ARTICLE IN PRESS

Journal of Crystal Growth 310 (2008) 51985203

Contents lists available at ScienceDirect

Journal of Crystal Growth

journal homepage: www.elsevier.com/locate/jcrysgro

MOCVD of thin lm photovoltaic solar cellsNext-generation

production technology?
S.J.C. Irvine , V. Barrioz, D. Lamb, E.W. Jones, R.L. Rowlands-Jones
Centre for Solar Energy Research, OpTIC Technium, St. Asaph Business Park, North Wales, UK

a r t i c l e in f o a b s t r a c t

Available online 13 August 2008 This paper will review the chalcogenide thin lm photovoltaic (PV) solar cells, based on cadmium
PACS: telluride (CdTe) and copper indium diselenide (CIS) and discuss the potential for metalorganic chemical
71.55.Gs vapour deposition (MOCVD) to enable more advanced devices in the second generation of CdTe module
73.40.Ei production. The current generation of production methods is based on physical vapour deposition (PVD)
73.61.Ga or close-spaced sublimation (CSS). This paper concentrates on the less well-known topic of MOCVD of
78.66.Jg thin lm chalcogenide cells, and in particular that of CdTe. Efcient CdTe PV solar cells (410% AM1.5)
have been demonstrated from deposition of the CdS, CdTe and CdCl2 lms in a single MOCVD chamber.
Keywords: The CdTe layer was doped with As and an additional high As concentration CdTe layer provides effective
A3. Metalorganic chemical vapour low resistance contacting without the need for wet etching the surface. The high level of exibility in
deposition
using MOCVD has been demonstrated where the CdS window layer has been alloyed with Zn to improve
B1. Cadmium compounds
the blue response of the PV device and improve AM1.5 efciency to 13.3%.
B2. Semiconducting IIVI materials
B3. Solar cells & 2008 Elsevier B.V. All rights reserved.

1. Introduction
The urgent concern over climate change is rapidly transform-
ing photovoltaic (PV) solar energy from a niche technology to one
which is entering the main stream of renewable energy genera-
tion. The market growth has been sustained at 3040% per annum
since the 1990s and last year grew by a staggering 62% [13]. Over
90% of this production is from crystalline silicon products, using
wafers of single-crystal or multi-crystalline silicon. The rest
comprises of various thin lm silicon materials, including
amorphous silicon, cadmium telluride (CdTe) and the copper
indium gallium diselenide (CIGS) family of materials. Over the
past 4 years the latter two classes of materials (the chalcogenides)
have started to take their place in volume manufacture and
increased market share. PV solar module production prices are
currently a factor of three too high to be cost competitive with
fossil fuel generation of electricity. Thin lm PV module
manufacture is inherently lower cost than for crystalline silicon
but the smaller production volume for thin lm PV has kept the
price relatively high. Pearce [1] predicts that multi-megawatt per
annum production scale will decrease the cost to be price
competitive, without the current subsidies operated in the major
PV adopter countries. Until recently, thin lm PV production
factories had outputs in the region of only 10 MW per annum.
 Corresponding author.
E-mail address: stuart.irvine@optictechnium.com (S.J.C. Irvine).
However, a new generation of thin lm plants are being built that
will have the order of 100 MW annual output. One of these
manufacturers, First Solar, is manufacturing CdTe PV modules
based on close-spaced sublimation (CSS) of the CdTe absorber.
This paper will report on recent research results using metalor-
ganic chemical vapour deposition (MOCVD) as an alternative
deposition method that could offer greater process exibility for
advanced structures but new MOCVD technology would be
required to enable large volume production of PV modules.
2. CdTe and CIS PV solar cells
In this section the application of MOCVD for laboratory scale
chalcogenide solar cells is reviewed. CdTe thin lm solar cells are
typically deposited onto transparent conducting oxide (TCO)-
coated glass substrates. This is known as the superstrate
conguration approach. Alternatively, the CIS family of cells is
deposited onto Mo-coated glass substrates, known as the sub-
strate conguration approach. The main difference between these
approaches are that for the superstrate the solar cell will receive
the solar radiation through the glass substrate whereas for the
substrate approach the solar radiation is received from the top of
the thin lm structure. These structures are represented schema-
tically in Fig. 1. In both cases the substrate material is polycrystal-
line or amorphous resulting in polycrystalline PV layers. A typical
layer structure for CdTe is shown in Fig. 1 and comprises of the
TCO for the front contact, a CdS n-type (window) layer and a CdTe

