Solar Water Splitting Cells
Michael G. Walter, Emily L. Warren, James R. McKone, Shannon W. Boettcher,†Qixi Mi, Elizabeth A. Santori,
and Nathan S. Lewis*
Division of Chemistry and Chemical Engineering, 210 Noyes Laboratory, 127-72 California Institute of Technology, Pasadena, California 91125
Received July 23, 2010
2.1. Theoretical Efficiencies for Water Splitting
2.2. Calculation of Solar-to-Chemical Conversion
2.3. Photoelectrolysis Cell Configurations
3. Semiconductor Photoelectrochemistry
3.1. Physics of Semiconductor/Liquid Contacts
3.2. Optimization of Interfacial Energetics for
3.3. Energetics of Semiconductor/Liquid Junctions
4. Photocathodes for Hydrogen Evolution
4.1. Photocathode Materials
4.2. Effects of Catalyst Particles on Photocathode
5. Photoanodes for Water Splitting
5.1. Transition Metal Oxides as Photoanodes
5.2. Photoanode-Based Unassisted Water Splitting 6458
5.3. Photoanodes with a Response to Visible
5.4. Factors Leading to Improved Photoanode
6. Dual Band Gap Solar Water Splitting Cells
6.1. p/n-Photoelectrolysis Cells
6.2. Photoanode-Photovoltaic Cells
6.3. Photocathode-Photovoltaic Cells
6.4. Photovoltaic-Photoelectrolysis Cells
7. Effects of Surface-Attached Catalysts on
Photoelectrodes for Water Splitting
7.1. Electrocatalysis for Solar Water Splitting
7.2. Mechanismand Theory of the Hydrogen
7.3. Catalyst Materials for Hydrogen Evolution
7.4. Stability of Catalysts for Hydrogen Evolution
7.5. Materials and Mechanismfor the Oxygen
7.6. Theory for the Activity of Oxygen Evolution
7.7. Catalyst Materials for Oxygen Evolution
7.8. Stability of Catalysts for Oxygen Evolution
7.9. Applications of Electrocatalysts to Solar Water
8. Micro- and Nanostructural Effects on the Efficiency
Energy harvested directly from sunlight offers a desirable
approach toward fulfilling, with minimal environmental
impact, the need for clean energy. Solar energy is a
decentralized and inexhaustible natural resource, with the
magnitude of the available solar power striking the earth’s
surface at any one instant equal to 130 million 500 MW
power plants.1However, several important goals need to be
met to fully utilize solar energy for the global energy demand.
First, the means for solar energy conversion, storage, and
distribution should be environmentally benign, i.e. protecting
ecosystems instead of steadily weakening them. The next
important goal is to provide a stable, constant energy flux.
Due to the daily and seasonal variability in renewable energy
sources such as sunlight, energy harvested from the sun needs
to be efficiently converted into chemical fuel that can be
stored, transported, and used upon demand. The biggest
challenge is whether or not these goals can be met in a cost-
effective way on the terawatt scale.2
Collecting and storing solar energy in chemical bonds, as
nature accomplishes through photosynthesis, is a highly
desirable approach to solving the energy challenge. The
“Holy Grail” of solar energy conversion and storage is
the photoelectrolysis of water using semiconductors as both
the light absorber and energy converter, to store solar energy
in the simplest chemical bond, H2.3Efficiently splitting water
into usable hydrogen could become a new industrial pho-
tosynthesis that would provide clean fuel whose only waste
product upon utilization is water. To accomplish this new
photosynthesis, an economically viable water splitting cell,
composed of stable semiconductors designed to split water
directly at the semiconductor surface, needs to be developed.4
Photoelectrolysis can also be accomplished using photovol-
taic (PV) modules connected directly to electrolyzers and/
or catalytic electrodes. Water splitting cells with direct
semiconductor/liquid contacts are attractive because they
avoid significant fabrication and systems costs involved with
the use of separate electrolyzers wired to p-n junction solar
cells.5Another attractive advantage of photoelectrochemical
water splitting directly at the semiconductor surface is the
ease with which an electric field can be created at a
* Corresponding author. E-mail: firstname.lastname@example.org.
