stoichiometric Si
oxide at the interface has been reported for ITO deposited on
HF-etched Si substrate by x-ray photoelectron spectroscopy
(XPS) and depth profiling with Arþ sputtering.39 Recent
studies employing a combination of transmission electron
microscopy (TEM) and XPS on thin and ultra-thin ITO films
deposited on Si have shown the formation of a mixed In-Si
oxide along the ITO-Si interface where Si does not participate in its formal stoichiometry (i.e., SiO2).31 Ex-situ XPS
and TEM studies41 on ITO films deposited on p-Si (100) substrates by dc magnetron sputtering revealed the presence of
amorphous In clusters at the interface associated with oxygen deficiency in the ITO film. A recent in-situ XPS study42
(similar to the one presented in this paper) of ITO grown on
J. Appl. Phys. 115, 083705 (2014)
p-Si showed the presence of elemental In and Sn as well as
Si sub-oxides.
C. Ab initio modelling
There have been a few previous first-principles modelling studies of ITO. Sn-doped In2O3 was studied with density
functional theory (DFT) for Sn contents of 12.5% by Odaka
et al.,43 6.25% by Mryasov and Freeman,44 between 3% and
37.5% by Brewer and Franzen45 while Tripathi et al. used
et al.
6.25% Sn in their DFT calculations on ITO.46 Agoston
used hybrid DFT-Hartree Fock potentials to calculate the
solubility of Sn in In2O3,47 which was in good correspondence with the experimental solubility of around 6%.
Inerbaev et al. studied defect clusters in ITO by DFT calculations,48 and a recent study by Bai et al. investigated several
different doping sites of Sn in In2O3 and concluded that the
preferable site is substituting for In.49 Some studies have
also focused at the atomistic properties of Si-oxide interfaces. The Si-SiO2 interface was the first one to be studied by
DFT.50,51 Examples of similar studies include descriptions of
band alignment and defect levels in the Si-HfO2 interface
studied with DFT,52 and band offsets in different Si-oxide
interfaces calculated with hybrid DFT-HF calculations.53 To
our knowledge, less activity has been devoted to the atomistic modelling of the ITO-Si interface.
In this work, in order to investigate the initial stages of
ITO growth on crystalline Si, we performed in-situ XPS
measurements upon magnetron sputtering of such layers on
polished crystalline Si. DFT modelling has been used to support experimentally observed facts concerning the formation
of ITO/Si interfaces.
II. EXPERIMENTAL
ITO ultra thin layers with different thicknesses were deposited onto a polished mono-crystalline p-type Si (100) oriented surface. Prior to deposition, 2% HF dip has been
applied to remove the native oxide. The substrate was rinsed
in deionized water and blown dry with nitrogen. Afterwards,
the substrate was transferred quickly into a magnetron sputtering chamber, which was integrated with a VG ESCALAB
210 spectrometer. The spectrometer was equipped with a
monochromatized Al Ka (1486.6 eV) radiation source. The
energy positions of the spectra were calibrated with reference to the Au 4f7/2 level of a clean gold sample positioned
at 84.0 eV binding energy.
ITO layers with thicknesses of 0.5, 1.5, and 3 nm were
deposited by RF magnetron sputtering at nominally RT. XPS
analysis was performed in-situ after deposition to each of the
above thicknesses. The in-situ measurements allowed us to
study the ITO film and the ITO-Si interfacial growth avoiding the complications present in ex-situ measurements due to
sample contamination or Ar sputtering effects. The thickness
of the deposited layers was controlled by a quartz microbalance system, which was calibrated by ellipsometry and
TEM measurements of the deposited ITO layers. An ITO
sintered target with In2O3 and SnO2 in a weight proportion
of 9:1 was used. The base pressure in the sputter system was
about 1.33 mPa (105 Torr). The total pressure of sputtering
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
129.240.152.50 On: Fri, 28 Feb 2014 15:08:32
083705-3
Løvvik et al.
gas mixture was adjusted to 0.399 Pa (3 mTorr) during the
film preparation. Only pure Ar gas was used upon the sputtering. The RF plasma power was about 50 W.
