Applied Surface Science 349 (2015) 970–982
Contents lists available at ScienceDirect
Applied Surface Science
journal homepage: www.elsevier.com/locate/apsusc
Electronic and surface properties of Ga-doped In2 O3 ceramics
A. Regoutz a,∗ , R.G. Egdell a , D.J. Morgan b , R.G. Palgrave c , H. Téllez d,1 , S.J. Skinner d ,
D.J. Payne d , G.W. Watson g , D.O. Scanlon e,f
a
Department of Chemistry, Inorganic Chemistry Laboratory, University of Oxford, South Parks Road, Oxford OX1 3QR, UK
Cardiff Catalysis Institute (CCI), School of Chemistry, Cardiff University, Park Place, Cardiff, CF10 3AT, UK
Department of Chemistry, University College London, 20 Gordon Street, London WC1H 0AJ, UK
d
Department of Materials, Imperial College London, Exhibition Road, London SW7 2AZ, UK
e
University College London, Kathleen Lonsdale Materials Chemistry, Department of Chemistry, 20 Gordon Street, London WC1H 0AJ, UK
f
Diamond Light Source Ltd., Diamond House, Harwell Science and Innovation Campus, Didcot, Oxfordshire OX11 0DE, UK
g
School of Chemistry and CRANN, Trinity College Dublin, Dublin 2, Ireland
b
c
a r t i c l e
i n f o
Article history:
Received 25 September 2014
Received in revised form 15 April 2015
Accepted 15 April 2015
Available online 22 April 2015
Keywords:
Transparent conducting oxide
Band bending
Dopant solubility
Ceramic
a b s t r a c t
The limit of solubility of Ga2 O3 in the cubic bixbyite In2 O3 phase was established by X-ray diffraction and
Raman spectroscopy to correspond to replacement of around 6% of In cations by Ga for samples prepared
at 1250 ◦ C. Density functional theory calculations suggest that Ga substitution should lead to widening of
the bulk bandgap, as expected from the much larger gap of Ga2 O3 as compared to In2 O3 . However both
diffuse reflectance spectroscopy and valence band X-ray photoemission reveal an apparent narrowing
of the gap with Ga doping. It is tentatively concluded that this anomaly arises from introduction of Ga+
surface lone pair states at the top of the valence band and structure at the top of the valence band in Gasegregated samples is assigned to these lone pair states. In addition photoemission reveals a broadening
of the valence band edge. Core X-ray photoemission spectra and low energy ion scattering spectroscopy
both reveal pronounced segregation of Ga to the ceramic surface, which may be linked to both relief of
strain in the bulk and the preferential occupation of surface sites by lone pair cations. Surprisingly Ga
segregation is not accompanied by the development of chemically shifted structure in Ga 2p core XPS
associated with Ga+ . However experiments on ion bombarded Ga2 O3 , where a shoulder at the top edge
of the valence band spectra provide a clear signature of Ga+ at the surface, show that the chemical shift
between Ga+ and Ga3+ is too small to be resolved in Ga 2p core level spectra. Thus the failure to observe
chemically shifted structure associated with Ga+ is not inconsistent with the proposal that band gap
narrowing is associated with lone pair states at surfaces and interfaces.
© 2015 Elsevier B.V. All rights reserved.
1. Introduction
Transparent conducting oxides (TCOs) combine the usually
orthogonal properties of optical transparency in the visible region
with a high electrical conductivity [1,2]. In2 O3 doped with Sn to
give so called indium tin oxide or ITO is a prototype TCO with
important applications as a transparent electrode in solar cells and
display devices. Despite the enormous technological importance
of ITO many of the basic physical properties of In2 O3 have aroused
controversy. For example the band gap of In2 O3 was for many years
∗ Corresponding author. Present address: Department of Materials, Imperial College London, 20 Gordon Street, London SW7 2AZ, UK. Tel.: +44 0 2075940976.
E-mail address: [email protected] (A. Regoutz).
1
Present address: Hydrogen Production Division, International Institute for
Carbon-Neutral Energy Research, Kyushu University, Fukuoka 819-0395, Japan.
http://dx.doi.org/10.1016/j.apsusc.2015.04.106
0169-4332/© 2015 Elsevier B.V. All rights reserved.
presumed to be 3.75 eV or thereabouts [3–5], even though a weak
optical onset at 2.69 eV in single crystals of In2 O3 was observed
in 1966 [6]. Recent work has established that In2 O3 has a direct
but dipole forbidden fundamental gap and the onset of strong optical absorption is shifted by almost 1 eV to higher energy [7]. The
lowest forbidden gap for In2 O3 therefore lies within the visible
region and bulk powder or ceramic samples have a bright yellow
or yellowish green colouration, whilst pronounced absorption in
the blue and green regions of the electromagnetic spectrum limits the possible thicknesses of transparent ITO thin films to around
200 nm–300 nm [8]. There has therefore been an increasing interest in In2 O3 -Ga2 O3 solid solutions as materials with potentially
enhanced visible region transparency.
Roy et al. [9] proposed that Ga2 O3 itself exists in at least
5 different polymorphs labelled ␣ to ␧, amongst which monoclinic ␤-Ga2 O3 is the most stable [9,10]. However the existence
of a cubic ␦ phase with the bixbyite structure has recently been
A. Regoutz et al. / Applied Surface Science 349 (2015) 970–982
called into question and an additional metastable ␬ phase with the
orthorhombic ␬-Al2 O3 structure has been identified [11]. The
optical onset in bulk single crystals of ␤-Ga2 O3 is at 260 nm,
corresponding to a band gap of around 4.8 eV [12]. Thus thin
films of ␤-Ga2 O3 are fully transparent across a window extending
well into the ultraviolet regime of the electromagnetic spectrum
[10,13–15]. On the other hand Ga2 O3 is intrinsically less dopable
than In2 O3 and it is not possible to achieve conductivities comparable with those of ITO. Ternary phases within the In Ga O
system were first investigated by Goldschmidt et al. [16] and later
by Schneider et al. [17], MacDonald et al. [18] and Shannon and
Prewitt [19]. Indium doping in Ga2 O3 is achieved with retention of the parent ␤ structure. A ternary compound GaInO3 with
the ␤-Ga2 O3 structure was identified in a number of the earlier
papers, although more recent investigation of the phase diagram
has established that the ␤-Ga2 O3 phase extends only to a limiting
composition of (Ga0.52 In0.48 )2 O3 in samples prepared by chemical
transport in sealed quartz tubes [20]. A stoichiometric hexagonal
Ga1.00 In1.00 O3 phase may however be prepared at high pressures
[19]. The solid solution range for ␤-Ga2 O3 was somewhat narrower in samples prepared by conventional solid state synthesis,
with limiting In doping levels of 44.0% at 1000 ◦ C and 41.4% at
1400 ◦ C [21]. On the In2 O3 side of the phase diagram the solubility limit for Ga in the cubic bixbyite phase is much more limited,
increasing from 4.8% in samples prepared at 1000 ◦ C to 10.5%
for samples prepared at 1400 ◦ C [21]. In the composition range
between the two doping limits bulk samples are biphasic, consisting of a mixture of cubic Ga-doped In2 O3 and monoclinic In-doped
␤-Ga2 O3 .
Ternary Ga In O oxide thin films (Gax In1−x )2 O3 have been
prepared by a wide range of techniques including direct current magnetron sputtering onto glass substrates (0 < x < 0.4)
[22]; metallo-organic chemical vapour deposition onto cubic
Y-stabilised ZrO2 (100) (0.1 < x < 0.9) [23] and MgO(1 0 0) substrates (0.63 < x < 1.00) [15]; O-plasma assisted molecular beam
epitaxy on ␤-Ga2 O3 (x = 1.00) [24] and Al2 O3 (0 0 0 1) substrates
(0.65 < x < 1.00) [25]; and by sol-gel spin-coating of Al2 O3 (0 0 0 1)
substrates (0 < x < 1) [26]. Most attention has been devoted to the
Ga-rich end of the phase diagram where it is clear that the band
gap decreases with increasing In doping [15,23,25,26]. However,
the variation of band gap with In doping level is not linear [26].
