Stable solar-driven oxidation of water by
semiconducting photoanodes protected by transparent
catalytic nickel oxide films
Ke Sun
a,b
, Fadl H. Saadi
b,c
, Michael F. Lichterman
a,b
, William G. Hale
b,d
, Hsin-Ping Wang
e
, Xinghao Zhou
b,c
,
Noah T. Plymale
a
, Stefan T. Omelchenko
b,c
, Jr-Hau He
e
, Kimberly M. Papadantonakis
a,b
, Bruce S. Brunschwig
b,f
,
and Nathan S. Lewis
a,b,f,g,1
Divisions of
a
Chemistry and Chemical Engineering and
c
Engineering and Applied Sciences, California Institute of Technology, Pasadena, CA 91125;
b
Joint
Center for Artificial Photosynthesis, California Institute of Technology, Pasadena, CA 91125;
d
Department of Chemistry, University of Southampton,
Southampton, Hampshire SO17 1BJ, United Kingdom;
e
Computer, Electrical, and Mathematical Sciences and Engineering Division, King Abdullah University of
Science & Technology, Thuwal 23955-6900, Saudi Arabia;
f
Beckman Institute Molecular Materials Research Center, California Institute of Technology,
Pasadena, CA 91125; and
g
Kavli Nanoscience Institute, California Institute of Technology, Pasadena, CA 91125
Edited by Michael L. Klein, Temple University, Philadelphia, PA, and approved February 10, 2015 (received for review December 3, 2014)
Reactively sputtered nickel oxide (NiO
x
) films provide transparent,
antireflective, electrically conductive, chemically stable coatings that
also are highly active electrocatalysts for the oxidation of water to
O
2
(g). These NiO
x
coatings provide protective layers on a variety of
technologically important semiconducting photoanodes, including
textured crystalline Si passivated by amorphous silicon, crystalline
n-type cadmium telluride, and hydrogenated amorphous silicon.
Under anodic operation in 1.0 M aqueous potassium hydroxide
(pH 14) in the presence of simulated sunlight, the NiO
x
films sta-
bilized all of these self-passivating, high-efficiency semiconducting
photoelectrodes for
>
100 h of sustained, quantitative solar-driven
oxidation of water to O
2
(g).
electrocatalysis
|
solar-driven water oxidation
|
photoanode stabilization
|
nickel oxide
T
he oxidation of water to O
2
(g) is a critical process for the
sustainable solar-driven generation of fuels, including the
generation of carbon-based fuels by the solar-driven reduction of
carbon dioxide as well as the generation of H
2
(g) by solar-driven
water splitting (1, 2). Many technologically important semi-
conductors, including silicon (Si), Group III
–
V materials such as
gallium arsenide (GaAs), and Group II
–
VI materials such as
cadmium telluride (CdTe), have optimal band gaps for use in an
integrated, dual light-absorber, solar-fuels generator (3). How-
ever, these materials are generally unstable and corrode or
passivate rapidly when operated under photoanodic conditions
in aqueous electrolytes. Furthermore, the efficient operation of
a passive and intrinsically safe water-splitting system requires the
use of either strongly alkaline or acidic electrolytes, presenting
additional constraints on the stability of the photoanodes and
electrocatalysts (4
–
6).
The search for new, stable compound semiconductors or
molecular systems for water oxidation has thus far yielded
materials with low efficiencies and/or limited stability (7, 8). An
alternative strategy for addressing the lack of materials known to
be stable under the conditions needed for the efficient oxidation
of water is to protect high-efficiency, technologically important
semiconductors to enable their use in integrated solar-fuels
generators by using buried semiconductor junctions to form
a photovoltaic (PV)-biased electrosynthetic cell (9). In this ap-
proach, the surface of the otherwise unstable semiconducting
light absorber is covered by a layer of a stable and partially
transparent conductive oxide or metal, which serves either as
a Schottky barrier or as a transparent conductive contact to
a photoelectrode that contains a buried junction to provide the
requisite charge separation. Metallized contacts to radial p-n
junctions in Si microspheres have been used to effect the un-
assisted solar-driven splitting of HI(aq) and HBr(aq) (10), and
metallized contacts have been used in conjunction with triple-
junction
–
based hydrogenated amorphous Si (
a
-Si:H) photovoltaics
for PV-biased electrosynthetic water splitting (11
–
13). The pro-
tective layers used in an integrated photoanode generally require
a separate electrocatalyst for the oxidation of water on the elec-
trode surface. Further, the entire assembly must be chemically
compatible with and stable in the
electrolytes and at the electrode
potentials associated with photoelectrochemical water oxidation
(14). Metals, metal alloys, semi
conductors, degenerately doped
transparent conducting oxides, catalytic transition-metal com-
pounds, organic polymers, and surface functionalization methods
have all been explored for this purpose, with only limited stability,
limited electrical properties at junctions, and/or limited activity for
water oxidation observed to date (14).
Sputtered nickel oxide (NiO
x
) films have been recently shown to
be optically transparent, antireflec
tive, conductive, stable, and highly
catalytically active while protecting n-Si and np
+
-Si photoanodes in
contact with aqueous 1.0-M KOH for the photoelectrochemical
oxidation of water (15, 16). For np
+
-Si photoanodes, such an ap-
proach allowed for the stable production of O
2
(g) for
>
1,200 h of
continuous operation under 1-Sun simulated solar illumination,
with a photocurrent-onset potential of
−
180 mV relative to the
equilibrium water-oxidation potential and current densities in
Significance
The development of efficient artificial photosynthetic systems,
designed to store solar energy in chemical bonds, requires the
pairing of stable light-absorbing electrodes for both the oxida-
tive and reductive half-reactions. The development of such sys-
tems has been hindered in part by the lack of semiconducting
photoanodes that are stable under the conditions required for
the production of O
2
(g) from water. We demonstrate herein that
a reactively sputtered NiO
x
layer provides a transparent, anti-
reflective, conductive, chemically stable, inherently catalytic
coating that stabilizes many efficient and technologically im-
portant semiconducting photoanodes under viable system oper-
ating conditions, thereby allowing the use of these materials in an
integrated system for the sustaina
ble, direct production of fuels
from sunlight.
