of 12
Photoelectrochemistry of core
shell tandem
junction n
p
+
-Si/n-WO
3
microwire array
photoelectrodes
Matthew R. Shaner,
ab
Katherine T. Fountaine,
ab
Shane Ardo,
a
Rob H. Coridan,
a
Harry A. Atwater
bc
and Nathan S. Lewis
*
ab
Tandem junction (n
p
+
-Si/ITO/WO
3
/liquid) core
shell microwire devices for solar-driven water splitting
have been designed, fabricated and investigated photoelectrochemically. The tandem devices exhibited
open-circuit potentials of
E
oc
¼
1.21 V
versus E
0
0
(O
2
/H
2
O), demonstrating additive voltages across the
individual junctions (n
p
+
-Si
E
oc
¼
0.5 V
versus
solution; WO
3
/liquid
E
oc
¼
0.73 V
versus
E
0
0
(O
2
/H
2
O)). Optical concentration (12

, AM1.5D) shifted the open-circuit potential to
E
oc
¼
1.27 V
versus E
0
0
(O
2
/H
2
O) and resulted in unassisted H
2
production during two-electrode measurements
(anode: tandem device, cathode: Pt disc). The solar energy-conversion e
ffi
ciencies were very low,
0.0068% and 0.0019% when the cathode compartment was saturated with Ar or H
2
, respectively, due to
the non-optimal photovoltage and band-gap of the WO
3
that was used in the demonstration system to
obtain stability of all of the system components under common operating conditions while also insuring
product separation for safety purposes.
Broader context
Direct photoelectrochemical conversion of sunlight into a storable, energy-dense fuel has the opportunity to provide a predictable, carbon-neutr
al energy source
to displace current carbon-based technologies. Solar hydrogen generation
via
water splitting is an important goal because the voltage requirements for this
process are well matched to the maximum power point of high-e
ffi
ciency tandem photovoltaics. In addition to including light-absorbing materials that provide
su
ffi
cient voltage for water splitting, an integrated solar fuels device requires catalysts connected to the light absorbers and an ionic transport pathw
ay between
the anode and cathode to complete the circuit while maintaining product separation below the lower explosive limits, for safety purposes. Integrati
on of these
di
ff
erent active materials is important to further development of this technology. Single-crystalline Si microwire arrays represent an architecture t
hat can allow
the system operation and integration requirements to be met, because photoactive Si microwires have been previously embedded into ionically conduc
tive, gas-
blocking membranes. However, Si microwires do not produce enough photovoltage for unassisted water splitting even in a tandem Si-based structure. W
e
describe a Si microwire based tandem junction device that produces su
ffi
cient photovoltage for unassisted water splitting, by use of WO
3
in a core
shell tandem
structure. This system provides a proof-of-principle for this design, which can be improved signi
cantly through the incorporation of higher e
ffi
ciency wide
band-gap semiconductors as they become available and are stable under the same conditions as the rest of the components of the device.
Introduction
Si microwire array photocathodes have been shown to generate
photovoltages in excess of 500 mV in acidic aqueous environ-
ments, and provide a preferred geometry, relative to planar
structures, for devices that e
ff
ect the unassisted generation of
fuels from sunlight.
1
3
Microwire arrays bene
t from orthogo-
nalization of the directions of light absorption and
minority-carrier collection,
4
8
as well as from light-trapping
e
ff
ects,
9,10
an increased surface area for catalyst loading per
unit of geometric area,
11,12
a small solution resistance as
compared to planar designs,
3,13
a reduced material usage
through reusable substrates,
14
and the ability to embed the
microwires into ion-exchange membranes that exhibit little
permeability to H
2
and O
2
,
15
thereby producing
exible devices
that persistently separate the products of the water-splitting
reaction. However, the voltage generated from single-junction
Si microwire arrays is much lower than the 1.23 V required for
solar-driven water splitting, so a wider band-gap partner light
absorber must be introduced electrically in tandem (Si/partner
tandem device), to generate useful current at voltages that
exceed the thermodynamically required values for fuel
production. Accordingly, tandem-junction devices o
ff
er the
highest theoretical
16
and experimentally realized
7
e
ffi
ciencies
a
Division of Chemistry and Chemical Engineering, California Institute of Technology,
Pasadena, CA, USA. E-mail: nslewis@caltech.edu
b
Joint Center for Arti
cial Photosynthesis, California Institute of Technology,
Pasadena, CA, USA
c
Thomas J. Watson, Sr. Laboratories of Applied Physics, California Institute of
Technology, Pasadena, CA, USA
Electronic supplementary information (ESI) available. See DOI:
10.1039/c3ee43048k
Cite this:
Energy Environ. Sci.
,2014,
7
,
779
Received 10th September 2013
Accepted 7th November 2013
DOI: 10.1039/c3ee43048k
www.rsc.org/ees
This journal is © The Royal Society of Chemistry 2014
Energy Environ. Sci.
,2014,
7
, 779
790 |
779
Energy &
Environmental
Science
PAPER
Published on 16 December 2013. Downloaded by California Institute of Technology on 20/03/2014 14:45:35.
View Article Online
View Journal
| View Issue
for solar-driven water splitting through additive voltages across
two photoabsorbers that e
ff
ectively utilize multiple regions of
the solar spectrum. Tandem device structures are also simpler
to fabricate and operate e
ff
ectively under a greater variety of
insolation conditions than more complex 3- or 4-junction
devices. Additionally, when kinetic overpotentials are consid-
ered in detail, water-splitting devices will most likely require a
tandem architecture to achieve appreciable current densities,
i.e.
10 mA cm

2
, thereby further motivating the use of tandem
structures in such applications.
In addition to band gap considerations for a Si/partner
tandem system, achieving the desired electronic behaviour at
the interface between Si and its tandem partner presents a
signi
cant challenge for production of an integrated solar fuels
generation device. The materials must be mutually compatible
and generally must operate in a batch reactor that contains a
single, concentrated (1.0 M)
3,17
aqueous electrolyte. Such
materials considerations are important to the performance of a
functioning device that consists of microwires embedded in a
gas impermeable, ion-exchange membrane, because both
semiconductors may need to be simultaneously in contact with
the electrolyte to produce a full solar-driven water-splitting
device. Tandem junction water-splitting devices using nano-
scopic or microscopic materials have focused on a single-
junction n
n heterojunction design in series with a liquid
second junction.
18,19
In contrast, the highest e
ffi
ciency water-
splitting devices
7,20
consist of planar tandem homojunction
photovoltaic cells that are physically isolated from the solution
and are electrically connected to the catalysts in contact with
solution. The materials currently used in high-e
ffi
ciency planar
tandem devices are not stable in concentrated aqueous elec-
trolyte environments. Nevertheless, the concept of buried p
n
homojunctions is a promising route to increase the e
ffi
ciency of
solar-driven water-splitting devices relative to systems that
utilize n
n heterojunctions.
1,18,19,21,22
To realize the advantages
of replacing the n
n heterojunction with a p
n homojunction,
ohmic behaviour at the Si/tandem partner interface is required.
This ohmic behaviour can be achieved in at least two ways: (i)
the Si tandem partner must have a proper band alignment
(type III, broken gap) such that upon direct contact, ohmic
behaviour is produced or (ii) a discrete intermediate third
material must be introduced that facilitates ohmic behaviour
between the Si and the tandem partner light absorbers.
TiO
2
,WO
3
, BiVO
4
and Fe
2
O
3
are stable in concentrated
aqueous electrolytes and form suitable tandem partners for Si.
However, Si is stable only in acidic aqueous environments,
limiting the presently available partner materials that are stable
under such conditions to only TiO
2
and WO
3
.WO
3
is the
preferred material because of its smaller band gap (
E
g

2.6 eV)
and signi
cant photocurrent response to visible-light illumi-
nation.
23
The electronic behaviour of the Si/WO
3
interface has
recently been shown to be non-ohmic, but addition of an
intermediate tin-doped indium oxide (ITO) layer has been
shown to provide low resistance, ohmic behaviour between
p-type, or p
+
-type, Si and WO
3
.
24
Thus, a Si/WO
3
microwire
device with an intermediate ITO layer presents an opportunity
to demonstrate an unassisted integrated solar-driven
water-splitting device that exploits the advantages of the
microwire-array architecture.
We describe herein a tandem core
shell photo-
electrochemical device that consists of a periodic array of buried
homojunction n
p
+
-Si microwires that have been sequentially
coated with a radial sheath of ITO and WO
3
. When immersed in
air-saturated 1.0 M H
2
SO
4
, the dual radial-junction microwire
structure enables e
ffi
cient carrier collection from both the Si
and WO
3
light absorbers, despite short minority-carrier di
ff
u-
sion lengths,
i.e.
,