†Current address: Department of Chemistry, University of Oregon, Eugene
Chem. Rev. 2010, 110, 6446–6473
© 2010 American Chemical Society
Published on Web 11/10/2010
In the nearly 40 years since Honda and Fujishima’s
electrochemical photolysis report using TiO2,7the approach
to solving the water splitting problem has been focused on
evaluating new materials for both anodic/cathodic processes
and integrating configurations that utilize photovoltaic cell
junctions, to increase the obtainable voltage for a single or
dual band gap device. The ultimate goal of these efforts is
an efficient photoelectrolysis cell design that can simulta-
neously drive, in an unassisted fashion, both the hydrogen
evolution and water oxidation reactions. Water splitting cells
require semiconductor materials that are able to support rapid
charge transfer at a semiconductor/aqueous interface, that
exhibit long-term stability, and that can efficiently harvest a
large portion of the solar spectrum. To achieve these
functions, multijunction configurations that use p- and n-type
semiconductors with differing band gaps, and surface bound
electrocatalysts, have been the predominant approach for the
development of efficient photoelectrolysis cells.8
This review will focus on the efforts to develop an efficient
semiconductor-based photoelectrochemical water splitting
device in which the only inputs are sunlight and water.9The
fundamental principles of photoinduced electron-transfer
mechanisms at a semiconductor/liquid junction, and opti-
mization of such processes, will be summarized. A wide
range of both p- and n-type semiconductor photoelectrode
materials that have been studied for photoelectrochemical
water splitting cells will then be described. In addition,
heterogeneous electrocatalysts used to facilitate the oxidation
and reduction reactions in electrochemical and in photoelec-
trochemical water splitting systems will be reviewed. Finally,
efforts to combine photoanode, photocathode, and photo-
voltaic-assisted coupled configurations in both single and
dual band gap photoelectrolysis cells will be examined in
terms of their solar-to-hydrogen (STH) efficiency, stability,
and expected viability. Strategies to optimize solar-to-
chemical energy-conversion efficiencies by optimization of
Michael G. Walter earnedaB.S. degreeinchemistryfromtheUniversity
syntheses at the Air Force Research Laboratory at Wright Patterson Air
ForceBase. HecompletedanM.S. degreein2004andPh.D. degreein
2008at PortlandStateUniversity andjoinedtheLewis groupat Caltech
in 2008. He is currently an NSF-ACCF postdoctoral fellow (2009) and
in contact with conductive polymers. His research interests include
molecular semiconductorsfor solar energyconversion, porphyrinmacro-
cycles for optoelectronic applications, and catalyst materials for photo-
Emily L. Warren received a B.S. in chemical engineering at Cornell
Development from Cambridge University in 2006. She is currently a
graduate student in Chemical Engineering at the California Institute of
Technology. Her research interests include semiconductor photoelectro-
chemistry, solar energy conversion, and semiconductor nanowires. She
iscurrentlyagraduatestudent inChemical Engineeringat theCalifornia
Institute of Technology, working under Nathan S. Lewis.
James R. McKoneis inhis thirdyear of graduatestudies intheDivision
of Chemistry and Chemical Engineering at the California Institute of
Technology, workingunder NathanS. LewisandHarryB. Gray. In2008
he graduated from Saint Olaf College with a Bachelor of Arts degree,
double-majoringinmusicandchemistry. Hiscurrent researchfocuseson
reaction using mixtures of earth-abundant transition metals.
Shannon W. Boettcher earned his B.A. degree in chemistry from the
University of Oregon, Eugene (2003), and, working with Galen Stucky,
his Ph.D. inInorganic Chemistry fromtheUniversity of California, Santa
Barbara (2008). Following postdoctoral work with Nate Lewis and Harry
totheUniversity of Oregontojointhefaculty as anAssistant Professor.
His research interests span synthesis and physical measurement with
the goal of designing and understanding solid-state inorganic material
architectures for use in solar-energy conversion and storage.