Author contributions: K.S. and N.S.L. designed resea
rch; K.S., F.H.S., M.F.L., W.G.H., H.-P.W., X.Z.,
N.T.P., and S.T.O. performed research; K.S., B.S.B., and N.S.L. analyzed data; and K.S., J.-H.H.,
K.M.P., B.S.B., and N.S.L. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
1
To whom correspondence should be addressed. Email: nslewis@caltech.edu.
This article contains supporting information online at
www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1423034112/-/DCSupplemental
.
3612
–
3617
|
PNAS
|
March 24, 2015
|
vol. 112
|
no. 12
www.pnas.org/cgi/doi/10.1073/pnas.1423034112
excess of 29 mA
·
cm
−
2
at the equilibrium water-oxidation po-
tential. We demonstrate herein that this protection strategy can be
extended to semiconducting materials used in commercial pho-
tovoltaic solar cells and in high-efficiency devices, specifically
textured crystalline Si passivated with a layer of amorphous Si,
known as a heterojunction Si (HTJ-Si) device with a typical
structure of (p
+
-
a
-Si
j
i-
a
-Si
j
n-c-Si
j
i-
a
-Si
j
n
+
-
a
-Si); n-i hydroge-
nated amorphous Si (
a
-Si:H) structures; and single-crystalline
n-type CdTe light absorbers. All of these materials have band
gaps appropriate (1
–
2 eV) for use in integrated solar fuel-gen-
eration devices, and this wo
rk demonstrates that the NiO
x
protection
strategy is effective on both singl
e-crystalline and noncrystalline
semiconductors.
Results
Fig. 1
A
shows the effect of the oxygen concentration (0
–
33.3%)
present during the sputter-deposition process on the electro-
chemical behavior of NiO
x
films deposited onto p
+
-Si substrates
(p
+
-Si
j
NiO
x
). The NiO
x
films were deposited from a Ni target
onto substrates via reactive radio-frequency (RF) sputtering in
a high-vacuum magnetron sputtering system with a maximum
base pressure of 8
×
10
−
8
Torr (
SI Appendix
,
Text S1
, provides
a detailed description of materials and methods). To minimize
effects due to electrochemical kinetic overpotentials, the be-
havior of the film-coated anode was evaluated in contact with
Fe(CN)
6
4
−
/3
−
(aq), an electrochemically reversible, one-electron
redox couple [0.35 M K
4
Fe(CN)
6
/0.05 M K
3
Fe(CN)
6
in 1.0 M
KCl(aq)]. Increases in the oxygen concentration during sputter-
ing produced more resistive films, as indicated by decreases in the
slopes of the linear portions of the current-density vs. potential
(
J
–
E
) data near zero current density, which could be due to the
increased oxidation of the Si substrate or to the decreased Ni(II)
to Ni(III) ratio in the films, as evidenced by X-ray photoelectron
spectroscopy (XPS) (
SI Appendix
,Fig.S2
A
).
Fig. 1
B
shows the effect of the oxygen concentration during
sputter deposition on the catalytic activity of the resulting NiO
x
films for the oxidation of water to O
2
(g) in 1.0 M KOH(aq). The
behavior depicted in Fig. 1
B
was observed after activation of the
film by 10 consecutive cyclic voltammetric scans between 0.93 V
and 1.93 V vs. the reversible hydrogen electrode (RHE), without
correction of resistance. Increases in the concentration of oxygen
during sputtering produced increases in the Tafel slope as well
as increases in the catalytic overpotential (
η
)forwateroxida-
tion (
SI Appendix
,Fig.S1andTableS1
). Over the range of
current densities investigated, the catalytic activity for water
oxidation of sputtered NiO
x
films prepared using an oxygen
concentration
≤
4.8% was in close accord with that observed for
the oxide/hydroxide derived from the oxidation of Ni metal
(black curve in Fig. 1
B
). Hence, these films exhibited an over-
potential of 330
±
7 mV to produce an anodic current density of
10 mA
·
cm
−
2
for water oxidation, within 70 mV of the lowest
overpotentials reported to date for water oxidation under these
conditions (17), with the other systems generally composed of
porous electrocatalyst films that have undesirably high optical
absorption and reflection properties. Consistently, no metallic Ni
was detected by XPS on sputtered NiO
x
films, even on films
prepared with oxygen concentrations of only 1% during sput-
tering (red curve in
SI Appendix
, Fig. S2
A
), and Ni
0
was de-
tectable by XPS only for films prepared in the absence of
oxygen (black curve in
SI Appendix
,Fig.S2
A
). On the surface of
KOH-treated NiO
x
and Ni metal films, a trace amount of Fe
was observed (
SI Appendix
,Fig.S3
), which could contribute to
the improvement in the catalytic activity (18
–
21), although
other activation mechanisms such as incorporation of other
impurities, thickening of the Ni(OH)
2
/NiOOH layers, and
changes to the crystallinity o
f the surface (22) could not be
excluded. Because the activatio
n process was unchanged for the
different samples, the observed differences in
J-E
behavior
were attributed to the difference
s in the oxygen concentration
during sputter deposition of the films.
Fig. 1
C
shows the effect of the deposition temperature on the
J-E
behavior of the Ni redox peak region (overpotential range
−
0.1 V to
∼
0.3 V) after 10 cycles of activation scans in 1.0 M
KOH(aq). The cathodic peaks at an overpotential of 0.12 V
corresponded to the
β
-NiOOH/
β
-Ni(OH)
2
transition in lower
oxidation states, which increased and became sharper as the
deposition temperature increased (20, 23, 24). Moreover, the
small reduction peak at an overpotential of 0.036 V developed
during the activation of NiO
x
and metallic Ni films prepared
without substrate heating, which corresponded to the
γ
-NiOOH/
α
-Ni(OH)
2
transition (25) (highlighted by black arrows in Fig.
1
C
). These observations suggest that substrate heating reduced
or delayed the formation of the disordered
α
-Ni(OH)
2
/
γ
-NiOOH
catalyst surface phases associated with the development of po-
rosity, presumably through a direct transition between highly
crystalline NiO sintered at higher temperatures and surficial
NiOOH (23, 26).
Fig. 1
D
shows the electrochemical behavior in Fe(CN)
6
3
−
/4
−
(aq)
of p
+
-Si
j
NiO
x
anodes before (solid curves) and after (dashed
curves) activation in 1.0 M KOH
(aq). When the substrate was
heated during sputter deposition of the NiO
x
film, no significant
change of conductivity, as measured by the
J
-
E
behavior, was
observed after activation (blue cur
ves); whereas samples deposited
without heating of the substrate showed a pronounced change in
J
-
E
behavior after activation, with a reduced slope at positive bias and
a decrease in the current density at negative bias (red curves).
