consequences of plasma oxidation and vacuum annealing on the …€¦ · reaction processes and...
TRANSCRIPT
-
Consequences of plasma oxidation and vacuum annealing on the chemical propertiesand electron accumulation of In2O3 surfaces
Theresa Berthold, Julius Rombach, Thomas Stauden, Vladimir Polyakov, Volker Cimalla, Stefan Krischok, OliverBierwagen, and Marcel Himmerlich
Citation: J. Appl. Phys. 120, 245301 (2016); doi: 10.1063/1.4972474View online: http://dx.doi.org/10.1063/1.4972474View Table of Contents: http://aip.scitation.org/toc/jap/120/24Published by the American Institute of Physics
http://aip.scitation.org/author/Berthold%2C+Theresahttp://aip.scitation.org/author/Rombach%2C+Juliushttp://aip.scitation.org/author/Stauden%2C+Thomashttp://aip.scitation.org/author/Polyakov%2C+Vladimirhttp://aip.scitation.org/author/Cimalla%2C+Volkerhttp://aip.scitation.org/author/Krischok%2C+Stefanhttp://aip.scitation.org/author/Bierwagen%2C+Oliverhttp://aip.scitation.org/author/Bierwagen%2C+Oliverhttp://aip.scitation.org/author/Himmerlich%2C+Marcel/loi/japhttp://dx.doi.org/10.1063/1.4972474http://aip.scitation.org/toc/jap/120/24http://aip.scitation.org/publisher/
-
Consequences of plasma oxidation and vacuum annealing on the chemicalproperties and electron accumulation of In2O3 surfaces
Theresa Berthold,1 Julius Rombach,2 Thomas Stauden,1 Vladimir Polyakov,3
Volker Cimalla,3 Stefan Krischok,1 Oliver Bierwagen,2 and Marcel Himmerlich1,a)1Institut f€ur Mikro- und Nanotechnologien MacroNano, Technische Universit€at Ilmenau, PF 100565,98684 Ilmenau, Germany2Paul-Drude-Institut f€ur Festk€orperelektronik, Hausvogteiplatz 5–7, 10117 Berlin, Germany3Fraunhofer-Institut f€ur Angewandte Festk€orperphysik, Tullastraße 72, 79108 Freiburg, Germany
(Received 23 August 2016; accepted 5 December 2016; published online 22 December 2016)
The influence of oxygen plasma treatments on the surface chemistry and electronic properties of
unintentionally doped and Mg-doped In2O3(111) films grown by plasma-assisted molecular beam
epitaxy or metal-organic chemical vapor deposition is studied by photoelectron spectroscopy. We
evaluate the impact of semiconductor processing technology relevant treatments by an inductively
coupled oxygen plasma on the electronic surface properties. In order to determine the underlying
reaction processes and chemical changes during film surface–oxygen plasma interaction and to
identify reasons for the induced electron depletion, in situ characterization was performed imple-menting a dielectric barrier discharge oxygen plasma as well as vacuum annealing. The strong
depletion of the initial surface electron accumulation layer is identified to be caused by adsorption
of reactive oxygen species, which induce an electron transfer from the semiconductor to localized
adsorbate states. The chemical modification is found to be restricted to the topmost surface and
adsorbate layers. The change in band bending mainly depends on the amount of attached oxygen
adatoms and the film bulk electron concentration as confirmed by calculations of the influence of
surface state density on the electron concentration and band edge profile using coupled
Schr€odinger-Poisson calculations. During plasma oxidation, hydrocarbon surface impurities areeffectively removed and surface defect states, attributed to oxygen vacancies, vanish. The recurring
surface electron accumulation after subsequent vacuum annealing can be consequently explained
by surface oxygen vacancies. Published by AIP Publishing. [http://dx.doi.org/10.1063/1.4972474]
I. INTRODUCTION
Indium oxide (In2O3) is traditionally known as active gas
sensor material or, if highly n-doped with Sn, as the transpar-
ent conducting oxide indium tin oxide (ITO) which has been
used as contact layer in displays and solar cells for decades.
Recent improvements in thin film deposition technology and
epitaxy with the possibility to produce high crystalline quality
In2O3 material with low impurity and defect densities stimu-
lated application of this large band gap material in transparent
semiconductor devices—see Ref. 1 and references therein.
The surface electron concentration and related processes for
its modification directly influence the characteristics of gas-
sensitive films based on In2O3 (Ref. 2) as well as the metal
contacts for electronic devices, either being of ohmic or recti-
fying nature.3 Based on improved material quality, the band
structure parameters and optical properties have been rede-
fined in recent years4–7 with consequences on the interpreta-
tion of the band edge profile at the surface, which was initially
interpreted in terms of an upward band bending.8 For In2O3films either grown by plasma-assisted molecular beam epitaxy
(PAMBE) or metal-organic chemical vapor deposition
(MOCVD), an enhanced electron concentration is found at
the surface compared to the bulk values,9–12 referred to as sur-
face electron accumulation layer (SEAL). In accordance,
the valence and conduction band (VB and CB) both bend
downward in the accumulation region at the surface and elec-
trons excited from the occupied CB states below the Fermi
level can be directly measured using photoelectron spectros-
copy (PES). The origin of this effect has been attributed to the
unique band structure and the related high charge neutrality
level in In2O3 (Ref. 9) which leads to donor like behavior of
many surface defects and impurities. For example surface oxy-
gen vacancies are predicted by density functional theory (DFT)
to be one major source for the SEAL13,14 and indications for
this aspect are also observed experimentally by PES.15
The bulk electron concentration nbulk in In2O3 is gener-ally adjustable by intentional doping, e.g., by Sn for high
nbulk values,16–18 while the incorporation of Mg acceptors is
not successful to generate p-type behavior,19,20 but signifi-
cantly lowers the film conductivity. Furthermore, implement-
ing a combination of Mg doping and subsequent annealing
in oxygen environment enables semi-insulating film charac-
teristics.20 The typically observed non-negligible electron
concentration in unintentionally doped (UID) epitaxial In2O3layers is caused by incorporated impurities and defects. The
related aspects and the extensive number of available studies
are reviewed in Refs. 1 and 21. Briefly, both incorporated
hydrogen atoms as well as oxygen vacancies are sources for
the unintentional n-type character of In2O3.
Like in the crystals bulk, defects (oxygen vacancies)
and impurities (adsorbates) are also a source for changes ina)[email protected]
0021-8979/2016/120(24)/245301/10/$30.00 Published by AIP Publishing.120, 245301-1
JOURNAL OF APPLIED PHYSICS 120, 245301 (2016)
http://dx.doi.org/10.1063/1.4972474http://dx.doi.org/10.1063/1.4972474http://dx.doi.org/10.1063/1.4972474http://dx.doi.org/10.1063/1.4972474http://dx.doi.org/10.1063/1.4972474mailto:[email protected]://crossmark.crossref.org/dialog/?doi=10.1063/1.4972474&domain=pdf&date_stamp=2016-12-22
-
electron concentration at the surface. Both, oxidizing and
reducing surface treatments can therefore affect the nature of
the band alignment (downward/upward bending) and elec-
tron concentration Ns (accumulation/depletion) at the sur-face. Consequently, adsorption of impurities or thermally/
light induced desorption processes are expected to influence
these quantities. Within this context, the influence of O2 and
CO interaction at different temperatures was investigated for
(111) and (001) oriented In2O3 films indicating a slight
reduction of the SEAL electron density upon their adsorp-
tion.22 Furthermore, a reversible transition between a
reduced In-adatom structure formed during annealing at
500 �C in ultra-high vacuum (UHV) and an oxidized surfaceconfiguration generated by annealing at 500 �C in 10�7 mbarO2 was found in scanning tunneling microscopy experi-
ments.23 The existence of stable reduced and oxidized (111)
surfaces was also predicted by density functional theory
(DFT) including calculations of the corresponding surface
structures (in this case, an oxygen vacancy structure in the
reduced form).24 In agreement, Tambasov et al. found ametal-semiconductor transition at low temperatures (about
100 K) after UV illumination, which was absent after expo-
sure to an oxygen environment.25 They suggest that the
observed effect is due to a change of the degree of disorder
(mainly oxygen vacancies) through a photoreduction effect
after UV illumination. The surface can be furthermore chemi-
cally reduced by UV light or annealing in vacuum as observed
for nanocrystalline,26 polycrystalline, and textured12 In2O3films, where oxidizing/reducing surface treatments severely
influence the electron concentration of the material27 leading
to, e.g., highly sensitive ozone detectors.28,29 Furthermore, for
In2O3 single crystals with low electron concentration, the
absence of the SEAL was reported for the (111) surface after
in situ vacuum cleavage.30 This aspect indicates the impor-tance of the used surface preparation method and the influence
of surface defect states or adsorbates on the formation of the
surface electron accumulation layer.