Rather less is known about the effects of Ga doping within the cubic
bixbyite phase of In2 O3 , especially with respect to the low energy
forbidden absorption onset.
The present study of the electronic and surface properties of Gadoped In2 O3 is partly motivated by our recent work on Sn-doped
rutile TiO2 where it is found that “alloying” wider gap SnO2 into
narrower gap TiO2 leads at low Sn doping levels to a small but significant reduction in the band gap [27,28]. So-called band bowing
effects of this sort (i.e. an initial reduction in band gap when a wider
gap material is alloyed into a narrower gap material at low doping
levels) are very well established in III–V semiconductor systems
[29] but have been less studied in oxides. In the few cases treated
by first principles calculations band bowing effects were predicted
to be small or non-existent [30,31]. On the other hand Chun et al.
[32] reported a small decrease in the band gap of In2 O3 nanowires
upon doping with 7% Ga.
The present work is also motivated by our recent interest in
growth of In2 O3 on cubic Y-stabilised ZrO2 substrates by oxygen
plasma assisted molecular beam epitaxy [33,34]. The face centred cubic fluorite structure of Y-stabilised ZrO2 (YSZ) belongs
to the space group Fm3̄m. For the minimum Y concentration
of around 17% required to stabilize a cubic phase, the lattice
parameter is given by a = 5.142 Å [35]. In2 O3 adopts a bodycentered cubic bixbyite structure belonging to space group Ia 3̄
and with lattice parameter ae = 10.1170 Å [36]. The structure is
971
derived from a (2 × 2 × 2) superstructure of fluorite with ordered
removal of ¼ of the oxygen ions. The mismatch m between 2as
for the YSZ and ae for the In2 O3 epilayer as defined by the
expression:
m = (ae − 2as )/2as
has a value m = −0.016 i.e. the mismatch is −1.6% [35]. Moreover
the two structures involve very similar cation arrays and thus YSZ
is an ideal substrate for epitaxial growth of In2 O3 films. However,
the epilayers are subject to significant tensile strain [37] which
leads in (1 1 1) oriented films to pronounced reduction in the band
gap as compared to bulk material [38,39]. On (0 0 1) and (1 1 0)
oriented substrates, strain coupled with large anisotropies in surface energies helps to drive formation of highly oriented island-like
nanostructures on (0 0 1) oriented substrates [34,40,41] or rod-like
nanostructures on (1 1 0) oriented substrates [41]. The anticipated
reduction in lattice parameter upon doping Ga onto In sites provides a simple means of increasing the lattice mismatch, thereby
enhancing strain at the interface and providing a way of tuning the
size of the nanostructures.
The organisation of the paper is as follows. The solubility limit
of Ga in the cubic bixbyite phase of In2 O3 is investigated by X-ray
diffraction and Raman spectroscopy. The band gaps in the doped
material are then studied by diffuse reflectance spectroscopy and Xray photoemission spectroscopy (XPS) and compared with results
from density functional theory calculations. This latter technique
provides evidence for pronounced segregation of Ga to the surface
of the doped material, and this segregation is further studied by
low energy ion scattering (LEIS). Finally, the mechanism for segregation is studied with reference to photoemission measurements
on Ga2 O3 itself.
2. Experimental
Samples of (Gax In1−x )2 O3 with (0.00 < x < 0.10) were prepared
by reaction between Ga2 O3 (Strem Chemicals 99.999%) and In2 O3
(Sigma–Aldrich 99.99%) supplied by Sigma–Aldrich Inc.). Stoichiometric mixtures of the oxide powders were ground in an agate
mortar for 30 min until the mixture was homogeneous. The powder mixtures were then pressed into pellets with a diameter of
1 cm and a thickness of approximately 1 mm under a loading
of 5 tons for 5 min. The samples were placed into recrystallised
alumina crucibles and sintered in a muffle furnace at 1250 ◦ C
for 72 h. The resulting pellets had a pale yellow colour. Reference pellets of In2 O3 and Ga2 O3 were prepared in a similar
fashion.
X-ray diffraction patterns were measured using a Phillips
PANanalytical X’Pert PRO diffractometer equipped with a Cu
K␣ X-ray source, which is monochromated with a curved Ge
(1 1 1) crystal. Raman spectra were recorded using a PerkinElmer
RamanStation 400F dispersive Raman spectrometer. This incorporates a near infrared (NIR) laser with a wavelength of 785 nm, which
delivers a maximum of 100 mW at the sample with a laser spot size
of 100 ␮m. It has a spectral range of 95–3500 cm−1 Raman shift with
a pixel resolution of 1 cm−1 across the entire range. A high sensitivity open electrode CCD detector with a 1024 × 256 pixel sensor is
used, which is air-cooled to −50 ◦ C during operation. The samples
were mounted on microscope slides and aligned with the help of a
CCD camera. Diffuse reflectance measurements were conducted on
a PerkinElmer Lambda 750S instrument incorporating double holographic grating monochromators and a 60 mm integrating sphere
and teflon reference samples.
Most of the high resolution X-ray photoemission spectra (XPS)
were recorded on a Scienta ESCA 300 spectrometer located in the
NCESS facility at Daresbury Laboratory, UK. This incorporated a
972
A. Regoutz et al. / Applied Surface Science 349 (2015) 970–982
rotating anode Al K␣ (h = 1486.6 eV) X-ray source, a seven crystal X-ray monochromator, a 300 mm mean radius spherical sector
electron energy analyser and a parallel electron detection system.
The X-ray source was run with 200 mA emission current and 14 kV
anode bias, while the analyzer operated at 150 eV pass energy
with 0.8 mm slits. The effective instrument resolution was about
400 meV as defined by the full width at half maximum height of
the derivative of the Fermi edge for a silver sample. Photoelectron
spectra were charge calibrated relative to a very weak but distinct
Fermi edge cut off observed for all Ga-doped In2 O3 samples. Further spectra were measured using a Kratos Axis Ultra DLD system
using a fixed anode monochromatic Al K˛ X-ray source operating
at 120 W and 125 mm mean radius spherical sector analyser. Data
were collected with pass energies of 160 eV for survey spectra, and
40 eV for the high resolution scans. The system was operated in the
hybrid mode, using a combination of magnetic immersion and electrostatic lenses with spectra acquired over an area approximately
300 × 700 ␮m2 . All spectra were taken with a 90◦ take off angle.
Magnetically confined charge compensation was used to minimize
sample charging and the resulting spectra were calibrated to the
C 1s line at 284.7 eV. A rastered Kratos minibeam-I ion source was
used for argon etching. The etch conditions were a 5 × 5 mm2 raster
area with the gun operating at 4 keV, 15 mA emission and an argon
pressure of 10−6 Torr. The experimental energy resolution in this
instrument was about 500 meV. A base pressure of ∼1 10−9 Torr
was maintained during collection of the spectra in both
systems.
Low energy ion scattering spectra were measured in a Qtac100
spectrometer (ION-TOF GmbH), located at Imperial College, London. The base pressure was 3 × 10−10 mbar, with a working
pressure in the range of 10−8 mbar. This instrument is fitted with
a double toroidal energy analyzer which collects backscattered
ions at an angle of 145◦ from all azimuth angles. Before the LEIS
measurements, sample surfaces were cleaned using Ar+ sputtering at 500 eV and 60◦ incidence angle, sputtered over an area of
2000 × 2000 ␮m2 . After cleaning, the surface was analysed with a
3 keV He+ beam at normal incidence to confirm the removal of any
surface contamination. For the actual analysis Ne+ ions at 5 keV normally incident on the surface were used. The Ne+ analysis beam was
rastered over the center of the sputtered area during depth profiling, using an analysis area of 1000 × 1000 ␮m2 to avoid crater-edge
effects. An electron flood gun was employed for charge compensation during the analysis.
4. Results and discussion
4.1. X-ray diffraction and Raman spectroscopy
The –2 X-ray diffraction profile of 1% Ga-doped In2 O3 is shown
in Fig. 1. All peaks can be indexed as belonging to the cubic bixbyite
phase of In2 O3 . Comparable data were obtained for doping levels up to 6%, beyond which extra reflections characteristic of new
phases were observed, as shown in Fig. 2 for a 9% Ga-doped sample. Cubic lattice parameters a for samples with doping levels up
to 7% Ga were obtained by Rietveld refinement of the diffraction
profiles using GSAS software. Beyond this doping level the Rietveld
refinements did not converge satisfactorily although there was no
indication of further shifts in the In2 O3 reflections to higher angle.