Moreover, a similarly pronounced change in the
J
-
E
behavior was
observed on a p
+
-Si electrode coated with Ni metal (
∼
2nm,black
Fig. 1.
(
A
and
B
) Effect of oxygen concentration during sputter deposition on
the conductivity (
A
) and the catalytic activity (
B
) of the resulting activated NiO
x
films for the oxidation of water to O
2
(g), with the overpotential measured
relative to the equilibrium water-oxidation potential (
E
o
′
(O
2
/H
2
O)). (
C
) Effect of
deposition temperature on the development of the cathodic double-current
peak. (
D
) Effect of deposition temperature on the conductivity of the
resulting films before (red and blue solid curves) and after (red and blue
dashed curves) activation in 1.0 M KOH(aq) under 4.8% oxygen concentra-
tion during sputtering and compared with a 2-nm metallic Ni film (black
solid and dashed curves). The conductivity of the NiO
x
films (
A
and
D
) was
evaluated in contact with an aqueous solution of 0.35 M Fe(CN)
6
3
−
/0.050 M
Fe(CN)
6
4
−
in 1.0 M KCl as supporting electrolyte. The catalytic activity (
B
and
C
) was evaluated following activation by 10 consecutive cyclic voltammetry
scans between 0.93 V and 1.93 V vs. the reversible hydrogen electrode (RHE),
without correcting for cell resistance.
Sun et al.
PNAS
|
March 24, 2015
|
vol. 112
|
no. 12
|
3613
CHEMISTRY
curves). For samples prepared at 300 °C, the Ni(II) peaks observed
in the XPS data were more pronounced both before and after
activation in KOH than the Ni(II) peaks observed on samples
coated with nonannealed NiO
x
or metallic Ni (
SI Appendix
, Fig.
S2
B
). Both electrically insulating Ni(OH)
2
and n-type electrically
conductive NiOOH remained on the surface after activation (27)
and their concentration is likely dependent on the substrate-
heating temperature and on the crystallinity of the starting oxide
(
SI Appendix
, Fig. S2
C
). Hence, high deposition temperatures
produced NiO
x
coatings that were more inert to the physical
(Fig. 1
C
) and chemical (
SI Appendix
, Fig. S2
B
) changes caused
by anodic activation, which thus affected the energy barriers for
hole and electron transfer in electrolytes that contained kineti-
cally rapid one-electron redox species (Fig. 1
D
).
The optical properties of sputtered NiO
x
films on polished
single-crystalline Si substrates were measured using variable-
angle spectroscopic ellipsometry (28) (
SI Appendix
,
Text S2
,
provides a detailed description of characterization methods).
Fig. 2
A
and
B
show the measured and fitted ellipsometric and
spectrophotometric data for 75-nm-thick NiO
x
films coated on
p
+
-Si, with the ellipsometric data obtained using angles of in-
cidence of 65°, 75°, and 85°, respectively. The optical dispersion
of the layer derived from a multilayer model exhibited behavior
expected for a dielectric oxide, and neither the refractive index
(
n
) nor the extinction coefficient (
k
) displayed behavior consis-
tent with Ni metal (increased
n
and
k
with a reduced photon
energy), consistent with the X-ray photoelectron spectroscopy
(XPS) and X-ray diffraction (XRD) analysis (
SI Appendix
, Fig.
S2
A
and
C
). The film exhibited a direction-dependent refractive
index (
x
is in-plane and
z
is normal to the film surface) as well as
a graded index along the z direction (Fig. 2
C
). The extinction
coefficient,
k
, was near zero for 1.0 eV
<
hv
<
3.5 eV, where the
nonzero value of
k
at
hv
>
3.5 eV was likely due to electron
interband excitation in the oxide. The increase in
k
at
hv
<
1.5 eV
can be ascribed to the free-carrier absorption typically existing
in transparent conducting oxides (29), consistent with the pres-
ence of a highly conductive film (similar conductivity to metallic
Ni shown in Fig. 1
D
) having a large (
∼
10
19
cm
−
3
) free-carrier
concentration (15). However, unlike the refractive index, the
extinction coefficient,
k,
did not show a strong direction de-
pendence or grading along the
z
direction. These observations
areinaccordwiththecross-se
ctional data obtained using
scanning-electron mic
roscopy (SEM) (Fig. 2
C
,
Inset
), which
showed a columnar structure typically associated with the
presence of lower refractive indexes at the bottom of the film due
to the nucleation, competitive growth, and coalescence processes.
These processes contributed to a
continuous change of morphology
and texture as a function of thickness (30). Therefore, the columnar
structure is consistently associated with the primary origin of
the anisotropy in both the refractive index and the electrical
conductivity of the films.
Fig. 2
D
shows a Tauc plot [
h
ν
vs. (
ah
ν
)
2
] for NiO
x
calculated
based on the extinction coefficient,
k
, using the relationship
α
=
4
π
k
/
λ
. The direct optical band gap of 3.74 eV was consistent with
the data obtained from the Tauc plot derived from the previously
reported optical transmission measurements (15).
Metallic electrocatalysts absorb and/or reflect sunlight. Hence,
the use of such electrocatalysts in an integrated photoanode
involves a trade-off between increasing the film thickness to in-
crease the catalytic activity and decreasing the film thickness to
achieve high optical transparency and low reflectance (31). Planar
light absorbers without surface texturing are typically reflective,
and the dielectric properties of transparent catalyst thin films can
therefore be used to minimize these reflections. These optical and
ellipsometric results indicate that nearly ideal quarter-wavelength
antireflective (AR) behavior can be obtained by integration of a
single NiO
x
layer with Si, CdTe, and
a
-Si. A calculation based
on effective-medium theory showed that coating these semi-
conductors with a 75-nm-thick layer of NiO
x
would result in an
effective suppression of reflection at the air
–
semiconductor in-
terface due to the high refractive indexes of these semiconductors,
especially in the visible wavelength region even with water as the
incident medium (
SI Appendix
,Fig.S4
). Therefore, a thickness of
75 nm was chosen for NiO
x
to optimize the antireflection properties
in air as opposed to optimizing the electrical conductivity along the
direction of the film thickness, th
e electrocatalytic activity, the
number of surface activation sites, or the electrochemical sta-
bility of the electrocatalyst, because these latter four properties
were unaffected by changes in the film thickness within the range
of 10
–
150 nm (
SI Appendix
, Fig. S5
).