10
m
minSi
25
and

1
m
minWO
3
. A necessary
feature of this tandem architecture is the incorporation of the
ITO layer between the Si and WO
3
light-absorbing materials.
This ohmic contact layer ensures facile, low-resistance carrier
transport between the Si and WO
3
and relaxes the requirements
for proper band alignment between the p
+
-Si emitter and the
WO
3
. Transparent conductive oxides, such as FTO or ITO, are
commonly used as back contacts to semiconductor metal
oxides; thus this design is expected to be robust towards
implementation of newly discovered materials, because the ITO
layer will be amenable to many di
ff
erent Si tandem partner
absorbers.
Experimental
Chemicals
All chemicals were used as received unless noted otherwise.
Water was
ltered using a MilliPore system and had a resistivity
>18 M
U
cm.
Si microwire array growth
Phosphorous-doped (
N
D
¼
3

10
17
cm

3
) and boron-doped
(
N
A
¼
1

10
17
cm

3
) Si microwire arrays were grown
via
a
Cu-catalyzed vapor
liquid
solid (VLS) process on As-doped n
+
-
Si or on B-doped p
+
-Si
h
111
i
wafers (<0.005
U
cm, Addison).
2,6,25
The n
+
-Si and p
+
-Si
h
111
i
growth wafers were received with a
400 nm thick thermal oxide (SiO
2
) that had been photolitho-
graphically patterned to produce 3
m
m diameter holes
lled
with Cu in a square lattice (7
m
m

7
m
m). The growth of Si
microwire arrays was performed in a chemical-vapor deposition
(CVD) furnace at atmospheric pressure using SiCl
4
(Strem,
99.9999+%) at 25 sccm
ow rate, H
2
(Matheson, research grade)
at 500 sccm
ow rate, and BCl
3
(Matheson, 0.25% in H
2
)at
1 sccm
ow rate for 20 min or PH
3
(Matheson, 100 ppm in H
2
)at
0.3 sccm
ow rate for 9 min. Following growth, the samples
were cooled to

200

C under a 500 sccm
ow of He.
Microwire array processing
Microwire arrays were cleaned using a 6 : 1 : 1 (by volume)
H
2
O : HCl(fuming, aqueous) : H
2
O
2
(30% in H
2
O) metal etch
(RCA 2) for 20 min at 60

C. The samples were then subjected to
a 15 s etch in bu
ff
ered HF(aq.) (BHF) etch, an H
2
O rinse, an
organic (piranha) etch in 3 : 1 H
2
SO
4
(99.6%, aque-
ous) : H
2
O
2
(30% in H
2
O) for 10 min at room temperature, and
an H
2
O rinse. Following a 30 s etch in 10% BHF and a rinse with
H
2
O, a 150 nm thick SiO
2
layer was grown
via
dry thermal
oxidation in a tube furnace at 1050

C under an O
2
atmosphere
780
|
Energy Environ. Sci.
,2014,
7
, 779
790
This journal is © The Royal Society of Chemistry 2014
Energy & Environmental Science
Paper
Published on 16 December 2013. Downloaded by California Institute of Technology on 20/03/2014 14:45:35.
View Article Online
for 2.5 h. A 15
m
m thick PDMS layer was deposited at the base of
the wires by spin coating a solution, consisting of 1.1 g of
polydimethylsiloxane (PDMS, Sylgard 185, Dow Corning) and
0.1 g of PDMS curing agent dissolved in 5 mL of toluene, on the
sample at 3000 rpm for 30 s, followed by a 30 min cure in a
vacuum oven at 150

C. These PDMS-in
lled arrays were
submerged in BHF for 5 min, to remove the SiO
2
on the exposed
microwire surfaces. The PDMS was removed by a 30 min soak in
3:1
N
-methyl-2-pyrrolidone (NMP):tetrabutylammonium
uo-
ride (TBAF, aq. 75 wt%), followed by a 30 s rinse with H
2
O. The
samples were then dried under a stream of N
2
(g). Residual
organics were then removed by a 10 min etch in a piranha
solution.
p
+
-Si emitter formation on n-Si
A boron-doped p
+
-Si radial emitter was formed on the n-Si
microwire arrays and on planar
h
111
i
n-Si wafers (Silicon Inc.,
0.7
U
cm) by exposure of the samples in a CVD furnace to a
20 : 400 sccm
ow of BCl
3
(Matheson, 0.25% in H
2
):H
2
(Matheson, research grade) at 950

C for 30 min, immediately
following a 30 s etch in 10% BHF. The samples were then rinsed
with H
2
O and dried under a stream of N
2
(g).
ITO deposition
Immediately following a 15 s etch in 10% BHF, a rinse in H
2
O
and drying under a stream of N
2
(g), 400 nm of In-doped tin
oxide was sputtered (48 W, 3 mTorr, 20:0.75 sccm Ar: 10% O
2
in
Ar) onto n
p
+
-Si microwire arrays and p-Si microwire arrays, by
DC magnetron sputtering under 10 W of substrate bias
(to facilitate conformal deposition on the microwire sidewalls).
The thickness of the ITO was determined by spectroscopic
ellipsometry measurements on a planar Si sample.
WO
3
deposition
n-WO
3
was electrodeposited from a tungstic peroxy-acid solu-
tion, as described previously.
26
Brie
y, 4.6 g of tungsten powder
(0.6
1
m
m, 99.99%, Sigma Aldrich) was dissolved in a molar
excess (60 mL) of H
2
O
2
(30% in H
2
O). Excess H
2
O
2
was
decomposed by addition of a trace amount of Pt black (99.9%,
Sigma Aldrich) for 24 h. The H
2
O
2
concentration was monitored
by peroxide test strips (EM Quant) until the
nal peroxide
concentration was <30 ppm. A concentrated stock solution was
made by addition of 80 mL of H
2
O and 60 mL of isopropyl
alcohol (IPA) to the as-made solution. To increase the lifetime of
the solution, the stock solution was protected from light and
stored at 2

C in a refrigerator. A 3 : 7 IPA : H
2
O mixture was
used to dilute the stock solution (3 : 2 IPA
H
2
O mix : stock
solution) to generate the deposition solution. Stock solutions
were used for one week and therea
er were freshly prepared. All
ITO-coated samples were used as prepared for deposition of
WO
3
, and were contacted directly to the ITO layer using a
at
alligator clip. Deposition of WO
3
on n
p
+
-Si and p-Si microwire
arrays was performed potentiostatically at

0.5 V
vs.
Ag/AgCl for
60 min. A
er deposition, all samples were annealed in air at
400

C for 2 h. This process formed monoclinic WO
3
,as
con
rmed by X-ray di
ff
raction data.
Electrode fabrication
Three types of electrodes were tested: n
p
+
-Si microwire arrays;
n
p
+
-Si/ITO/WO
3
microwire arrays; and p-Si/ITO/WO
3
micro-
wire arrays. To make ohmic contact to Si substrates that sup-
ported the microwire arrays, In-Ga (99.99%, Alfa-Aesar) eutectic
was scratched into the back side of the samples. Exposed In-Ga
(Si electrodes) was a
ffi
xed to a coiled Cu
Sn wire with Ag paint
(SPI 05001-AB). The active area (

0.12 cm
2
for Si microwire
arrays) was de
ned with epoxy (Loctite Hysol 9460) and the
entire electrode was sealed with epoxy to the bottom of a glass
tube (6 mm O.D.). The electrode orientation, down- or side-
facing, was determined by the orientation of the coiled wire
that protruded from the glass tube. Geometric areas were
measured by scanning the active area, and using so
ware
(ImageJ) to calculate the area.
Non-aqueous photoelectrochemistry
Bottom-facing electrodes that contained n
p
+
-Si microwire
arrays were etched for 30 s in 10% BHF immediately prior to
introducing the samples into a glove box. Solutions for photo-
electrochemical measurements consisted of CH
3
CN (anhy-
drous, 99.8%, Sigma Aldrich) distilled under N
2
(g) (ultra-high
purity, Air Liquide) from calcium hydride (CaH
2
), 1 M LiClO
4
(battery grade, 99.99%, Sigma Aldrich), and either 25 mM
bis(cyclopentadienyl) iron(
II
) (ferrocene, FeCp
2
0
, Sigma Aldrich)
and 3 mM bis(cyclopentadienyl) iron(
III
) tetra
uoroborate
(ferrocenium, FeCp
2
+
$
BF
4