Most significantly, for crystalline thin films, a drastic
reduction of Ns was realized for UID indium oxide films11 if
the films were modified by an oxygen plasma treatment,
severely depleting the surface from electrons. This effect was
directly monitored by a strong shift of the VB edge towards
the Fermi level EF at the surface and the depopulation of theCB electrons. Consequently, since the SEAL only allows the
production of ohmic metal contacts, a reactive (oxygen
enriched) plasma deposition process is required to produce rec-
tifying Schottky contacts on In2O3.3 This trend is a conse-
quence of the plasma-induced surface electron depletion,
which is a drawback if these layers are designated for sensing
applications of oxidizing gases,2 but might be beneficial for
the detection of reducing species. However, the origin of this
drastic reduction of Ns and the underlying compositional sur-face changes during oxygen plasma treatment are not fully
clarified so far. Possible reasons for this effect were sug-
gested11 such as the generation of crystal defects or the incor-
poration of interstitial oxygen atoms,31 which both could be
reversed by thermal treatments inducing SEAL re-occurrence.
Within this context, it has to be mentioned that for ITO
films, the influence of oxygen plasma processing on the
surface electronic properties has been reported much earlier,
demonstrating an increase of the work function by
0.1–0.5 eV,32–34 which is beneficial for display device tech-
nology, as well as changes in the surface plasmon energy,
band bending, and Fermi energy.32
In this study, In2O3 films were characterized before and
after two different oxygen plasma treatments using PES. The
influence of Mg-doping, plasma processing conditions, and
vacuum annealing on the chemical and electronic surface
properties are analyzed in order to determine the interrela-
tions between the presence of surface adsorbates and defects
and the value of band alignment and surface dipole energy
and to determine the reasons for the distinct surface electron
depletion after oxygen plasma processing of In2O3 films.
II. EXPERIMENTAL
About 350 nm thick UID as well as Mg-doped (Mg con-
centration up to 1021 cm�3) single crystalline In2O3(111) films
were grown on (111)-oriented Y-stabilized zirconia (YSZ)
substrates by PAMBE at �800 �C.2,35 The Mg-doped sampleswere not annealed in oxygen atmosphere thus exhibiting only
a slight reduction of the electron concentration due to the
effect of the compensating oxygen vacancy formation.
Oxygen-annealed Mg-doped samples typically approach a
semi-insulating character,20 which results in charging prob-
lems in PES measurements. In addition 330 nm thick textured
UID In2O3(111) films were deposited onto Al2O3(0001) sub-
strates by PAMBE (at �800 �C)2 or MOCVD. MOCVDgrowth was performed in a close-coupled showerhead system
(AIXTRON), using trimethylindium (In(CH3)3) and N2O as
precursors and N2 as carrier gas. For film growth, a �22 nmthick nucleation layer was deposited at 500 �C, followed bythe growth of a high-quality In2O3 epilayer of 300 nm thick-
ness at 1000 �C.12
After growth, the samples were exposed to ambient con-
ditions and cut into 6� 8 mm2 pieces. Room temperatureHall effect measurements in van-der-Pauw geometry as
described in Ref. 36 were performed to characterize the sheet
resistance Rsh and sheet electron concentration Nsh withinthe In2O3 epilayers after growth and after subsequent induc-
tively coupled plasma (ICP) oxygen plasma treatment.
Dividing Nsh by the film thickness yielded the apparent bulkelectron concentration nbulk.
Oxygen plasma modification was performed ex situusing a 13.56 MHz inductively coupled plasma (ICP) reac-
tive ion-etching (RIE) system (Samco – RIE-400iP, process
pressure 0.025 mbar, oxygen flow 10 sccm) operated at a
ICP power of 100 W and a RIE power of 50 W applied to the
sample for 5 min. The as-grown or plasma-treated samples
were mounted onto grounded Ta sample holders and inserted
into the ultra-high vacuum (UHV) surface preparation and
analysis system. A dielectric barrier discharge (DBD) was
generated by a high voltage (11 kV, 10 kHz), an oxygen pres-
sure of 200 mbar and quartz as dielectric barrier using a vac-
uum device on the basis of the design presented in Ref. 37.
The plasma was ignited for 5 min between the grounded
sample and quartz barrier having a distance of approximately
1 mm. The DBD plasma generator is part of a UHV surface
245301-2 Berthold et al. J. Appl. Phys. 120, 245301 (2016)
-
analysis system together with a sample heating stage and the
PES unit. Hence, in situ oxygen plasma preparation and anal-ysis could be realized to exclude the influence of gas adsorp-
tion under ambient conditions.
Thermal annealing of the samples in vacuum was per-
formed by radiative heating from the sample’s backside. The
substrates used for MBE film growth had a �1 lm thick Tibackside coating to provide good absorption of the heat radi-
ation. During vacuum annealing, a temperature of 500 �C, asdetermined by pyrometry from the sample front side, was
sufficient to efficiently remove the surface adsorbates as con-
firmed by the PES analyses. Since the substrates used for
MOCVD growth lacked a backside metallization, radiative
heating in the UHV system was less efficient, and pyrometry
was impossible due to the partial transparency of the sample
to infrared radiation. In order to compensate the effect of
lower heating efficiency, the heater filament current was
increased accordingly. In a power-dependent annealing
series, we determined that an increase of the heater power by
40% resulted in comparable effects of desired adsorbate
removal and surface cleaning and thus a comparable sample
temperature.
Photoelectron spectroscopy measurements have been
performed in normal emission using monochromated AlKa(h�¼ 1486.7 eV) as well as HeI (h�¼ 21.2 eV) radiation anda hemispherical electron analyzer operated with an electron
acceptance angle of 8� for standard measurements and 8� aswell as 1� for angle-dependent experiments. More detailsabout the setup and the used experimental conditions can be
found in Ref. 38. The binding energy (BE) scale and the
position of the Fermi level are regularly calibrated using a
silver reference sample.
III. RESULTS AND DISCUSSION
A. Initial surface characteristics after growth
Textured UID In2O3 films grown by MOCVD or
PAMBE on Al2O3(0001) substrates were analyzed in com-
parison to single crystalline, UID, and Mg-doped In2O3grown by PAMBE on YSZ(111) substrates in order to iden-
tify the influence of Mg concentration and crystallinity on
the electronic surface properties. Table I lists the general
characteristics of the investigated samples. After growth, all
samples were transported in ambient conditions before they
were introduced into the surface analysis system and subse-
quently characterized by PES. In this stage of surface
preparation (as loaded after growth), the In3d5=2 and O1s
core levels of the In2O3 structure are located at a binding
energy of 444.6 and 530.2 eV (Fig. 1(a)), respectively, and
an additional O1s component at higher BE (532.2–532.7 eV)
can be observed for all samples (see Fig. 2 for sample III).