Most electrocatalysts of the oxygen-evolution reaction (OER),
including transition metals such as Cr, Mn, Fe, Co, Ni, and Ir, as
well as their oxides and mixed oxides thereof, are electrochromic
(32). The electrochromism results in a change in the optical
absorption (effective extinction coefficient) under positive bias in
aqueous, especially alkaline, media due to surface adsorption of
ions (33). NiO
x
films prepared at high temperature exhibited
negligible electrochromism (15), in accord with previous results
in which high-temperature processing was found to suppress the
electrochromic darkening (34, 35). We note that the antireflec-
tive performance of NiO
x
coatings in water could be further
optimized by consideration of the changes in the layered struc-
ture of the electrode under anodic conditions. Under anodic
conditions, the effective refractive index is reduced due to the
presence of electrochemically formed NiOOH (36, 37). This
compensation is further supported by the high light-limited cur-
rent density as well as the high external quantum yield of such
electrodes (see below).
Although a 75-nm layer of NiO
x
on polished crystalline Si
substrates is an effective antireflective coating (
SI Appendix
,
Fig. S4
), a single material without a properly graded refractive
index is not capable of producing broadband antireflection (14)
Fig. 2.
(
A
and
B
) Representative ellipsometric data (curves) and fits (circles)
showing the change in polarization as light reflects from a 75-nm NiO
x
-
coated crystalline Si substrate including the amplitude ratio,
Ψ
(
A
) and the
phase difference,
Δ
(
B
) vs. incident photon energy. (
C
) Graded indexes of the
NiO
x
films in the
z
direction.
Inset
displays a cross-sectional SEM image
obtained at a 45° tilt angle showing the columnar structure of the sputtered
NiO
x
film. (
D
) Tauc plot based on the absorption coefficient calculated from
the measured extinction coefficient (
k
).
Inset
displays optical images of NiO
x
films with various thicknesses and of 15-nm metallic Ni on quartz substrates.
3614
|
www.pnas.org/cgi/doi/10.1073/pnas.1423034112
Sun et al.
(Fig. 3
A
, blue curve). Hence, random surface textures with
a smooth transition from ambient to high-index substrates are
normally required for reduced sensitivity to the incident wave-
length as well as to the angle of incidence. The textured surface
of the HTJ-Si cell (Fig. 3
A
, red curve) reduced the reflectance
and thus improved the absorptance from
∼
60% for a planar
crystalline substrate to
∼
80% for the textured sample (15). Ad-
dition of the NiO
x
layer to the textured Si surface reduced the
reflectance to below 10%, with nearly 100% absorptance at 550
nm (Fig. 3
A
, black curve). For comparison, a HTJ-Si cell coated
with a 2-nm-thick metallic Ni electrocatalyst film (
SI Appendix
,
Fig. S6
A
, black curve) exhibited an optical absorptance of
∼
70%
that was decreased relative to that of a bare HTJ-Si cell (
SI
Appendix
, Fig. S6
A
, red curve), due to the higher reflection of
this structure.
Fig. 3
B
depicts the
J
-
E
behavior of a NiO
x
-coated HTJ-Si
photoanode (HTJ-Si
j
NiO
x
) in the presence and absence of illu-
mination, respectively. Under 100 mW
·
cm
−
2
of simulated air mass
(AM) 1.5-G illumination, the HTJ-Si
j
NiO
x
/1.0 M KOH(aq) con-
tact produced photocurrent-onset potentials of
−
280
±
16 mV
relative to the equilibrium wa
ter-oxidation potential,
E
o
′
(O
2
/H
2
O),
along with current densities at
E
o
′
(O
2
/H
2
O) of 20.7
±
5mA
·
cm
−
2
,
and a solar-to-O
2
(g) ideal regenerative cell efficiency of 1.45
±
0.17%. A load-line analysis using an equivalent-circuit model con-
sisting of a photodiode connected in
series with a dark electrolysis
cell indicated that obtaining shifts in the
J
-
E
properties equivalent
to those observed for the champion HTJ-Si
j
NiO
x
photoanode in
1.0 M KOH(aq) relative to that of the p
+
-Si
j
NiO
x
would require the
use of a 9.0% efficient Si PV cell with a
V
oc
=
600 mV, a short-
circuit current density (
J
sc
) of 34.0 mA
·
cm
−
2
, and a fill factor of 0.42
(1, 38). The performance of the PV-biased electrosynthetic pho-
toelectrode was lower than that o
f the solid-state PV device (39,
40), likely due to nonideality of the back contact in conjunction
with additional front-contact resi
stance in the amorphous Si layer
and surface passivation that was partially degraded during the de-
position of the NiO
x
,bothofwhichledtoincreasedrecombination
(reduced carrier lifetime) at the front and back interfaces as
revealed by a time-resolved microwave conductivity measurement
(TRMC) (
SI Appendix
,Fig.S7
).
Fig. 4
A
shows chronoamperometric data for HTJ-Si
j
NiO
x
,
n-CdTe
j
NiO
x
, and
a
-Si:H
j
NiO
x
photoanodes in contact with
1.0 M KOH(aq) and under simulated 1-Sun illumination. The
current density at 1.73 V vs. RHE decreased very slightly over
the 100 h of testing for the HTJ-Si
j
NiO
x
photoelectrode, but
remained constant for the n-CdTe
j
NiO
x
and
a
-Si:H
j
NiO
x
photo-
electrodes. In the absence of protective coatings, these materials
showed rapid surface oxidation (
SI Appendix
, Fig. S8
A
–
D
) when
positively biased under the same conditions, a behavior consis-
tent with the reported anodic oxidation of Si and CdTe in
alkaline electrolytes (41, 42). Furthermore, passage of Faradaic
current from these bare electrodes under illumination did not yield
significant amounts of O
2
(g) and only effected oxidation of the
electrode. In contrast, all of the NiO
x
-coated electrodes displayed
more than 100 h of stability unde
r water-oxidation conditions
under simulated 1-Sun illumination in 1.0 M KOH(aq).