III. MODELLING: METHODOLOGY AND APPROACHES
DFT calculations were performed using the Vienna
ab initio Simulation Package (VASP).54,55 The generalized
gradient approximation (GGA) was employed, using the
PBE potential.56 It is well-known that this level of theory often significantly under-estimates band gaps, and it is possible
to remedy this by using more expensive hybrid calculations
including Hartree-Fock exchange or by self-energy GW calculations.57 However, the primary output from this work is
based on molecular dynamics (MD), geometric structures,
and relative energies, none of which depend on the band gap.
Band gaps beyond GGA are thus beyond the scope of this
work. The valence electron configuration of the potentials
was 5s25p1 (In), 5s25p2 (Sn), 2s22p4 (O), and 3s23p2 (Si). The
criterion for self-consistency was a difference of less than
105 eV of the calculated total electronic energy Etot between
two consecutive electronic iterations. A k-point distance of
0.20 Å1 was sufficient to achieve numeric convergence of
Etot within 1 meV. A similar convergence was obtained for a
plane-wave cut-off energy of 800 eV; this value was only
used for single point high-accuracy calculations. A cut-off
energy of 400 eV was used for all relaxations and molecular
dynamics calculations. This was found to be sufficient for
numeric convergence of forces within 0.05 eV/Å; the criterion for structural relaxation was similarly that all calculated
forces should be below 0.05 eV/Å. A quasi-Newton formalism was used for the force relaxation, using the residual minimization method direct inversion in iterative subspace
(RMM-DIIS) technique. When the number of electrons was
even, the calculations were spin-restricted, since no magnetic
material was included. Some of the Sn-doped In2O3 (ITO)
models were exceptions; then it was necessary to allow for
spin polarization to accommodate the odd number of
J. Appl. Phys. 115, 083705 (2014)
electrons. In these models, a background charge was also
added, to compensate for the extra electrons added to the system by Sn. DFT-based MD calculations were performed on
an interface model using the following temperature profile:
heating for 2.5 ps at a temperature T ¼ 1000 K followed by
quenching from T ¼ 1000 K to 100 K during 90 ps and then
relaxed. The time step was 2.5 fs.
IV. RESULTS AND DISCUSSION
A. In situ XPS measurements for initial stages of ITO
on Si growth
Fig. 1 displays high resolution Si 2p, In 3d, and Sn 3d
spectra acquired during deposition of ITO on polished p-type
crystalline Si. The absence or limited amount of Si oxide is
seen prior to deposition (lowest Si 2p spectrum on the left).
In the early stages of deposition (0.5 nm), SiOx starts forming
accompanied with formation of metallic In and Sn as seen by
the low binding energy shoulders of the In and Sn 3d peaks.
It should be noted that the energy separation between the Si
2p main peak and the oxide component is 3.0 eV which is
less than that corresponding to SiO2. This should be attributed to the small interfacial oxide thickness58 and/or to an
oxidation state less than 4þ. There is an enhanced presence
of metallic In and Sn in the early stages of deposition
(0.5 nm). The presence of metallic In and Sn indicates that at
the early stages of deposition the deposited oxidized In and
Sn species from the ITO target are reduced upon deposition
on the Si substrate. The Gibbs free energies DfG0 for the formation of In2O3, SnO2, and SiO2 are 827, 516, and
855 kJ/mol, respectively.59 There is thus a thermodynamic
driving force to oxidize Si on behalf of Sn and to a less
extent In. Therefore, from a thermodynamic point of view,
an oxide free Si substrate could take up oxygen and reduce
the above species by forming SiOx when in contact with
ITO. However, kinetic factors influenced by deposition parameters such as the deposition (substrate) temperature and
the angle of incidence (Refs. 35 and 36) are expected to play
FIG. 1. High resolution Si 2p (left), In 3d (middle), and Sn 3d peaks (right) of ITO deposited on polished crystalline Si substrate.