This additional feature is caused by the existence of –OH
and –CO bonds,39 originating from the adsorption of surface
impurities during sample transport through air.
Consequently, also a C1s signal is detected including states
from the adsorbed hydrocarbon and carbon oxide molecules.
In Fig. 1 the influence of growth method, substrate as
well as Mg-doping of In2O3(111) films on the energy posi-
tion of core levels, the valence band maximum (VBM) and
on the work function U is shown together with the measuredVB spectra excited by X-ray radiation. A higher Mg
TABLE I. Growth method, Mg doping concentration NMg, substrate and the
crystal structure of the investigated In2O3(111) films. Textured films were
grown on Al2O3(0001) and single crystalline layers on YSZ(111)
substrates.2,12
No NMg (cm�3) Substrate/crystallinity Growth method
I … Al2O3(0001)/textured MOCVD
II … Al2O3(0001)/textured PAMBE
III … YSZ(111)/single crystalline PAMBE
IV 1020 YSZ(111)/single crystalline PAMBE
V 1021 YSZ(111)/single crystalline PAMBE
FIG. 1. (a) Influence of the growth method, substrate, and Mg concentration
on the position of core levels, valence band maximum (VBM), and work
function U after growth. (b) Illustration of the related VB edge emissionshift as measured by XPS (mon. AlKa radiation). The samples are numberedas described in Table I and were measured without any additional surface
preparation (as loaded).
FIG. 2. Surface cleaning effect of oxygen DBD plasma treatment and UHV
annealing at 500 �C demonstrated by the reduction of the C1s and O1sadssignals.
245301-3 Berthold et al. J. Appl. Phys. 120, 245301 (2016)
-
concentration results in a shift of the core levels and the
VBM to lower BE, while the work function is unaffected.
Mg acts as acceptor in In2O3, resulting in a movement of the
Fermi level (EF) from the CB edge deeper into the band gap.This aspect is reflected in the PES results that reveal a reduc-
tion of the distance between the occupied states and EF forincreasing Mg concentration. The observed effect is stronger,
if the binding energy of the considered core level peak is
lower (higher kinetic energy) and thus the information depth
is higher. In comparison, the results of textured UID
In2O3(111) films grown by MOCVD and PAMBE are also
shown in Fig. 1. In the initial state, no difference to single
crystalline UID In2O3 can be observed, except for the higher
work function in the case of MOCVD grown In2O3. This
effect might be due to a different composition of the surface
adsorbate layer for films that are removed from the MOCVD
growth reactor, where the reactive precursors can possibly
passivate the film surfaces.
Table II lists the quantitative results of the Hall measure-
ments considering the potential uncertainties of the apparent
bulk electron concentration and sheet resistance as discussed
next. The high electron concentration of the SEAL could theo-
retically lead to a slight overestimation of nbulk while for sam-ples that underwent an ICP oxygen plasma treatment, the
existing surface electron depletion layer would result in a
slight underestimation. The deviations in nbulk values for dif-ferent samples cut from one epitaxy wafer are below 20%.
For the UID In2O3 samples, the values are �2� 1018cm�3 orslightly below and were practically identical after additional
ICP oxygen plasma treatment, indicating that for these sam-
ples, the SEAL had no influence on the performed analysis.
The determined bulk electron concentration of the Mg-doped
sample IV with NMg¼ 1020cm�3 was almost one order ofmagnitude lower after growth (2.9� 1017 cm�3) and was fur-ther reduced to 2.2� 1017 cm�3 after the ICP oxygen plasmatreatment. Consequently, for the lower bulk concentration in
Mg-doped In2O3 films, the SEAL does have a slight influence
on the electron density analysis based on Hall experiments.
Overall, these findings indicate that, although a surface elec-
tron channel exists after growth, the Hall measurements are
generally suitable to determine the bulk electron concentration
in the samples examined in this study, which is an essential
parameter for further analysis of the PES measurements and
the surface band alignment.
The determined bulk electron concentration nbulk wasused to calculate the associated position of the Fermi energy
EF with respect to the conduction band minimum (CBM) inthe bulk of the material implementing recent band structure
parameters by Feneberg et al.7 that include the nonparabolic-ity of the conduction band and hence, the electron concentra-
tion dependence of the effective electron mass. The
determined position of the bulk Fermi energy is shown of
Table II. For all considered samples, EF is located close tothe CBM, i.e., slightly above the CBM for the UID samples
and 49 meV below the CBM after growth and 56 meV after
ICP oxygen plasma treatment for the Mg-doped sample IV.
Consequently, the changes in the bulk Fermi energy in the
investigated samples are relatively small and have no consid-
erable influence on the band bending analysis discussed
below and conclusions drawn from it.
In combination with the determined position of the
VBM at the surface via linear extrapolation of the measured
XPS-VB spectra and assuming a band gap of Eg¼ 2.7 eV,6we have calculated the band bending Vbb at the surface usingVbb ¼ Eg þ ðEF � CBMÞbulk � ðEF � VBMÞsurface, wherenegative Vbb values account for downward band bending andelectron accumulation. All experimentally determined quan-
tities are compiled in Table II, negative values of EF�CBMindicate that the Fermi level is below the CBM. In the
preparation state directly after insertion into the vacuum sys-
tem (as loaded), all samples exhibit a slight downward bend-
ing of the bands at the surface of Vbb¼�0.2 eV for UIDIn2O3 and �0.1 eV if Mg acceptors are incorporated.Consequently, for the Mg-doped In2O3 films, the larger shift
of the VBM at the surface than in the bulk indicates a reduc-
tion of the downward band bending compared to the UID
films. In this context, it has to be mentioned that band gap
renormalization effects as discussed in Ref. 15 can be
neglected for the fairly low electron concentration of the
samples under investigation,7 since they are in the range of
the experimental accuracy of determining absolute BE val-
ues in PES (100 meV).
B. Chemical changes and adsorbates
Changes of the chemical surface composition are ana-
lyzed in in sections B and C for different sample treatments.
As already mentioned, all samples initially exhibit a surface
adsorbate layer due to their transport in ambient conditions.
After both investigated in situ processes (vacuum annealingas well as DBD oxygen plasma treatment), changes in the
surface adsorbate layer are observed. Fig. 2 exemplarily
TABLE II. Electrical characteristics (sheet resistance Rsh and electron concentration nbulk) of the UID and Mg-doped In2O3 films after growth and the calcu-
lated location of the Fermi energy EF in the bulk based on recent band structure parameters.7 Location of the surface valence band maximum (VBM) and the
work function U of the In2O3 films as determined by PES measurements after different surface preparation processes: AL; as loaded, DBD; DBD oxygenplasma treatment, ICP; ICP oxygen plasma treatment, T; vacuum annealing (500 �C), T! DBD – DBD oxygen plasma treatment after vacuum annealing.
No Rsh (X) nbulk (cm�3) (EF – CBM)bulk (meV)
(EF – VBM)surface (eV) U (eV)
AL DBD ICP T T! DBD AL DBD ICP T T! DBD
I (1.2 6 0.2)� 103 (2.1 6 0.2)� 1018 10 2.9 2.5 2.0 2.9 2.5 4.1 5.1 4.2 4.2 5.1II (1.8 6 0.6)� 103 (1.6 6 0.5)� 1018 1 2.9 2.2 2.0 2.9 2.3 3.9 5.0 4.3 4.2 5.1III (1.3 6 0.1)� 103 (1.8 6 0.3)� 1018 5 2.9 2.2 2.0 2.9 2.2 3.9 4.9 4.2 4.2 5.3IV (3.6 6 2.3)� 104 (2.9 6 0.9)� 1017 �49 2.8 2.2 2.0 2.8 2.2 3.9 5.1 4.1 4.2 4.7
245301-4 Berthold et al. J. Appl. Phys. 120, 245301 (2016)
-
documents the changes of the C1s and O1s spectra for an
UID In2O3/YSZ(111) film (sample III), that were also recog-
nized for all other samples. Noticeably, the carbon content is
drastically reduced after the vacuum annealing or the DBD
oxygen plasma process. The reactive oxygen species, which
are generated in the plasma region right in front of the sur-
face efficiently oxidize the carbon impurities leading to vola-
tile species that desorb from the surface. At the same time,
reactive oxygen species attach at the In2O3 surface as
directly identified by the side feature O1sads (2.0 eV higher
BE compared to the oxygen atoms in the crystal lattice) of
the O1s state. The shift of absolute binding energy indicated
by vertical dashes in Fig. 2 is due to band bending effects
that will be discussed below. Hence, the treatment of the
indium oxide films by oxygen plasma effectively cleans the
surface from hydrocarbons with the effect of replacing the
species that saturate free surface sites by oxygen.
