of 8
This journal is
©
The Royal Society of Chemistry 2015
Energy Environ. Sci.,
2015,
8
, 2977--2984 |
2977
Cite this:
Energy Environ. Sci.,
2015,
8
,2977
Functional integration of Ni–Mo electrocatalysts
with Si microwire array photocathodes to
simultaneously achieve high fill factors and
light-limited photocurrent densities for
solar-driven hydrogen evolution
Matthew R. Shaner,
ab
James R. McKone,
a
Harry B. Gray
bc
and Nathan S. Lewis*
abc
An n
+
p-Si microwire array coupled with a two-layer catalyst film consisting of Ni–Mo nanopowder and
TiO
2
light-scattering nanoparticles has been used to simultaneously achieve high fill factors and light-
limited photocurrent densities from photocathodes that produce H
2
(g) directly from sunlight and water.
The TiO
2
layer scattered light back into the Si microwire array, while optically obscuring the underlying
Ni–Mo catalyst film. In turn, the Ni–Mo film had a mass loading sufficient to produce high catalytic
activity, on a geometric area basis, for the hydrogen-evolution reaction. The best-performing microwire
array devices prepared in this work exhibited short-circuit photocurrent densities of

14.3 mA cm

2
,
photovoltages of 420 mV, and a fill factor of 0.48 under 1 Sun of simulated solar illumination, whereas the
equivalent planar Ni–Mo-coated Si device, without TiO
2
scatterers, exhibited negligible photocurrent due
to complete light blocking by the Ni–Mo catalyst layer.
Broader context
Solar-driven photoelectrochemical water splitting is a promising approach to enable the large-scale conversion and storage of solar energy. Few in
tegrated
systems have been realized that use earth-abundant semiconductor and catalyst materials for the half-reactions involved in solar-driven water-sp
litting,
i.e.
hydrogen evolution and oxygen evolution, while also achieving high energy-conversion efficiencies. We describe herein a hydrogen-evolving Si-base
d
photoelectrode that exhibits high light-limited photocurrent densities, as well as high catalytic activities, while using a high mass loading of an
earth-abundant
electrocatalyst. The design is reminiscent of a membrane-electrode assembly as used for stand-alone fuel cell and electrolysis systems. The approa
ch exploits
the high aspect ratio of the absorber layer to avoid parasitic optical absorption normally associated with a thick catalyst layer on the surface of an i
lluminated
photocathode.
I. Introduction
Photon management is an important attribute of photoelectrodes
used for solar-driven water-splitting, especially for device archi-
tectures that incorpor
ate optically opaque electrocatalyst coatings
on the surface of a light absorbing material.
1–4
Specifically, the fill
factor (ff) is generally negatively correlated with the light-limited
photocurrent density (
J
ph
), because increases in catalyst loading
increase the ff, but also produce la
rger parasitic optical absorp-
tion losses and thus decrease the value of
J
ph
(Fig. 1).
1
Such
behavior is especially apparent for earth-abundant electrocatalysts
for the hydrogen-evolution reac
tion (HER), which are generally
optically opaque and require large mass loadings (
Z
1mgcm