As shown in Fig. 3, incorporation of Ga into the bulk of the samples
leads to a contraction in lattice parameters, as expected from the
smaller size of Ga3+ (rGa3+ = 0.62 Å) as compared to In3+ (rIn3+ =
0.79 Å). This trend in a can be compared with both a Vegard plot
and with a line derived by interpolation between values of lattice
parameters for pure and Ga-doped In2 O3 from DFT calculations (see
Fig. 3). Vegard’s law states that a linear relationship between a of
an alloy solid solution and its constituents exists [52]. This usually
requires both constituents have the same crystal structure. However, this method can still be adopted for the present case by using
cell volumes rather than lattice parameters for In2 O3 and Ga2 O3 .
Fig. 1. –2 X-ray diffraction profile of 1% Ga-doped In2 O3 with all peaks indexed
to reflections characteristic of bixbyite phase of In2 O3 .
3. Computational methods
All our DFT calculations were performed using the VASP
code [42] with interactions between the cores (In:[Kr], Ga:[Ar]
and O:[He]) and the valence electrons described using the PAW
method [43]. The calculations were performed using the HSE06
hybrid functional as proposed by Krukau et al. [44] In the HSE06
approach, a value of exact nonlocal exchange, ˛, of 25%, and
screening parameter of ω = 0.11 bohr−1 are added to the PBE formalism. The HSE approach has been proven to result in structural
and band gap data in better agreement with experiment than
standard DFT functionals [45–48] and, importantly, to provide an
excellent description of the electronic structure of In2 O3 [49]. A
planewave cutoff of 400 eV and a k-point sampling of Gammacentered 3 × 3 × 3 for the primitive cell of In2 O3 was used, with
the structure deemed to be converged when the forces on all the
atoms were less than 0.01 eV Å−1 . The optical transition matrix
elements and the optical absorption spectrum were calculated
within the transversal approximation [50]. Within this methodology, the adsorption spectrum is summed over all direct VB
to CB transitions and therefore ignores indirect and intraband
adsorptions [51].
Fig. 2. Expanded –2 X-ray diffraction profile of 1%, 6% and 9% Ga-doped In2 O3 in
the region between (2 1 1) and (2 2 2) reflections showing appearance of reflections
due to an impurity phase (highlighted with *) for the 9% doped sample.
A. Regoutz et al. / Applied Surface Science 349 (2015) 970–982
In2 O3 (222)
(a)
0% 3% 5% 7%
30.4
30.6
973
30.8
31.0
2θ (degrees)
Cubic lattice parameter (A)
10.12
(b)
o
10.10
10.08
Fig. 4. Raman spectrum of polycrystalline In2 O3 with the peaks labelled following
the assignments of Garcia-Domene et al. [53].
10.06
10.04
0
2
4
6
8
Ga doping level (%)
Fig. 3. (a) XRD scans across (2 2 2) diffraction peak of 3%, 5% and 7% Ga-doped
In2 O3 in comparison with scan for undoped material. (b) Variation of cubic lattice
parameter of Ga-doped In2 O3 as a function of Ga doping level derived from Rietveld
refinements of X-ray powder patterns in the range of Ga doping levels up to 7%. The
two lines represent a Vegard plot (top, red) and values from DFT calculations (bottom, blue), the latter scaled such that the lattice parameter for In2 O3 itself coincides
with the experimental value (correction factor of 0.995). (For interpretation of the
references to color in this figure legend, the reader is referred to the web version of
this article.)
The decrease in a observed from X-ray diffraction lies in between
the variation derived from computed values and from Vegard’s law
applied as described above. This provides strong evidence that Ga is
incorporated onto In sites up to the solubility limit discussed above
and that the composition of the sintered ceramic material matches
that of the starting mixture.
The range of phase stability was also studied by Raman spectroscopy. The body centred cubic structure of In2 O3 belongs to the
space group Ia 3̄ with Z = 16 and the primitive cell contains 8 formula units or 40 atoms. The 120 vibrational modes span irreducible
representations in the Th factor group as follows:
= 4Ag + 4Eg + 14Tg + 5Au + 5Eu + 17Tu
The Au and Eu modes are silent. One of the Tu modes corresponds
to acoustic vibrations. The remaining 16 Tu modes are infrared
active whilst all 22 gerade modes are Raman active. It is possible
to identify 17 of these modes in the experimental Raman spectrum
(Fig. 4), with the frequencies in excellent agreement with those
identified by Garcia-Domene et al. [53]. Corresponding Raman
spectra of the Ga-doped samples are shown in Fig. 5. The vibrational
frequencies of all of the modes increase with increasing Ga doping
level up to around 7%, beyond which the frequencies remain constant. Fig. 6 shows representative data for the strong Ag mode close
to 307 cm−1 . At the same time additional peaks appear at 171 cm−1
and 429 cm−1 for doping levels beyond 6%, again signalling the
appearance of a second phase. The Raman spectrum of a reference sample of ␤-Ga2 O3 was measured for comparison with the
data for the Ga-doped In2 O3 samples and is shown in Fig. 7. The
Fig. 5. Raman spectra of polycrystalline Ga-doped In2 O3 samples for the doping
levels indicated. Peaks related to an impurity phase for higher Ga-doping levels are
highlighted with *.
monoclinic structure of this oxide belongs to the monoclinic space
group C2/m with two formula units per cell. The 30 vibrations
span the following irreducible representations of the group C2h
[54,55]:
= 10Ag + 5Bg + 5Au + 10Bu
The acoustic modes are Au + 2Bu leaving 12 IR active ungerade
modes and 15 Raman active gerade modes. As is evident in Fig. 7 the
strongest vibrational mode for ␤-Ga2 O3 is found close to 200 cm−1 .
All modes for ␤-Ga2 O3 could be identified with the frequencies
in excellent agreement with those identified by Dohy et al. [53].
The absence of an obvious peak close to this frequency in samples
beyond the doping limit suggests that the phase that co-exists with
the Ia3̄ In2 O3 phase is not the ␤-Ga2 O3 phase as has been suggested
elsewhere [20,21]. It is beyond the scope of this paper to identify
the new phase.
974
A. Regoutz et al. / Applied Surface Science 349 (2015) 970–982
309.0
-1
Raman shif t (cm )
308.5
308.0
307.5
307.0
0
2
4
6
8
Ga doping level (%)
10
Fig. 6. Variation in frequency of Ag vibrational mode near 307 cm−1 as a function of
Ga doping level.
Fig. 7. Raman spectrum of ␤-Ga2 O3 with the peaks labelled following the assignment of Dohy et al. [54,55].
4.2. Diffuse reflectance spectra.
Fig. 8(a) shows plots of (F(R)h)2 against photon energy h␯ for
pellets of In2 O3 and Ga2 O3 , where the Kubelka-Munk function F(R)
is derived from diffuse reflectance spectra. Linear fits to the absorption edge give band gaps of 2.95 eV for In2 O3 and 4.60 eV for Ga2 O3 .
The former value corresponds to the direct but dipole forbidden
absorption onset for In2 O3 , as discussed recently [7]. The larger gap
for Ga2 O3 is in agreement with values determined elsewhere [13].
Fig. 8(b) shows corresponding data for 1%, 6% and 9% Ga-doped
In2 O3 samples in comparison with an expanded scan of the absorption edge for In2 O3 itself. The band gap decreases with 1% and 6%
Ga doping but this trend does not extend to 9% doping. The gaps
derived from the complete series of samples studied in the present
work are shown in Fig. 8(c). It can be seen that the decrease in
band gap with doping does indeed level off at about 6%. Clearly
then the band gap at low Ga doping levels does not follow the
trend expected from interpolation between In2 O3 and Ga2 O3 . The
behaviour is not expected from previous work on In2-2x Ga2x O3 thin
films, although in most previous work relatively coarse steps in x
of about 0.1 have been investigated [23,25,26]. On the other hand
Fig. 8. (a) Plots of (F(R)h␯)2 against photon energy for Ga2 O3 and In2 O3 where the
Kubelka-Munk function F(R) is derived from diffuse reflectance spectra. (b) Corresponding plots on expanded scale for undoped In2 O3 and 1%, 6% and 9% Ga-doped
In2 O3 . (c) Variation of band gap with Ga doping level. The solid curve is for guidance
only.
band bowing effects at low Sn doping level in Ti1-x Snx O2 have been
recently observed leading to a narrowing of the gap when wider
gap SnO2 is alloyed with narrower gap TiO2 [27,28].