, Sigma Aldrich), or 20 mM bis(cy-
clopentadienyl) cobalt(
II
) (cobaltocene, CoCp
2
0
, Sigma Aldrich)
and 2 mM bis(cyclopentadienyl) cobalt(
III
) hexa
uorophosphate
(cobaltocenium, CoCp
2
+
$
PF
6

, Sigma Aldrich). Cobaltocene
and ferrocene were puri
ed by vacuum sublimation at room
temperature, and cobaltocenium and ferrocenium were recrys-
tallized prior to use. An ELH-type W
halogen lamp with a
dichroic rear re
ector was used for illumination, and was set to
produce the same current density on a calibrated Si photodiode
as was obtained from 100 mW cm

2
of 1 Sun AM1.5G illumi-
nation. Three-electrode photoelectrochemical data were
obtained in a single-compartment cell by use of a Gamry
potentiostat (Reference 600), with a Pt counter electrode and a
Pt quasi-reference electrode at a scan rate of 20 mV s

1
.
Aqueous photoelectrochemistry
Side-facing planar and microwire array (p-Si/ITO/WO
3
,n
p
+
-
Si/ITO/WO
3
) devices were tested in 1 M H
2
SO
4
(aq.) (trace metal
grade, Fischer Scienti
c) saturated with air. Multiple devices
were tested (>5) with the reported results representing the best
performing devices. Two-electrode and three-electrode
measurements were conducted using a Biologic (SP-200)
potentiostat in a two-compartment cell (whose compartments
are referred to as the anode compartment and the cathode
compartment) that contained an epoxied (Loctite Hysol 1C)
quartz window on the anode side and a Na
on® (0.05 mm
thick, Alfa Aesar) membrane separator between the compart-
ments. Illumination was produced by a Xe lamp (Oriel 67005,
Newport Instruments) with an AM1.5G
lter (Newport
This journal is © The Royal Society of Chemistry 2014
Energy Environ. Sci.
,2014,
7
, 779
790 |
781
Paper
Energy & Environmental Science
Published on 16 December 2013. Downloaded by California Institute of Technology on 20/03/2014 14:45:35.
View Article Online
Instruments 81094) that produced light intensities of either
1 (100 mW cm

2
) or 12 (1080 mW cm

2
) Suns on a calibrated Si
photodiode. The concentration value (12

) was determined
using the AM1.5D spectrum to be consistent with what would be
expected in outdoor testing, although an AM1.5G source was
used. Calibration was performed such that the cited light
intensity was the highest light intensity anywhere in the cell and
the position of the photodiode at this light intensity was
marked to assure that the sample was positioned at the same
point as the photodiode. Optical concentration was achieved
using a plano-convex lens (Thorlabs LA4984). The spot size
(

1cm
2
) over
lled all of the samples tested (

0.1 cm
2
in area).
Three-electrode measurements
Three-electrode measurements were conducted at a scan rate of
20 mV s

1
with the working electrode in the anode compart-
ment open to air, an SCE reference electrode (CH Instruments,
CHI150) in the anode compartment, and a Pt mesh counter
electrode in the cathode compartment. Pt disc (0.0314 cm
2
)
three-electrode measurements were performed with the Pt disc
in the cathode compartment and with the SCE reference and the
Pt mesh counter electrodes in the anode compartment. For Pt
disc measurements, the cathode compartment was saturated
with either Ar(g) (research grade, Air Liquide) or H
2
(g) (research
grade, Air Liquide) by bubbling the gas through the solution for
15 min before testing, as well as throughout the experiment. All
of the measurements were referenced to the potential of the
regular hydrogen electrode (RHE) (
E
(H
+
/H
2
)), obtained empiri-
cally using the Pt disc electrode under 1 atm of H
2
(g),

0.247 V
vs.
SCE.
Two-Electrode measurements
Two-electrode measurements were made in the same two-
compartment cell as used for the three-electrode measure-
ments and with the electrodes in the same physical location.
The cathode compartment was purged with Ar(g) or H
2
(g). A Pt
disc electrode was used as the cathode, to simulate the expected
catalyst area for an integrated device. The anode compartment
contained the working electrode, which was illuminated
through a quartz window and was open to air. Chro-
noamperometric measurements were taken under potentio-
static control at 0 V applied bias between the photoanode and
cathode.
Load-line analysis
The Pt disc current versus potential (
I
E
) data were mirrored
about the abscissa, to facilitate straightforward evaluation of
the projected electrochemical operating conditions of the
device. The Pt disc data included the solution and membrane
resistances, because during the measurement the reference
electrode was in the opposite cell from the Pt electrode. Hence,
the Pt disc data should account for all of the expected cell
resistances in the system of interest. The predicted operating
currents from the load-line analysis were calculated based on
the average of the forward and reverse scans and were
6.7

10

3
mA (6.1

10

2
mA cm

2
) and 2.6

10

3
mA
(2.1

10

2
mA cm

2
) for the Ar(g)- and H
2
(g)-saturated solu-
tions, respectively. The current densities were calculated per
geometric surface area of the electrode and were determined
using Fig. 5a. E
ffi
ciencies were calculated using the Gibbs
free energy for water splitting (
D
E
0
¼
1.23 V) into H
2
(g) and O
2
(g)
at standard conditions to obtain the energy content of the
fuel produced, and the known irradiance (12 suns AM1.5D,
1080 mW cm

2
).
Product analysis
Oxidation products (peroxydisulfate (S
2
O
8
2

)) generated at the
WO
3
/1.0 M H
2
SO
4
interface were detected using an ultraviolet-
visible spectrophotometer (Agilent 8453, 1 cm quartz cuvette)
as reported previously.
26
Calibration curves were determined
using potassium peroxydisulfate (K
2
S
2
O
8
).
Reduction products (H
2
(g)) generated at the Pt disc/1.0 M
H
2
SO
4
interface were detected using a mass spectrometer (Hiden
Analytical HPR-20 QIC). Current was passed through the Pt disc
electrode for 40 min under identical conditions (identical elec-
trochemical cell, current (

6.5
m
A), solution (trace metal grade
1.0 M H
2
SO
4
) and Ar(g) purge) as used in the two-electrode
experiment described above. The experiment was started only
a
er obtaining a steady background signal for
m
/
z
¼
2.
Light absorption simulation
1D and 2D light absorption was simulated in Lumerical FDTD, a
commercially available Maxwell
s equation solver that uses the
FDTD method. The so
ware requires material-speci
c refractive
index data (Table S1 and Fig. S1
). The experimentally fabricated
microwire structures were reproduced in the Lumerical work-
space in 2D. Bloch boundary conditions were used to model an
in
nite planar structure and an in
nite 2D microwire array.
Each structure was illuminated with single-wavelength plane
waves with the electric
eld polarized in the 2D structured plane,
at wavelengths ranging from 350 to 1100 nm in 50 nm intervals.
Partial spectral averaging was used to remove simulation arti-
facts that were caused by the use of single-wavelength simula-
tions. The structure was meshed with 20 mesh boxes per
wavelength. The spatially resolved electric
eld,
E
, and complex
refractive index (
3
) were recorded and then used to calculate the
spatially resolved carrier generation rate,
C
gen
(eqn (1)):
C
gen
¼
p
j
E
j
2
imag
ð
3
Þ
h
(1)
where
h
is Planck
s constant. The spatially resolved carrier
generation rate was used as the optical input for the electronic
simulations. The power absorbed in each material was calcu-
lated by integrating the spatially resolved absorbed power,
P
abs
(eqn (2)).
P
abs
¼
0.5
u
|
E
|
2
imag(
3
)
(2)
The absorbed photon
ux in each material as a function of
wavelength was weighted with the AM1.5G spectrum, integrated
over wavelength, and multiplied by Faraday
s constant to obtain
a short-circuit current density assuming unity internal quantum
782
|
Energy Environ. Sci.
,2014,
7
, 779
790
This journal is © The Royal Society of Chemistry 2014
Energy & Environmental Science
Paper
Published on 16 December 2013. Downloaded by California Institute of Technology on 20/03/2014 14:45:35.
View Article Online
yield (IQY). The concentrated illumination modeling was per-
formed at 11 Suns using the AM1.5G spectrum to match the
experimental photon
ux of 12

of AM1.5D.
n
p
+
-Si junction modeling
Electronic device simulations were performed in Synopsys
Sentaurus, a commercially developed so
ware package that
solves the dri
di
ff
usion equation for charge carriers using a
nite-element method. For simplicity, a single n
p
+
-Si homo-
junction and an n-WO
3
/(O
2
/H
2
O) liquid junction were modeled
separately. The built-in materials parameter
le for Si was used
with modi
ed time constants (
s
n
(
N
A
¼
10
20
cm