Solar Water Splitting Cells Chemical Reviews, 2010, Vol. 110, No. 11 6447
light harvesting structured semiconductors, photoinduced
electron transfer and transport, and surface catalysis will also
The free energy change for the conversion of one molecule
of H2O to H2and 1/2 O2under standard conditions is ∆G )
237.2 kJ/mol, which, according to the Nernst equation,
corresponds to ∆E° ) 1.23 V per electron transferred. To
use a semiconductor and drive this reaction with light, the
semiconductor must absorb radiant light with photon energies
of >1.23 eV (equal to wavelengths of ∼1000 nm and shorter)
and convert the energy into H2and O2. This process must
generate two electron-hole pairs per molecule of H2(2 ×
1.23 eV ) 2.46 eV) or four electron-hole pairs per molecule
of O2(4 × 1.23 eV ) 4.92 eV). In the ideal case, a single
semiconductor material having a band gap energy (Eg) large
enough to split water, and having a conduction band-edge
energy (Ecb) and valence band-edge energy (Evb) that
straddles the electrochemical potentials E° (H+/H2) and E°
(O2/H2O), can drive the hydrogen evolution reaction (HER)
and oxygen evolution reaction (OER) using electrons/holes
generated under illumination (Figure 1).
To carry out one or both reactions without recombination,
photoinduced free charge carriers (electrons and holes) in
the semiconductor must travel to a liquid junction, and must
react only with solution species directly at the semiconductor
surface. The electron-transfer processes at semiconductor/
liquid junctions produce losses due to the concentration and
kinetic overpotentials needed to drive the HER and the OER.
The energy required for photoelectrolysis at a semiconductor
photoelectrode is therefore frequently reported as 1.6-2.4
eV per electron-hole pair generated, to account for these
losses.5,10The surface properties of these liquid/semiconduc-
tor interfaces, and how these properties affect the energetics
of photoelectrolysis, will be explored further in section 3.
The practical need for 1.6-2.4 eV to effectively drive water
splitting motivates the use of multiple semiconductors with
different energy gaps, as described in the next section.
2.1. Theoretical Efficiencies for Water Splitting
Bolton classified solar water splitting themes, and esti-
mated their overall efficiencies, based on the number of
photosystems (semiconductor materials) and the minimum
number of absorbed photons per H2molecule. For example,
a single semiconductor material with a band gap of 1.6 eV
(threshold wavelength ) 775 nm) has an ideal maximum
conversion efficiency at 1 Sun of 30%. This system is given
the classification of S2 to indicate a single band gap device
Qixi Mi obtained his B.S. at Peking University in 2003 and Ph.D. at
Northwestern University in 2009. He is currently a postdoctoral fellowof
theNSFCenter for Chemical Innovation(CCI Solar) at Caltech, working
on the development of novel photoanode materials for water oxidation
using visible light, under the supervision of Profs. Nathan S. Lewis and
Harry B. Gray.
Elizabeth Santori received a B.S. in chemistry from the University of
Chicagoin2007. Sheiscurrentlyagraduatestudent inchemistryat the
CaliforniaInstituteof TechnologyinProf. Lewis’slab. Researchinterests
Dr. Nathan Lewis, George L. Argyros Professor of Chemistry, has been
on the faculty at the California Institute of Technology since 1988 and
at Caltechsince1992, isthePrincipal Investigator of theJoint Center for
Artificial Photosynthesis, the DOE’s $122 MMEnergy Innovation Hub in
asanassistant professor from1981to1985andasatenuredAssociate
Professor from1986to1988. Dr. Lewis receivedhis Ph.D. inChemistry
fromthe Massachusetts Institute of Technology. He has been an Alfred
P. Sloan Fellow, a Camille and Henry Dreyfus Teacher-Scholar, and a
the ACSAward in Pure Chemistry in 1991, the Orton Memorial Lecture
Faraday Medal of the Royal Society of Electrochemistry in 2008. He is
currentlytheEditor-in-Chief of Energy&Environmental Science. Hehas
students and postdoctoral associates. His research interests include
artificial photosynthesis and electronic noses. Technical details of these
surfaces and in transition metal complexes, surface chemistry and
organic polymers and polymer/conductor composites, and development
mimicking the mammalian olfaction process.