4.3. Apparent band gap variations from valence band XPS.
Valence band XPS for undoped In2 O3 and a 5% Ga-doped sample
are shown in Fig. 9. In each case binding energies may be referenced
to weak structure at the Fermi energy associated with occupation
of conduction band states. The figure also shows expanded scans
around the valence band edges and a plot of difference in spectral
intensity between the two samples. It can be seen that Ga doping
leads to a broadening of the valence band edge and to increased
intensity in the bandgap region above the valence band edge where
weak structure extends down to the conduction band feature. Similar features are found in spectra of the other doped samples.
A. Regoutz et al. / Applied Surface Science 349 (2015) 970–982
975
3.00
(a)
2.95
x15
2.90
Eg (eV)
x = 0.00
x = 0.05
10
(b)
8
6
4
2
0
2.85
2.80
2.75
x15
2.70
0
3
2
1
0
-1
(c)
3
2
1
0
-1
Binding energy (eV)
Fig. 9. (a) Overlaid valence band XPS of In2 O3 and (In0.95 Ga0.05 )2 O3 (b) Expanded
view of valence band edge and conduction band structure in an energy range extending to 3.5 eV binding energy. Note broadening of the valence band edge and the
appearance of enhanced intensity above the valence band edge as a consequence of
Ga doping. (c) Difference spectra between (In0.95 Ga0.05 )2 O3 and In2 O3 over the same
range as in (b).
2
4
6
8
Ga doping level (%)
10
Fig. 10. Variation in position of valence band edge in XPS of Ga-doped In2 O3 as
a function of Ga doping level. The error bars correspond to the step size used in
measurement of photoemission spectra. The data point show a wide scatter with a
Pearson correlation coefficient r = −0.59 for the linear regression shown in the figure.
the band gap from photoemission are susceptible to these effects,
it is nonetheless the case that changes in the position of the valence
band edge are qualitatively consistent with a decrease in band gap
with doping similar to that found in the diffuse reflectance spectra,
despite the large scatter in the data.
XPS provides important information about the change in shape
of the valence band edge and how this influences the magnitude
of the apparent bandgap. The new structure straddling the valence
band edge and extending into the bandgap contributes to broadening at the band edge. However other effects such as broadening
due to disorder in the Ga dopant sites and modification of phonon
broadening at the band edge may also contribute to the broadening and thus lead to a reduction in the bandgap determined
by a simple linear extrapolation. This will be discussed further in
Section 5.
4.4. HSE06 calculated fundamental and optical band gap
Using a simple linear extrapolation procedure it is possible to
identify the positions of the valence band onsets relative to the
Fermi energies discussed above. The results of this analysis for the
complete series of samples are shown in Fig. 10. The error bars
are rather large as the step size in the XPS experiments is 0.05 eV
and the scatter in the data points is large, with a Pearson correlation coefficient r = −0.59 in the linear regression. Nonetheless the
separation between the Fermi level and the position of the valence
band edge follows the same downward trend as in the optical measurements of the bandgap. A complication in interpreting these
changes in terms of a reduction in band gap that the position of
the band edge may be influenced by band bending at the surface
[56–58]. Downward band bending leads to carrier accumulation at
the surface over a length range which may be comparable with the
electron inelastic mean free path (20 Å in photoemission measurements) with partial occupation of the conduction band leading to
the valence band edge lying further below the Fermi level than in
the bulk. This effect is cancelled to some extent by experimental and
phonon broadening of the valence band edge shifting the valence
band edge to lower binding energies. Even if the absolute value of
We have tested the effects of incorporation of one Ga into 40
atom primitive cell of In2 O3 (i.e. replacing 6.25% of the In atoms)
on both the 8b and 24d sites. Ga is 0.096 eV more stable on the 8b
site, indicating that it will have a strong preference at room temperature for taking up this site. The HSE06 calculated fundamental
band gap and optical band gaps for In2 O3 and In1.875 Ga0.125 O3 are
shown in Fig. 11. It is immediately obvious that the incorporation
of Ga causes the fundamental band gap to increase slightly from
2.75 eV to 2.81 eV, which is not in agreement with the experimental
results. The optical band gap is also increased. Clearly, the incorporation of the much smaller Ga onto an In site is not sufficient to
break the symmetry of the disallowed VB to CB transitions, and
so does not hugely alter the optical properties. We have trialled
many symmetry distortions with the Ga ion placed off-centre from
both distinct In sites, and without fail the Ga returns to the In lattice
site. In summary then density functional theory calculations do not
predict a reduction in bandgap at low bulk Ga doping levels and we
must consider other possible reasons for apparent narrowing of the
bandgap.
976
A. Regoutz et al. / Applied Surface Science 349 (2015) 970–982
replacement of a host cation by a dopant cation of a different size
incurs a penalty of elastic strain energy which may be relieved to
some extent if the dopant cation segregates to the surface. Following Bataille et al. [60] the elastic strain energy per dopant Ga3+
cation may be estimated as:
Eelastic =
24KGa2 O3 GIn2 O3 rGa3+ r 3 3+
In
3KGa2 O3 rGa3+ + 3GIn2 O3 rIn3+
×
2
rGa3+
−1
rIn3+
where KGa2 O3 is the bulk modulus of Ga2 O3 , GIn2 O3 is the shear modulus of In2 O3 and the r are ionic radii. We take KGa2 O3 = 228 GPa [61]
and estimate GIn2 O3 from the bulk modulus KIn2 O3 = 194 GPa [62]
through the relationship:
GIn2 O3 = KIn2 O3 ×
3 (1 − 2)
2 (1 + )
where the Poisson ratio has been measured to be 0.31 [37,63].
In this way GIn2 O3 is found to be 84 GPa, leading to an elastic strain
energy of 0.185 eV/Ga ion. This is sufficient to drive pronounced
surface segregation.
Alternatively, segregation of post transition metal dopants may
be driven by a change in valence state at the surface, as found in
systems such as Sn-doped TiO2 [27,28], Sb-doped SnO2 [64,65] and
Sn-doped In2 O3 [66]. In these systems the dopant ion substitutes
into bulk sites in the group oxidation state N where the dopant
ion has an electron configuration 5s0 5p0 . However, the segregated
ions are in the N-2 oxidation state where the free ion would have
an electron configuration 5s2 5p0 . The non-centrosymmetric crystal
field necessarily present at surface sites allows mixing between 5s
and 5p states mediated by mutual interactions with O 2p valence
band states [67], thus lowering the internal electronic energy of the
segregated ions and providing a driving force for segregation. This
mechanism may drive segregation even when the ion in the group N
oxidation state in the bulk does not induce strain. However the two
mechanisms may obviously operate in tandem with strain leading
to surface enrichment and the necessarily asymmetric coordination
environment associated with surface sites facilitating reduction
from the N to the N-2 valence state.
Extending these ideas to Ga-doped In2 O3 segregation could be
understood in terms of occupation of surface sites by Ga+ ions. If
Fig. 11. HSE06 calculated optical absorption spectra for (In1−x Gax )2 O3 for x = 0.0000
and 0.0625. Dashed lines indicate fundamental band gap, full lines indicate optical
absorption.
4.5. Core level XPS and Ga surface segregation
In 4d
In 4s
Ga 3p
In 4p
Ga 3s
x 10
C 1s
10% Ga-doped In2 O3
Ga LMM
O 1s
In 3p
In 3s
O KLL
Ga 2p
In MNN
In 3d
Widescan XPS of 1% and 10% Ga-doped In2 O3 are shown in
Fig. 12. The strong Ga 2p core lines at high binding energy and the
associated Ga LMM Auger structure provide the most obvious signatures of Ga doping. Weaker Ga 3p core level peaks are observed
in the dip between the very broad In 4p and In 4s structure characteristic of In2 O3 itself, but the Ga 3d core line overlaps the In 4d
and O 2s peaks of the host material. Fig. 13(a) shows the Ga 2p
core level structure in greater detail, with intensities normalised
relative to the In 3d peaks of the host material. The apparent surface Ga content ratio obtained by correction of Ga 2p3/2 and In 3d5/2
intensities with the relevant atomic sensitivity factors (Ga 2p3/2 5.4;
In3d5/2 3.9) are shown in Fig. 13(b). There is pronounced surface
enrichment in Ga, especially at low doping levels.