3
)
¼
3

10

6
s;
s
p
(
N
D
¼
10
17
cm

3
)
¼
1

10

3
s), with
s
n
and
s
p
being the
electron and hole lifetime, respectively, and
N
A
and
N
D
being
the acceptor and donor concentrations, respectively. The silicon
n
p
+
junction was constructed with a 100
m
m thick n-region
with
N
D
¼
10
17
cm

3
and a 0.2
m
m thick p
+
-region with
N
A
¼
10
20
cm

3
. In the quasi-neutral bulk of the n-Si, a standard
mesh size of 500 nm and 5
m
m was used transverse and parallel
to the junction, respectively. Near the ohmic contact with n-Si,
the mesh was re
ned to 500 nm and 100 nm, and near the
junction, the mesh was re
ned to 500 nm and 20 nm, to accu-
rately model the band bending in these regions. The
J
E
(current density
vs.
potential) characteristics of this structure
were obtained by
rst solving for the
V
¼
0 case in the dark.
Subsequently, the voltage was stepped at 0.010 V intervals in
both the positive and negative directions, to obtain the dark
J
E
behaviour. The carrier-generation rate from Lumerical was then
applied to extract the
J
E
characteristics in the presence of
illumination. Similarly, the
V
¼
0 case in the light was solved
rst, and then the voltage was stepped at 0.010 V intervals, to
obtain the light
J
E
performance. Shockley
Read
Hall recom-
bination was used for all simulations.
WO
3
/liquid junction modeling
The built-in
oxide as semiconductor
materials parameter
le
was used to model WO
3
, with the following parameters and
their values in parentheses: modi
ed band gap (
E
g
¼
2.6 eV),
work function (
c
¼
4.4 eV), relative permittivity (
3
r
¼
5.76),
e
ff
ective conduction- and valence-band density of states (
N
C
¼
1.8

10
19
cm

3
,
N
V
¼
7.1

10
19
cm

3
), recombination time
constant (
s
n
¼
s
p
¼
1

10

8
s) and mobility (
m
n
¼
m
p
¼
40 cm
2
V

1
s

1
). The band gap was experimentally measured from
absorption measurements using an integrating sphere and a
Tauc plot. The relative permittivity was calculated from ellip-
sometric measurements of the complex refractive index. The
work function was chosen based on reports in the literature.
27
The density of states can be calculated from
m
*
, the e
ff
ective
mass of holes in the valence band and of electrons in the
conduction band,
k
B
, the Boltzmann constant,
h
, the Planck
constant and
T
, the temperature (eqn (3)):
N
c
¼
2

2
p
m
*
e
k
B
T
h
2

3
=
2
(3)
E
ff
ective masses in the conduction band have been reported
to be

0.8
m
0
,
28
where
m
0
is the mass of a free electron. Density
functional theory calculations of the band structure of WO
3
indicate that the valence band has less curvature than the
conduction band, indicating heavier holes and leading to an
estimate of 2
m
0
for the hole e
ff
ective mass.
29
The mobility
values were also taken from the literature.
28
Preliminary exper-
imental measurements indicated a di
ff
usion length of 1
m
m,
thereby determining the time constant.
The WO
3
/liquid junction was modeled as a Schottky junc-
tion, with the metal work function equal to the water oxidation
redox potential,
c
¼
5.68 eV, which was in contact with a 1
m
m
thick slab of WO
3
. A value of
N
D
¼
10
15
cm

3
was chosen to
match the experimentally observed short-circuit current density
and open-circuit voltage. Mesh sizes of 10 nm and 250 nm were
used perpendicular and parallel to the junction, respectively.
The method to obtain the dark and light
J
E
behaviour was
identical to that used for modeling the Si junction. Shockley-
Read-Hall recombination and thermionic emission physics
were used for these simulations.
Hydrogen-evolution catalysis modeling
Butler
Volmer kinetics in the absence of mass-transport limi-
tations were used with a Tafel slope,
a
¼
1 and an exchange
current density,
j
0
¼
10

3
Acm

2
, to simulate the cathodic
overpotential (eqn (4)) of platinum (Pt) for hydrogen evolution
in 1.0 M H
2
SO
4
:
h
¼
RT
F
ln

j
j
0
þ
1

(4)
Where
R
is the gas constant,
T
is the absolute temperature,
and
F
is Faraday
s constant. The overpotential,
h
, was added to
the n
p
+
-Si homojunction
J
E
data at the same current density
to yield a simulated hydrogen generation device curve in the
absence of mass transport.
Results
Fig. 1a
f depicts the process used to fabricate the on-wafer
devices used herein. Fig. 1g displays an image of a completed
wire-array device, while Fig. 1h shows a cross-section of a single
wire demonstrating the layered device structure. The Si micro-
wires were 40
70
m
m in length, had a diameter of

2
m
m and
had doping densities on the order of 10
17
cm

3
. Secondary-ion
mass spectrometry data from planar samples (Fig. S2
) indi-
cated that the p
+
-Si emitter thickness was

200 nm. The
sequential, conformal layers of ITO and WO
3
were

100 nm and

400 nm, respectively. Fig. S3
and Fig. S4
further con
rm the
layered structure of the tandem microwire device using EDX
analysis.
Device design
Fig. 2a shows the core
shell Si/WO
3
microwire array design, and
Fig. 2b shows an individual two-dimensional unit cell of the
design. The highly doped p
+
-Si sheath surrounds a moderately
doped n-Si microwire core, creating a radial n
p
+
-Si buried
This journal is © The Royal Society of Chemistry 2014
Energy Environ. Sci.
,2014,
7
, 779
790 |
783
Paper
Energy & Environmental Science
Published on 16 December 2013. Downloaded by California Institute of Technology on 20/03/2014 14:45:35.
View Article Online
junction. A conformal layer of ITO surrounds the p
+
-Si, forming
a transparent ohmic contact between the p
+
-Si and the
conformal layer of n-WO
3
. A liquid junction is formed between
the WO
3
and the solution redox couple (shown as the potential
of the oxygen-evolution reaction (OER)).
Fig. 2c and d depict the device electronic band structures in
the absence and presence of illumination, respectively. Illumi-
nation (Fig. 2d) results in splitting of the quasi-Fermi levels at
both junctions, generating two voltage sources in series.
Photoexcited majority-carrier electrons in the n-Si core are
transported axially to the back contact through the degenerately
doped substrate (
n
+
-Si) to perform the hydrogen-evolution reac-
tion (HER) at a Pt counter electrode, while photoexcited minority-
carrier holes are collected radially in the p
+
-Si sheath. The holes
in Si recombine with photoexcited majority-carrier electrons
from the n-WO
3
at the ITO contact, while minority-carrier holes
that are photoexcited in the n-WO
3
are collected at the liquid
interface and drive the oxidation of water or anolyte.
26
Fig. 1
(a) Photolithographically patterned n
+
-Si
h
111
i
wafer with a SiO
2
mask layer and Cu catalyst in the desired growth pattern. (b) VLS Cu-
catalyzed growth of n-type Si microwires on an n
+
-Si substrate followed by a metal etch (RCA 2). (c) SiO
2
di
ff
usion barrier (boot) formation
via
SiO
2
growth, PDMS in
fi
ll, HF etch and PDMS removal. (d) p
+
-Si emitter drive-in from BCl
3
precursor at 950