0022-0248/$ - see front matter & 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jcrysgro.2008.07.121
 ARTICLE IN PRESS
S.J.C. Irvine et al. / Journal of Crystal Growth 310 (2008) 51985203
Back contact
CdTe layer Front contact
CdS layer
TCO layer
Glass
substrate
Direction of solar irradiation for
superstrate configuration
Direction of solar irradiation for
substrate configuration
TCO layer
CdS layer
CIS layer
Mo layer
Glass
substrate
Fig. 1. Schematic of the superstrate conguration for CdTe and substrate
conguration for CIS PV cells: (a) CdTe solar cell and (b) CuInSe2 (CIS) solar cell.
p-type (absorber) layer. Current module manufacture involves
each of these being a separate deposition process followed by a
chlorine treatment and anneal to obtain efcient (410% AM1.5)
cells [4].
Early work by Nouhi et al. [5] on MOCVD deposition of CdTe
solar cells achieved 9.4% for a 0.079 cm2 contact area. The CdTe
layer was deposited onto commercial CdS/SnO2/glass super-
strates. In 1994, again using a MOCVD Te-rich CdTe, Chou and
Rohatgi [6] achieved a device efciency of 11.5% in a study on Cu
diffusion for CdTe of different grain sizes. Both results used Au and
Cu back contacts but with no intentional doping of the CdTe layer.
However, these elements have been shown to diffuse rapidly into
the device structure and can cause poor long-term stability.
Ferekides et al. [7] used MOCVD to deposit the CdTe layer and
studied extrinsic doping and in situ homojunction formation.
The difculty of p-type doping CdTe using As or Sb was described.
Again high efciency around 10% (AM1.5) was achieved, but
without being able to produce the entire structure was unlikely to
provide any obvious advantage over cheaper deposition methods.
Success around the same time with the CSS method using the
same CdS/SnO2:F/glass superstrates made this the preferred
method, in the absence of any clear advantage for the more
complex MOCVD. In 1992, Chu et al. [8] reported on the then
world record performance of 14.6% AM1.5 using CSS and that has
set the scene for the next decade and a half. This record was
broken in 2001 by Wu et al. [9], increasing this maximum
efciency to 16.5%. This has remained the record for the past
7 years and is well below the maximum theoretical efciency of
27% (AM1.5) for these cells. The reasons for this are the
summation of a number of factors, including the absence of
independent doping control in the CdS and CdTe layers, blue
absorption by the CdS window layer, polycrystalline structure and
grain boundary passivation, difculty in forming ohmic contacts
to p-type CdTe and material non-uniformity.
Attempts of growing the entire device structure by MOCVD
have been few and far between, although this is the logical
approach when using MOCVD. Early work reported by Berrigan
5199
et al. in 1998 [10] was the rst to report on an attempt at growing
the CdS/CdTe structure by MOCVD in a single process and doping
both CdS and CdTe layers. The paper highlighted the problems in
achieving sufcient control over the materials to achieve efcient
devices and only achieved conversion efciencies of around 1%.
More recently, Barrioz et al. [11] carried out an empirical
investigation into improving device performance of MOCVD-