There are two possible driving forces for Ga surface segregation in this system. As recognised by Friedel [59], substitutional
x 10
1% Ga-doped In2 O3
1200
1000
800
600
400
Binding energy (eV)
Fig. 12. Widescan XPS of 1% and 10% Ga-doped In2 O3 .
200
0
A. Regoutz et al. / Applied Surface Science 349 (2015) 970–982
977
10% Ga-doped In2 O3
Ga 2p3/2
1% Ga-doped In2 O3
1124
1122
1120
1118
1116
1114
Binding energy(eV)
Fig. 14. Curve fits to Ga 2p3/2 core level for 1% Ga-doped In2 O3 and 10% Ga-doped
In2 O3 .
Table 1
Parameters from curve fits to Ga 2p3/2 core lines of 1% and 10% Ga doped In2 O3 . The
curve fits were performed using pseudo-Voigt functions consisting of the product
of Gaussian and Lorentzian components.
Fig. 13. (a) Ga 2p core level spectra of Ga-doped In2 O3 for the doping levels indicated. The intensities have been normalised to the In 3d intensity. (b) Apparent
surface doping level as a function of bulk doping level derived from In 3d5/2 and Ga
2p3/2 XPS intensities. The solid line is for guidance only.
this situation does indeed pertain one might expect to observe a
chemical shift between surface and bulk substituted Ga ions with
a greater contribution from surface segregated ions at the lower
doping level. However as shown in Fig. 14 the Ga 2p3/2 core lines
for 1% and 10% Ga-doped In2 O3 are well fitted by single pseudoVoigt functions with characteristics as given in Table 1. It should be
noted that attempts to introduce further components in the curve
fits generally led to the second peak converging toward zero intensity or zero chemical shift from the first peak, while constrained
fits using two peaks with fixed separation invariably gave two
peaks of roughly equal intensity, which has no obvious physically
meaningful interpretation. We thus conclude that either there is no
measurable chemical shift between bulk Ga3+ and surface Ga+ ; or
that this mechanism does not pertain. We explore this topic further
in the next section.
4.6. Influence of ion bombardment on valence and core XPS of
Ga2 O3
Ga2 O3 was found to be highly insulating and satisfactory spectra could not be obtained in the Scienta spectrometer. However,
Sample
Binding energy (eV)
FWHM (eV)
% Lorentzian
1% Ga-doped In2 O3
10% Ga-doped In2 O3
1118.69
1118.56
1.73
1.65
71
71
meaningful data were obtained in the Kratos XPS system, which
employs a superior flood gun charge neutralisation system coupled into the magnetic immersion lens. In contrast to the doped
In2 O3 samples, no conduction band features were observed in the
experimental spectra and the Fermi level was apparently pinned
mid gap around 3 eV above the valence band edge. The valence
band X-ray photoemission spectrum of “as-presented” ␤-Ga2 O3 is
shown in Fig. 15, along with spectra obtained after two successive
periods of ion etching for 20 min as described in the experimental section. The spectrum of the sample before ion bombardment
contains two main features: a broad band at lower binding energy
apparently corresponding to bands I-III in the spectrum of In2 O3
and a sharper band labelled IV. By analogy with the spectrum of
In2 O3 discussed earlier, band IV must correspond to O 2p states
hybridised with Ga 4s states to give strongly bonding states at the
bottom of the valence band, whilst II and III relate to similar states
involving hybridisation between Ga 4p and O 2p states. The top
of the valence band will be dominated by states of more purely O
2p character but with some interaction with shallow core Ga 3d
states. After ion bombardment new intensity appears above the
top of the O 2p valence band. This gap state structure labelled
GS in Fig. 15 grows in intensity with more prolonged ion bombardment. The gap state structure is very similar to that found in
valence band XPS of vacuum annealed or otherwise reduced SnO2
[68] or indeed in the valence band spectrum of SnO [69]. Here the
structure is assigned to Sn2+ states arising from new occupied antibonding states of mixed O 2p–Sn 5s character above the top of the
valence band of SnO2 , which further hybridise with Sn 5p states to
give a directional lone pair [67,70,71]. By analogy with this earlier
work we take the appearance of the gap state structure to be the
definitive signature of reduction of Ga2 O3 under ion bombardment
by preferential sputtering of O to give reduced Ga+ . In agreement
with this idea the O 1s/Ga 2p3/2 intensity ratio falls from a value of
0.235 for the as-presented sample to 0.221 after 40 min of ion bombardment. Assuming that the topmost ionic layer contributes about
978
A. Regoutz et al. / Applied Surface Science 349 (2015) 970–982
GS
(d)
(c)
GS
GS
I-III
IV
(b)
(a)
12
10
8
6
4
2
0
-2
Binding energy (eV)
Fig. 15. (a) Valence band XPS of as-presented Ga2 O3 (b) valence band spectrum after
ion bombardment for 20 min (c) valence band spectrum after ion bombardment for
40 min (d) difference spectrum between (c) and (a) emphasising growth of gap-state
(GS) structure above the valence band edge as a consequence of ion bombardment.
The spectra have been aligned with the valence band edge of Ga2 O3 set at 0.0 eV.
Spectra measured in Kratos XPS system.
10% of the total XPS intensity, this change suggests that around 75%
of the Ga ions in this layer are reduced from Ga3+ to Ga+ . If (as is
probable) removal of oxygen extends beyond the first ionic layer,
the degree of chemical reduction within each layer would obviously
be lower.
The emergence of the gap state structure GS is not accompanied by growth of structure to low binding energy on the Ga 2p3/2
core peak as might be expected if a lower oxidation state were
present. In fact after the initial tranche of ion bombardment there
is small shift to higher binding energy with very little change in
the peak width (Table 2) which remains roughly constant with a
full width at half maximum height (FWHM) of just over 1.7 eV.
Dealing first with the linewidth, our peaks are slightly narrower
than in earlier studies of bulk Ga2 O3 powders and crystals [72,73]
and Ga2 O3 oxide films on GaN [74] and GaAs [75]: the large width
of this comparatively deep core level must include significant contributions from phonon broadening [76] and lifetime broadening
[77]. Thus phonon broadening for the Mg 1s core hole state in MgO
has been estimated as 0.84 eV [76], and even greater broadening
is expected for the comparably deep Ga 2p levels for Ga2 O3 owing
to the higher phonon frequencies. Lifetime broadening for the L3
core level in atomic Ga is around 0.76 eV [77]. The analyser and
X-ray source contributions to the core linewidth are negligible,
although experimental broadening in XPS of insulators can arise
from imperfections in charge neutralisation using a flood gun. In
agreement with the idea that this last effect may make a significant contribution to the linewidth it is noticeable that the Ga core
lines are narrower for Ga-doped In2 O3 which is much less effected
by sample charging. The curve fitting procedure used in the current manuscript used pseudo-Voigt functions where a product of
Gaussian and Lorentzian contributions is used. Further analysis of
core levels using true Voigt functions which are a convolution of
the Gaussian and Lorentzian components would be interesting but
is beyond the scope of the current paper. Turning to the apparent
shift, this is most likely due to graphitisation of the carbon contamination used as the energy reference. In the raw C 1s spectra a shift
to low binding energy is in fact observed (supplementary data) so
that if other peaks are referenced to a C 1s position assumed to
be constant, a spurious shift to high binding energy will be found.
Moreover the Ga 2p3/2 peak shape and the separation from the O
1s line remain essentially unchanged between the first and second cycles of ion bombardment (Fig. 16 and Table 2) even though
the structure linked to Ga+ in the valence band grows markedly
in intensity. This therefore suggests that there is no measurable
chemical shift between Ga3+ in Ga2 O3 and the Ga+ species responsible for the gap state structure. Similar analysis of the Ga 3p core
levels (where the measurements are less surface sensitive) lead to
the same conclusion. There was no significant change in the core
line as a result of ion bombardment.