The increased binding energy of O1sads compared to the
chemical state of the bulk O atoms indicates that the addi-
tional oxygen provided by the plasma treatment is not
attached at crystal sites, but in a different configuration.
Possible bonding sites could be saturated In surface bonds or
interstitial positions. The (111) orientation of In2O3 is the
most stable surface configuration with fivefold and sixfold
coordinated In sites24 and negligible surface relaxation com-
pared to the bulk crystal structure. It is expected that the
reactive oxygen atoms can attach to free bonds at the surface
without significant structural reconfiguration as for example,
possible for the (100) surface where at high oxygen partial
pressures, formation of a peroxide phase is suggested24,40
corresponding to an oxygen-rich surface configuration.
Nevertheless, it has been shown that the oxygen content at
the (111) oriented In2O3 surface can be modified by thermal
treatments switching between an oxidized and a reduced sur-
face configuration.23 These experiments demonstrate the
capability of the surface to adsorb oxygen atoms at the sur-
face. In addition, the bixbyte crystal structure might allow
incorporation of oxygen at interstitial sites and possibly
enable implantation in subsurface layers.41
In order to identify whether the DBD oxygen plasma
treatment enables a penetration of activated species into
deeper regions beneath the top surface layer, we have per-
formed angular-dependent XPS measurements. Fig. 3 includes
a series of O1s spectra in dependence of emission angle H.Generally, if H is increased from normal emission (0�) tograzing emission, the depth of information is reduced by a fac-
tor of cos(H) within some experimental limitations. As a con-sequence, for a possible surface/adsorbate layer, the relative
contribution of adatom signals would be drastically increased
while any substrate signal would be even stronger attenuated,
resulting in a typical H-dependence of adsorbate and filmintensity. As other extremum, a uniform intermixing (constant
depth profile) would result in no angular dependence.
The chemical component of the In2O3 oxygen lattice
atoms (O1sbulk) and the oxygen adsorbate signal (O1sads)
have been fitted using Gaussian-Lorentzian profiles and their
area ratio is plotted in the right part of Fig. 3. The angular
dependence was furthermore modeled using different
assumptions to account for a possible distribution of
incorporated oxygen species from the plasma. All samples
exhibit ordered and smooth surface topographies with a low
rms roughness (0.5 nm for MBE-grown In2O3/YSZ and
MOCVD-grown In2O3/Al2O3 and 1.2 nm for MBE-grown
In2O3/Al2O3 (Refs. 2 and 12)). Therefore, the surface of the
films has been regarded as ideally sharp without any struc-
tural deviations. The isolines in Fig. 3 (right) correspond to
the scenario that oxygen is only adsorbed at the free bonds
of the surface without any penetration into deeper layers. A
fairly good agreement to the experiment is found for an
effective surface coverage c of 0.7 ML. We have also consid-ered different implantation profiles of oxygen atoms that
might be incorporated into deeper layers (exponential decay,
Gaussian decay profiles, linear decay, and step function) and
fitted these profiles to the measured data. Although these pro-
files include one further free parameter, none of the extracted
angle-dependencies was able to better describe the experi-
mental data compared to the model of a simple surface
adsorbate. For comparison, the isolines and experimental
values for these four model distributions are shown in the
supplementary material exemplarily assuming a width dads(width at half maximum of the individual decay profile) of
0.5 nm. When fitting these profiles in dependence of the
parameter dads and surface coverage c, dads always convergedto the value of a single In–O bond length. This result
matches the situation of a classic adsorbate contribution.
As a consequence, these results show strong evidence
that no significant incorporation of larger amounts of oxygen
into deeper layers occurs for the DBD oxygen plasma treat-
ment. Only surface O adatoms and incorporation of O atoms
in the top In2O3 surface layer accounts for the angular depen-
dence of the O1s spectra in Fig. 3. In this context, it has to
be mentioned that the discussion is only valid within the
FIG. 3. Left: angular-dependent measurements (variation of emission angle
H from 0� to 70�) of the O1s state after DBD oxygen plasma treatment of anUID In2O3 film (sample II). Right: comparison of the experimentally deter-
mined signal ratio between oxygen adsorbate component (O1sads) and the
oxygen bulk signal (O1sbulk) from oxygen lattice sites (dots) with calcula-
tions based on a two-layer model with In2O3 film and oxygen adatoms
(isolines–variation of surface coverage from 0.1 to 1.0 ML). In this model,
one monolayer (1 ML) is equivalent to an effective oxygen surface coverage
of 1.8� 1015 atoms/cm2. The datapoints represent the experimental valuesof the DBD treatment in red, a single ICP treatment in blue and the average
value for several ICP treated samples measured in normal emission as the
orange dot.
245301-5 Berthold et al. J. Appl. Phys. 120, 245301 (2016)
ftp://ftp.aip.org/epaps/journ_appl_phys/E-JAPIAU-120-020648
-
sensitivity limit of the used experimental method (�0.1 at. %for XPS). Considering the experimental and modeling
aspects of the performed analysis, it needs to be noted that
this approach is not capable to exclude the existence of rela-
tively low oxygen concentrations, which might diffuse into
deeper layers and occupy vacancies or interstitial sites.
Nevertheless, the measurements demonstrate that after the
DBD plasma treatment, the majority of attached oxygen spe-
cies are directly located at the In2O3 surface. Thus, the
resulting variations in the electronic properties, that will be
discussed in the following, are a result of the adsorbed oxy-
gen species that saturate the free bonds at the outermost sur-
face. The binding energy of the O1s adsorbate component is
reproducibly �2.0 eV higher than the O1s state of the O lat-tice atoms. Previous experiments on gas interaction of poly-
crystalline In2O3 have shown that the O� species are
adsorbed at the surface after molecule dissociation.26 Since
the spectral fingerprints are comparable for the DBD treat-
ment, a similar reaction mechanism could be possible for the
reactive oxygen species that are generated in the plasma
excitation.
The oxygen adsorbate component O1sads is also found
for samples that were processed in an ICP plasma etching
reactor, but broadened and not as well defined as after
the DBD oxygen plasma treatment. In this case, the subse-
quent removal from the vacuum recipient allows readsorp-
tion of surface contaminants. Hence, the O1sads signal could
also contain contributions from other species that typically
adsorb in air such as water, hydroxyl groups or oxygen con-
taining carbon-based molecules. In consistence, a consider-
able uptake of carbon is also detected, comparable to the
amount for samples after growth. Such additional surface
adsorbates certainly influence the angular dependence of the
XPS signal. Nevertheless, we have characterized the angular
distribution of the O1sads/O1sbulk ratio also for ICP oxygen
plasma treated samples. The results are included in Fig. 3 as
blue datapoints for a MBE grown UID In2O3 sample on
Al2O3 (equivalent to sample II) and the orange dot represents
the mean value for several UID In2O3 films grown by either
MOCVD or MBE on different substrates. This large set of
samples was only characterized in normal emission. As for
the DBD oxygen plasma treatment, the experimental data of
the ICP treatment follow the h-dependence of the model iso-lines for a simple oxygen adatom layer; in this case for a sur-
face coverage of �1 ML. One might expect a strongerphysical impact of the high density ICP plasma, since the
reactive oxygen species reach kinetic energies of up to a few
hundred eV in this case compared to the gentle �10 eV ionsin a DBD discharge that typically only act at the surface. In
consistence, the total amount of adsorbed oxygen is higher,
which agrees with the fact that the ICP treatment leads to a
stronger electron depletion at the surface (see Section III D).