C for 30 min in a CVD furnace. (e)
Conformal DC sputter coating of ITO. (f) Conformal n-WO
3
electrodeposition and annealing at 400

C for 2 h. (g) Fully assembled tandem
junction device array SEM (scale bar
¼
10
m
m). (h) Cross-sectional SEM of a fully assembled tandem junction single wire demonstrating the
layered structure of the device (scale bar
¼
500 nm).
Fig. 2
(a) Tandem junction microwire array with a buried homojunction (n
p
+
-Si) coated by ITO and n-WO
3
. (b) 2D cross-section of an
individual tandem junction array unit cell. (c) Electronic structure of the tandem device in the dark showing the buried n
p
+
-Si junction, ohmic Si/
ITO/n-WO
3
junction, and n-WO
3
/liquid junction. The device is shown equilibrated with the oxygen-evolution potential. (d) Steady-state
electronic structure of the tandem device under illumination with the carrier movement directions shown. Both the oxygen- and hydrogen-
evolution potentials are shown, with the overpotentials accounted for by the di
ff
erence between the respective quasi-Fermi level and reaction
potential. Electrons and holes are collected radially in the n-WO
3
. Holes are collected radially in the n
p
+
-Si and electrons are collected axially at
the back contact.
784
|
Energy Environ. Sci.
,2014,
7
, 779
790
This journal is © The Royal Society of Chemistry 2014
Energy & Environmental Science
Paper
Published on 16 December 2013. Downloaded by California Institute of Technology on 20/03/2014 14:45:35.
View Article Online
During the course of this work, a similar, yet distinct, device
that performs unassisted solar-driven water splitting was
reported that incorporates a p
n
+
-Si nanowire junction con-
nected to TiO
2
by the p-Si nanowire core (referred to as Si/
TiO
2
).
22
One major di
ff
erence between the device design pre-
sented here (referred to as Si/WO
3
) and the Si/TiO
2
design is the
collection probability of excess charge carriers in Si. Excess
minority carriers are collected radially throughout the Si wire
for the Si/WO
3
design, maintaining collection lengths (<2
m
m)
shorter than reported minority-carrier di
ff
usion lengths
(

10
m
m).
25,30
Conversely, excess minority carriers are collected
axially in the top half of the Si/TiO
2
design, requiring collection
lengths much longer than the largest di
ff
usion lengths
measured in Si microwires (

10
m
m).
25,30
Optical absorption
modeling of Si microwires indicates that a majority of the
incident light is absorbed near the top of the wire, emphasizing
the need for e
ffi
cient minority-carrier collection in this region.
31
Si nanowires also have excessive junction area that leads to high
rates of carrier recombination relative to Si microwires.
4
These
di
ff
erences are apparent when comparing the open-circuit
voltage values for the Si/TiO
2
device (370 mV) and the Si/WO
3
device (480 mV). The Si/TiO
2
device utilized two masking steps
to de
ne the structure and a top-down fabrication process that
began with a high-quality photo-active p-Si wafer, whereas
fabrication of the Si/WO
3
device consisted of sequential depo-
sition of the active materials with a single masking step and
featured a bottom-up fabrication process from a reusable
photo-inactive n
+
-Si substrate.
14
n
p
+
-Si microwire non-aqueous photoelectrochemistry
The performance of individual buried junction n
p
+
-Si micro-
wire array devices was investigated through non-aqueous pho-
toelectrochemical measurements in contact with a series of one-
electron, outer-sphere redox couples (Fig. 3). The cobaltocene
+/0
(CoCp
2
+/0
) and ferrocene
+/0
(FeCp
2
+/0
) redox species were used to
determine the quality of the buried junction, by probing the n-
and p-type character at the Si
liquid interface. Fig. 3b displays
the redox potentials of CoCp
2
+/0
and FeCp
2
+/0
with respect to the
potentials of the Si conduction and valence bands. The p
+
-Si
radial sheath is expected theoretically, and was observed
experimentally, to form an ohmic contact to FeCp
2
+/0
and to
form a tunnel junction to CoCp
2
+/0
. For a fully buried junction,
similar performance should therefore be observed in contact
with both of these redox systems. However exposed n-Si is
expected theoretically and, is observed experimentally, to form
an electrical short (
i.e.
, an ohmic contact) to CoCp
2
+/0
and a
rectifying contact to FeCp
2
+/0
.
In contact with FeCp
2
+/0
under 100 mW cm

2
of simulated
AM1.5G illumination, the n
p
+
-Si microwire array devices
exhibited an open-circuit potential of
E
oc
¼
0.5 V
versus
solution,
and a short-circuit current density of
J
sc
¼
8.7 mA cm

2
,witha
ll
factor of 0.44 (Fig. 3a). Similar performance was observed for
microwire structures in contact with CoCp
2
+/0
indicating that the
performanceisduetotheburiedn
p
+
-Si junction.
Fig. 3
(a) Non-aqueous photoelectrochemical (forward scan, scan rate
¼
20 mV-s

1
) results using ferrocene
+/0
((Cp)
2
Fe
+/0
) and cobaltocene
+/0
((Cp)
2
Co
+/0
) as redox couples to probe the n
p
+
-Si buried junction performance in microwire arrays. The dark scans are dashed lines and the
light scans are shown as solid lines. (b) Redox potentials of cobaltocene and ferrocene with respect to the potentials of the conduction-band
edge and the valence-band edge of Si.
Fig. 4
Three electrode photoelectrochemical (forward scan, scan
rate
¼
20 mV s

1
) performance for single (black) and tandem (blue)
junction microwire devices in contact with 1.0 M H
2
SO
4
(aq.). The
single junction microwire device consisted of WO
3
supported on p-Si
microwires that had been coated with ITO. Here the p-Si/ITO contact
is ohmic so the only rectifying junction is at the WO
3
/liquid junction.
These data demonstrate the presence of an additive voltage from each
junction, with 0.73 V and 0.5 V produced by the WO
3
/liquid and n
p
+
-
Si buried junctions, respectively. The
E
oc
for the single junction device
was de
fi
ned as the point at which the dark current, due to capacitive
charging, and the illuminated current separated. The
E
oc
for the
tandem junction device was de
fi
ned as the point at which no current
was
fl
owing because no positive dark current existed in this region.
This journal is © The Royal Society of Chemistry 2014
Energy Environ. Sci.
,2014,
7
, 779
790 |
785
Paper
Energy & Environmental Science
Published on 16 December 2013. Downloaded by California Institute of Technology on 20/03/2014 14:45:35.
View Article Online
The 0.1 V decrease in
E
oc
observed for the CoCp
2
+/0
contact
compared to the FeCp
2
+/0
contact is consistent with the pres-
ence of exposed n-Si near the SiO
2
/Si boundary (boot) creating
an electrical short to solution.
12
This is important to note,
however it has been shown that the n-Si/ITO interface produces
a barrier rather than an electrical short like n-Si/CoCp
2
+/0
.
24
Thus, the n
p
+
-Si-microwire part of the ultimate tandem
structure is expected to contribute
V
oc
¼
0.5 V and is not
expected to limit the current of the tandem device, because the
maximum possible
J
sc
from WO
3
is

5mAcm

2
.
32
Single (WO
3
) and tandem (Si/WO
3
) junction
photoelectrochemistry
Fig. 4 shows the photoelectrochemical behaviour of single
junction (p-Si/ITO/n-WO
3
/1.0 M H
2
SO
4
) and tandem junction
(n-Si/p
+
-Si/ITO/n-WO
3
/1.0 M H
2
SO
4
) microwire array devices
under simulated one Sun illumination conditions. The p-Si/ITO
and p
+
-Si/ITO contacts have been shown to produce ohmic
behaviour allowing isolation of the n-WO
3
/1.0 M H
2
SO
4
liquid
junction performance in the single junction case and e
ffi
cient
use of the buried n
p
+
-Si junction in the tandem junction case.
24
The single- and tandem-junction microwire devices exhibited
J
¼
0.50 mA cm