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
129.240.152.50 On: Fri, 28 Feb 2014 15:08:32
083705-4
Løvvik et al.
J. Appl. Phys. 115, 083705 (2014)
a significant role. As an example, surface diffusion would
determine the time available for reduction of In and /or Sn
depositing species and subsequent oxidation of Si substrate
as well as influencing clustering of the atoms of the reduced
species. The energy separation between Si 2p and Si oxide
upon deposition and after a film thickness 1.5 nm remains at
about 3.3 eV which is smaller than those reported for thin
SiO2 films on Si. We attribute this low value to the mixed nature of the interfacial oxide where the contribution of the trivalent In is significant. The presence of a mixed interfacial
oxide or the presence of SiOx (x < 2) has been also reported
in previous studies.31,42
To clarify the mechanism of SiOx and pure metallic In
and Sn species formation at the ITO/Si interface, several atomistic models of ITO as well as ITO/Si interfaces are proposed and presented below.
B. Tin-doped In2O3 (ITO)
We first need to establish models of tin-doped indium
oxide (ITO) in order to create Si/ITO surfaces in the next
section. The focus of this was to achieve an appropriate geometric structure, without regarding electronic band properties beyond what is relevant for the geometry. The atomistic
models were thus made in correspondence with previous
first-principles studies on ITO.43–49 The experimental crystal
structure of ITO is in the cubic Ia 3 space group (No. 206),60
and most of the sites are partially filled. Even the average
number of atoms in each unit cell (UC) differs from whole
numbers, as can be seen in Table I. This means that we have
to choose whole numbers as closely as possible to the average number of atoms when creating supercell models of
ITO. In practice, this means that we choose the number of
atoms according to the rightmost column of Table I. We substitute, e.g., one of the In1 and one of the In2 atoms with Sn
to create a 6.25% Sn filled model. Also, we neglect the minutely filled O2 position in our calculations, which may serve
to compensate the charge carried by tin substitutions under
oxidizing conditions. We thus create three such models, with
1–3 Sn atoms substituting for In. We first substitute on the
Sn2 site, then Sn1, and finally on Sn2. The Sn atoms are
placed by maximizing the Sn-Sn distance, assuming effective repulsive interaction between Sn atoms in the lattice,
according to test calculations not presented here. The atomistic model with 6.25% Sn is shown in Fig. 2.
TABLE I. Atomic positions in ITO with 6.25% Sn.60 The number of atoms
used in the model UC is compared to the average number of atoms in the experimental unit cell, which is obtained from the occupancy of the different
positions.
Type
Wyckoff
position
In1
Sn1
In2
Sn2
O1
O2
8b
8b
24d
24d
48e
16c
Occupancy
Number of
atoms in expt. UC
Number of
atoms in model UC
0.9
0.1
0.95
0.05
1
0.032
7.2
0.8
22.8
1.2
48
0.5
7
1
23
1
48
0
FIG. 2. The ITO model with 6.25% Sn, that is with two Sn atoms per unit
cell. In and Sn atoms are shown as large grey and yellow balls, while O
atoms are shown as small red balls.
Relaxation of the ITO models is relatively straightforward, since cubic symmetry is maintained even when all
lattice parameters are allowed to relax without any restrictions (the maximal deviation from cubic symmetry is 0.03%
of the lattice constants and 0.08˚ of the angles). The structures are thus fully relaxed; the results of those relaxations
are shown in Table II.
We see from Table II that both the total electronic
energy and the average lattice constant vary almost linearly
with the Sn content. This is expected due to the difference in
binding energy between Sn-O and In-O and due to the larger
size of the Sn atom.