However, one would also expect a penetration of additional
oxygen into deeper layers for the ICP plasma. A clear indica-
tion for an implantation scenario cannot be found from
the performed measurements but we want to point out that
these results cannot provide proof that these effects do not
happen due to the mentioned uncertainties caused by the
additional adsorbates and due to the limited sensitivity of
XPS (�0.1 at. %).
C. Effects of vacuum annealing
As already mentioned, after vacuum annealing, both
adsorbate signals (C1s and O1sads) have vanished, indicating a
successful removal of surface impurities (Fig. 2). In a temper-
ature dependent series, we have determined that heating to
500 �C is sufficient, in consistence with the comparable resultsof bulk In2O3 samples and sputtered thin films.
41 Therefore,
this routine is effective to prepare clean In2O3(111) surfaces.
In addition, if such a vacuum annealing is performed after
oxygen DBD plasma modification, a complete removal of the
oxygen adatoms (O1sads signal) is also achieved. This allows
to study the repeated cycles of oxidation and reduction experi-
ments at this specific surface.
Figures 4(a) and 5 include the O1s and VB spectra of
films I–IV after annealing the samples in vacuum. In com-
parison with the measurements of the samples in their initial
FIG. 4. O1s core level spectra of (a) different vacuum-annealed In2O3 films
(clean surface) and (b) after subsequent oxygen DBD plasma treatment. The
binding energies of the O1s components of the In2O3 crystal structure as
well as the plasma-induced oxygen adsorbate component are indicated.
FIG. 5. Comparison of the XPS-VB spectra of UID In2O3(111) films after
vacuum annealing (solid lines) and after subsequent DBD oxygen plasma
treatment (dotted lines): (a) sample I; In2O3(111)/Al2O3(0001) by MOCVD,
(b) sample II; In2O3(111)/Al2O3(0001) by PAMBE, (c) sample III;
In2O3(111)/YSZ(111) by PAMBE and (d) sample IV; Mg-doped
(NMg¼ 1020 cm�3) In2O3(111)/YSZ(111) by PAMBE.
245301-6 Berthold et al. J. Appl. Phys. 120, 245301 (2016)
-
condition after introduction into the UHV chamber (as
loaded), the core level BE and VBM values do not change
after vacuum annealing although the desorption of oxygen
adatoms and other surface impurities occurs, and the emis-
sion of photoelectrons of the SEAL near the Fermi energy
clearly remains. In addition, solely for the In2O3 films on
YSZ(111) (samples III and IV), a broad emission close to the
VB edge extending up to �1 eV below EF is observed afterannealing. This feature is assigned to the thermally induced
formation of surface defects (oxygen vacancies42,43) caused
by desorption of lattice oxygen atoms. Interestingly, the oxy-
gen DBD plasma treatment completely saturates this defect
state emission, indicating that the reactive oxygen species in
the plasma can heal intrinsic surface defects in the material.
We assign these changes for films on YSZ to the greater
differences in thermal expansion coefficients and lattice mis-
match between the substrate and the epilayer in the case of
YSZ compared to Al2O3. Considering the room temperature
lattice constants and thermal expansion coefficients of
In2O3,44–46 Al2O3 (Refs. 47 and 48) and YSZ,
44,49 we expect
a slight compressive strain of �0.3% at room temperatureand �0.4% at 500 �C for In2O3 films on sapphire which iseasily accommodated within the film without causing struc-
tural damage. However, In2O3 epilayers on YSZ are esti-
mated to be tensile strained (1.6% at room temperature and
1.8% at 500 �C). We anticipate that the increasing tensilestrain during heating up in vacuum could cause a deteriora-
tion of the structural integrity of the crystal and possibly a
defect formation at higher temperatures.
Furthermore, annealing indium oxide thin films in vac-
uum is known to increase the electron concentration of the
layers due to formation of oxygen vacancies.20 In order to
quantify the influence of the annealing/surface cleaning pro-
cedure on the electron transport properties, UID and Mg-
doped In2O3 films on Al2O3(0001) and YSZ(111) substrate
have been characterized in air by Hall measurements after
performing the vacuum heating procedure. Afterwards, the
samples exhibited a bulk electron concentration slightly
above 1019 cm�3. As a consequence, the position of the bulk
Fermi energy is shifted to �200 meV above the CBM. It isremarkable that these changes in (EF – CBM)bulk do not havean influence on the measured BE position of the core levels
and the VBM at the surface if one compares the situation
before and after annealing or the oxygen DBD plasma treat-
ment of an untreated vs. an annealed sample. This observa-
tion might be explained by a pinning of the surface Fermi
level for both sample conditions, (i) in case of electron accu-
mulation and (ii) in case of the oxygen plasma treated sur-
face which exhibits a surface electron depletion layer. The
origin of this effect is not clarified yet and is the subject of
ongoing studies. Nevertheless, these variations of the bulk
Fermi level do not significantly influence the observed
changes in band bending after oxygen plasma treatments that
will be analyzed in the successive section.
D. Changes in electronic properties
The O1s and VB edge spectra of samples I–IV after vac-
uum annealing and after a subsequent oxygen DBD plasma
treatment are compared in Figs. 4 and 5, respectively.
Comparable measurements of the samples after growth (as
loaded) and after direct oxygen DBD processing, which
exhibit very similar characteristics, can be found in the sup-
plementary material. After the DBD oxygen plasma process,
a reproducible shift of all core level states as well as the
VBM by 0.2–0.4 eV towards lower BE is observed for sam-
ple I grown by MOCVD, while all samples prepared by
MBE exhibit an even stronger shift of 0.6–0.8 eV after the
DBD treatment. In addition, we have also analyzed samples
that were prepared ex situ by an oxygen plasma in an ICPreactor. In this case, the occupied states shift up to 0.9 eV
towards the Fermi level for all samples under investigation
(spectra not shown, compare to Fig. 4 in Ref. 2 and the num-
bers given in Table II). Furthermore, the emission close to
the Fermi energy of electrons from the initial surface elec-
tron accumulation layer10 is completely vanished after the
oxygen DBD plasma (Fig. 5) or the oxygen ICP plasma treat-
ment, indicating a removal of the surface electron channel
by these plasma treatments. Similar spectral changes have
been observed in an earlier PES study.11
The shifts in core level and the VB binding energy are a
result of variations of the surface band alignment, i.e., the
surface band bending Vbb for the different surface treatments,since the bulk Fermi level is not expected to be strongly
affected by oxygen plasma modifications.11 We have ana-
lyzed the surface band structure based on the information
from PES characterization and Hall measurements. As
already discussed, the samples initially exhibit a slight down-
ward band bending and a related emission from electrons of
the SEAL. Based on the shifts of the core level BE and
VBM, we have calculated the variation of Vbb and comparethese quantities with the work function U values as deter-mined from UPS measurements that are included in Table II.
Vacuum annealing causes no shift of core levels and the
VBM with respect to the initial state (as loaded), resulting in
the preservation of the surface electron accumulation as can
be directly monitored by the emission of electrons close to
EF. This aspect points out that the existence of the surfaceaccumulation layer cannot solely be caused by surface adsor-
bates. We anticipate that the existence of surface states of
unsaturated bonds and surface defects (e.g., oxygen vacan-
cies) play an important role for the SEAL formation, since
the plasma-induced surface saturation by O adsorbates leads
to a complete depletion of electrons at the surface. This sce-
nario might be different for thin films grown in a vacuum
chamber or samples that have been annealed in UHV, where
the formation of surface defects is expected at higher temper-
atures, compared to clean (111) surfaces that have been pre-
pared by crystal cleavage at room temperature,30 where no
SEAL layer was found.