2
and
J
¼
0.58 mA cm

2
, respectively, at the
formal potential for oxidation of water to O
2
,
E
o
0
(O
2
/H
2
O). The
rst peak in photocurrent density is a dark redox process that
results in the photochromism of WO
3
, whereupon reverse scans
the WO
3
lm is reduced through proton intercalation, and is
subsequently oxidized on the forward scan. The second peak is
associated with photocurrent that results in actual solution
redox reactions. The slightly lower current density exhibited by
the single junction is consistent with decreased absorption due
to the use of shorter microwire arrays. For comparison, Fig. S5
displays data from the planar equivalent of these microwire
devices.
The open-circuit potentials were
E
oc
¼
0.73 V
vs.
E
o
0
(O
2
/H
2
O) and
E
oc
¼
1.21 V
vs. E
o
0
(O
2
/H
2
O) for the single-
and tandem-junction devices, respectively. The
E
oc
for the
WO
3
/liquid contact is in accord with expectations for WO
3
photoanodes operating under these conditions.
19
The 0.48 V
shi
in
E
oc
of the tandem junction device relative to the single
junction device is therefore attributable to the presence of the
n
p
+
-Si buried junction in the tandem device (Fig. 3). This
voltage shi
demonstrates that the buried n
p
+
-Si junction
increases the voltage generated by Si as compared to n-Si/n-type
metal
oxide heterojunction devices.
18,19
Load-line analysis
Under modest optical concentration (12 Suns, AM1.5D),
tandem junction microwire-array devices exhibited
E
oc
¼

1.27 V
vs. E
0
0
(O
2
/H
2
O), which exceeds the 1.23 V potential
di
ff
erence necessary for unassisted water splitting under
standard-state conditions (Fig. 5a and S6
). The operating
current for this device under modest optical concentration can
be predicted using a load-line analysis.
33
Fig. 5a shows the
I
E
behaviour of an illuminated tandem microwire device, along
with the
I
E
behaviour, mirrored about the abscissa, of a Pt disc
electrode of similar projected area, in contact with either a
saturated Ar(g) or H
2
(g) solution at 1 atm. In an Ar(g)-saturated
solution, the onset potential for the HER was shi
ed positive
compared to that observed in a H
2
(g)-saturated solution, in
accordance with Le Ch
ˆ
atelier
s principle. These data were
obtained using a two-compartment cell (Fig. S7
) with a Na
on
membrane separating the anode and cathode compartments.
The
I
E
behaviour of the Pt disc includes the solution and
membrane resistances of the electrochemical cell, because the
reference electrode was placed in the opposite (anode)
compartment. This type of measurement provides a robust
prediction of the unassisted operating current that should be
obtained between an illuminated Si/WO
3
microwire array device
and a Pt button electrode in the same geometry and physical
location. The anolyte was not purged with O
2
because the
primary oxidation product from WO
3
under these conditions
has been shown to be peroxydisulfate.
21
Fig. 5b displays the chronoamperometric response from a
two-electrode experiment at 0 V applied bias between an illu-
minated tandem junction WO
3
/Si microwire array device and a
Pt disc electrode. The devices produced solar-to-hydrogen
energy-conversion e
ffi
ciencies of 0.0068% (6.5

10

3
mA,
0.060 mA cm

2
) and 0.0019% (1.9

10

3
mA, 0.017 mA cm

2
)
when the Pt disc was in contact with Ar(g)- and H
2
(g)-saturated
solutions, respectively, which agree with the predicted oper-
ating points from the load-line analysis (dots at intersection
points in Fig. 5 insets). The peroxydisulfate/sulfate redox system
has a formal reduction potential that is

0.75 V positive of
E
o
0
(O
2
/H
2
O), indicating that the tandem core
shell microwire
device generated

1.8 V of photopotential under these condi-
tions. Device photostability was demonstrated for over 10 min
by the H
2
(g)-purged device. Thus, the decrease in current for the
Ar(g)-purged device is attributable to an increasing H
2
concen-
tration in solution from the HER at the Pt disc electrode.
Fig. 5
(a)
J
E
curves and load-line analysis of the tandem junction (n
p
+
-Si/ITO/WO
3
) microwire device at 12 suns (AM1.5D) plotted against
the dark HER curves (mirrored about the abscissa) using a Pt disc
electrode in an Ar(g)- or H
2
(g)-saturated solution. These measure-
ments were conducted in a two-electrode cell with 1 M H
2
SO
4
in both
compartments separated by a Na
fi
on membrane to maintain product
separation. The Pt disc HER curves include solution and membrane
resistances because the reference electrode (SCE) was placed in the
opposite cell at the same location used for the tandem microwire array
device. The inset is a zoomed-in view around the operational points,
which are indicated by the red (Ar) and blue (H
2
) circles. (b) Two-
electrode measurements at 0 V applied bias between the tandem
junction device (concentrated illumination) and Pt disc electrode in
either Ar(g)- or H
2
(g)-saturated solution. Turning the light o
ff
,as
indicated, demonstrated that the positive current was photoinduced.
786
|
Energy Environ. Sci.
,2014,
7
, 779
790
This journal is © The Royal Society of Chemistry 2014
Energy & Environmental Science
Paper
Published on 16 December 2013. Downloaded by California Institute of Technology on 20/03/2014 14:45:35.
View Article Online
Additionally, the chopped-light response demonstrated that the
observed current was photo-induced. The negative current
observed in the dark for the H
2
(g)-purged device is consistent
with O
2
(g) and H
2
(g) recombination to form water, similar to
fuel cell operation. This behaviour demonstrates that the
operating voltage of the device can be tuned by changing the
partial pressures of the photoelectrochemical reaction prod-
ucts,
i.e.
by changing the chemical load across the device, which
can be calculated using the Nernst equation.
Product analysis
Product analysis was performed separately on the oxidation and
reduction products in 1.0 M H
2
SO
4
. As reported previously,
electrolyte bu
ff
er species are oxidized preferentially at the WO
3
/
liquid interface, relative to the oxidation of water.
26
Thus, in
contact with H
2
SO
4
(aq.), sulfate (SO
4
2

) is preferentially
oxidized to peroxydisulfate (S
2
O
8
2

), which was con
rmed as an
oxidative product by ultraviolet-visible absorption spectroscopy
as published previously.
26
Although direct oxygen evolution was
not realized due to the slow O
2
evolution kinetics of WO
3
, per-
oxydisulfate has been shown to stoichiometrically evolve O
2
using Ag
+
as a catalyst.
26
At the Pt disc cathode, H
2
(g) production was detected by
mass spectrometry of the reaction products when the
operational current density was passed at the Pt disc electrode
(Fig. S8
). Due to the small amount of H
2
(g) produced, direct
quanti
cation of the faradaic e
ffi
ciency was not performed,
however no other products are expected due to the use of trace
metal grade H
2
SO
4
.
1D optoelectronic model
Fig. 6 shows the
J
E
behaviour and simulated load-line analyses
of the one-dimensional device architecture (Fig. S9
) in the bulk
recombination limit under both unconcentrated (1 Sun) and
concentrated illumination su
ffi
cient to match the experimental
photon
ux (see Fig. S10
for the individual
J
E
behaviour of the
Si and WO
3
junctions). Here, the Si homojunction performance
is shown in the absence of mass-transport limitations, with and
without the incorporation of catalytic overpotentials (
h
) asso-
ciated with the HER, which are present in any actual water-
splitting device. The maximum predicted operating points
from this analysis are 0.5 mA cm

2
(Fig. 6a inset) and 5.7 mA
cm

2
(Fig. 6b inset) for unconcentrated (one Sun) and concen-
trated illumination, respectively. This corresponds to an
E
oc
¼

1.29 V
vs. E
0
0
(O
2
/H
2
O) for unconcentrated illumination and
E
oc
¼
1.44 V
vs. E
0
0
(O
2
/H
2
O) for concentrated illumination.
Fig. S11
compares the modeled and experimentally measured
J
E
behaviour for a Pt electrode performing the HER under
1 atm of H
2
. The di
ff
erences arise from the exclusion of mass
transport and series resistance in the model.
2D microwire optical modeling
Table 1 shows the light-limited photocurrent densities (
J
ph
) that
were calculated assuming unity internal quantum yield (IQY)
for the tandem structure for varying WO
3
coating thicknesses:
(i) the entire device (Si and WO
3
), (ii) only WO
3
, and (iii) WO
3
coating the microwire sidewalls only. Fig. S12
shows the
simulated carrier-photogeneration-rate maps for photons with
energy larger than the WO
3
band-gap energy (

