Toward Annealing-Stable Molybdenum-Oxide-Based
Hole-Selective Contacts For Silicon Photovoltaics
Item Type Article
Authors Essig, Stephanie; Dréon, Julie; Rucavado, Esteban; Mews,
Mathias; Koida, Takashi; Boccard, Mathieu; Werner, Jérémie;
Geissbühler, Jonas; Löper, Philipp; Morales-Masis, Monica;
Korte, Lars; De Wolf, Stefaan; Balllif, Christophe
Citation Essig S, Dréon J, Rucavado E, Mews M, Koida T, et al. (2018)
Toward Annealing-Stable Molybdenum-Oxide-Based Hole-
Selective Contacts For Silicon Photovoltaics. Solar RRL: 1700227.
Available: http://dx.doi.org/10.1002/solr.201700227.
Eprint version Post-print
DOI 10.1002/solr.201700227
Publisher Wiley
Journal Solar RRL
Rights This is the peer reviewed version of the following article: Toward
Annealing-Stable Molybdenum-Oxide-Based Hole-Selective
Contacts For Silicon Photovoltaics, which has been published in
final form at http://doi.org/10.1002/solr.201700227. This article
may be used for non-commercial purposes in accordance With
Wiley Terms and Conditions for self-archiving.; This file is an
open access version redistributed from: http://www.helmholtz-
berlin.de/pubbin/oai_publication?VT=1&ID=96014
Download date 09/08/2022 14:15:25
1
Towards annealing-stable molybdenum-oxide-based hole-selective
contacts for silicon photovoltaics
Stephanie Essig
1
, Julie Dréon
1
, Esteban Rucavado
1
, Mathias Mews
2
, Takashi Koida
3
,
Mathieu Boccard
1a
, Jérémie Werner
1
, Jonas Geissbühler
4
, Philipp Löper
1
, Monica
Morales-Masis
1
, Lars Korte
2
, Stefaan De Wolf
5
, Christophe Ballif
1
1
École Polytechnique Fédérale de Lausanne (EPFL), Institute of Microengineering (IMT), Photovoltaics and Thin Film
Electronic Laboratory (PV-Lab), Rue de la Maladière 71b, 2002 Neuchâtel, Switzerland
2
Helmholtz-Zentrum Berlin for Materials and Energy (HZB), Institute of Silicon Photovoltaics, Kekuléstraße 5,
12489 Berlin, Germany
3
National Institute of Advanced Industrial Science and Technology (AIST), 1-1-1 Umezono, Tsukuba, 305-8568, Japan
4
CSEM PV-center, Rue Jaquet-Droz 1, 2002 Neuchâtel, Switzerland
5
King Abdullah University of Science and Technology (KAUST), KAUST Solar Center (KSC), Thuwal 23955-6900,
Saudi Arabia
Molybdenum oxide (MoO
X
) combines a high work function with broadband optical transparency.
Sandwiched between a hydrogenated intrinsic amorphous silicon passivation layer and a transparent
conductive oxide, this material allows a highly efficient hole-selective front contact stack for crystalline
silicon solar cells. However, hole extraction from the Si wafer and transport through this stack degrades
upon annealing at 190 °C, which is needed to cure the screen-printed Ag metallization applied to typical
Si solar cells. Here, we show that effusion of hydrogen from the adjacent layers is a likely cause for this
degradation, highlighting the need for hydrogen-lean passivation layers when using such metal-oxide-
based carrier-selective contacts. Pre-MoO
X
-deposition annealing of the passivating a-Si:H layer is
shown to be a straightforward approach to manufacturing MoO
X
-based devices with high fill factors
using screen-printed metallization cured at 190 °C.