2
)
to achieve the requisite catalytic activity.
5–7
Similar issues can also
preclude optimal functional incorporation of currently available
electrocatalysts for the oxygen-evolution reaction (OER) into
integrated photoanode structures
for the solar-driven oxidation
of H
2
OtoO
2
(g).
8,9
For planar photoelectrode architectures, various options to
mitigate this deleterious tradeoff between catalytic activity and
optical transparency have been developed. For discrete photovoltaic
a
Division of Chemistry and Chemical Engineering, California Institute of
Technology, 1200 East California Blvd, Pasadena, California 91125, USA
b
Joint Center for Artificial Photosynthesis, California Institute of Technology,
1200 East California Blvd, Pasadena, California 91125, USA
c
Beckman Institute and Kavli Nanoscience Institute, California Institute of
Technology, 1200 East California Blvd, Pasadena, California 91125, USA.
E-mail: nslewis@caltech.edu
Electronic supplementary information (ESI) available: Complete materials and
methods, Fig. S1–S8 and associated discussion. See DOI: 10.1039/c5ee01076d
Current address: Department of Chemistry and Chemical Biology, Cornell
University, Ithaca, NY, USA.
Received 5th April 2015,
Accepted 13th July 2015
DOI: 10.1039/c5ee01076d
www.rsc.org/ees
Energy &
Environmental
Science
PAPER
Published on 07 August 2015. Downloaded by California Institute of Technology on 29/10/2015 17:27:02.
View Article Online
View Journal
| View Issue
2978
|
Energy Environ. Sci.,
2015,
8
, 2977--2984
This journal is
©
The Royal Society of Chemistry 2015
(PV)-biased electrolysis systems,
1–4,10
the PV device can be connected
electrically to discrete, catalytic electrodes that do not physically
obscure incoming light from the PV cell, with the interfacial
reactions being performed by majority carriers.
1,11
Alternatively,
in photoelectrode structures comprised of a single photoabsorber,
a transparent back contact can be used in conjunction with
‘‘backside’’ illumination so that the catalyst layer is not in the
optical path of the semiconductor.
5–7,12
For integrated photo-
electrodes in which the interfacial reactions can be performed
by photogenerated minority carriers (as well as by majority
carriers, for structures that contain a buried junction
8–10
), the
thickness of the electrocatalyst film can be adjusted to obtain
an optimum compromise between the optical density and
activity of the electrocatalyst film. This type of optimization
favors the use of an extremely thin (
o
5 nm) catalytic layer, at
the expense of catalytic activity.
9
In certain instances, nano-
structuring can produce optically transparent, highly active
films of noble metal electrocatalysts.
4
Yet another method
involves optimization of the spatial location and areal coverage
ofislandsofanactive,butopticallyabsorbingelectrocatalyst
film, by use of a photolithographic or shadow mask.
13
This
approach is analogous to the use of grid-line top contacts in
photovoltaics.
Absorbers that are structured in three dimensions, such as
Si microwire arrays
14–17
provide an alternative, general approach
to ameliorating the negative correlation between the ff and
J
ph
.
In such systems, a relatively high loading of catalyst can be
positioned at the base of a high-aspect-ratio microwire array,
leaving exposed a large proportion of the array (Fig. 2). For
example, earth-abundant metal catalysts such as Ni–Mo,
5
Ni or
Co phosphide,
6,7
or Mo chalcogenides
18
with mass loadings on
the order of 1 mg cm

2
(
i.e.
several
m
m thick) exhibit compar-
able geometric HER activity to a planar
4
10 nm thick Pt film.
By comparison, the depth over which light can be completely
absorbed in a sparse silicon microwire array with appropriate
scattering elements is on the order of 100
m
m. The relatively
low proportional volume occupied by high catalyst mass
loadings can therefore enable high overall catalytic activity
while preserving optical accessibility to the majority of the
light-absorber material.
A recent study has shown that a CoP HER electrocatalyst
could be deposited at the base of a Si microwire array, yielding
activity toward photoelectrochemical hydrogen evolution com-
parable to that of a control sample that instead used Pt as
the HER electrocatalyst.
19
Si microwire arrays, however, do not
absorb a large proportion of incoming, normally incident photons
on the ‘‘first pass’’ of such photons through the structure of the
device.
8
As a result, significant parasitic optical absorption occurs
in such devices even with the cata
lyst film deposited exclusively
at the base of the array. An improved device structure, which
we denote as the ‘‘MEA’’ architecture, resembles a fuel cell or an
electrolyzer membrane-electro
de assembly (MEA) (Fig. 2). Such a
structure is characterized by a Si microwire array on the order of
100
m
minheightwitha5–10
m
m thick porous catalyst layer
consisting of an earth-abundant electrocatalyst, such as Ni–Mo,
Mo sulfide, or a transition-metal phosphide, in addition to a
1–3
m
m thick porous light-scattering layer, such as particulate
TiO
2
or another high-dielectric scattering material. The catalyst
layer is analogous to the one found in a conventional MEA,
whereas the light-scattering layer serves as an optical scattering
element as well as a gas-diffusion layer. The microwires act as
current collectors by absorbing solar photons and producing
sufficient electrochemical potential to effect the HER.
In this approach, if the catalyst and scattering layers are
both sufficiently porous, reactant species (
e.g.
, water or H
+
)
can diffuse through and hydrogen gas can diffuse out of the
internal volume of the device. However, when incident on
the scattering layer, photons will be reflected laterally and will
therefore pass multiple times through the Si microwires. An
analogous light-scattering design was proposed and validated
for Si microwire photovoltaics, and high light-limited photo-
current densities (
4
15 mA cm