The issue of chemical shifts between the group N valence states
and reduced N-2 states in post transition metal oxides has aroused
widespread discussion over many years. For example the NIST
data base [78] lists 18 values for the Sn 3d5/2 binding energy in
SnO2 ranging between 486.10 eV and 487.13 eV, with an average
of 486.6 eV. For SnO there are 5 values ranging between 485.60 eV
and 487.00 eV with an average of 486.4 eV i.e. there is a shift in
the average of only 0.2 eV. On the other hand Themlin et al. report
a chemical shift of 0.7 eV between Sn 3d5/2 binding energies in
Ga 2p3/2
(c)
Table 2
Parameters from curve fits to Ga 2p3/2 core lines of as-presented and ion bombarded
Ga2 O3 . The curve fits were performed using pseudo-Voigt functions consisting of the
product of Gaussian and Lorentzian components.
Sample
Binding
energy (eV)
O 1s–Ga
2p3/2 shift
(eV)
FWHM (eV)
% Lorentzian
Ga2 O3
as-presented
Ga2 O3 after
20 min ion
bombardment
Ga2 O3 after
40 min ion
bombardment
1118.10
587.30
1.74
68
1118.44
587.00
1.75
56
(b)
(a)
1124
1122
1120
1118
1116
1114
Binding energy
1118.57
587.10
1.77
57
Fig. 16. (a) Ga 2p3/2 core XPS of as-presented Ga2 O3 , (b) after 20 min ion bombardment, and (c) after 40 min ion bombardment. Spectra measured in Kratos XPS
system.
A. Regoutz et al. / Applied Surface Science 349 (2015) 970–982
In
Ga
10% Ga-doped In 2 O3
1% Ga-doped In 2 O3
1000
1500
2000
2500
3000
Energy (eV)
Fig. 17. Low energy ion scattering spectra of 1% and 9% Ga-doped In2 O3 measured
with 5 keV Ne ions after removal of initial carbon contamination layer.
10% Ga-doped In2 O3 LEIS
979
SnO2 and SnO [79]. However it is clear from valence band spectra in this paper that the shift is strongly influenced by a shift in
the Fermi level from just below the conduction band minimum in
SnO2 to just above the valence band edge in p-type SnO. In agreement with the idea that the shift here is not a chemical state effect
but reflects a change in Fermi energy it was found that the O 1s
level experienced almost the same shift as the Sn 3d5/2 level so
the separation between the two core lines in SnO2 (48.77 eV) is
essentially the same as in SnO (48.68 eV). This agrees with earlier
work by Wertheim [69]. Moreover where the N and N-2 oxidation states exist within the same oxide matrix chemical shifts are
always small or unobservable, as in the mixed valence oxide Pb3 O4
where despite the structurally characterised existence of Pb4+ and
Pb2+ unsplit and symmetric Pb 4f peaks are observed [80]. Similarly in reduced SnO2 showing the clear signature of Sn2+ in valence
band XPS sharp and symmetric Sn 3d peaks are observed [68]. The
absence of a chemical shift between N and N-2 oxidation states can
be understood in terms of a cancellation of the changes in the free
ion potentials and the changes in Madelung site potentials associated with the fact that whereas the N ions occupy 6-coordinate
centrosymmetric sites, the N-2 cations occupy asymmetric 3- or
4-coordinate sites [81,82]. We do note, however, that it has been
suggested that there is a pronounced chemical shift of around 1.5 eV
between Ga2 O3 and Ga2 O [83]. This large shift is completely out
of line with other “N versus N-2” systems as discussed above and
not consistent with our current results. Elsewhere a smaller shift
of 0.65 eV between Ga2 O and Ga2 O3 present in native oxide layers on GaAs has been reported [84]. Attempts to fit our own data
with two components constrained to be 0.65 eV apart did not lead
to any significant improvements in the curve fits as compared to
a single component fit. In particular these fits did not reveal an
increase in the intensity of the lower binding energy component
with increasing ion bombardment, as expected from the growth in
In
In
Ga
Ga
1555 s
55 s
0s
1400
1600
1800
2000
115 s
2200
2400
2600
1400
1600
1800
2000
2200
2400
2600
Energy (eV)
Fig. 18. Low energy ion scattering spectra of 10% Ga-doped In2 O3 measured with 5 keV Ne ions as a function of etching time. Left hand panel: spectra measured at 5 s intervals
between 0 s and 55 s. Right hand panel: spectra measured at 120 s intervals between 115 s and 1555 s.
980
A. Regoutz et al. / Applied Surface Science 349 (2015) 970–982
100
In
80
60
1% Ga -doped In2 O3
40
20
Ga
0
0
100
200
300
400
500
100
80
Atomic %
In
60
6% Ga -doped In2 O3
css of the samples were derived. Surface segregation was found
in all samples as presented with cis values bigger than the nominal bulk concentration with Ga fractions of 0.050, 0.221 and 0.290
for nominal doping levels of 0.010, 0.060 and 0.100. However the
steady state concentrations after prolonged etching had respective
values of less than 0.001, 0.014 and 0.031 that were significantly
less than nominal bulk values. Taken in conjunction with the evidence that the variation in lattice parameters is consistent with
the nominal bulk doping level, these observations suggest that in
the steady state established in a LEIS experiment the surface is
depleted of the lighter Ga atoms. Comparing XPS and LEIS data
it is apparent that the initial surface Ga concentration seen with
the latter technique is higher than with the former. LEIS is sensitive to atomic composition in the topmost ionic layer, whereas XPS
probes atomic concentrations over a length range determined by
the inelastic electron mean free path. Taken together these observations suggest that segregation is strongly enhanced in the topmost
ionic layer. However when dealing with rough polycrystalline samples of the sort studied here it is difficult to quantify these ideas
further.
40
Ga
20
5. Concluding discussion
0
0
100
200
300
400
500
400
500
100
80
In
60
10% Ga -doped In2O3
40
Ga
20
0
0
100
200
300
Etch time (seconds)
Fig. 19. Variation of In and Ga surface atomic composition for 1%, 6% and 10%
Ga-doped In2 O3 with etching time as determined by low energy ion scattering
spectroscopy.
intensity of the gap state feature in the valence band. Overall we
must therefore conclude that it is not possible to establish the existence of two different oxidation states using our Ga 2p3/2 core level
data.
4.7. Low energy ion scattering spectroscopy
Ion scattering spectra of 5 keV Ne ions for 1% and 10% Ga-doped
In2 O3 are shown in Fig. 17. Both spectra were taken after mild argon
ion bombardment for 10 s, which is sufficient to remove surface carbon contamination, followed by analyses with the probe ion beam.
The scattering peak associated with collision with the lighter Ga is
found to low energy of the peak associated with scattering from
In. Fig. 18 shows a series of spectra for 10% Ga-doped In2 O3 measured as a function of etching time. A progressive attenuation of
the Ga peak is observed. Fig. 19 shows the variation with etch time
in the relative surface Ga and In concentrations. From these data
the initial surface and steady state surface concentrations cis and
Both diffuse reflectance spectroscopy and X-ray photoemission
spectroscopy reveal an apparent narrowing of the band gap of In2 O3
with Ga doping up to the solubility of limit of Ga in the cubic
bixbyite phase of In2 O3 . Taken at face value the results suggest an
unusually large band bowing effect. However, state of the art hybrid
density functional theory calculations on bulk Ga-doped In2 O3 fail
to reproduce a narrowing of the band gap with doping and instead
predict a monotonic increase of gap with increasing Ga content, as
expected from the respective band gaps of the parent compounds.