Cyclic voltammograms were collected every 10 h during the
chronoamperometric stability tests, and Fig. 4
B
shows represen-
tative
J
-
E
data for HTJ-Si
j
NiO
x
,n-CdTe
j
NiO
x
,and
a
-Si:H
j
NiO
x
photoanodes in contact with 1.0 M KOH(aq) under 100 mW
·
cm
−
2
of simulated illumination. The photocurrent-onset potentials
referenced to
E
o
′
(O
2
/H
2
O) were
−
70 mV and
−
116 mV and the
photocurrent densities at
E
o
′
(O
2
/H
2
O) were 5.2 mA
·
cm
−
2
and
3.3 mA
·
cm
−
2
, for the n-CdTe
j
NiO
x
and
a
-Si:H
j
NiO
x
photoanodes,
respectively, under simulated 1
-Sun illumination. These values
remained unchanged after 100 h. Decreases were observed
for the n-CdTe
j
NiO
x
photoelectrode after 1,000 h of continuous
operation, where the photocurrent onset potential dropped to
−
30 mV and the photocurrent density at the equilibrium water-
oxidation potential dropped to 1.4 mA
·
cm
−
2
. The Ni(II/III)
redox peaks in general increas
ed in current and separated in
potential during the stability tests, suggesting an increase in
catalytically active sites, loss of reversibility, and/or an increase
in resistance.
Fig. 3.
(
A
) Optical absorptance of a textured HTJ-Si device without (red
curve) and with (black curve) a 75-nm NiO
x
coating compared with a NiO
x
-
coated planar crystalline Si photoelectrode without textures (blue curve).
Inset
displays an SEM image of the textured HTJ-Si surface without NiO
x
.
(
B
)
J
-
E
data for a HTJ-Si
j
NiO
x
photoanode in contact with 1.0 M KOH(aq)
under 100 mW
·
cm
−
2
of simulated AM 1.5 illu
mination attenuated by
various neutral density filters with different optical densities (OD) and in
the dark. The dashed line indicates the equilibrium water-oxidation potential,
E
o
′
(O
2
/H
2
O).
Fig. 4.
(
A
) Chronoamperometry of HTJ-Si
j
NiO
x
,n-CdTe
j
NiO
x
,and
a
-Si:H
j
NiO
x
photoelectrodes in contact with 1.0 M KOH(aq) under 100 mW
·
cm
−
2
of sim-
ulated AM 1.5-G solar illumination from an ENH-type tungsten-halogen lamp.
The electrodes were maintained potentiostatically at 1.73 V vs. RHE. (
B
)
J
-
E
data for HTJ-Si
j
NiO
x
,n-CdTe
j
NiO
x
,and
a
-Si:H
j
NiO
x
photoanodes before and
after 100 h of chronoamperometric stability testing in 1.0 M KOH(aq).
J-E
data
for n-CdTe
j
NiO
x
after 1,000 h are also included.
J-E
data for an HTJ-Si
j
ITO
j
CoO
x
photoanode before (red dashed curve) and after a 20-h (black dashed curves)
stability test in KOH, as well as for an HTJ-Si
j
Ni (red dashed-dotted curve)
photoanode, are also included for comparison purposes. (
C
) Behavior of
the external quantum yield vs. wavelength for HTJ-Si
j
NiO
x
,n-CdTe
j
NiO
x
,and
a
-Si:H
j
NiO
x
photoelectrodes in 1.0 M KOH(aq). (
D
)FaradaicefficiencyforO
2
(g)
evolution at HTJ-Si
j
NiO
x
,n-CdTe
j
NiO
x
,and
a
-Si:H
j
NiO
x
photoelectrodes under
100 mW
·
cm
−
2
of AM 1.5-G simulated solar illumination in 1.0 M KOH(aq) with
the electrode held at a potential sufficient to maintain a constant current density
of 1 mA
·
cm
−
2
or lower for 40 min. The mass of O
2
(g) that would be produced by
an electrode operating at 100% Faradaic efficiency calculated based on the total
charge passed (red dotted lines) and the mass of O
2
(g) measured experimentally
(blue solid lines) using a calibrated O
2
probe are shown.
Sun et al.
PNAS
|
March 24, 2015
|
vol. 112
|
no. 12
|
3615
CHEMISTRY
A more significant degradation of the photoanodic perfor-
mance was observed for the HTJ-Si
j
NiO
x
photoelectrodes, which
resulted from a decrease in the photocurrent density at the
equilibrium water-oxidation potential from 24.3 mA
·
cm
−
2
to
16.3 mA
·
cm
−
2
, whereas a small change (
∼
40 mV anodic shift)
was noted in the photocurrent-onset potential after 100 h. The deg-
radation of the HTJ-Si
j
NiO
x
photoelectrodes during 100 h of oper-
ation could be caused by oxidation of the thin p
+
-
a
-Si(6 nm)
j
i-
a
-Si
(5 nm) on c-Si, because this junction is sensitive to the changes caused
by oxidation. Oxidation of either of the thin
a
-Si layers can result in a
reduction in band bending due to the degraded p
+
-
a
-Si emitter and/or
a reduction in the surface passivation provided by the i-
a
-Si layer.
For comparison, a thin (2 nm) Ni metal-coated HTJ-Si (HTJ-
Si
j
Ni) photoanode showed strong signs of oxidation during the
first five cyclic voltammetric scans (dashed-dotted red line in Fig.
4
B
), indicating that the buried junctions were completely destroyed
due to anodic Si oxidation, with diode behavior then restored due
to the formation of interfacial silicon oxide during subsequent
cyclic voltammetric cycles (
SI Appendix
,Fig.S6
B
). For additional
comparison, In-doped SnO
2
(ITO)-coated HTJ-Si electrodes were
coated with electrodeposited Co-Pi (phosphate-containing CoO
x
,
HTJ-Si
j
ITO
j
Co-Pi) and characterized electrochemically (
SI
Appendix
,
Text S2
, provides a detailed description of characteri-
zation methods). A HTJ-Si
j
ITO
j
Co-Pi photoanode in 1.0 M
KOH(aq) (HTJ-Si
j
ITO
j
CoO
x
when in alkaline media,
SI Appendix
,
Fig. S9
) exhibited lower catalytic activity than the NiO
x
presented
herein and also yielded a continuous degradation in performance
during operation. Thus, the ITO/CoO
x
films produced a lower
solar-to-O
2
(g) ideal regenerative cell efficiency than that observed
for the stable, essentially transparent, protective and highly cata-
lytic NiO
x
films reported herein.