2
at 100 mW cm

2
of Air Mass
(AM) 1.5G simulated illumination intensity) were indeed obtained
in that system.
8,20
In principle, the MEA approach allows for
simultaneous maximization of the catalytic activity and optical
Fig. 1
Schematic of a water-splitting device incorporating two planar
light absorbers in a tandem configuration, two electrocatalysts, and a
single tunnel junction to allow serial addition of the photovoltages
produced by each semiconducting light-absorber unit. The opaque catalyst
on the light-incident surface (left side) is the source of the trade-off between
the fill factor (ff) and the light-limited current density (
J
ph
).
Fig. 2
Schematic depiction of the membrane-electrode assembly (MEA)-
type hydrogen-evolving photocathode based on arrays of Si microwires
embedded in a membrane or supported on a substrate, with a layer of a
non-noble catalyst deposited at the base of the array and covered by a
light-scattering layer.
Paper
Energy & Environmental Science
Published on 07 August 2015. Downloaded by California Institute of Technology on 29/10/2015 17:27:02.
View Article Online
This journal is
©
The Royal Society of Chemistry 2015
Energy Environ. Sci.,
2015,
8
, 2977--2984 |
2979
absorption in the semiconducting material, even when a relatively
large mass loading of electrocatalyst is required. In this work, we
have modeled and validated the MEA design using Si microwire-
array photocathodes to drive the HER in conjunction with a Ni–Mo
nanopowder catalyst and a TiO
2
light-scattering layer.
II. Modeling and experimental
A brief summary of the modeling and experimental details are
provided below, with additional details included in the ESI.
Modeling
A zero-dimensional model was constructed to predict the maxi-
mum performance expected from the MEA-type photocathode
device, based on a previously derived analytical expression for a
buried junction in series with a catalyst and an additional
resistor.
21
Two-dimensional full-wave electromagnetic simula-
tions were used to determine the light-limited current density
of the microwire structure, to provide input into the device
model. The catalytic performance was determined by modeling
experimental data using the Butler–Volmer equation and by
incorporation of a series resistance as determined experimentally
by impedance measurements.
Experimental
Si (p-type) microwires were grown in a custom-built chemical-
vapor deposition (CVD) reactor
via
a vapor–solid–liquid process.
22
The growth wafers were patterned with 3
m
m diameter Cu catalysts
located
via
photolithography in a square pattern with a 7
m
mpitch.
Highly doped n
+
-Si emitters were formed by use of a phosphorous-
doped spin-on-glass.
Ni–Mo nanopowder was synthesized following a previously
reported procedure.
5
A Ni–Mo catalyst supported on carbon
(Ni–Mo/C) was synthesized similarly to the Ni–Mo nanopowder
except for use of an additional step that involved thoroughly
grinding the Ni–Mo oxide with carbon black to form the supported
structure. TiO
2
and Ni–Mo nanopowders were suspended in iso-
propanol at concentrations of 100 mg mL

1
and 1–2 mg mL

1
,
respectively. The Ni–Mo nanopowder was deposited onto Ti foil
electrodes or onto Si photoelectrodes (both planar and microwire
arrays) by centrifugal flocculation from the nanoparticle inks,
to achieve mass loadings of
B
1mgcm

2
. The Ni–Mo, but not
Ni–Mo/C catalysts, were actived after deposition by an anneal at
450
1
C in forming gas. 15
m
L of the TiO
2
nanopowder solution
was deposited in a nominally identical fashion to that used to
deposit Ni–Mo nanopowders onto S
i samples. Platinum catalysts,
used for control experiments, we
re deposited ele
ctrolessly on Si
substrates by use of a solution that consisted of 1 mM Pt and 2%
HF in H
2
O.
Hydrogen-saturated, trace-metal-grade sulfuric acid (0.5 or
1.0 M) solutions were used to determine the cyclic voltammetric
performance of the electrodes using a three-electrode electro-
chemical configuration. The electrolyte was constantly bubbled
with research grade H
2
(g) (AirLiquide) to maintain a constant
reversible hydrogen electrode (RHE) potential, as determined
by measurement of the open-circuit potential of an electro-
chemically cleaned Pt button electrode. Simulated 1-Sun illu-
mination was provided by an ELH-type tungsten-halogen lamp
or by a Xe lamp with an AM 1.5G filter. Spectral response
measurements with the electrodes maintained potentiostatically
at