2.6 eV, 476 nm).
Table S2
lists the geometric parameters used, based on the
structure depicted in Fig. 2b.
Discussion
Tandem junction performance
The
E
oc
¼
1.27 V
vs. E
o
0
(O
2
/H
2
O) exhibited by the tandem
junction n
p
+
-Si/ITO/n-WO
3
microwire array device indicates
that the structure provides enough voltage to drive unassisted
Fig. 6
Modeled
J
E
data and load-line analysis for unconcentrated (a)
and concentrated (b) illumination conditions that match the experi-
mental photon
fl
uxes at 1 and 12 Suns (AM 1.5D). The Si homojunction
is shown with (green) and without (black) inclusion of realistic
hydrogen-evolution catalytic overpotentials. Butler
Volmer kinetics
with
a
¼
1 and
j
0
¼
10

3
Acm

2
was used to calculate the catalytic
overpotentials in the absence of mass-transport limitations. The insets
show the operating point for unconcentrated illumination (black dot)
and for concentrated illumination, with (green dot) and without (black
dot) overpotentials due to HER catalysis included. The rate of catalysis
is not expected to a
ff
ect the WO
3
J
E
behaviour due to the large band
gap of WO
3
and the proximity of
E
o
0
(O
2
/H
2
O) to the potential of the
conduction band of WO
3
.
Table 1
Dependence of the light-limited photocurrent density (
J
ph
) on the WO
3
coating thickness for light above the WO
3
band gap (<476 nm)
in: (i) all photoactive material in the core
shell tandem structure (Si + WO
3
), (ii) WO
3
and (iii) WO
3
on the microwire sidewalls only (WO
3
on
microwire).
J
ph
is calculated assuming unity IQY.
WO
3
Thickness
(nm)
Geometric
lling
fraction (%)
Si + WO
3
J
ph
<476 nm (mA cm

2
)
WO
3
J
ph
(mA cm

2
)
WO
3
on microwire
J
ph
(mA cm

2
)
300
11.7
3.34
0.56
0.41
500
15.4
3.43
0.7
0.54
700
19.6
3.47
0.8
0.65
1000
26.9
3.53
0.89
0.79
1500
41.7
3.52
1.08
1.07
This journal is © The Royal Society of Chemistry 2014
Energy Environ. Sci.
,2014,
7
, 779
790 |
787
Paper
Energy & Environmental Science
Published on 16 December 2013. Downloaded by California Institute of Technology on 20/03/2014 14:45:35.
View Article Online
solar-driven water splitting under standard-state conditions
and 12 Suns illumination (AM1.5D). Additionally, H
2
(g) was
detected as a reduction product con
rming unassisted
hydrogen production. To realize a complete, direct water-
splitting device, an oxygen-evolution catalyst coupled to the
WO
3
surface would be required. Two-electrode operation with a
Pt HER electrode demonstrated stable operation of the device as
well as validation of the operating point determined by the load-
line analysis. This demonstration therefore provides a proof-of-
concept for the development of core
shell high-aspect ratio
tandem junction devices for fuel formation directly from
sunlight. Such a device could be embedded in a gas-
impermeable, ion-selective membrane
15
and removed from a
reusable substrate
14
to form a free-standing device. This archi-
tecture minimizes the distance ions must travel to complete the
fuel-forming circuit, thus minimizing the potential drop due to
solution resistance e
ff
ects.
1D optoelectronic modeling
The performance of the tandem junction n
p
+
-Si/ITO/n-WO
3
device described herein is fundamentally limited in two ways by
WO
3
: (i) its wide band gap (

2.6 eV) limits the maximum
current density to

5mAcm

2
under AM1.5G illumination
conditions; and (ii) the barrier height between WO
3
and
E
o
0
(O
2
/H
2
O) limits the
E
oc
to less than half of the band gap.
Given these two limitations, a one-dimensional optoelectronic
model (Fig. S9
) for the tandem structure was developed to
investigate the maximum performance that could be expected
from the Si/WO
3
device and can be compared to the experi-
mental results.
The modeled
E
oc
values are 80 mV and 170 mV larger,
respectively, than the experimentally observed
E
oc
values, which
can be ascribed to an increased junction area (

9

)inthe
experiment as compared to the planar model. These open-circuit
potential di
ff
erences manifest themselves as large di
ff
erences
between the modeled and experimental two-electrode operating
points due to the proximity of the modeled operating point to
the maximum power point in the WO
3
J
E
behaviour; any
unaccounted resistances and/or open-circuit potential losses
between the model and experiment, such as junction area, will
therefore cause a precipitous decrease in the experimental
operating current density. This e
ff
ect is observed in both
experimental illumination conditions, where under one Sun
illumination the
E
oc
is less than 1.23 V such that the device
cannot perform unassisted water splitting at standard condi-
tions. Under concentrated illumination the experimental
E
oc
exceeds 1.23 V, but is lower than the modeled value and explains
the cause for the di
ff
erence in operating points.
2D optical modeling
As shown in Fig. 2a and b, WO
3
forms a conformal coating on
the Si microwires and the planar, degenerately doped Si base, all
of which can result in photocurrent. To investigate the photo-
current contribution from WO
3
on the microwire sidewalls
relative to that from WO
3
on the planar, degenerately doped Si
base, two-dimensional
nite-di
ff
erence time domain (FDTD)
electromagnetic modeling was performed on the structure
presented in Fig. 2b. Table 1 demonstrates that for thin WO
3
coatings (300 nm), only 17% (0.56 mA cm

2
) of the above band-
gap light absorbed by the device is absorbed in the WO
3
. The
WO
3
absorption fraction increases to 31% (1.08 mA cm

2
) for
thicker WO
3
coatings (1.5
m
m). The carrier-generation-rate
maps (Fig. S12
) demonstrate that the majority of the WO
3
absorption is within the top 10
m
m of the device, with many
photons whose energies are larger than the energy of the WO
3
band gap transmitted through the WO
3
to the underlying Si,
where absorption is not useful due to the current-limiting
absorption in the WO
3
. This behaviour implies that an alter-
native WO
3
geometry is desired to enhance the WO
3
absorption.
However, a device geometry designed to increase the WO
3
absorption should optimally accommodate the

1
m
m
minority-carrier di
ff
usion length of WO
3
, implying the bene
-
cial use of WO
3
layers <1
m
m thick.
For 300 nm thick WO
3
coatings, 73% of the total WO
3
optical
absorption occurred in WO
3
on the sidewalls of the Si micro-
wires. This absorption fraction increased to 99% for 1.5
m
m
thick WO
3
coatings; the 500 nm thick WO
3
coatings used
experimentally are projected at

76%. This substantial fraction
of absorption along the sidewalls
versus
at the bottom of the
device architecture indicates that similar performance is
expected for on-wafer microwires compared to free-standing
microwire array devices that have been removed from the
growth wafer, which will be a crucial step for integration of this
tandem device into a fully functional solar fuels generator.
Additionally, scattering particles could be introduced to redirect
more light toward the microwire sidewalls for enhanced WO
3
absorption.
1,10
Toward higher e
ffi
ciency devices
Integration of new photoanode materials in place of WO
3
has
the potential to increase the performance of the tandem device
by producing more negative
E
oc
values as well as much larger
values of the current density at
E
¼
E
o
0
(O
2
/H
2
O). To produce a
more negative value of
E
oc
, the potential of the conduction band
of the anode material must be more negative than the potential
of the conduction band of WO
3
,
i.e.
closer to the vacuum level,
thereby increasing the barrier height at the semiconductor/
liquid junction. Recent studies of mixed-metal oxides have
demonstrated photoanode materials with smaller electron
a
ffi
nities than WO
3
.
34
36
The production of increased current
density at
E
¼
E
o
0
(O
2
/H
2
O) will require lowering the recombi-
nation rates, by improving the material quality and passivating
surface states, as well as the discovery of narrower band-gap
materials that are stable under oxidizing conditions. Addition-
ally the anodes must be stable under conditions where the
cathode and membrane materials are stable, and under
conditions where the membrane exhibits high transference
numbers for protons, to allow for e
ff
ective, passive neutraliza-
tion of the pH gradient between the sites of water oxidation and
water reduction while maintaining product separation for
intrinsically safe operation of the system under varying levels of
illumination.
788
|
Energy Environ. Sci.
,2014,
7
, 779
790
This journal is © The Royal Society of Chemistry 2014
Energy & Environmental Science
Paper
Published on 16 December 2013. Downloaded by California Institute of Technology on 20/03/2014 14:45:35.
View Article Online
Conclusions
A Si/WO
3
integrated tandem junction device capable of unas-
sisted solar-driven water-splitting has been developed and used
to demonstrate unassisted hydrogen evolution under moderate
light concentration. This system provides proof-of-principle for
the design. The approach is attractive because it provides
materials
exibility for the Si tandem partner absorber, an
optimized electrochemical geometry, embedment in a
exible,
gas-impermeable, ion-exchange membrane and enhanced
absorption and carrier-collection properties relative to planar
designs. The Si/WO
3
described herein demonstrated additive
voltages across the tandem device resulting in an
E
oc
¼
1.21 V
vs. E
o
0
(O
2
/H
2
O). Modest optical concentration (12 Suns)
produced a shi
in
E
oc
to potentials negative of
E
o
0
(H
2
O/H
2
),
indicating that the device could split water in an unassisted
fashion. Two-electrode measurements performed with no
applied bias between the photoanode and a Pt disc cathode
resulted in hydrogen production at current densities of
0.060 mA cm