Fig. 4
C
shows the wavelength (
λ
)-dependent external quantum
yield (
Φ
ext
) for HTJ-Si
j
NiO
x
, n-CdTe
j
NiO
x
, and
a
-Si:H
j
NiO
x
photoelectrodes maintained potentiostatically at 1.93 V vs. RHE,
at which the light-limited current density was produced. The
quantum yields of the HTJ-Si
j
NiO
x
structure were higher than
the values reported for analogous HTJ-Si
–
based solid-state de-
vices (39). The higher value of
Φ
ext
can be ascribed primarily to
the antireflective behavior and br
oadband suppression of reflection
by the NiO
x
film on the textured HTJ-Si surface maintained under
working conditions. The decrease in
Φ
ext
at low and high wave-
lengths, respectively, was primarily attributable to recombination
at the textured front and back interfaces between the crystalline Si
and the
a
-Si surface passivation layer (39). The low blue response
was also associated with absorption by the
a
-Si heterogeneous
coating. The
Φ
ext
observed for the n-CdTe
j
NiO
x
photoanode is
comparable to the behavior reported previously for n-CdTe pho-
toanodes (43). The low
Φ
ext
values for
λ
>
500 nm can be further
attributed to the higher reflection (
SI Appendix
, Fig. S4
C
)atthese
wavelengths. The quantum yield data for the
a
-Si:H
j
NiO
x
pho-
toelectrode were in agreement with previous measurements on
a
-Si:H
j
CH
3
OH
–
1.0 M LiClO
4
–
0.10 mM 1,1
′
-dimethylferrocene
(Me
2
Fe)
–
0.0010 mM Me
2
Fe
+
interfaces (44). The value of
Φ
ext
for
a
-Si:H can be further improved to match the absorption
spectrum to its highest internal quantum yield (typically in the
wavelengths of 400
–
550 nm) by reducing the thickness of NiO
x
to
an optimum calculated value of 42
–
53 nm.
Fig. 4
D
shows the Faradaic efficiency for the production of
O
2
(g) by the HTJ-Si
j
NiO
x
,n-CdTe
j
NiO
x
,and
a
-Si:H
j
NiO
x
pho-
toelectrodes operated for 50 min in 1.0 M KOH(aq) at a bias
sufficient to maintain a photocurrent density of 1 mA
·
cm
−
2
. The
mass of O
2
(g) that would be produced based on the total charge
passed and assuming 100% Faradaic efficiency was in excellent
agreement with the measured mass of O
2
determined using a
calibrated O
2
probe. Hence, all three electrodes evolved O
2
(g)
with
∼
100% Faradaic efficiency under these conditions.
Defects in the sputtered NiO
x
films are produced primarily by
textures and arcs that are present during sputtering. These
defects do not allow complete isolation of the underlying light
absorbers from contact with the electrolyte. To demonstrate this
behavior, NiO
x
-coated Si(100) and Si(111) electrodes were im-
mersed for 240 h in 10.0 M KOH(aq). The anisotropic etching of
Si(100) was clearly visible and resulted in the formation of
inverted pyramid structures (
SI Appendix
, Fig. S10
A
), whereas no
significant change in morphology was observed on either Si(111)
or NiO
x
-coated Si(111) surfaces immersed in 10.0 M KOH(aq).
Despite this access to the electrolyte, the NiO
x
coatings
allowed for the stable, continuous oxidation of water in 1.0 M
KOH(aq) for
>
1200 h on both n-Si(111) and n-Si(100) surfaces
(15). A reasonable hypothesis explaining the ability to realize
long-term stability using this protection scheme can be based on
anodic oxidation of the underlying light absorber by the forma-
tion of nondissolvable oxides. During the initial operation of the
photoelectrode under anodic bias, SiO
2
can be formed on crys-
talline Si, as well as on the HTJ-Si cell and on the
a
-Si:H sur-
faces, through pinholes and/or defects in the NiO
x
film. Similarly,
a passivating layer of CdTeO
3
can be formed on CdTe photo-
anodes (41, 43). After a passivating oxide forms at sites of defects
in the NiO
x
film, the reaction kinetics for continuous oxide
growth are significantly slower relative to those for the oxidation
of water by the NiO
x
film, and thus oxidation of water then
becomes the dominant anodic process. The slow etching rate of
these oxides in 1.0 M KOH(aq) is also critical to producing
a near-unity Faradaic efficiency for O
2
(g) evolution.
The ability of the electrocatalytic sputtered NiO
x
coating to sta-
bilize the underlying semiconductor under these conditions thus
depends on the ability of the photoanode to form a nondissolvable
self-passivating oxide. Consistently, a NiO
x
-coated n
+
-GaAs pho-
toanode showed a continuously increasing corrosion/dissolution
current below the equilibrium water-oxidation potential in 1.0 M
KOH(aq) during cyclic voltammetric scans (
SI Appendix
, Fig.
S10
B
). Catastrophic surface damage was observed on GaAs (
SI
Appendix
,Fig.S10
B
,
Inset
), mainly because of the dissolution of
the oxidized film followed by a lifting off of the NiO
x
.
Protection of photoanodes by use of a coating of NiO
x
has the
advantage of integrating the transparent conductive catalyst onto
planar and textured semiconductors that can form insoluble
oxides through the inherent presence of defects in the sputtered
catalyst film. In contrast, amorphous TiO
2
films deposited by
atomic-layer deposition (ALD) that are conformal and pinhole-
free can protect corroding materials like GaAs and other Group
III
–
V and Group II
–
VI semiconductors (45), as well as a high
–
aspect-ratio Si microwire-array photoanode (46), whereas sput-
tered NiO
x
does not provide conformal films.