In addition, the desorption of surface impurities affects
the work function which is found to be 4.2 eV for a clean
In2O3(111) surface. The DBD oxygen plasma treatment sig-
nificantly changes the electronic surface properties: after-
wards an upward band bending (formation of a surface
depletion layer) is observed with Vbb¼�0.2 eV for theMOCVD sample and �0.5 eV for the samples prepared byPAMBE (for the previously annealed samples according to
245301-7 Berthold et al. J. Appl. Phys. 120, 245301 (2016)
ftp://ftp.aip.org/epaps/journ_appl_phys/E-JAPIAU-120-020648ftp://ftp.aip.org/epaps/journ_appl_phys/E-JAPIAU-120-020648
-
the change in bulk EF, the values are �0.2 eV larger). It isinteresting to note, that for the Mg-doped sample, the ener-
getic shifts are slightly smaller, resulting in the same final
Vbb value of 0.5 eV.A clear explanation of the differences measured for
In2O3 films prepared by the two different growth methods
cannot be given, but from several repetitions of the oxygen
plasma treatments, we have found a correlation between the
adsorbate coverage and change in Vbb. Fig. 6(a) comparesthe change in surface band bending DVbb after oxygen DBDplasma treatment of an initially clean surface in correlation
with the calculated oxygen adatom coverage c based on theanalysis of the O1sads/O1sbulk ratio for each measurement. A
slight scattering in the dependence of cðDVbbÞ is found, but ageneral trend of monotonic increase is observed. For the
MOCVD samples, the values are located in the lower left
corner and the reduced change in band bending correlates
with a lower surface adatom density. For the MBE samples,
DVbb scatters around 0.7 6 0.2 eV for c¼ 0.7 6 0.1 ML.Hence, the degree of upward band bending is increasing with
the uptake of reactive oxygen from the plasma, and the
experimental data-points can be fitted by an empirical power
function resulting in DVbb [eV]¼ 1.17 � (c [ML])1:38. In con-clusion, while Mg incorporation or the film crystallinity
(grain boundaries) seem to have a minor influence on oxygen
adsorption, a strong effect of the preparation method was
observed. For the MOCVD samples, only values slightly
above flat band conditions are found, but the SEAL is effec-
tively depleted in these cases as well.
For the ICP oxygen plasma treatment, the surface
upward band bending is even higher and found to be inde-
pendent of the sample and growth method at Vbb¼ 0.7 eV.This indicates a broadening of the depletion layer width in
this case. We anticipate that due to the higher plasma power
and density of this ICP reactor, a stronger oxidation of the
In2O3 surface occurs, possibly leading to a higher surface
coverage and a stronger degree of surface passivation.
Unfortunately, due to additional surface adsorbates originat-
ing from the exposure to air, a direct analysis of the correla-
tion between adsorbed oxygen species from the plasma and
upward band bending is not possible for these samples.
To get a deeper insight into the correlations between
bulk/surface electron concentrations induced by different
oxygen plasma treatments of as grown In2O3 samples with-
out vacuum annealing, we calculated the depth profiles of
the conduction/valence band edges as well as of the electron
density at various surface potentials and bulk electron con-
centrations. To obtain these depth distributions, the
Schr€odinger and Poisson equations are self-consistentlysolved. For the Poisson equation, the Dirichlet boundary con-
ditions (i.e., fixed surface potential relative to the bulk Fermi
level position) are used, and we assume the full ionization of
donors present in the samples. The effect of the conduction
band nonparabolicity is also accounted for and the band
structure parameters as well as the dielectric parameters
have been taken from Ref. 7. The net surface charge is calcu-
lated from the charge neutrality condition applied to the
whole sample. Fig. 6(b) compares the variation of band
bending in dependence of the density of surface states Nssfor different bulk electron concentrations nbulk between1� 1017 cm�3 and 1� 1019 cm�3. Generally, if the localizednegative charges exist at the surface, they induce a depletion
of electrons (upward band bending) in the subsurface region
for charge compensation. In contrast, a positive surface net
charge is compensated by a delocalized electron accumula-
tion layer (downward band bending) at the In2O3 surface.
Such localized surface charges can either be caused by filled
or empty electron states of unsaturated bonds at the outer-
most surface atoms, by missing atoms (vacancies) in the near
surface layers or by adatoms that bond to these free electrons
and extract or donate charge from/towards the semiconduc-
tor. Obviously, the surface Fermi level and the surface band
bending Vbb depend on the surface donor/acceptor concentra-tion Nss as well as the bulk electron concentration nbulk in anonlinear manner. If we consider the effect of the oxygen
DBD plasma treatment, we start with an initial slight down-
ward band bending followed by adsorption of oxygen species
that extract electron density from the semiconductor surface
and represent a negatively charged adatom at the surface.
This aspect is also confirmed by the strong rise in work func-
tion from 4.2 eV to (5.0 6 0.3) eV (see Table II). The changein work function is a sum of both effects: a possible variation
of surface band bending DVbb and the modification or emer-gence of a surface dipole DUdip (DU ¼ DUdip þ DVbb). Afterthe oxygen DBD plasma process, the increase in work func-
tion is larger than the variation of Vbb which can beexplained by an effective electric dipole (DUdip) formedbetween the In2O3 surface and the attached negatively
FIG. 6. (a) Influence of the surface O adsorbate coverage on band bending
after oxygen DBD plasma treatment for different In2O3 samples with a fit of
the experimental data by a power function in grey and (b) dependence of the
density of localized surface states Nss on the bulk electron and band bendingVbb from the accumulation (Vbb< 0 eV) to the depletion range (Vbb> 0 eV)as determined from Schr€odinger-Poisson calculations.
245301-8 Berthold et al. J. Appl. Phys. 120, 245301 (2016)
-
charged O adatoms, which increases the barrier for electron
emission. The fact that the absolute work function is rather
independent of O coverage, while Vbb is gradually increasingfor higher adsorbate densities, indicates that the effective
surface dipole induced by the attached O species is decreas-
ing for a higher coverage, maybe due to collective phenom-
ena or coulomb interactions between the negatively charged
oxygen adatoms. The consequence of the attachment of these
electronegative adsorbates is a cumulative saturation of posi-
tive surface states resulting in a change from electron accu-
mulation towards depletion at the In2O3(111) surface. In
qualitative agreement with the results from Schr€odinger-Poisson calculations, the experimentally determined change
in band bending follows this trend in dependence of cover-
age (compare Figs. 6(a) and 6(b)).
Besides the experimental evaluation of surface band
bending and surface states, we have used the Schr€odinger-Poisson calculations to determine the distribution of electrons
close to the surface as well as the corresponding band profiles.
We have used the bulk electron concentration and the value
of Vbb, determined as described above, to calculate n(z) andCBM(z)/VBM(z) with respect to EF at 0 eV. In Fig. 7, theresults for one of the UID films (sample III), which all exhibit
an nbulk of �2� 1018cm�3, as well as the Mg-doped sampleIV with an almost one order of magnitude lower bulk electron
concentration are presented. For the samples after growth (as
loaded), the electron accumulation layer is found to be only a
few nm wide and the SEAL sheet electron concentration
NSEAL is found to be as high as 2.9� 1012cm�2 for the UIDsample, while expectedly, the value is significantly reduced
for the Mg-doped sample, since both nbulk and Vbb are lower
in this case. After the two different oxygen plasma treatments,
the near surface region is depleted from electrons with a space
charge layer width of several 10 nm, hence significantly
broader than that in the case of electron accumulation. In both
cases of electron depletion or accumulation, the net surface
charge has to be screened by sub-surface charges. In case of
the positive surface charge (downward band bending),
because of the high accumulation electron density, only a nar-
row sub-surface region is required to fully screen the surface
charge. In the opposite case of upward band bending, the neg-
ative surface charge is screened by fixed positively charged
donors. Due to the much lower donor concentration, the
screening length is substantially larger in this case. Hence, for
the Mg-doped sample with a lower bulk electron concentra-
tion, the depletion layer is almost three times as broad com-
pared to the UID samples, while only slight variations in the
electron and band profiles are found if the two oxygen plasma
treatments are compared.