0.15 V
versus
the reversible hydrogen electrode, RHE, were
performed with a custom-built ap
paratus under electrochemical
testing conditions that were nomin
ally identical to those used for
collection of the cyclic voltammetric data.
III. Results
Fig. 3 shows the modeling results (see ESI
for details of the
model parameters) for a photocathode consisting of an n
+
p-Si
microwire array in conjunction with a Ni–Mo nanopowder
catalyst covered by a porous TiO
2
light-scattering overlayer. To
provide a first-order approximation to the maximum expected
device performance for such a structure, the modeling treated
the integrated photoelectrode as a zero-dimensional, series-
connected n
+
p-Si junction coupled to a catalyst layer and to a
resistive circuit element. The increased dark-current junction
area (
g
in eqn (S1), ESI
), and the series resistance required to
transport electrons down the length of the microwire through
the n
+
-Si emitter to the catalyst, were considered explicitly.
The light-limited current density was obtained through two-
dimensional full-wave electromagnetic simulations of a Si
microwire array architecture.
3,4
The modeling indicated that
utilization of a stand-alone n
+
p-Si microwire array PV device
having a solar energy-conversion efficiency of 12.9% could, in
principle, yield a maximum ideal regenerative cell efficiency
(
Z
IRC
)
23
of 11.2% based on the hydrogen-evolution half-reaction
being performed at such a photocathode. This value is essen-
tially the same as the value that would be obtained by wiring the
respective photovoltaic device to
a discrete, catalytic electrode
Fig. 3
Current-density
versus
potential (
J
E
) behavior obtained from the
modeling of an n
+
p-Si microwire-array membrane-electrode assembly
(MEA, magenta curve) and for a stand-alone n
+
p-Si microwire-array
photovoltaic device wired to a discrete, purely catalytic electrode (red
curve). Ni–Mo was used as the hydrogen-evolution catalyst in both cases,
and the activity of the catalyst was modeled according to experimental
observations (blue curve).
Energy & Environmental Science
Paper
Published on 07 August 2015. Downloaded by California Institute of Technology on 29/10/2015 17:27:02.
View Article Online
2980
|
Energy Environ. Sci.,
2015,
8
, 2977--2984
This journal is
©
The Royal Society of Chemistry 2015
obtained by using a high mass-loading of the Ni–Mo HER catalyst
on an inert conducting substrate.
Fig. 4a and b show a scanning-electron micrograph of a Si
microwire array coated at the base by a layer of Ni–Mo nano-
powder, with the Si and Ni–Mo bound together by a small
quantity of poly-tetrafluoroethylene (PTFE). The structure also
contained an overlayer of TiO
2
nanoparticles bound together
with a Nafion ionomer, in accord with the full device structure
depicted schematically in Fig. 2.
24
Fig. 4b has been highlighted
to delineate the layers of Ni–Mo and TiO
2
, respectively, and
clearly demonstrates the successful fabrication of the desired
MEA photocathode structure. Fig. 4c shows an optical image of
four microwire samples. A bare Si microwire array was iridescent
and reflective, similar to the starting wafer prior to microwire
growth, because of the highly ordered, sparse microwire array.
However, the addition of a Ni–Mo catalyst layer altered the
appearance to a matte black, due to the optical absorption in
both the microwires and the Ni–Mo electrocatalyst. Deposition
of a nanoparticulate TiO
2
layer directly onto the base of the Si
microwire array, or over a Ni–Mo film, yielded a gray photo-
electrode, consistent with scattering of a large proportion of the
incoming light into the vertically oriented Si microwires. The
sample that contained an underlying Ni–Mo layer was slightly
darker, due to imperfect Ni–Mo deposition leaving some residual
optically absorbing catalyst on the sidewalls of the microwires.
Fig. 5 shows the
J
E
behavior for dark HER electrocatalysis by
Ni–Mo nanopowders deposited on planar Ti substrates, as well
as for Ni–Mo nanopowders deposited at the bases of degener-
ately doped p-type Si microwire arrays (p
+
-Si MWs), with and
without overlayers of TiO
2
particles, respectively. Before catalyst
deposition, the Si microwire arrays were metallized with Ag and
were annealed to minimize any effects of interfacial contact
resistance with the catalyst materials. As indicated in Fig. 5,
Fig. 4
(a) Scanning-electron micrographs of a silicon microwire array onto which a layer of Ni–Mo nanopowder and a layer of TiO
2
light-scattering
particles had been deposited sequentially. (b) Detail of the boxed area in a, highlighted to delineate the Ni–Mo nanopowder layer beneath the layer of
TiO
2
particles. (c) Optical image of four microwire arrays with different nanoparticulate depositions within the array.
Fig. 5
Dark
J
E
behavior for electrodes with the noted compositions.
The Si microwire samples were degenerately doped p-type and were
metallized with Ag to minimize the interfacial contact resistance. All of the
samples were evaluated in 1.0 M H
2
SO
4
(aq), and Ni–Mo samples had mass
loadings of
B
1mgcm

2
, except for the Ti foil Ni–Mo/TiO
2
sample which
had a mass loading of
B
3mgcm

2
.
Paper
Energy & Environmental Science
Published on 07 August 2015. Downloaded by California Institute of Technology on 29/10/2015 17:27:02.
View Article Online