2
Passivating carrier-selective contacts offer a simple way to approach the theoretical efficiency limit of silicon (Si) solar
cells without the need for expensive layer patterning, and they can offer superior performance under outdoor conditions
[1,2]. In conventional Si homo-junction solar cells, carrier separation is ensured by highly doped regions within the Si
wafer, and electrical contacts with low resistivity are obtained by applying metal electrodes directly onto the highly
doped Si surfaces. Despite being extensively manufactured worldwide, such solar cells are limited in their efficiency
potential due to defect-assisted and Auger recombination of charge carriers, respectively at the metal/silicon interface
and in the doped Si regions. The use of passivating contacts—like the ones used in silicon heterojunction (SHJ) solar
cells [3]—suppresses these recombination routes by separating the metal from the surface of the Si wafer and omitting
heavy doping of the Si wafer. To reduce further optical losses in passivating contacts, the application of wide-bandgap
materials like molybdenum-, tungsten-, titanium- or nickel-oxide (MoO
X
[4-6], WO
X
[7, 8], TiO
2
[9-11], NiO [10]),
as well as lithium- or magnesium-fluoride (LiF [12], MgF
2
[13]) has received much attention in recent years. These
materials feature a work function that is either higher than the ionization energy, or lower than the electron affinity of
crystalline Si (c-Si). Therefore, when in contact with Si and depending on the band lineup and defect density of the
interface, these materials may induce an electrical potential at the Si surface, which promotes the collection of either
only holes or only electrons. As a specific example, the transition metal oxide MoO
X
(x ≈3) combines a wide optical
bandgap energy of 2.8-3.1 eV [14] with a high work-function of 4.8 to 6.9 eV [15, 16]; when deposited on Si surfaces,
it may thus promote hole collection. Practically, conversion efficiencies up to 22.5% [5] were demonstrated for Si solar
cells employing a MoO
X
(x ≈3) based hole-selective front emitter stack. The higher transparency of MoO
X
compared
to p-type hydrogenated amorphous silicon layers (a-Si:H) led to ~0.3 mA/cm
2
gain in photo current density compared
to the reference SHJ cells [5]. To ensure passivation of the silicon surface, an additional a-Si:H buffer layer was inserted
underneath this MoO
X
film, similar as in conventional SHJ technology. The MoO
X
film was capped with a transparent
conductive oxide (TCO), either hydrogen doped indium oxide (IO:H) and indium tin oxide (ITO) stack [17] or simply
ITO, to minimize resistive losses and maximize light in-coupling [18]. To finish the devices, a Cu front grid electrode
[19 was formed by electroplating. This metallization technique notably does not require any thermal treatment above
125 °C. Despite its high work function, MoO
X
is an n-type material [20]. As a consequence, efficient carrier extraction
requires that photogenerated holes in the valence band of c-Si recombine with electrons present in the MoO
X
conduction band; the latter electrons are injected from the degenerately n-doped TCO [4, 6]. Efficient charge-carrier
transport through this contact stack depends on the thickness, defect density (for trap-assisted transport) and work
function of MoO
X
[21, 22], as well as on the line-up with the band edge energies of the surrounding layers.
Despite these promising results, industrial implementation of such contacts demands their compatibility with
contemporary high-throughput grid metallization techniques, which currently consist mainly of screen-printing Ag
paste, followed by a moderate-temperature cure at about 190 °C. Unfortunately, applying this to our current
implementation of MoO
X
-based solar cells results in light current-voltage (JV) curves that are S-shaped near the open-
circuit voltage (V
OC
), resulting in reduced fill factors (FF) below 70% [5]. We present here an experimental
investigation by means of thermal desorption spectroscopy (TDS) and surface photovoltage spectroscopy (SPV) of the
underlying degradation mechanisms leading to this FF degradation. These results hint at hydrogen effusion from the
a-Si:H layer being the origin of the FF degradation, and we propose a solution similar to the approach described in
[23] to mitigate this effect by annealing the a-Si:H-coated silicon wafers prior to MoO
x
deposition: This reduces the
hydrogen content in the film, leading to reduced H effusion upon the final annealing step of the finished device.
For all experiments, 240 µm thick, ~3 Ωcm phosphorus-doped, float-zone (100) Si wafers were textured and
cleaned. During solar cell fabrication at EPFL, a 5-nm-thick intrinsic, a-Si:H layer was applied by plasma enhanced
chemical vapor deposition (PECVD) on the wafer’s front side, and a ~15 nm thick stack of intrinsic and n-type doped
a-Si:H layers was deposited as the rear-electron collecting contact. On the front side, a ~8 nm thick MoO
X
layer was
thermally evaporated from MoO
3
powder using a deposition rate of 0.05 nm/sec. The active solar cell areas were
defined by sputter deposition of ~70 nm thick ITO layers on the MoO
X
. A full area ITO/Ag rear contact stack was
deposited by sputtering and finally a front metal grid was prepared by Ag screen printing. Current-voltage (JV)
characteristics were measured after stepwise curing of the Ag paste at 130 °C and 190 °C. SPV measurements were
performed at Helmholtz-Zentrum Berlin (HZB) on cells with Cu-plated front contacts using a home-built setup with
905 nm laser excitation [24]. TDS was performed at the National Institute of Advanced Industrial Science and
Technology (AIST) with a constant heating rate of (20.0 ± 0.1) K/min at a base pressure lower than 10
-9
mbar.
Figure 1a shows the recorded TDS spectra of H
2
from single a-Si:H and MoO
X
films as well as a-Si:H/MoO
X
and a-Si:H/MoO
X
/IO:H stacks. A single a-Si:H layer releases hydrogen (H
2
) already at low temperatures around 100 °C
with a desorption peak at 360 °C, similarly to literature [25-28]. The presence of a MoO
X
layer on top of the a-Si:H
leads to an earlier and also more pronounced release of H
2
(at temperatures as low as 150 °C) and shifts the effusion
peak to ~320 °C. A similar effect was obtained from identical measurements employing doped a-Si:H overlayers [28].