One cannot rule out the possibility that DFT fails in some way to capture the essential physics of this system. However, the observation
that there is pronounced Ga surface segregation leads us to consider whether the effects observed could be attributed to surface
segregated Ga because experiments on Ga2 O3 reduced by in situ ion
bombardment demonstrate that surface Ga+ introduces new “lone
pair” electronic states above the top of the valence band in XPS. Similar but weaker structure is found in valence band XPS of Ga-doped
In2 O3 . The lower intensity of the structure in the bandgap seen for
5% Ga doped In2 O3 in Fig. 9 as compared to that of ion bombarded
Ga2 O3 seen in Fig. 15 is to be expected as we estimate that 75%
of the surface Ga ions in the latter are reduced to Ga+ . The surface
Ga/In percentage as probed by XPS for the doped sample is only
around 15% and we cannot assume that all surface segregated Ga is
reduced to the univalent state as size mismatch without reduction
can also drive segregation. No direct evidence for Ga+ can be found
in core XPS of Ga-segregated In2 O3 . However, the core lineshape
for pure Ga2 O3 changes little after ion bombardment, despite the
change in surface stoichiometry and the emergence of the bandgap
structure characteristic of Ga+ . This leads us to conclude that the
chemical shift between Ga+ and Ga3+ in this oxide system is too
small to allow reliable curve fitting to Ga+ and Ga3+ components
in the Ga 2p core spectra and so the absence of chemically shifted
components for the doped In2 O3 samples does not rule out the
hypothesis of surface Ga+ . Structure associated with Ga+ overlapping the valence band edge contributes to broadening of the edge.
However other effects such as disorder associated with random
occupation of In sites by Ga and enhanced phonon broadening at
the band edge may also contribute to the broadening and associated apparent reduction in bandgap. These effects are not included
in the periodic calculations but phonon broadening in particular
is important for materials such as oxides with high phonon frequencies and provides the mechanism for the isotope changes in
A. Regoutz et al. / Applied Surface Science 349 (2015) 970–982
bandgaps that have now been characterised recently for a number of systems [85]. The local Ga O stretching frequencies will be
higher than In-O frequencies owing to the smaller size of Ga as
compared with In. Further theoretical work is needed to explore
these issues in more detail.
Acknowledgements
The work at TCD was supported by SFI through the PI program
(06/IN.I/I92). Calculations at TCD were performed on the Lonsdale
and Kelvin clusters as maintained by TCHPC, and the Stokes and
Fionn clusters as maintained by ICHEC. The UCL/Diamond work
presented here made use of the UCL Legion HPC Facility, the IRIDIS
cluster provided by the EPSRC funded Centre for Innovation (Grants
No. EP/K000144/1 and No. EP/K000136/1), and the ARCHER supercomputer through support by the UK’s HPC Materials Chemistry
Consortium, which is funded by EPSRC Grant No. EP/L000202.
RGP acknowledges EPSRC Grant EP/K014099/1. DJP, DOS and RGP
acknowledge the Materials Design Network. D.J.P. acknowledges
support from the Royal Society (UF100105). HT acknowledges
financial support from the Marie Curie Intra European Fellowship within the 7th European Community Framework Programme
the European Union (PIEF-GA-2010-274999). X-ray photoelectron
spectroscopy was provided through the EPSRC “Access to Research
Equipment Initiative: Cardiff XPS” (Grant No. EP/F019823/1). AR
thanks Trinity College Oxford for financial support.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.apsusc.2015.04.
106
References
[1] D.S. Ginley, C. Bright, MRS Bull. 25 (2000) 15.
[2] D.S. Ginley, H. Hosono, D.C. Paine (Eds.), Handbook of Transparent Conductors,
Springer, New York, 2010.
[3] I. Hamberg, C.G. Granqvist, K.F. Berggren, B.E. Sernelius, L. Engstrom, Phys. Rev.
B 30 (1984) 3240.
[4] P. Erhart, A. Klein, R.G. Egdell, K. Albe, Phys. Rev. B 75 (2007) 153205.
[5] Y. Gassenbauer, R. Schafranek, A. Klein, S. Zafeiratos, M. Havecker, A. KnopGericke, R. Schlogl, Phys. Rev. B 73 (2006) 245312.
[6] R.L. Weiher, R.P. Ley, J. Appl. Phys. 37 (1966) 299.
[7] A. Walsh, et al., Phys. Rev. Lett. 100 (2008) 167402.
[8] R.J. Cava, J.M. Phillips, J. Kwo, G.A. Thomas, R.B. Vandover, S.A. Carter, J.J.
Krajewski, W.F. Peck, J.H. Marshall, D.H. Rapkine, Appl. Phys. Lett. 64 (1994)
2071.
[9] R. Roy, V.G. Hill, E.F. Osborn, J. Amer. Chem. Soc. 74 (1952) 719.
[10] L.Y. Kong, J. Ma, C.N. Luan, W. Mi, Y. Lv, Thin Solid Films 520 (2012) 4270.
[11] H.Y. Playford, A.C. Hannon, E.R. Barney, R.I. Walton, Chem. Eur. J. 19 (2013)
2803.
[12] Z. Galazka, R. Uecker, K. Irmscher, M. Albrecht, D. Klimm, M. Pietsch, M.
Brutzam, R. Bertram, S. Ganschow, R. Fornari, Cryst. Res. Technol. 45 (2010)
1229.
[13] M. Orita, H. Ohta, M. Hirano, H. Hosono, Appl. Phys. Lett. 77 (2000) 4166.
[14] M. Orita, H. Hiramatsu, H. Ohta, M. Hirano, H. Hosono, Thin Solid Films 411
(2002) 134.
[15] L.Y. Kong, J. Ma, C.N. Luan, Z. Zhu, J. Solid State Chem. 184 (2011) 1946.
[16] V.M. Goldschmidt, T. Barth, G. Lunde, Skr. Norske Vidensk. Akad. Maturv. Kl. 7
(1925).
[17] S.J. Schneider, R.S. Roth, J.L. Waring, J. Res. Natl. Stand. 65A (1961) 345.
[18] J. Macdonald., J.A. Gard, F.P. Glasser, J. Inorg. Nucl. Chem. 29 (1967) 661.
[19] R.D. Shannon, C.T. Prewitt, J. Inorg. Nucl. Chem. 30 (1968) 1389.
[20] G. Patzke, M. Binnewies, Solid State Sci. 2 (2000) 689.
[21] D.D. Edwards, P.E. Folkins, T.O. Mason, J. Am. Ceram. Soc. 80 (1997) 253.
[22] T. Minami, Y. Takeda, T. Kakumu, S. Takata, I. Fukuda, J. Vacuum Sci. Technol. A
15 (1997) 958.
[23] L.Y. Kong, J. Ma, F. Yang, C.N. Luan, Z. Zhu, J. Alloys Compd. 499 (2010) 75.
[24] T. Oshima, N. Arai, N. Suzuki, S. Ohira, S. Fujita, Thin Solid Films 516 (2008)
5768.
[25] T. Oshima, S. Fujita, in: Y. Hirayama, T. Sogawa (Eds.), Physica Status Solidi C –
Current Topics in Solid State Physics, 5, 2008, p. 3113.
[26] Y. Kokubun, T. Abe, S. Nakagomi, Physica Status Solidi A 207 (2010) 1741.
981
[27] M.H. Harunsani, F.E. Oropeza, R.G. Palgrave, R.G. Egdell, Chem. Mater. 22 (2010)
1551.
[28] F.E. Oropeza, B. Davies, R.G. Palgrave, R.G. Egdell, Phys. Chem. Chem. Phys. 13
(2011) 7882.
[29] I. Vurgaftman, J.R. Meyer, L.R. Ram-Mohan, J. Appl. Phys. 89 (2001) 5815.
[30] X.F. Fan, H.D. Sun, Z.X. Shen, J.L. Kuo, Y.M. Lu, J. Phys. Condens. Matter 20 (2008)
235221.
[31] E.P. Nakhmedov, E. Nadimi, M. Bouhassoune, C. Radehaus, K. Wieczorek, Phys.
Rev. B 75 (2007) 115204.
[32] H.J. Chun, Y.S. Choi, S.Y. Bae, H.C. Choi, J. Park, Appl. Phys. Lett. 85 (2004)
461.
[33] A. Bourlange, D.J. Payne, R.G. Palgrave, J.S. Foord, R.G. Egdell, R.M.J.
Jacobs, A. Schertel, J.L. Hutchison, P.J. Dobson, Thin Solid Films 517 (2009)
4286.
[34] A. Bourlange, D.J. Payne, R.M.J. Jacobs, R.G. Egdell, J.S. Foord, A. Schertel, P.J.