In summary, NiO
x
films prepared using reactive sputtering are
able to protect not only crystalline Si photoelectrodes, but also
a HTJ-Si cell, as well as n-CdTe and
a
-Si:H photoanodes, against
corrosion or anodic oxidation, while maintaining the perfor-
mance of the photoanodes during extended operation (
>
100 h)
under water-oxidation conditions in 1.0 M KOH(aq) under
simulated 1-Sun illumination. In addition to imparting prolonged
stability to the Si photoanodes, the NiO
x
coatings also serve as
antireflective coatings that increase the amount of light absorbed
and thus increase the solar-to-O
2
(g) ideal regenerative cell effi-
ciency of such electrodes. Optimization of the conditions used to
deposit the NiO
x
films is critical to obtaining the observed per-
formance of the photoelectrodes, particularly with respect to
minimization of the resistivity, maximization of the transparency,
and minimization of the electrochromic darkening of the resulting
films. This work demonstrates that coatings of p-type transparent
conducting oxides can be successfully used to protect a high-effi-
ciency, small-band-gap semiconducting photoanode in a water-
splitting device. Hence, NiO
x
coatings are a promising approach to
the long-term stabilization of self-passivating semiconductors for
use in solar-fuels applications. The coating is especially useful for
PV-biased electrosynthetic systems that either use or form buried
3616
|
www.pnas.org/cgi/doi/10.1073/pnas.1423034112
Sun et al.
junctions in situ, but the method is not amenable to materials with
active majority-carrier
–
based grain-boundary shunts such as direct
photoelectrochemical cells based on semiconductor/liquid junc-
tions. Improvements in the performance of the in situ Schottky
contacts in each case can be expected based on band-edge and
interfacial engineering of the energetics at the absorber/protective
layer interface. Hence, the sputtered NiO
x
protective layer allows
a variety of photoanode materials to be considered as options for
PV-biased electrosynthetic cells that involve water oxidation in
alkaline media, where intrinsically s
afe, efficient sol
ar-driven elec-
trolysis systems can be constructed (4
–
6).
Materials and Methods
The NiO
x
films were deposited by reactive radio-frequency sputtering onto
substrates under various heating temperatures, using a high-vacuum mag-
netron sputtering system in a chamber with a maximum base pressure of 8
×
10
−
8
Torr. The O
2
/Ar ratio was varied from 0 to 0.33 with a constant Ar flow
of 20 sccm (standard cubic centimeters per minute) whereas the working
pressure was held at 5 mTorr.
SI Appendix
,
Text S1
, provides details on the
preparation of the p
+
-Si, n-Si,
a
-Si:H, HTJ-Si, n-CdTe, and n
+
-GaAs substrates and
on the sputtering conditions used to grow the NiO
x
films. Electrochemical
experiments, including measurements of the electrocatalytic performance of
the NiO
x
films and chronoamperometric sta
bility testing of electrodes, were
performed using a three-electrode configuration under potential control with
a Hg/HgO/1.0-M KOH reference electrode and a carbon-cloth counter electrode.
SI Appendix
,
Text S2
, provides details on characterization of electrodes, in-
cluding stability-measurement protocols, preparation of the HTJ-Si
j
ITO
j
Co-Pi
photoelectrode, X-ray photoelectron spectroscopy, X-ray diffraction spectroscopy,
time-resolved microwave conductivity, spe
ctroscopic ellipsometry, Faradaic effi-
ciency measurements, and total reflectance calculations.
ACKNOWLEDGMENTS.
This material is based on work performed by the
Joint Center for Artificial Photosynthesis, a Department of Energy (DOE)
Energy Innovation Hub, supported through the Office of Science of the US
DOE under Award DE-SC0004993. N.T.P. acknowledges support from the
Graduate Research Fellowship Program of the US National Science Founda-
tion. B.S.B. was supported by the Beckman Institute of the California Institute
of Technology. This work was also supported by the Gordon and Betty Moore
Foundation under Award GBMF1225.
1. Walter MG, et al. (2010) Solar water splitting cells.
Chem Rev
110(11):6446
–
6473.
2. Lewis NS, Nocera DG (2006) Powering the planet: Chemical challenges in solar energy
utilization.
Proc Natl Acad Sci USA
103(43):15729
–
15735.
3. Hu S, Xiang C, Haussener S, Berger AD, Lewis NS (2013) An analysis of the optimal
band gaps of light absorbers in integrated tandem photoelectrochemical water-
splitting systems.
Energy Environ Sci
6:2984
–
2993.
4. Jin J, et al. (2014) An experimental and modeling/simulation-based evaluation of the
efficiency and operational performance characteristics of an integrated, membrane-free,
neutral pH solar-driven water-splitting system.
Energy Environ Sci
7(10):3371
–
3380.
5. Haussener S, et al. (2012) Modeling, simulation, and design criteria for photoelectro-
chemical water-splitting systems.
Energy Environ Sci
5:9922
–
9935.
6. McKone JR, Gray HB, Lewis NS (2013) Will solar-driven water-splitting devices see the
light of day?
Chem Mater
26(1):407
–
414.
7. Jia Q, Iwashina K, Kudo A (2012) Facile fabrication of an efficient BiVO
4
thin film
electrode for water splitting under visible light irradiation.
Proc Natl Acad Sci USA
109(29):11564
–
11569.
8. Alibabaei L, et al. (2013) Solar water splitting in a molecular photoelectrochemical
cell.
Proc Natl Acad Sci USA
110(50):20008
–
20013.
9. Nielander AC, Shaner M, Papadantonakis KM, Francis SA, Lewis NS (2014) A taxonomy
for solar fuels generators.
Energy Environ Sci
8:16
–
25.
10. Kilby JS, Lathrop JW, Porter WA (1977) Solar energy conversion.
US Patent
4,021,323.
11. Rocheleau RE, Miller EL, Misra A (1998) High-efficiency photoelectrochemical hy-
drogen production using multijunction amorphous silicon photoelectrodes.
Energy
Fuels
12(1):3
–
10.
12. Ayers W (1984) Photolytic production of hydrogen.
US Patent Application
06/523,251.
13. Reece SY, et al. (2011) Wireless solar water splitting using silicon-based semi-
conductors and earth-abundant catalysts.
Science
334(6056):645
–
648.
14. Sun K, et al. (2014) Enabling silicon for solar-fuel production.
Chem Rev
114(17):
8662
–
8719.
15. Sun K, et al. (2015) Stable solar-driven water oxidation to O
2
(g) by Ni-oxide coated
silicon photoanodes.
J Phys Chem Lett
6:592
–
598.
16. Mei B, et al. (2014) Iron-Treated NiO as a Highly Transparent p-Type Protection Layer
for Efficient Si-Based Photoanodes.
J Phys Chem Lett
5(20):3456
–
3461.
17. Hunter BM, et al. (2014) Highly active mixed-metal nanosheet water oxidation cat-
alysts made by pulsed-laser ablation in liquids.