6448 Chemical Reviews, 2010, Vol. 110, No. 11Walter et al.
requiring two photons to produce one molecule of H2.10The
theoretical efficiency is based upon the conversion of incident
solar energy to chemical energy, eq 1:
where Jg is the absorbed photon flux, µex is the excess
chemical potential generated by light absorption, φconvis the
quantum yield for absorbed photons, and S is the total
incident solar irradiance (mW cm-2).
A dual band gap configuration is given the classification
of (D2) or (D4), indicating a system that requires two or
four photons to produce one molecule of H2, respectively.
A dual band gap (D4) solar water splitting system, realized
by stacking two materials in tandem, has an ideal theoretical
efficiency of 41% using the same theoretical basis used for
S2.10The chemical conversion efficiency for a dual band
gap (D4) system is 27% when including losses due to the
fraction of unused energy per absorbed photon (about 0.8
eV). In this scenario, the two semiconductor materials have
complementary absorption characteristics, such that the top
layer has a band gap of 1.7 eV (730 nm) and the bottom
layer absorbs longer wavelengths, 1.1 eV (1120 nm). If the
two materials were illuminated side-by-side, an increase in
the illuminated area lowers the overall photocurrent density
resulting in an overall efficiency that reflects the area-
weighted average of two S2 photosystems.10
The theoretical limits presented for S2 and D4 systems
do not account for losses in current efficiencies that are
caused by nonradiative recombination of photogenerated
electron-hole pairs in the semiconductor bulk, or for losses
due to the overpotentials required to drive the HER and OER
at the electrode surfaces. The overpotentials required to drive
water oxidation and reduction are especially important for
placement of the valence-band and conduction-band edge
positions of a semiconductor relative to the potentials for
the oxygen and hydrogen evolution reactions, respectively.11
A valence band whose potential is not sufficiently positive,
for water oxidation, or whose conduction band is not
sufficiently negative, for proton reduction, can lead to slow
or negligible water splitting.11Overpotentials of 400 mV at
10 mA cm-2for water oxidation and 50 mV at 10 mA cm-2
for H2evolution at a semiconductor/liquid junction are useful
starting points for estimating how efficiently a photoelectrode
can drive the OER or HER.11
Figure 2 presents the band-edge positions vs the normal
hydrogen electrode (NHE) for several common p- and n-type
semiconductor materials. The semiconductor band-edge
positions are plotted versus their integrated maximum
photocurrent under Air Mass 1.5 illumination, in the
conventional Shockley-Quiesser limit for solar energy
conversion.12,13Some of the large band gap n-type semicon-
ductors in Figure 2 straddle the potentials of the O2/H2O
redox couple, but these materials are not capable of produc-
ing high photocurrent densities under AM 1.5 illumination.
In contrast, the smaller band gap (higher photocurrent) p-type
materials shown in Figure 2 have more negative conduction/
valence bands that are well suited to effect reactions at the
H+/H2potential. A dual band gap (D4) cell configuration
would connect, in series, an n-type material having a
sufficiently positive valence band to drive water oxidation
with a reasonable photocurrent, with a smaller band gap,
p-type semiconductor that would drive the hydrogen evolu-
tion reaction. The use of two semiconductor materials (D4)
remains an attractive option for capturing a large portion of
the solar spectrum, with the two band gaps tuned to absorb
complementary portions of the solar spectrum.14Solar-to-
hydrogen efficiencies of 15% have been calculated for a
model dual band gap tandem p/n photoelectrolysis cell (p/
n-PEC), including losses associated with mass-transport and
kinetic overpotentials.11Energy conversion devices that
utilize multiple semiconductors with different band gaps can
achieve higher efficiencies, as will be discussed in section
In addition to exhibiting an optimal band gap for solar
absorption, semiconductor photoelectrodes must exhibit
excellent oxidative/reductive stability in contact with aqueous
electrolyte solutions. For thermodynamic stability, a semi-
conductor’s reductive and oxidative decomposition potentials
must be more negative than the semiconductor’s conduction
band-edge for water reduction or more positive than the
semiconductor valence band-edge potential for water oxida-
evolution reaction (HER) for overall water splitting (under acidic
conditions); ideal semiconductor material for splitting water at its
surface under illumination with absolute energy scale represented
[left vertical axis (-) and (+)] for Ecband Evband the electrochemi-
cal potentials given by -qE°, where E° is the reduction potential
for both (H+/H2) and (O2/H2O) redox couples.