IV. CONCLUSIONS
The oxygen plasma induced changes of chemical and
electronic surface properties of In2O3(111) thin films are
identified. The reactive oxygen species of the plasma excita-
tion remove the adsorbed carbon impurities and saturate free
surface bonds as well as oxygen vacancies of the outermost
surface layer. The chemical changes and incorporation of
oxygen are restricted to the adsorbate and topmost layer for
an oxygen DBD plasma treatment. A penetration/implanta-
tion effect of O species into subsurface layers was not
observed and hence, the changes in electronic properties are
solely induced by surface adsorbates in the case of the DBD
oxygen plasma modification. The attached electronegative
oxygen adsorbates extract electrons from the substrate form-
ing a surface dipole that enlarges the barrier for electron
emission (work function). These charge transfer mechanisms
induce a strong depletion of the initially accumulated surface
electrons that manifests itself by a strong upward bending of
the CBM and VBM at the surface. A direct correlation
between the O adsorbate coverage and quantity of surface
band bending and depletion layer width is observed, which
highlights a tunability of the surface electronic properties of
In2O3 semiconductor thin films. Surface oxygen plasma
treatments in an ICP reactor induce a slightly stronger elec-
tron depletion and a higher amount of incorporated surface
oxygen species. The SEAL depletion was confirmed for
unintentionally n-doped and Mg-doped In2O3 layers. The
effects of oxygen plasma modification can be fully reversed
by a vacuum annealing procedure above 500 �C resulting inthe desorption of the attached oxygen species and an enrich-
ment of the surface by electrons, i.e., recovery of the SEAL.
This result confirms that surface oxygen vacancies are one
possible origin of the SEAL. The observed effects of surface
electron depletion and enhancement of work function
induced by oxygen plasma processing are of great relevance
for any subsequent device processing for applications of
In2O3 in modern semiconductor devices or of ITO films in
solid-state lighting devices and displays as well as solar
cells.
FIG. 7. Electron density distribution n(z) (left) and corresponding VBM/CBM profiles (right) based on Schr€odinger-Poisson calculations using theexperimental bulk electron concentration and band bending values after dif-
ferent surface preparation processes as input parameter for the simulations.
The related Vbb values are indicated by the corresponding color. Top: UIDIn2O3/YSZ(111) (sample III) and bottom: Mg-doped (NMg¼ 1020 cm�3)In2O3/YSZ(111) (sample IV). In the latter case, the profiles for a virtual
band bending of �0.2 eV are also included. The gray shaded areas representthe sheet electron density NSEAL of the electron accumulation region, beingthe difference between the n(z) profiles for the samples with downward bandbending (Vbb< 0 eV) and the hypothetical case of flat band conditions(Vbb¼ 0 eV).
245301-9 Berthold et al. J. Appl. Phys. 120, 245301 (2016)
-
SUPPLEMENTARY MATERIAL
See supplementary material for the XPS data of samples
after growth (as loaded) and after direct oxygen DBD plasma
treatment as well as for detailed results of modeling the oxy-
gen depth distribution from angular dependent XPS
measurements.
ACKNOWLEDGMENTS
We are grateful for the financial support by Deutsche
Forschungsgemeinschaft (DFG) within the project “Seebeck
Gas Sensors” (Grant Nos. HI 1800/1-1, AM 105/31-1 and BI
1754/1-1). We thank O. H€offt and M. Marschewski(Technische Universit€at Clausthal) for support and fruitfuldiscussions on the design of DBD plasma systems.
1O. Bierwagen, Semicond. Sci. Technol. 30, 024001 (2015).2J. Rombach, A. Papadogianni, M. Mischo, V. Cimalla, L. Kirste, O.
Ambacher, T. Berthold, S. Krischok, M. Himmerlich, S. Selve, and O.
Bierwagen, Sens. Actuators, B 236, 909 (2016).3H. von Wenckstern, D. Splith, F. Schmidt, M. Grundmann, O. Bierwagen,
and J. S. Speck, APL Mater. 2, 046104 (2014).4F. Fuchs and F. Bechstedt, Phys. Rev. B 77, 155107 (2008).5A. Walsh, J. L. F. Da Silva, S.-H. Wei, C. K€orber, 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, 167402 (2008).6K. Irmscher, M. Naumann, M. Pietsch, Z. Galazka, R. Uecker, T. Schulz,
R. Schewski, M. Albrecht, and R. Fornari, Phys. Status Solidi A 211, 54(2014).
7M. Feneberg, J. Nixdorf, C. Lidig, R. Goldhahn, Z. Galazka, O.
Bierwagen, and J. S. Speck, Phys. Rev. B 93, 045203 (2016).8A. Klein, Appl. Phys. Lett. 77, 2009 (2000).9P. D. C. King, T. D. Veal, D. J. Payne, A. Bourlange, R. G. Egdell, and C.
F. McConville, Phys. Rev. Lett. 101, 116808 (2008).10P. 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, 205211 (2009).11O. Bierwagen, J. S. Speck, T. Nagata, T. Chikyow, Y. Yamashita, H.
Yoshikawa, and K. Kobayashi, Appl. Phys. Lett. 98, 172101 (2011).12C. Y. Wang, L. Kirste, F. M. Morales, J. M. M�anuel, C. C. R€ohlig, K.
K€ohler, V. Cimalla, R. Garc�ıa, and O. Ambacher, J. Appl. Phys. 110,093712 (2011).
13A. Walsh, Appl. Phys. Lett. 98, 261910 (2011).14S. Lany, A. Zakutayev, T. O. Mason, J. F. Wager, K. R. Poeppelmeier, J.
D. Perkins, J. J. Berry, D. S. Ginley, and A. Zunger, Phys. Rev. Lett. 110,016802 (2012).
15K. H. L. Zhang, R. G. Egdell, F. Offi, S. Iacobucci, L. Petaccia, S.
Gorovikov, and P. D. C. King, Phys. Rev. Lett. 110, 056803 (2013).16R. Bel Hadj Tahar, T. Ban, Y. Ohya, and Y. Takahashi, J. Appl. Phys. 83,
2631 (1998).17A. Bourlange, D. J. Payne, R. G. Palgrave, H. Zhang, J. S. Foord, R. G.
Egdell, R. M. J. Jacobs, T. D. Veal, P. D. C. King, and C. F. McConville,
J. Appl. Phys. 106, 013703 (2009).18O. Bierwagen and J. S. Speck, Phys. Status Solidi A 211, 48 (2014).19Y. Kanai, Jpn. J. Appl. Phys., Part 2 24, L361 (1985).
20O. Bierwagen and J. S. Speck, Appl. Phys. Lett. 101, 102107 (2012).21P. D. C. King and T. D. Veal, J. Phys.: Condens. Matter 23, 334214
(2011).22V. Brinzari, B. K. Cho, M. Kamei, and G. Korotcenkov, Appl. Surf. Sci.
324, 123 (2015).23M. Wagner, S. Seiler, B. Meyer, L. A. Boatner, M. Schmid, and U.
Diebold, Adv. Mater. Interfaces 1, 1400289 (2014).24P. Agoston and K. Albe, Phys. Rev. B 84, 045311 (2011).25I. A. Tambasov, V. G. Maygkov, A. S. Tarasov, A. A. Ivanenko, L. E.