J Am Chem Soc
136(38):13118
–
13121.
18. Smith RDL, et al. (2013) Photochemical route for accessing amorphous metal oxide
materials for water oxidation catalysis.
Science
340(6128):60
–
63.
19. Miller EL, Rocheleau RE (1997) Electrochemical and electrochromic behavior of re-
actively sputtered nickel oxide.
J Electrochem Soc
144(6):1995
–
2003.
20. Trotochaud L, Young SL, Ranney JK, Boettcher SW (2014) Nickel-iron oxyhydroxide
oxygen-evolution electrocatalysts: The role of intentional and incidental iron in-
corporation.
J Am Chem Soc
136(18):6744
–
6753.
21. Friebel D, et al. (2015) Identification of highly active Fe sites in (Ni,Fe)OOH for elec-
trocatalytic water splitting.
J Am Chem Soc
137(3):1305
–
1313.
22. Alsabet M, Grde
n M, Jerkiewicz G (2014) Electrochemical growth of surface oxides on
nickel. Part 3: Formation of
β
-NiOOH in relation to the polarization potential, po-
larization time, and temperature.
Electrocatalysis
6(1):60
–
71.
23. Bode H, Dehmelt K, Witte J (1966) Contribution to the Knowledge of Nickel Hy-
droxide Electrodes 1. About The Nickel(II) Hydroxide Hydrate.
Electrochim Acta
11(8):
1079
–
1087.
24. Louie MW, Bell AT (2013) An investigation of thin-film Ni-Fe oxide catalysts for the
electrochemical evolution of oxygen.
J Am Chem Soc
135(33):12329
–
12337.
25. Guzmán RSS, Vilche JR, Arvía AJ (1979) Non-equilibrium effects in the nickel hy-
droxide electrode.
J Appl Electrochem
9(2):183
–
189.
26. Xing W, Li F, Yan Z-f, Lu GQ (2004) Synthesis and electrochemical properties of
mesoporous nickel oxide.
J Power Sources
134(2):324
–
330.
27. Tuomi D (1965) The forming process in nickel positive electrodes.
J Electrochem Soc
112(1):1
–
12.
28. Herzinger CM, Johs B, McGahan WA, Woollam JA, Paulson W (1998) Ellipsometric
determination of optical constants for silicon and thermally grown silicon dioxide via a
multi-sample, multi-wavelength, multi-angle investigation.
JApplPhys
83(6):3323
–
3336.
29. Woollam JA, McGahan WA, Johs BD (1994) Ellipsometric measurement of the optical
properties and electrical conductivity of indium tin oxide thin films. Optical inter-
ference coatings.
Proc SPIE
2253:962
–
968.
30. Petrov I, Barna PB, Hultman L, Greene JE (2003) Microstructural evolution during film
growth.
J Vac Sci Technol A
21(5):S117
–
S128.
31. Trotochaud L, Mills TJ, Boettcher SW (2013) An optocatalytic model for semi-
conductor
–
catalyst water-splitting photoelectrodes based on in situ optical mea-
surements on operational catalysts.
J Phys Chem Lett
4(6):931
–
935.
32. Granqvist CG (1995) Preface.
Handbook of Inorganic Electrochromic Materials
,ed
Granqvist CG (Elsevier, Amsterdam), pp vii
–
viii.
33. Hackwood S, Dautremont-Smith WC, Beni G, Schiavone LM, Shay JL (1981) Volume
changes induced by the electrochromic process in sputtered iridium oxide films.
J Electrochem Soc
128(6):1212
–
1214.
34. Andersson AM, Estrada W, Granqvist C-G, Gorenstein A, Decker F (1990) Character-
ization of electrochromic dc-sputtered nickel-oxide-based films. Optical materials
technology for energy efficiency and solar energy conversion IX.
Proc SPIE
1272:
96
–
110.
35. Koro
š
ec RC, Bukovec P (2006) Sol
–
gel prepared NiO thin films for electrochromic
applications.
Acta Chim Slov
53:136
–
147.
36. Wronkowska AA (1989) An ellipsometric investigation of nickel oxide films in alkaline
solution: The multilayer film approach.
Surf Sci
214(3):507
–
522.
37. Ord JL (1976) An optical study of the deposition and conversion of nickel hydroxide
films.
Surf Sci
56(0):413
–
424.
38. Mills TJ, Lin F, Boettcher SW (2014) Theory and simulations of electrocatalyst-coated
semiconductor electrodes for solar water splitting.
Phys Rev Lett
112(14):148304.
39. Wang H-P, et al. (2013) Realizing high-efficiency omnidirectional n-type Si solar cells
via the hierarchical architecture concept with radial junctions.
ACS Nano
7(10):
9325
–
9335.
40. Taguchi M, et al. (2014) 24.7% record efficiency HIT solar cell on thin silicon wafer.
IEEE J Photovoltaics
4(1):96
–
99.
41. Ahearn JS, Davis GD, Byer NE (1982) Mechanism of anodic oxidation of Hg
0.8
Cd
0.2
Te.
J Vac Sci Technol
20(3):756
–
759.
42. Philipsen HGG, Kelly JJ (2005) Anisotropy in the anodic oxidation of silicon in KOH
solution.
J Phys Chem B
109(36):17245
–
17253.
43. Lichterman MF, et al. (2014) Stabilization of n-cadmium telluride photoanodes for
water oxidation to O
2
(g) in aqueous alkaline electrolytes using amorphous TiO
2
films
formed by atomic-layer deposition.
Energy Environ Sci
7:3334
–
3337.
44. Kumar A, Lewis NS (1990) Short-wavelength spectral response properties of semi-
conductor/liquid junctions.
J Phys Chem
94(15):6002
–
6009.
45. Hu S, et al. (2014) Amorphous TiO
2
coatings stabilize Si, GaAs, and GaP photoanodes
for efficient water oxidation.
Science
344(6187):1005
–
1009.
46. Shaner MR, Hu S, Sun K, Lewis NS (2015) Stabilization of Si microwire arrays for solar-
driven H
2
O oxidation to O
2
(g) in 1.0 M KOH(aq) using conformal coatings of amor-
phous TiO
2
.
Energy Environ Sci
8:203
–
207.
Sun et al.
PNAS
|
March 24, 2015
|
vol. 112
|
no. 12
|
3617
CHEMISTRY