Oxygen evolution reaction (OER) and hydrogen
Figure 2. Conduction band (left bar) and valence band (right bar)
positions vs NHE of common semiconductors used in photoelec-
trolysis cells. The band gap value is in parentheses, and the ordinate
indicates the maximum theoretical photocurrent under Air Mass
1.5 illumination. The dotted lines indicate the thermodynamic
potentials for water reduction and oxidation, respectively.
Solar Water Splitting Cells Chemical Reviews, 2010, Vol. 110, No. 11 6449
tion, respectively. Very few semiconductor materials exhibit
the necessary requirements for electrode stability in aqueous
electrolyte solutions simultaneously for both water oxidation
2.2. Calculation of Solar-to-Chemical Conversion
Efficiencies for water splitting photoelectrode devices that
require an external bias in order to drive water electrolysis
can be calculated using eq 2 (assuming no corrosion reaction
at the photoelectrodes and a Faradaic efficiency of unity for
both reactions). In order to obtain a true systems efficiency,
these measurements should be done in a two-electrode
configuration as opposed to a three-electrode electrochemical
cell. The efficiency (η) can be calculated from J-V data
using eq 2:
with Vapp as the applied voltage measured between the
oxygen-evolving photoanode and the hydrogen-evolving
photocathode, Jmpas the externally measured current density,
and Pin as the power density of the illumination. In this
scenario, products must be separated to ensure that the
reverse reactions are suppressed, e.g. oxidation of H2at the
photoanode leading to erroneous cell measurements. For an
integrated device (back-to-back photoelectrodes) with no
external wiring, the efficiency must be calculated by physi-
cally collecting the generated products, e.g. hydrogen and
oxygen, and relating the free energy contained in the
chemical products to the energy of the incoming light.14b,16
Using eq 2, a true solar-to-hydrogen (STH) production
efficiency, where the only inputs are sunlight and water, can
be calculated by measuring the photocurrent (Jmp) at Vapp)
0 V (under short-circuit conditions).
The solar conversion efficiency of individual candidate
photoelectrode materials that might be used in a multiple
band gap photoelectrolysis cell to drive either the HER or
OER can be calculated from current-voltage data obtained
using a potentiostat in an illuminated three-electrode cell. It
is useful to calculate efficiencies of a photoanode or
photocathode separately from the other half of the water
splitting reaction, to allow for optimization of the materials
independently. It is important to recognize that characterizing
individual photoelectrodes by the power produced represents
only a portion of the Gibbs free energy needed to split
water.8c,17Therefore, the open-circuit voltage (Voc) and short-
circuit current density (Jsc) are referenced to the thermody-
namic potential of the water splitting reactions (H+/H2, O2/
H2O) at a specific pH. The efficiency (η) of a photoelectrode
can be calculated from its current-voltage data using eq 3,
where Jmpis the current density at the maximum power point
Vmpis the voltage at the maximum power point, and Pin(in
W cm-2) is the power of the incoming illumination. The fill
factor (ff) is calculated using the open-circuit potential Voc
and the value of Jscwith respect to the desired half-reaction.
The ff is defined as the maximum power output divided by
the product of the open circuit voltage and the short-circuit
current density. Further details examining the physical
phenomena associated with the ff characteristics for a
photoelectrochemical cell have been rigorously studied and
are most closely related to surface recombination of electrons/
holes, bulk recombination in the semiconductor, and un-
compensated solution resistances.18These topics will be
covered in greater detail in section 3.