Dobson, J.L. Hutchison, Chem. Mater. 20 (2008) 4551.
[35] R.A. Cowley, A. Bourlange, J.L. Hutchison, K.H.L. Zhang, A.M. Korsunsky, R.G.
Egdell, Phys. Rev. B 82 (2010) 165312.
[36] M. Marezio, Acta Cryst. 20 (1966) 723.
[37] K.H.L. Zhang, A. Regoutz, R.G. Palgrave, D.J. Payne, R.G. Egdell, A. Walsh, S.P.
Collins, D. Wermeille, R.A. Cowley, Phys. Rev. B 84 (2011) 233301.
[38] A. Walsh, C.R.A. Catlow, K.H.L. Zhang, R.G. Egdell, Phys. Rev. B 83 (2011)
161202.
[39] K.H.L. Zhang, V.K. Lazarov, T.D. Veal, F.E. Oropeza, C.F. McConville, R.G. Egdell,
A. Walsh, J. Phys. Condens. Matter 23 (2011) 334211.
[40] K.H.L. Zhang, A. Bourlange, R.G. Egdell, S.P. Collins, R.J. Bean, I.K. Robinson, R.A.
Cowley, ACS Nano 6 (2012) 6717.
[41] K.H.L. Zhang, A. Walsh, C.R.A. Catlow, V.K. Lazarov, R.G. Egdell, Nano Lett. 10
(2010) 3740.
[42] G. Kresse, J. Hafner, Phys. Rev. B 49 (1994) 14251.
[43] G. Kresse, D. Joubert, Phys. Rev. B 59 (1999) 1758.
[44] A.V. Krukau, O.A. Vydrov, A.F. Izmaylov, G.E. Scuseria, J. Chem. Phys. 125 (2006)
224106.
[45] M. Burbano, D.O. Scanlon, G.W. Watson, J. Am. Chem. Soc. 133 (2011)
15065.
[46] D.O. Scanlon, A.B. Kehoe, G.W. Watson, M.O. Jones, W.I.F. David, D.J. Payne, R.G.
Egdell, P.P. Edwards, A. Walsh, Phys. Rev. Lett. 107 (2011) 246402.
[47] J.P. Allen, D.O. Scanlon, G.W. Watson, Phys. Rev. B 81 (2010) 161103 (R).
[48] D.O. Scanlon, G.W. Watson, J. Mater. Chem. 21 (2011) 3655.
[49] P. Agoston, K. Albe, R.M. Nieminen, M.J. Puska, Phys. Rev. Lett. 103 (2009)
245501.
[50] M. Gajdos, K. Hummer, G. Kresse, J. Furthmuller, F. Bechstedt, Phys. Rev. B 73
(2006) 045112.
[51] B. Adolph, J. Furthmuller, F. Bechstedt, Phys. Rev. B 63 (2001) 125108.
[52] A.R. Denton, N.W. Ashcroft, Phys. Rev. A 43 (1991) 3161.
[53] B. Garcia-Domene, et al., J. Appl. Phys. 112 (2012) 123511.
[54] D. Dohy, G. Lucazeau, J. Mol. Struct. 79 (1982) 419.
[55] D. Dohy, G. Lucazeau, A. Revcolevschi, J. Solid State Chem. 45 (1982) 180.
[56] K.H.L. Zhang, R.G. Egdell, F. Offi, S. Iacobucci, L. Petaccia, S. Gorovikov, P.D.C.
King, Phys. Rev. Lett. 110 (2013) 056803.
[57] P.D.C. King, T.D. Veal, D.J. Payne, A. Bourlange, R.G. Egdell, C.F. McConville, Phys.
Rev. Lett. 101 (2008) 116808.
[58] P.D.C. King, et al., Phys. Rev. B 79 (2009) 205211.
[59] J. Friedel, Adv. Phys. 3 (1954) 446.
[60] C.C. Battaile, R. Najafabadi, D.J. Srolovitz, J. Am. Ceram. Soc. 78 (1995)
3195.
[61] H. Wang, Y. He, W. Chen, Y.W. Zeng, K. Stahl, T. Kikegawa, J.Z. Jiang, J. Appl.
Phys. 107 (2010) 033520.
[62] A. Walsh, C.R.A. Catlow, A.A. Sokol, S.M. Woodley, Chem. Mater. 21 (2009)
4962.
[63] A. Regoutz, K.H.L. Zhang, R.G. Egdell, D. Wermeille, R.A. Cowley, J. Mater. Res.
27 (2012) 2447.
[64] P.A. Cox, R.G. Egdell, C. Harding, A.F. Orchard, W.R. Patterson, P.J. Tavener, Solid
State Commun. 44 (1982) 837.
[65] P.A. Cox, R.G. Egdell, C. Harding, W.R. Patterson, P.J. Tavener, Surf. Sci. 123 (1982)
179.
[66] P.A. Cox, W.R. Flavell, R.G. Egdell, J. Solid State Chem. 68 (1987) 340.
[67] A. Walsh, D.J. Payne, R.G. Egdell, G.W. Watson, Chem. Soc. Rev. 40 (2011)
4455.
[68] R.G. Egdell, J. Rebane, T.J. Walker, D.S.L. Law, Phys. Rev. B 59 (1999) 1792.
[69] C.L. Lau, G.K. Wertheim, J. Vac. Sci. Technol. 15 (1978) 622.
[70] A. Walsh, G.W. Watson, Phys. Rev. B 70 (2004) 235114.
[71] A. Walsh, Appl. Phys. Lett. 98 (2011) 212110.
[72] C.C. Surdu-Bob, S.O. Saied, J.L. Sullivan, Appl. Surf. Sci. 183 (2001) 126.
[73] T. Kamimura, K. Sasaki, M.H. Wong, D. Krishnamurthy, A. Kuramata, T. Masui,
S. Yamakoshi, M. Higashiwaki, Appl. Phys. Lett. 104 (2012) 192104.
[74] W. Priyantha, G. Radhakrishnan, R. Droopad, M. Passlack, J. Cryst. Growth 323
(2011) 103.
[75] S.D. Wolter, B.P. Luther, D.L. Waltemeyer, C. Onneby, S.E. Mohney, R.J. Molnar,
Appl. Phys. Lett. 70 (1997) 2156.
[76] C.J. Nellin, P.S. Bagus, M.A. Brown, M. Sterrer, H.J. Freund, Angew. Chem. Int. Ed.
50 (2011) 10174.
[77] M.O. Krause, J.H. Oliver, J. Phys. Chem. Ref. Data 8 (1979) 329.
[78] NIST X-Ray Photoelectron Spectroscopy Database, NIST Standard Reference
Database 20, Version 4.1. Data compiled and evaluated by A. V. Naumkin,
A. Kraut-Vass, S. W. Gaarenstroom and C. J. Powell. Measurement Services
982
A. Regoutz et al. / Applied Surface Science 349 (2015) 970–982
Division of the National Institute of Standards and Technology Materials Measurement Laboratory, Gaithersberg, MD 20899-2600, USA, 2010.
[79] J.M. Themlin, M. Chtaib, L. Henrardm, P. Lambin, J. Darville, J-M. Gilles, Phys.
Rev. B 46 (1992) 2460.
[80] J.M. Thomas, M.J. Tricker, J. Chem. Soc. Faraday Trans. II 71 (1974) 329.
[81] J.Q. Broughton, P.S. Bagus, J. Electron. Spectroscop. Relat. Phenom. 20 (1980)
261.
[82] D.E. Parry, J. Chem. Soc. Faraday Trans. II 71 (1975) 337.
[83] R. Carli, C.L. Bianchi, Appl. Surf. Sci. 74 (1994) 99.
[84] C.L. Hinkle, M. Milojevic, B. Brennan, A.M. Sonnet, F.S. Aguirre-Tostado,
G.J. Hughes, E.M. Vogel, R.M. Wallace, Appl. Phys. Lett. 94 (2009)
162101.
[85] J. Bhosale, A.K. Ramdas, A. Burger, A.A. Munoz, A.H. Romero, M. Cardona, R.
Lauck, R.K. Kremer, Phys. Rev. B 86 (2012) 195208.