grown devices with efciencies reaching 6%. The complexity of
making devices in polycrystalline materials requires far greater
understanding of issues not encountered in the planar counter-
parts. This paper reports on the progress made since the early
attempts at growing the entire structure by MOCVD to achieve
efcient devices and pave the way for device structures not
readily achievable by the conventional CSS process.
The approach in making CIS devices and the related quaternary
(Cu(In,Ga)Se2) (CIGS) has been different to that of CdTe. However,
it has tracked the rise in performance of polycrystalline chalco-
genide PV devices and has overtaken CdTe in maximum
performance [12]. The synthesis of the absorber layer is normally
a multi-step process where the metal components are deposited
in the rst step, by vacuum evaporation, followed by selenisation
with H2Se to create the chalcopyrite phase. Various MOCVD
processes have been used to synthesis the absorber layer and
some have followed the multi-step approach. Lee et al. [13]
deposited CIS lms using a two-step MOCVD process, where a Cu
lm was deposited in the rst step followed by a combined InSe
precursor for the second step. A challenge for MOCVD deposition
is the low volatility of Cu precursors, which normally requires
working at source temperatures above 80 1C. Sagnes et al. [14]
investigated precursors for lower temperature deposition and
used (hfa)2Cu as the rst step in synthesising CIS lms. A two-step
process was also used by Gallon et al. [15] where In2Se3 lms were
deposited at 350 1C in the rst step followed by Cu deposition at
the same temperature. The nal step was a heat treatment in a Se
vapour to obtain a single-phase lm.
Recently, work has been reported on the formation of the
quaternary alloy CIGS by Choi and Lee [16] using a three-step
process. The rst step deposited InSe onto a Mo lm substrate
using a single source precursor ([(Me)2In(m-SeMe)]2), followed by
Cu deposition. The third step entailed the deposition of alternate
layers of GaSe and InSe where the period of each determined the
relative alloy composition in the annealed structure. Single-phase
CIGS was obtained by this method but no device results have been
reported. The remainder of this paper will solely concentrate on
the recent developments in MOCVD of CdTe PV solar cells.
3. High-efciency CdTe/CdS solar cells by MOCVD
The research carried out by the authors has concentrated on
obtaining high-efciency PV solar cells (dened as having AM1.5
efciency 410%) using a single MOCVD chamber. The aim has
been to both simplify the PV cell deposition process and to obtain
greater materials control for more advanced devices. The device
simplication can be seen in Table 1, where a comparison is made
between the conventional approach and the MOCVD approach.
Note that the CdCl2 step appears in both, although earlier
attempts at avoiding this step were unsuccessful in yielding the
required device efciency.
3.1. CdS and CdZnS window layer
Early work on MOCVD of CdS window layers onto indium tin
oxide (ITO)-coated glass substrates indicated that low-temperature
deposition was required for good lm transmission [17]. Using the
precursors dimethylcadmium (DMCd) and ditertiarybutylsulphide
 ARTICLE IN PRESS