Calculations based on eq 3 are identical to an efficiency
calculation for a photovoltaic cell. However, because the
potentiostat is operated in three-electrode mode, polarization
losses associated with driving the reaction at the counter
electrode are not taken into account in the calculation. Hence,
calculated efficiencies represent only photoanode or photo-
cathode efficiencies and not overall efficiencies for water
By independently characterizing photoanodes and photo-
cathodes, the expected performance of an integrated system
with no external electronics can be directly calculated.
Overall water splitting efficiencies (STH) for photoelectroly-
sis cells can be estimated by overlapping the individually
tested J-V data for each photocathode/anode (Figure 3).11,19
The intersection of the two curves indicates the maximum
operating current density (Jop) for the complete cell. The
highest efficiency for a p/n-photoelectrochemical cell (PEC)
will be obtained when the two curves intersect closest to
their individual maximum power points (PPC or PPA )
Jmp·Vmp, maximum power for photoanode PPAor photocath-
ode PPC). A theoretical p/n-PEC photoanode/photocathode
device is depicted in Figure 3, illustrating the power
generated for each component of the cell (red shaded area)
and the power generated at the maximum operating current
density (blue shaded area).
Increasing the fill factor for one or both of the photoelec-
trodes has a dramatic effect on the solar-to-hydrogen
efficiencies, due to the steep increase in operating photo-
current densities that are produced by small changes in the
fill factor.11,20For a p/n dual band gap photoelectrolysis cell,
increases in the fill factor (from 0.5 to 0.7) for both the
photocathode and the photoanode increased the calculated
overall solar-to-hydrogen efficiency from 10% to 15%.11If
the two curves do not cross, i.e. Jop e 0, the serial
combination of electrodes will not drive the photoelectrolysis
Jmp(1.23V - Vapp)
Figure 3. Overlaid current density-potential behavior for a p-type
photocathode and an n-type photoanode, with overall efficiency
projected by the power generated PSTH) Jop(1.23 V) by the cell
for splitting water.
6450 Chemical Reviews, 2010, Vol. 110, No. 11Walter et al.
Because the voltage generated under illumination is a key
parameter for describing the efficiencies of semiconductor
photoanodes and photocathodes, proper measurement of the
photovoltage is important. Several practical steps should be
taken when measuring the Vocof an individual photoelectrode
to be used in a water splitting cell. For a photocathode
system, e.g. p-InP decorated with platinum catalyst
particles,8cmeasurement of Vocis achieved by submerging
the photoelectrode in electrolyte bubbled with pure hydrogen,
and measuring the voltage under illumination versus a clean
Pt electrode that serves as a reversible hydrogen electrode
(RHE) reference in a three-electrode electrochemical cell.
To accurately measure Vocfor an H2-evolving photocathode,
it is important to have both the oxidized and reduced species
present at the electrode; therefore, H2must be present in the
solution. Because platinum is a good catalyst for both H+
reduction and H2oxidation, the Pt reference rapidly equili-
brates to the thermodynamic potential for hydrogen evolution
in the specific electrolyte of interest. A standard calomel or
silver chloride reference electrode can be used as well,
provided that the reversible hydrogen evolution potential is
first measured in the same solution using a clean Pt electrode,
and the final data are referenced to the explicitly measured
RHE potential. Bubbling hydrogen at 1 atm during J-E data
collection maintains constant H+/H2concentrations in solu-
tion, ensuring a well-defined Nernstian potential with which
the semiconductor can equilibrate. This situation also ap-
proximates that of an actual working device, because under
operating conditions, H2 is generated continuously at the
electrode surface, causing the solution to locally saturate with
the evolved gas. Measurement of Vocwithout bubbling H2,
even versus a standard reference such as the saturated
calomel, will produce an incorrect photovoltage, due to a
poorly defined solution potential (i.e., with no hydrogen
present, the potential of the H+/H2redox couple is undefined,
per the Nernst equation.
A similar method should be used for the measurement of
the Vocfor n-type photoanodes for oxygen evolution. In this
case, the electrode should be submerged in the electrolyte,
O2should be bubbled through the solution, and a standard
reference electrode should be used (due to the relative
difficulty in using the oxygen couple as a reference). The
voltage is measured under these standard conditions versus
the reference, and then converted to a scale based upon the
actual O2/H2O potential, which can be calculated from the
pH of the solution. Current-voltage data should be recorded
under 1 atm of bubbling O2to yield correct Vocvalues.