C. Interface between ITO and Si
We then create an atomistic model of the ITO/Si interface using the ITO models discussed above as the starting
point for the oxide. There is always a challenge to find
matching lattice constants when two crystalline surfaces are
to be joined at an interface. A very good fit was found
between the hexagonal (111) surface of Si and a sublattice of
the hexagonal (111) surface of cubic ITO. In order to fit the
number of atoms, a part of the (111) surface unit cell was
chosen, containing seven In atoms and one Sn atom. Three
such layers were added, with one Sn atom in each layer. Six
Si layers were attached to this, giving a unit cell with 108
atoms and orthorhombic symmetry. No vacuum was added
between any layers, so this actually represents a repeating
structure with two interfaces defined by 8 and 9 Å thick ITO
and Si slabs, respectively, as shown in Fig. 3. The Sn content
TABLE II. The calculated total electronic energy of ITO with varying Sn
content is given in eV relative to that of pure In2O3 (0% Sn). The relaxed lattice constant a in Å is also shown.
Sn content (%)
0
3.1
6.3
9.4
Total energy (eV)
Average a (Å)
0.00
1.59
2.69
3.47
10.341
10.347
10.355
10.360
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
129.240.152.50 On: Fri, 28 Feb 2014 15:08:32
083705-5
Løvvik et al.
FIG. 3. The atomistic interface model between Si and ITO before (left) and
after (right) a molecular dynamics simulation. The temperature profile of the
simulation is described in Sec. III. The vertical axis corresponds to the (111)
axis of both Si and ITO. The Si, In, Sn, and O atoms are represented as light
blue, grey, orange, and red balls, respectively.
is 12.5%, which is slightly higher than the models discussed
in the previous section. The O content is similar to that of
undoped In2O3.
This model is then used as input to molecular dynamics
calculations. Molecular dynamic calculations can be used to
probe and predict interfacial changes occurring upon, e.g.,
thermal treatments such as deposition at elevated temperatures, post deposition air annealing, etc., of ITO-Si structures. Figure 3 shows the model before and after a molecular
dynamics calculation consisting of heat treatment followed
by quenching and relaxation (see Sec. III for details). The
calculations were performed spin restricted without extra
background charge, since it was found that this did not influence the calculated forces significantly. It is seen that significant changes in the geometric structure occurred during the
simulations, particularly at the lower interface in Fig. 3.
Several Si-O bonds were created, and some oxygen atoms
even moved into the Si slab; the number of Si-O bonds (with
distance <1.8 Å) is increased from 6 to 18 during the MD
simulation. This was accompanied by a reduction of the
nearby metal atoms, of which several oxygen bonds were
broken. The number of In-O bonds (with distance <2.8 Å)
was reduced from 5.2 per atom to 4.2 per atom; this means
that 21 bonds between In and O were broken, primarily at
the interface. The number of Sn-O bonds (with distance
<2.8 Å) was already 5.0 per atom and was not reduced any
further during MD simulation. (The number of metal-oxygen
bonds is 6 per atom in bulk ITO.) While there is not yet metallic In or Sn in the model after 90 ps, there is a clear tendency towards reduction of the metal oxides, particularly the
In oxide. This is consistent with the experimental findings,
where a similar tendency is found—in that case reduction
of both oxidised Sn and In is seen, only more pronounced
for In.
This result is apparently not consistent with the relative
stability of the oxide phases cited above (tin oxide being the
least stable and SiO2 only slightly more stable than In2O3).
One could speculate that this is due to kinetics: if the kinetics
of Sn diffusion is slower than that of In, this could explain
J. Appl. Phys. 115, 083705 (2014)
the difference. Since the atomic radius of metallic Sn is
smaller than that of metallic In, it cannot be explained by diffusion of metal atoms forming clusters. The remaining
option is that Sn-O clusters diffuse faster than In-O clusters
towards the Si surface when arriving onto the Si substrate, so
that the reduction rate of Sn oxide is slower than that of In
oxide.