This effect was explained by reduction of the defect-formation energy when the Fermi-level inside the intrinsic a-Si:H
is shifted closer to its band edges. Our experiments on MoO
X
-based devices support well these earlier findings since a
Fermi-level shift in the intrinsic a-Si:H closer to the valence band is also expected in this case, though the surface from
which H effuses is different in our case. Next, we find that the H
2
effusion peak of the a-Si:H/MoO
X
/IO:H stack is
3
significantly lower in intensity, with an onset at increased temperatures (~230 °C vs. ~170 °C for the a-Si:H/MoO
X
stack). This suggests that H
2
from the a-Si:H/MoO
X
stack is partially absorbed in the IO:H, and released in the form of
H
2
O, as clearly observed in Fig. 2b. The H
2
O desorption from IO:H (Figure 2b) is in good agreement with refs. [29,30].
The H
2
O effusion spectra of c-Si with MoO
X
layers and a-Si.H/MoO
X
stacks have a sharp rise at temperatures close to
75 °C and indicate the thermal decomposition (reduction) of MoO
X.
(x~3), which is partly triggered by the presence of
hydrogen [31]. The effect of the H
2
effusion on the solar cell performance will be discussed below, based on SPV and
IV measurements of completed solar cells.
FIG. 1. H
2
and H
2
O thermal desorption spectra of test structures consisting of a single layer of MoO
X
or a-Si:H on c-Si wafers or of stacks of MoO
X
/a-Si:H or a-Si:H/ MoO
X
/IO:H on c-Si wafers. H
2
effusion spectra: the existence of a MoO
X
layer on top of a-Si:H (red curve) leads to a shift of the
main hydrogen effusion peak to lower temperatures as compared to a-Si:H alone (blue curve). H
2
O
effusion spectra: the spectra show the release of H
2
O from the top IO:H film (green curve) and
indicate the decomposition (reduction) of the MoO
X
films (red and black curves).
Figure 2a shows the changes in the MoOx[/a-Si:H]/c-Si band bending of our solar cells throughout annealing
(temperatures of 100-250 °C, each step 5 min), as extracted from SPV measurements. Note that although the absolute
value obtained with the setup used in this study are typically lower compared to the ones measured in other places [23],
relative differences between samples can be discussed and correlate well to device properties [8]. The band bending of
the MoO
X
-based solar cell with the a-Si:H(i) buffer layer is significantly reduced by annealing, whereas the changes
in the MoO
X
cell without the a-Si:H(i) layer are much smaller (especially between 170 °C and 210 °C). The similar
onset temperatures for band-bending reduction and H effusion from a-Si:H / MoO
X
stacks suggests that these effects
are linked. We surmise that, similar to what has been found in tungsten oxide (WO
X
)/a-Si heterojunctions [8], hydrogen
effuses from the a-Si:H layer, partially reducing MoO
X
and lowering its workfunction, leading to reduced c-Si band
bending, thus degrading the hole-selectivity of our contact. This is reflected in the IV curves of corresponding devices
at various annealing temperature, where a strong S-shape characteristic appears after annealing at 190 °C for the MoO
X
-
based device incorporating an a-Si:H layer but not for the one not incorporating this a-Si:H layer. Also, in case the
degradation of workfunction is accompanied with a lowering of the electron affinity and ionization potential, the
energetic gap between the a-Si:H valence band and the TCO conduction band is increased, as sketched in Figure 2b,
deteriorating transport by reducing the probability of (trap-assisted) tunneling within the MoO
X
layer. Notably,
reduction of the oxidation state of the MoO
X
layer—and possibly hydrogenation—can- affect the workfunction by
moving the Fermi-level only without affecting the band structure, but can also reduce the workfunction and activation
energy and ionization potential similarly, thus shifting the band structure towards the vacuum level energy [15,32].
Both effects have been reproduced in Fig. 2b to keep it general. A reduced band bending close to the Si surface also
decreases the potential drop in the a-Si:H layer. In turn, the thermionic field emission over the band offset at the a-
Si:H/c-Si interface is reduced, eventually leading to a transport barrier for holes, resulting in S-shaped JV curves as
reported in our earlier work [5]. The dominant transport limitation at stake in our device is unclear—the role of eventual
trap states and dipoles possibly bringing additional contributions, though the origin being a drop of workfunction
through H-enhanced reduction is a likely cause in all three cases which correlates well with effusion measurement. As
a consequence, buffer layers which do not effuse hydrogen upon annealing up to 200 °C are desirable to obtain
annealing-stable MoO
X
-based hole-selective contacts. The next section discusses how annealing the a-Si:H passivating
layers prior to MoO
X
deposition could lead to such H-effusion-free buffer layer, as we observe good passivation and
efficient carrier transport when using such layer—even after annealing the finished device at 190 °C.