5200 S.J.C. Irvine et al. / Journal of Crystal Growth 310 (2008) 51985203

Table 1
A comparison of process steps between conventional processing of CdS/CdTe solar 18
cells and those produced by MOCVD in a single chamber 16
14

Jsc (mA.cm-2)
Conventional processing MOCVD
12
Deposit CdS from chemical bath Deposit CdS 10
Anneal Deposit As-doped CdTe 8
Deposit CdTe by CSS Deposit high As cap CdTe layer 6
Deposit CdCl2 layer Deposit CdCl2 layer
4
Anneal Anneal
Chemical etch Deposit metal back contact 2
Deposit metal back contact 0
Anneal 1.E+15 1.E+16 1.E+17 1.E+18 1.E+19 1.E+20
As concentration (atoms cm-3)

(DTBS) it was found that the best lm properties were obtained at 6

290 1C. More recent work has obtained good transmission over the

Device efficiency (%)

range of growth temperatures from 300 to 350 1C. Although, Cl
doping has been used, using the precursor t-butylchloride
(tBuCl) [10] and n-hexylchloride (nHexCl) [11], it was found
that the undoped CdS was sufciently n-type for junction
formation.
A thickness of 240 nm was necessary to avoid high-current loss
due to shunting, however, high-efciency CdTe devices generally
require a thin CdS window layer to increase the blue response of
the cell [18]. This creates a compromise between the maximising
on the blue response and obtaining good IV characteristics. One
way round this problem is to use MOCVD to control the alloying of
a ternary compound using Cd(1x)Zn(x)S to broaden the band gap
while keeping the thickness constant. With x 10% zinc content,
the ternary alloy, with a band gap of 2.7 eV (compared with 2.4 eV
with CdS) shift the response of the cell from 500 down to 400 nm,
resulting in an improved generation of photo-current.
3.2. Arsenic-doped CdTe layer
The CdTe absorber layer was arsenic doped, in situ with tris-
dimethylaminoarsenic (TDMAAs) using a double dilution line.
The cadmium and tellurium precursors are DMCd and diisopro-
pyltellurium (DIPTe), respectively. In the early work by Berrigan
et al. [10] the arsenic concentration was very high, in the region of
6  1019 cm3. The subsequent introduction of the double dilution
circuit has enabled improved control of the arsenic concentration
over a wide range from 1 1016 to over 1 1019 cm3 [19].
The work by Rowlands et al. [19] has shown that the reaction
mechanism for the decomposition of the TDMAAs follows second-
order kinetics and is probably occurring due to the formation of an
intermediate dimer complex. The CdTe layer was deposited under
Cd-rich conditions to promote the As doping. However, for
polycrystalline CdTe layers grown onto glass, the donor concen-
tration (around 1 1014 cm3) is far lower than the As concentra-
tion measured using calibrated SIMS [19]. Fig. 2 shows the device
measurements on a series of CdTe layers with different
As concentrations in the CdTe absorber layer. These layers were
simple CdS/CdTe structures without any CdCl2 treatment.
The corresponding SIMS analysis for these layers shows that the
threshold for achieving a measurable photo-response is 1 1017
cm3 and the maximum photo-response occurs at a concentration
of 2  1018 cm3 with a decrease for higher concentrations. In view
of the observation that the p-type carrier concentration does not
vary much over this range of arsenic concentrations, the effect of
the optimum concentration is clearly complex and might be
related to grain boundary segregation. Although, the efciency
results are much improved compared with the earlier reported
devices [10], it still falls short of the 410% target required for
efcient PV devices. The spectral quantum efciency shows a
5
4
3
2
1
0
1.E+15 1.E+16 1.E+17 1.E+18 1.E+19 1.E+20
As concentration (atm cm-3)
Fig. 2. A plot of the Jsc (a) and efciency (b) of CdTe PV cells with different arsenic
concentrations in the CdTe layer. These layers did not have any CdCl2 treatment.
decrease to the longer wavelength end of the spectrum which can
be ascribed to incomplete passivation of the grain boundaries. The
mechanism would be that longer wavelength photons have a
lower absorption coefcient and will typically generate minority
carriers further from the junction (i.e. deeper into the CdTe layer).
The lower quantum efciency for longer wavelength photons
would then be due to minority electrons generated further from
the junction having a higher probability of recombination before
they reach the junction, possibly at the grain boundaries.
Other optimisation studies of the CdTe absorber layer, using a
statistical design of experiments approach, have shown that the
largest contribution to improved PV efciency was an increase in
growth temperature from 350 to 390 1C. The most signicant
observation was the increase in grain size and again would
imply that grain boundaries are a key limiting factor in these
devices [20].
3.3. Contact layer
Forming an ohmic contact to the CdTe absorber layer, referred
to as the back contact is very difcult due to the high work
function of CdTe. High series resistance can limit the solar
conversion efciency of the cell and avoiding this often entails
further processing steps requiring etching of the CdTe to produce a
Te-rich surface and alloying with copper or gold [21].
The MOCVD CdTe growth was terminated with an additional,
sub-micron thick, highly arsenic-doped CdTe cap layer with
concentrations in excess of 1 1019 cm3 [22]. The purpose of
this layer was to create similar conditions to etching and alloy
formation but with the advantage of removing the need to carry
out any of these steps after the lm is grown. The measured series
resistance in PV cells with varying high As, CdTe cap thickness, is
shown in Fig 3. This study looked at the effect of varying the
thickness of the CdTe cap layer. For the results shown with zero
thickness, the layers were given a conventional Br/MeOH etch
 ARTICLE IN PRESS