Bykova, I. V. Nemtsev, E. V. Eremin, and E. V. Yozhikova, Semicond.
Sci. Technol. 29, 082001 (2014).26M. Himmerlich, A. Eisenhardt, T. Berthold, C. Y. Wang, V. Cimalla, O.
Ambacher, and S. Krischok, Phys. Status Solidi A 213, 831 (2016).27N. Siedl, P. Guegel, and O. Diwald, J. Phys. Chem. C 117, 20722
(2013).28M. Bender, N. Katsarakis, E. Gagaoudakis, E. Hourdakis, E. Douloufakis,
V. Cimalla, and G. Kiriakidis, J. Appl. Phys. 90, 5382 (2001).29C. Y. Wang, V. Cimalla, T. Kups, C.-C. R€ohlig, T. Stauden, O. Ambacher,
M. Kunzer, T. Passow, W. Schirmacher, W. Pletschen, K. K€ohler, and J.Wagner, Appl. Phys. Lett. 91, 103509 (2007).
30M. Nazarzahdemoafi, F. Titze, S. Machulik, C. Janowitz, Z. Galazka, R.
Manzke, and M. Mulazzi, Phys. Rev. B 93, 081303 (2016).31Y. Gassenbauer, A. Wachau, and A. Klein, Phys. Chem. Chem. Phys. 11,
3049 (2009).32V. Christou, M. Etchells, O. Renault, P. J. Dobson, O. V. Salata, G.
Beamson, and R. G. Egdell, J. Appl. Phys. 88, 5180 (2000).33M. G. Mason, L. S. Hung, C. W. Tang, S. T. Lee, K. W. Wong, and M.
Wang, J. Appl. Phys. 86, 1688 (1999).34D. J. Milliron, I. G. Hill, C. Shen, A. Kahn, and J. Schwartz, J. Appl. Phys.
87, 572 (2000).35P. Vogt, A. Trampert, M. Ramsteiner, and O. Bierwagen, Phys. Status
Solidi A 212, 1433 (2015).36N. Preissler, O. Bierwagen, A. T. Ramu, and J. S. Speck, Phys. Rev. B 88,
085305 (2013).37L. Wegewitz, S. Dahle, O. H€offt, F. Voigts, W. Vi€ol, F. Endres, and W.
Maus-Friedrichs, J. Appl. Phys. 110, 033302 (2011).38M. Himmerlich, S. Krischok, V. Lebedev, O. Ambacher, and J. A.
Schaefer, J. Cryst. Growth 306, 6 (2007).39V. M. Bermudez, A. D. Berry, H. Kim, and A. Piqu�e, Langmuir 22, 11113
(2006).40V. Golovanov, M. A. M€aki-Jaskari, T. T. Rantala, G. Korotcenkov, V.
Brinzari, A. Cornet, and J. Morante, Sens. Actuators, B 106, 563 (2005).41S. P. Harvey, T. O. Mason, Y. Gassenbauer, R. Schafranek, and A. Klein,
J. Phys. D: Appl. Phys. 39, 3959 (2006).42I. Tanaka, F. Oba, K. Tatsumi, M. Kunisu, M. Nakano, and H. Adachi,
Mater. Trans. 43, 1426 (2002).43P. Agoston, P. Erhart, A. Klein, and K. Albe, J. Phys.: Condens. Matter
21, 455801 (2009).44A. Bourlange, D. J. Payne, R. G. Egdell, J. S. Foord, P. P. Edwards, M. O.
Jones, A. Schertel, P. J. Dobson, and J. L. Hutchison, Appl. Phys. Lett. 92,092117 (2008).
45R. L. Weiher and R. P. Ley, J. Appl. Phys. 34, 1833 (1963).46K. D. Kundra and S. Z. Ali, J. Appl. Cryst. 3, 543 (1970).47Y. V. Shvyd’ko, M. Lucht, E. Gerdau, M. Lerche, E. E. Alp, W. Sturhahn,
J. Sutter, and T. S. Toellner, J. Synchrotron Radiat. 9, 17 (2002).48H. Chikh, F. S. I. Ahmed, A. Afir, and A. Pialoux, J. Alloy Compd. 654,
509 (2016).49H. Hayashi, T. Saitou, N. Maruyama, H. Inaba, K. Kawamura, and M.
Mori, Solid State Ionics 176, 613 (2005).
245301-10 Berthold et al. J. Appl. Phys. 120, 245301 (2016)
ftp://ftp.aip.org/epaps/journ_appl_phys/E-JAPIAU-120-020648http://dx.doi.org/10.1088/0268-1242/30/2/024001http://dx.doi.org/10.1016/j.snb.2016.03.079http://dx.doi.org/10.1063/1.4870536http://dx.doi.org/10.1103/PhysRevB.77.155107http://dx.doi.org/10.1103/PhysRevLett.100.167402http://dx.doi.org/10.1002/pssa.201330184http://dx.doi.org/10.1103/PhysRevB.93.045203http://dx.doi.org/10.1063/1.1312199http://dx.doi.org/10.1103/PhysRevLett.101.116808http://dx.doi.org/10.1103/PhysRevB.79.205211http://dx.doi.org/10.1063/1.3583446http://dx.doi.org/10.1063/1.3658217http://dx.doi.org/10.1063/1.3604811http://dx.doi.org/10.1103/PhysRevLett.108.016802http://dx.doi.org/10.1103/PhysRevLett.110.056803http://dx.doi.org/10.1063/1.367025http://dx.doi.org/10.1063/1.3153966http://dx.doi.org/10.1002/pssa.201330224http://dx.doi.org/10.1143/JJAP.24.L361http://dx.doi.org/10.1063/1.4751854http://dx.doi.org/10.1088/0953-8984/23/33/334214http://dx.doi.org/10.1016/j.apsusc.2014.10.072http://dx.doi.org/10.1002/admi.201400289http://dx.doi.org/10.1103/PhysRevB.84.045311http://dx.doi.org/10.1088/0268-1242/29/8/082001http://dx.doi.org/10.1088/0268-1242/29/8/082001http://dx.doi.org/10.1002/pssa.201532458http://dx.doi.org/10.1021/jp4069834http://dx.doi.org/10.1063/1.1410895http://dx.doi.org/10.1063/1.2779971http://dx.doi.org/10.1103/PhysRevB.93.081303http://dx.doi.org/10.1039/b822848ehttp://dx.doi.org/10.1063/1.1312847http://dx.doi.org/10.1063/1.370948http://dx.doi.org/10.1063/1.371901http://dx.doi.org/10.1002/pssa.201431889http://dx.doi.org/10.1002/pssa.201431889http://dx.doi.org/10.1103/PhysRevB.88.085305http://dx.doi.org/10.1063/1.3611416http://dx.doi.org/10.1016/j.jcrysgro.2007.04.014http://dx.doi.org/10.1021/la061578ahttp://dx.doi.org/10.1016/j.snb.2004.07.026http://dx.doi.org/10.1088/0022-3727/39/18/006http://dx.doi.org/10.2320/matertrans.43.1426http://dx.doi.org/10.1088/0953-8984/21/45/455801http://dx.doi.org/10.1063/1.2889500http://dx.doi.org/10.1063/1.1702698http://dx.doi.org/10.1107/S0021889870006842http://dx.doi.org/10.1107/S0909049501019203http://dx.doi.org/10.1016/j.jallcom.2015.09.131http://dx.doi.org/10.1016/j.ssi.2004.08.021
s1ln1s2s3s3At1f1f2s3Bt2f3s3Cf4f5s3Df6s4f7s5c1c2c3c4c5c6c7c8c9c10c11c12c13c14c15c16c17c18c19c20c21c22c23c24c25c26c27c28c29c30c31c32c33c34c35c36c37c38c39c40c41c42c43c44c45c46c47c48c49