Another possible explanation is associated with the thermodynamics of the ternary and quaternary systems which
may be drastically different from that of the binary oxides;
this may lead to the simplified arguments based on the stability of binary oxides not being valid anymore. Finally, there
may be hydrogen atoms present in the experiments, which
can change the thermodynamics or kinetics of this system
significantly. The Si substrate has been subjected to HF treatment and H has passivated dangling bonds. One possible
effect is that H preferentially reduces oxidised In and to a
lesser excess Sn. However, even if H plays such a role this is
not applicable in the simulations as H was not taken into
account. All the explanations above remain speculative,
since there is no experimental or theoretical support for
anyone.
We should furthermore keep in mind that the DFT-MD
models in this study do not exactly simulate the growth of
the interface upon impingement of oxidised species on Si
substrates. It rather simulates the changes of an initially
formed ITO-Si interface upon temperature variations. The
presence of metallic In and Sn is associated with nanoclusters (as has been shown, e.g., in Ref. 41) which form
most probably mainly by surface diffusion. The DFT-MD
model in the present study refers to bulk rather than surface
diffusion. This may explain why we do not see formation of
metallic bonds (clustering) within the duration of the simulation as bulk diffusion is slower than surface. Moreover, the
MD calculations were only performed for a few picoseconds,
so many effects on a larger timescale may have been left out
by the calculations.
In any case, it is interesting that the experiments and calculations consistently demonstrate that the stability of binary
oxides is not sufficient information to predict the qualitative
behaviour of complicated systems like the present ones. In
addition, the experiments and calculations in the present
study clearly show that there is a tendency to form SiOx at
the interface, which is also consistent with previously
reported work.31,35,36,42,61
V. CONCLUDING REMARKS
Initial stages of ITO growth on a polished Si substrate
upon magnetron sputtering were studied experimentally
using in-situ XPS measurements. The presence of pure indium and tin as well as Si bonded to oxygen related peaks at
the ITO/Si interface was detected confirming previous results
from ex-situ and a recent in-situ study. Several atomistic
models of ITO as well as an ITO/Si interface were constructed in order to interpret the experimental findings.
Atomistic models of ITO were created by substituting
Sn at the In sublattice, using the cubic In2O3 structure as
starting point. It was found that the predicted lattice constant
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
129.240.152.50 On: Fri, 28 Feb 2014 15:08:32
083705-6
Løvvik et al.
changed almost linearly with the Sn content, and that cubic
symmetry was preserved. These models were used to construct an atomistic model of the Si/ITO interface. The Sn
content in this model was 12.5% to ensure a homogeneous
distribution of Sn atoms, and the model contained 108 atoms.
Molecular dynamics calculations on this structure revealed
that oxygen from ITO was attracted to the Si phase, initiating
formation of SiOx. Similarly, the first stage of metallic In
and Sn formation was indicated via several metal-O bonds
being broken in the MD models.
It is important to note that processes at the ITO/c-Si
interface discussed and simulated above should not depend
on the type of conductivity of the Si substrate. Formation of
a thin SiOx layer at the ITO/c-Si interface produces a MIS
nanostructure. It should be noted that formation of a thin
SiOx layer, as well as pure In at the ITO/Si interface upon
in-situ ITO deposition has been reported very recently in
Ref. 42. Formation of a MIS structure provides an explanation for the long time ago experimentally established fact
that degenerate n-type semiconductor ITO being deposited
on n-type Si can be used for fabrication of high-efficiency
ITO/n-Si heterojunction solar cells. These phenomena can be
extremely useful for current activities, related to processing
of Si nanowire based solar cells, in which an ultrathin emitter
is an important issue.57,62,63 In particular, processing technology of such cells can be essentially simplified, if instead
of a-Si:H based ultrathin emitter, which has been used in
Refs. 57, 62, and 63, only an ITO layer will be deposited at
special conditions, as discussed in Ref. 4.
ACKNOWLEDGMENTS
This work was funded, in part, by the European commission 7th framework program under Grant Agreement No.
246331 (NanoPV). Computation time through the NOTUR
supercomputer consortium is gratefully acknowledged.