S.J.C. Irvine et al. / Journal of Crystal Growth 310 (2008) 51985203 5201

prior to metallisation with Au. These conventionally processed
layers had series resistance of around 10 O cm2, which is too high
for an efcient device. Much higher series resistance was
measured for thick CdTe cap layers, probably due to the resistivity
of the layer itself. For thinner layers, approximately 200 nm, the
series resistance was reduced to below 2 O cm2.
The mechanism for the reduced series resistance is not simply
related to producing a p+layer although the carrier concentration
does rise towards the surface. The need for high arsenic
concentrations 42  1019 cm3 could be related to the formation
of a second phase material near the surface. The remarkable result
of this process is the ease for which an evaporated gold contact,
without any prior etching, can produce a low-resistance contact,
removing the wet etching process step.
20
Series resistance/Ohm cm2
3.4. CdCl2 anneal
Arsenic doping of the CdTe layer was insufcient to achieve
high-efciency CdTe/CdS PV devices, so the CdCl2 treatment step
was introduced. With the intention of carrying out all the
necessary steps in an MOCVD chamber, CdCl2 was deposited from
the precursors DMCd and tBuCl [23]. The MOCVD process differs
from the conventional Cl treatment in being carried out in a single
MOCVD chamber and does not involve any ex situ treatment.
In contrast to the conventional, ex situ, Cl treatment, the As
doping is still needed in order to obtain a high photo-current, to
date, the results achieving 410% efciency, used a substrate
temperature of 200 1C during deposition and 420 1C for 10 min for
the annealing. The choice of optimised parameters for the CdCl2
treatment will depend heavily on the thickness of the CdTe
absorber layer. It appears that most of the CdCl2 layer is consumed
while annealing, diffusing along the grain boundaries.

18
16 4. PV device results
14
12 A summary of device performance made using 5  5 mm2 Au
10 contacts onto MOCVD-grown PV structures is shown in Table 2.
8 The values from the table are taken over an average of three
6 devices (i.e. contact areas) as shown in Fig. 4, showing a
photograph of a test sample (25  17 mm2), ready for testing in
4
an AM1.5 solar simulator. Using the high-As concentration contact
2 layer, described in Section 3.3, there is no etching of the CdTe
0 surface prior to evaporation of the Au contacts.
0 200 400 600 800 1000 The results in Table 2 show average results for two different
Thickness of the heavily doped CdTe:As layer / nm device structures, one with the CdS window layer and the other
Fig. 3. Series resistance measured in PV cells with 5  5 mm2 Au contacts onto the
with the CdZnS window layer. Both achieved AM1.5 efciencies
highly doped As cap layer [23]. Estimated concentration of As in the cap layer is higher than the 10% target but it can be seen that the CdZnS
2  1019 cm3. window layer makes a substantial improvement in efciency

Table 2
Summary of device performance for 5  5 mm2 PV devices using all the MOCVD process steps and comparing CdS with Cd0.9Zn0.1S window layers

Device structures Efciency (%) Jsc (mA/cm2) FF (%) Voc (mV)

+
CdS (240 nm)/CdTe:As (2 mm)/CdTe:As (250 nm) 10.3 24 66 650
Cd0.9Zn0.1S (240 nm)/CdTe:As (2 mm)/CdTe:As+ (250 nm) 13.3 26 74 690

Fig. 4. Photograph of a test sample (25  17 mm2) ready for measurement of device PV performance. The three gold contact pads represent three separate devices that can
be used to test consistency and uniformity of the device structure.