2
and 0.017 mA cm

2
when the catholyte was
saturated with Ar(g) and H
2
(g), respectively. These operating
points agreed well with the values that were predicted from the
load-line analysis based on separate measurements of the
performance of the cathodic and photoanodic electrodes. The
low energy-conversion e
ffi
ciencies result from a highly non-
optimal band gap and photovoltage of the WO
3
/liquid
contact, and much higher e
ffi
ciencies could be obtained if an
alternative suitable photoanode system were identi
ed that was
also stable under conditions where the remainder of the system
was stable.
Acknowledgements
This material is based upon work performed by the Joint Center
for Arti
cial Photosynthesis, a DOE Energy Innovation Hub,
supported through the O
ffi
ce of Science of the U.S. Department
of Energy under Award Number DE-SC0004993. M.S. acknowl-
edges the Resnick Sustainability Institute for a graduate
fellowship. K.F. is supported by the National Science Founda-
tion Graduate Research Fellowship under Grant No. DGE-
1144469. S.A. acknowledges support from a U.S. Department
of Energy, O
ffi
ce of Energy E
ffi
ciency and Renewable Energy
(EERE) Postdoctoral Research Award under the EERE Fuel Cell
Technologies Program. The authors would like to thank Dr Shu
Hu for assistance in boron doping, Rick Gerhart for fabrication
of the electrochemical cells used and Dr Andrew Leenheer for
the WO
3
refractive index data.
References
1 S. W. Boettcher, E. L. Warren, M. C. Putnam, E. A. Santori,
D. Turner-Evans, M. D. Kelzenberg, M. G. Walter,
J. R. McKone, B. S. Brunschwig, H. A. Atwater and
N. S. Lewis,
J. Am. Chem. Soc.
, 2011,
133
, 1216
1219.
2 S. W. Boettcher, J. M. Spurgeon, M. C. Putnam, E. L. Warren,
D. B. Turner-Evans, M. D. Kelzenberg, J. R. Maiolo,
H. A. Atwater and N. S. Lewis,
Science
, 2010,
327
, 185
187.
3 S. Haussener, C. Xiang, J. M. Spurgeon, S. Ardo, N. S. Lewis
and A. Z. Weber,
Energy Environ. Sci.
, 2012,
5
, 9922
9935.
4 B. M. Kayes, H. A. Atwater and N. S. Lewis,
J. Appl. Phys.
, 2005,
97
, 114302.
5 M. H. Lee, K. Takei, J. Zhang, R. Kapadia, M. Zheng,
Y.-Z. Chen, J. Nah, T. S. Matthews, Y.-L. Chueh, J. W. Ager
and A. Javey,
Angew. Chem., Int. Ed.
, 2012,
51
, 10760
10764.
6 J. R. Maiolo, B. M. Kayes, M. A. Filler, M. C. Putnam,
M. D. Kelzenberg, H. A. Atwater and N. S. Lewis,
J. Am.
Chem. Soc.
, 2007,
129
, 12346
12347.
7 O. Khaselev and J. A. Turner,
Science
, 1998,
280
, 425
427.
8 H. J. Lewerenz,
Photoelectrochemical Materials and Energy
Conversion Processes
, Wiley-VCH Verlag GmbH & Co. KGaA,
Weinheim, Germany, 2010, vol. 12.
9 J. M. Rubi and S. Kjelstrup,
J. Phys. Chem. B
, 2003,
107
,
13471
13477.
10 M. D. Kelzenberg, S. W. Boettcher, J. A. Petykiewicz,
D. B. Turner-Evans, M. C. Putnam, E. L. Warren,
J. M. Spurgeon, R. M. Briggs, N. S. Lewis and
H. A. Atwater,
Nat. Mater.
, 2010,
9
, 239
244.
11 A. G. Mu
̃
noz, C. Heine, M. Lublow, H. W. Klemm, N. Szabo,
T. Hannappel and H. J. Lewerenz,
ECS J. Solid State Sci.
Technol.
, 2013,
2
, Q51
Q58.
12 E. L. Warren, J. R. McKone, H. A. Atwater, H. B. Gray and
N. S. Lewis,
Energy Environ. Sci.
, 2012,
5
, 9653.
13 C. Xiang, A. C. Meng and N. S. Lewis,
Proc. Natl. Acad. Sci. U.
S. A.
, 2012,
109
, 15622
15627.
14 J. M. Spurgeon, K. E. Plass, B. M. Kayes, B. S. Brunschwig,
H. A. Atwater and N. S. Lewis,
Appl. Phys. Lett.
, 2008,
93
,
032112.
15 J. M. Spurgeon, M. G. Walter, J. Zhou, P. A. Kohl and
N. S. Lewis,
Energy Environ. Sci.
, 2011,
4
, 1772.
16 J. Bolton, S. Strickler and J. S. Connolly,
Nature
, 1985,
316
,
495
500.
17 Y. Zhao, E. S. Smotkin and T. Mallouk,
Energy Environ. Sci.
,
2012,
5
, 7582
7589.
18 Y. J. Hwang, A. Boukai and P. Yang,
Nano Lett.
, 2009,
9
, 410
415.
19 M. T. Mayer, C. Du and D. Wang,
J. Am. Chem. Soc.
, 2012,
134
,
12406
12409.
20 S. Licht, B. Wang, S. Mukerji, T. Soga, M. Umeno and
H. Tributsch,
J. Phys. Chem. B
, 2000,
104
, 8920
8924.
21 E. L. Warren, S. W. Boettcher, M. G. Walter, H. A. Atwater
and N. S. Lewis,
J. Phys. Chem. C
, 2011,
115
, 594
598.
22 C. Liu, J. Tang, H. M. Chen, B. Liu and P. Yang,
Nano Lett.
,
2013,
13
, 2989
2992.
23 X. Liu, F. Wang and Q. Wang,
Phys. Chem. Chem. Phys.
, 2012,
14
, 7894.
24 R. H. Coridan, M. Shaner, C. Wiggenhorn, B. S. Brunschwig
and N. S. Lewis,
J. Phys. Chem. C
, 2013,
117
, 6949
6957.
25 M. D. Kelzenberg, D. B. Turner-Evans, M. C. Putnam,
S. W. Boettcher, R. M. Briggs, J. Y. Baek, N. S. Lewis and
H. A. Atwater,
Energy Environ. Sci.
, 2011,
4
, 866.
26 Q. Mi, A. Zhanaidarova, B. S. Brunschwig, H. B. Gray and
N. S. Lewis,
Energy Environ. Sci.
, 2012,
5
, 5694
5700.
27 A. Subrahmanyam and A. Karuppasamy,
Sol. Energy Mater.
Sol. Cells
, 2007,
91
, 266
274.
This journal is © The Royal Society of Chemistry 2014
Energy Environ. Sci.
,2014,
7
, 779
790 |
789
Paper
Energy & Environmental Science
Published on 16 December 2013. Downloaded by California Institute of Technology on 20/03/2014 14:45:35.
View Article Online