1
C. G. Granqvist, Sol. Energy Mater. Sol. Cells 91(17), 1529–1598 (2007).
S. J. Fonash, Solar Cell Device Physics (Academic Press, 1981).
3
A. K. Ghosh, C. Fishman, and T. Feng, J. Appl. Phys. 49(6), 3490–3498
(1978).
4
T. Feng, A. K. Ghosh, and C. Fishman, J. Appl. Phys. 50(7), 4972–4974
(1979).
5
J. B. Dubow, D. E. Burk, and J. R. Sites, Appl. Phys. Lett. 29(8), 494–496
(1976).
6
H. Kobayashi, T. Ishida, Y. Nakato, and H. Mori, J. Appl. Phys. 78(6),
3931–3939 (1995).
7
Y. Nakato, K. Kai, and K. Kawabe, Sol. Energy Mater. Sol. Cells 37(3–4),
323–335 (1995).
8
H. Kobayashi, Y. Kogetsu, T. Ishida, and Y. Nakato, J. Appl. Phys. 74(7),
4756–4761 (1993).
9
M. Tanaka, M. Taguchi, T. Matsuyama, T. Sawada, S. Tsuda, S. Nakano,
H. Hanafusa, and Y. Kuwano, Jpn. J. Appl. Phys., Part 1 31(11),
3518–3522 (1992).
10
M. Taguchi, K. Kawamoto, S. Tsuge, T. Baba, H. Sakata, M. Morizane, K.
Uchihashi, N. Nakamura, S. Kiyama, and O. Oota, Prog. Photovoltaics
8(5), 503–513 (2000).
11
A. G. Ulyashin, R. Job, M. Scherff, M. Z. Gao, W. R. Fahrner, D.
Lyebyedyev, N. Roos, and H. C. Scheer, Thin Solid Films 403, 359–362
(2002).
12
T. H. Breivik, S. Diplas, A. G. Ulyashin, A. E. Gunnaes, B. R. Olaisen, D.
N. Wright, A. Holt, and A. Olsen, Thin Solid Films 515(24), 8479–8483
(2007).
2
J. Appl. Phys. 115, 083705 (2014)
13
A. Ulyashin, M. Scherff, R. Hussein, M. Z. Gao, R. Job, and W. R.
Fahrner, Sol. Energy Mater. Sol. Cells 74(1–4), 195–201 (2002).
14
V. D. Akhmetov, A. G. Ulyashin, A. Holt, and M. Kittler, Mater. Sci.
Eng., B 159–160, 182–185 (2009).
15
A. Ulyashin, R. Bilyalov, A. Bruck, M. Scherff, R. Job, W. Fahrner, and J.
Poortmans, in Proceedings of 3rd World Conference on Photovoltaic
Energy Conversion, Tokyo (2003), Vols. A-C, pp. 1088–1091.
16
R. Bilyalov, A. Ulyashin, M. Scherff, K. Meusinger, J. Poortmans, and W.
Fahrner, in Proceedings of 3rd World Conference on Photovoltaic Energy
Conversion (2003), Vols. A-C, pp. 1139–1142.
17
A. M. Froitzheim, M. L. D. Scherff, A. Ulyashin, O. Milch, M. Schmidt,
W. R. Fahrner, and W. Fuhs, in Proceedings of 3rd World Conference on
Photovoltaic Energy Conversion (2003), Vols. A-C, pp. 180–183.
18
L. Carnel, G. Agostinelli, A. Ulyashin, C. Solanki, H. Kim, S. De Wolf, G.
Beaucarne, and J. Poortmans, in Conference Record of the Thirty-First
IEEE Photovoltaic Specialists Conference (2005), pp. 1157–1160.
19
A. G. Ulyashin, A. Bentzen, S. Diplas, A. Suphellen, A. E. Gunnaes, A.
Olsen, B. G. Svensson, E. S. Marstein, A. Holt, D. Grambole, and E.
Sauar, in Conference Record of the 2006 IEEE 4th World Conference on
Photovoltaic Energy Conversion (2006), Vols. 1 and 2, pp. 1354–1357.