5202 S.J.C. Irvine et al. / Journal of Crystal Growth 310 (2008) 51985203
14
12
AM 1.5 efficiency
10
0
Undoped/ low As
no cap (1E17 cm-3) As (2E18 CdCl2 cap As (2E18 As (2E18
/ no cap cm-3)/ no
Fig. 5. Comparison of AM1.5 PV efciency for different MOCVD process steps, as indicated.
above 13%. This is higher than any previous result for MOCVD-
grown CdTe absorber material. Looking at the device parameters it
can be seen that the improvements have occurred not only from
the higher short circuit current (Jsc) with enhanced collection of
blue photons, but also in open circuit voltage (Voc) and ll factor
(FF). The latter 2 parameters would indicate that the junction
quality has improved, indicating a lower defect density in this
region. This effect is being investigated further and will be
included in a more detailed paper on the alloy window layer.
A summary of all the MOCVD process steps is shown in Fig 5.
This shows the combination of optimum As doping, high As
contact layer, CdCl2 cap layer and the Cd0.9Zn0.1S window layer.
This demonstrates the importance of controlled As doping in
combination with the CdCl2 cap and anneal. The absence of
signicant photocurrent with just the CdCl2 anneal might be
due to other changes that occur with ex situ annealed material in
the native defect concentration to make the CdTe layer p-type.
These MOCVD process steps demonstrate the ability to engineer
the device using MOCVD that cannot be readily achieved by the
conventional annealing procedures to reach high-efciency
devices. This process avoids the need for chemical bath deposition
(CBD) and wet etch processes and is carried out at atmospheric
pressure, which is attractive for large scale production processes.
5. Conclusions
Recent advances in the deposition of CdTe solar cells using
MOCVD have enabled a signicant increase in AM1.5 efciency to
over 10%. This is the rst time that all the signicant steps in
forming a CdTe solar cell, including the CdCl2 treatment have been
carried out in a single MOCVD chamber. MOCVD has the potential
to considerably reduce the processing complexity and could
become compatible with very large scale manufacture if these
processes could be transferred to an in-line CVD system. This
paper reviews some of the critical processing steps and how
MOCVD has been used to increase PV conversion efciency
through in situ As doping of both the absorber layer and a contact
layer, and in situ deposition of a CdCl2 cap layer to passivate the
grain boundaries. The intentional doping of CdTe with arsenic
is unusual but has enabled optimisation of the characteristics of
the absorber layer and provides a new type of back contact
layer that avoids the need for wet etching before metallisation.
ARTICLE IN PRESS
Optimum No As/ Optimum Optimum CdZnS
window
cm-3)/ cm-3)/ high layer
cap CdCl2 cap As contact/
CdCl2 cap
Device structure
The concentrations of elemental As and Cl are far higher than the
active acceptor concentration and it is not clear at this stage what
role the grain boundaries play in this process. Finally the ability to
change the composition of the CdS window layer to the CdZnS
alloy has given a further boost in efciency to 13.3% and illustrates
how MOCVD can now be used to engineer the CdTe PV device
structure.
The future potential for MOCVD as a volume production