20
N. Mori, S. Ooki, N. Masubuchi, A. Tanaka, M. Kogoma, and T. Ito, Thin
Solid Films 411(1), 6–11 (2002).
21
Y. Gassenbauer and A. Klein, Solid State Ionics 173(1–4), 141–145
(2004).
22
G. B. Gonzalez, J. S. Okasinski, T. O. Mason, T. Buslaps, and V.
Honkimaki, J. Appl. Phys. 104(4), 043520 (2008).
23
P. Erhart, A. Klein, R. G. Egdell, and K. Albe, Phys. Rev. B 75(15),
153205 (2007).
24
F. Fuchs and F. Bechstedt, Phys. Rev. B 77(15), 155107 (2008).
25
P. D. C. King, T. D. Veal, F. Fuchs, C. Y. Wang, D. J. Payne, A.
Bourlange, H. Zhang, G. R. Bell, V. Cimalla, O. Ambacher, R. G. Egdell,
F. Bechstedt, and C. F. McConville, Phys. Rev. B 79(20), 205211 (2009).
26
A. Walsh, J. L. F. Da Silva, S. H. Wei, C. Korber, A. Klein, L. F. J. Piper,
A. DeMasi, K. E. Smith, G. Panaccione, P. Torelli, D. J. Payne, A.
Bourlange, and R. G. Egdell, Phys. Rev. Lett. 100(16), 167402 (2008).
27
I. Hamberg, A. Hjortsberg, and C. G. Granqvist, Appl. Phys. Lett. 40(5),
362–364 (1982).
28
T. Minami, Y. Takeda, S. Takata, and T. Kakumu, Thin Solid Films 308,
13–18 (1997).
29
A. V. Mudryi, A. V. Ivaniukovich, and A. G. Ulyashin, Thin Solid Films
515(16), 6489–6492 (2007).
30
M. Purica, F. Iacomi, C. Baban, P. Prepelita, N. Apetroaei, D. Mardare,
and D. Luca, Thin Solid Films 515(24), 8674–8678 (2007).
31
S. Diplas, O. M. Lovvik, H. Nordmark, D. M. Kepaptsoglou, J. M. Graff,
C. Ladam, F. Tyholdt, J. C. Walmsley, A. E. Gunnaes, R. Fagerberg, and
A. Ulyashin, Surf. Interface Anal. 42(6–7), 874–877 (2010).
32
S. Diplas, A. Ulyashin, K. Maknys, A. E. Gunnaes, S. Jorgensen, D.
Wright, J. F. Watts, A. Olsen, and T. G. Finstad, Thin Solid Films
515(24), 8539–8543 (2007).
33
Y. Gassenbauer, R. Schafranek, A. Klein, S. Zafeiratos, M. Havecker, A.
Knop-Gericke, and R. Schlogl, Phys. Rev. B 73(24), 245312 (2006).
34
G. Guenther, G. Schierning, R. Theissmann, R. Kruk, R. Schmechel, C.
Baehtz, and A. Prodi-Schwab, J. Appl. Phys. 104(3), 034501 (2008).
35
T. Ishida, H. Kobayashi, and Y. Nakato, J. Appl. Phys. 73(9), 4344–4350
(1993).
36
H. Kobayashi, T. Ishida, K. Nakamura, Y. Nakato, and H. Tsubomura,
J. Appl. Phys. 72(11), 5288–5293 (1992).
37
K. Maknys, A. G. Ulyashin, H. Stiebig, A. Y. Kuznetsov, and B. G.
Svensson, Thin Solid Films 511, 98–102 (2006).
38
J. S. Christensen, A. G. Ulyashin, K. Maknys, A. Y. Kuznetsov, and B. G.
Svensson, Thin Solid Films 511, 93–97 (2006).
39
H. Kobayashi, T. Ishida, Y. Nakato, and H. Tsubomura, J. Appl. Phys.
69(3), 1736–1743 (1991).
40
S. M. Goodnick, J. F. Wager, and C. W. Wilmsen, J. Appl. Phys. 51(1),
527–531 (1980).
41
A. Thogersen, M. Rein, E. Monakhov, J. Mayandi, and S. Diplas, J. Appl.
Phys. 109(11), 113532 (2011).
42
M. H. Rein, M. V. Hohmann, A. Thogersen, J. Mayandi, A. O. Holt, A.
Klein, and E. V. Monakhov, Appl. Phys. Lett. 102(2), 021606 (2013).
43
H. Odaka, Y. Shigesato, T. Murakami, and S. Iwata, Jpn. J. Appl. Phys.,
Part 1 40(5A), 3231–3235 (2001).
44
O. N. Mryasov and A. J. Freeman, Phys. Rev. B 64(23), 233111 (2001).
45
S. H. Brewer and S. Franzen, Chem. Phys. 300(1–3), 285–293
(2004).
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
129.240.152.50 On: Fri, 28 Feb 2014 15:08:32
083705-7
46
Løvvik et al.
M. N. Tripathi, K. Shida, R. Sahara, H. Mizuseki, and Y. Kawazoe,
J. Appl. Phys. 111(10), 103110 (2012).
47
P. Agoston, C. Korber, A. Klein, M. J. Puska, R. M. Nieminen, and K.
Albe, J. Appl. Phys. 108(5), 053511 (2010).
48
T. M. Inerbaev, R. Sahara, H. Mizuseki, Y. Kawazoe, and T. Nakamura,
Solid State Commun. 150(1–2), 18–21 (2010).
49
L. N. Bai, Y. P. Wei, J. S. Lian, and Q. Jiang, J. Phys. Chem. Solids 74(3),
446–451 (2013).
50
A. Pasquarello, M. S. Hybertsen, and R. Car, Nature 396(6706), 58–60
(1998).
51
A. Bongiorno and A. Pasquarello, Appl. Phys. Lett. 83(7), 1417–1419 (2003).
52
P. Broqvist, A. Alkauskas, and A. Pasquarello, Appl. Phys. Lett. 92(13),
132911 (2008).
53
A. Alkauskas, P. Broqvist, F. Devynck, and A. Pasquarello, Phys. Rev.
Lett. 101(10), 106802 (2008).
54
G. Kresse and J. Hafner, Phys. Rev. B 47, 558–561 (1993).
55
G. Kresse and J. Furthm€
uller, Phys. Rev. B 54, 11169–11186 (1996).
J. Appl. Phys. 115, 083705 (2014)
56
J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77(18),
3865–3868 (1996).
57
M. Steglich, A. Bingel, G. Jia, and F. Falk, Sol. Energy Mater. Sol. Cells
103, 62–68 (2012).
58
S. Iwata and A. Ishizaka, J. Appl. Phys. 79(9), 6653–6713 (1996).
59
Collaboration: Scientific Group Thermodata Europe (SGTE), The LandoltB€
ornstein Database (Springer-Verlag, Berlin, 1999).
60
N. Nadaud, N. Lequeux, M. Nanot, J. Jove, and T. Roisnel, J. Solid State
Chem. 135(1), 140–148 (1998).
61
D. M. Kepaptsoglou, S. Diplas, Q. M. Ramasse, A. E. Gunnæs, and A.
Ulyashin, “A high resolution electron microscopy and spectroscopy study
of interfaces in Si-based solar cell heterojunctions,” J. Appl. Phys.
(unpublished).
62
G. B. Jia, M. Steglich, I. Sill, and F. Falk, Sol. Energy Mater. Sol. Cells
96(1), 226–230 (2012).
63
G. Jia, B. Eisenhawer, J. Dellith, F. Falk, A. Thøgersen, and A. Ulyashin,
J. Phys. Chem. C 117, 1091–1096 (2013).
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
129.240.152.50 On: Fri, 28 Feb 2014 15:08:32