technique for thin lm solar cells depends on the ability to scale
this process to carry out each of the deposition steps as a
continuous in-line process. This could be used for deposition onto
glass substrates or onto exible substrates as a roll-to-roll process.
Growth rates are currently too low to achieve the desired in-line
process speed above 10 cm/min and new developments would be
needed in the deposition kinetics. The advantages in using
MOCVD would be in having an atmospheric pressure process,
with no wet chemical steps. The materials exibility in, for
example, changing alloy composition could lead to higher-
efciency PV modules in production that are currently not
achievable with CSS and CBD deposition techniques.
Acknowledgements
The authors gratefully acknowledge the nancial support of
EPSRC for funding the PV21 Supergen project and to Qinetiq for
their continued support with the use of their MOCVD reactor.
References
[1] J.M. Pearce, Renew. Energy 33 (2008) 1101.
[2] Solar Buzz Industry Report /http://www.solarbuzz.comS.
[3] W. Hoffman, Sol. Energy Mater. Sol. Cells 90 (2006) 3285.
[4] D. Bonnet, P. Meyers, J. Mater. Res. 13 (1998) 2740.
[5] A. Nouhi, P.V. Meyers, R.J. Stirn, C.H. Liu, J. Vac. Sci. Technol. A 7 (1989) 833.
[6] H.C. Chou, A. Rohatgi, J. Electron. Mater. 23 (1994) 31.
[7] T.L. Chu, S.S. Chu, C. Ferekides, J. Britt, C.Q. Wu, J. Appl. Phys. 69 (1991) 7651.
[8] T.L. Chu, S.S. Chu, J. Britt, C. Ferekides, C. Wang, C.Q. Wu, H.S. Ullal, IEEE
Electron Device Lett. 13 (1992) 303.
[9] X. Wu, J.C. Keane, R.G. Dhere, C. Dehert, D.S. Albin, A. Dude, T.A. Gessert,
S. Asher, D.H. Levi, P. Sheldon, in: Proceedings of the 17th European
Photovoltaic Solar Energy Conference, vol. II, Munich, Germany, 2001, p. 995.
[10] R.A. Berrigan, N. Maung, S.J.C. Irvine, D.J. Cole-Hamilton, D. Ellis, J. Crystal
Growth 195 (1998) 718.
[11] V. Barrioz, R.L. Rowlands, S.J.C. Irvine, in: Proceedings of the 20th European
Photovoltaic Solar Energy Conference, Barcelona, 2005, June 610, p. 1918.
 ARTICLE IN PRESS
S.J.C. Irvine et al. / Journal of Crystal Growth 310 (2008) 51985203
[12] I. Repins, A. Miguel, A. Contreras, B. Egaas, A. Clay DeHart, J. Scharf,
C.L. Perkins, B. To, R. Nou, Prog. Photovoltaics: Res. Appl. 16 (2008) 235.
[13] S.S. Lee, K.W. Seo, J.P. Park, S.K. Kim, Ii-W. Shim, Inorg. Chem. 46 (2007) 1013.
[14] B. Sagnes, A. Salesse, M.C. Artaud, S. Duchemin, J. Bougnot, J. Crystal Growth
124 (1992) 620.
[15] P. Gallon, F. Ouchen, M.C. Artaud, S. Duchemin, Inst. Phys. Conf. Ser. 152
(1998) 365.
[16] I.H. Choi, D.H. Lee, Thin Solid Films 515 (2007) 4778.
[17] R.A. Berrigan, S.J.C. Irvine, A. Stafford, D.J. Cole-Hamilton, D. Ellis, J. Mater. Sci.:
Mater. Electron. 9 (1998) 267.
5203
[18] B.E. McCandless, K.D. Dobson, Sol. Energy 77 (2004) 839.
[19] R.L. Rowlands, S.J.C. Irvine, V. Barrioz, E.W. Jones, D.A. Lamb, Semicond. Sci.
Technol. 23 (2008) 15017.
[20] R.L. Rowlands, V. Barrioz, E.W. Jones, S.J.C. Irvine, D.A. Lamb, J. Mater. Sci.:
Mater. Electron. 19 (2008) 639.
[21] A.L. Fahrenbruch, Mater. Res. Soc. Symp. Proc. 1012 (2007) Y07.
[22] V. Barrioz, Y. Proskuryakov, E.W. Jones, J. Major, S.J.C. Irvine, K. Durose,
D. Lamb, Mater. Res. Soc. Symp. Proc. 1012 (2007) Y12.
[23] V. Barrioz, S.J.C. Irvine, E.W. Jones, R.L. Rowlands, D.A. Lamb, Thin Solid Films
515 (2007) 5808.