In addition, for measuring both photocathodes and pho-
toanodes, ultrahigh purity electrolytes should be used, and
can be produced from standard reagent grade salts and
buffers, via overnight pre-electrolysis using two large, inert
electrodes such as platinum or carbon.21Overnight pre-
electrolysis can remove organic impurities and trace metals
through oxidation/reduction, respectively. Note, however, that
if Pt electrodes are used for pre-electrolysis, small amounts
of dissolved Pt can enter the electrolyte, which can become
problematic when studying non-noble catalysts.
2.3. Photoelectrolysis Cell Configurations
A basic photoelectrochemical water splitting device can
be constructed from a single p-type or n-type semiconductor,
i.e. as a single band gap device (S2) or from two semicon-
ductors connected in series, i.e. a dual band gap device -
p/n-PEC (D4). As mentioned in section 2.2, a single band
gap device requires, at a minimum, a semiconductor with a
1.6 to 1.7 eV band gap in order to generate the Vocrequired
to split water, though once other voltage-loss mechanisms
(i.e., catalysis) are accounted for, a band gap above 2 eV is
generally necessary.11To obtain efficient water splitting
devices using currently available semiconductor materials,
a D4 photoelectrolysis cell configuration is advantageous,
due to the ability to explore various combinations of smaller
band gap semiconductor materials that have complementary
absorption and stability characteristics.10,11A dual band gap
photoanode/cathode configuration also allows for a higher
obtainable photovoltage, while partitioning the water splitting
half-reactions between two semiconductor/liquid interfaces.
Another approach to a dual band gap cell uses a single
photocathode/anode in series with a PV cell layer that
provides the additional bias needed to drive the water
splitting reactions. Regardless of the configuration, semi-
conductor/liquid junctions in these devices present unique
challenges to sustain the necessary photovoltages for driving
the reactions at the photocathode and the photoanode.
Multiple semiconductor layers within the cell also incur
losses due to reflected or scattered photons at junctions,
thereby lowering the overall system performance.
Figure 4 depicts the basic configurations with a single
(Figure 4a) or dual band gap (Figure 4b-d) photoelectro-
chemical cell device structure utilizing a back-to-back
“wireless” design that was first pursued by Nozik in a
“photochemical diode” using a p/n-PEC device configura-
tion.14bThese configurations can also be evaluated in separate
electrochemical cells, allowing for the testing and charac-
terization of individual photoelectrodes. A back-to-back
wireless configuration has the potential to be incorporated
into a low-cost, manufacturable device structure. The single
band gap cell depicted in Figure 4a can be constructed using
a p-type photocathode for hydrogen evolution, electrically
connected to an oxygen-evolving metal electrode. Many
known p-type semiconductors that have high absorption in
the visible region of the solar spectrum do not have valence
band-edge potentials that are sufficiently positive to oxidize
water. Therefore, photoelectrolysis cells using one type of
photoelectrode (photocathode or photoanode) can include a
PV cell, connected in series to supply the extra needed bias
voltage. Figure 4c depicts an n-type photoanode electrolysis
cell with a buried p-n junction PV layer connected directly
to the photoelectrode, along with the relative band energetics
of these components with respect to the hydrogen and oxygen
In the cell configurations in (Figure 4a-c), minority charge
carriers generated under illumination (holes in n-type semi-
conductor photoanodes and electrons in p-type semiconductor
photocathodes) are driven to the semiconductor/aqueous
solution interface as a result of the electric field formed at
the semiconductor/liquid contact. Majority carriers recombine
at ohmic contacts that connect the photoelectrodes, or are
transferred to a metal cathode/anode and carry out the
complementary photoelectrolysis step. In the n-type photo-
anode/PV cell (Figure 4c), majority carriers generated in the
PV cell reduce protons in solution on a metal cathode and
minority holes generated in the n-type photoelectrode oxidize
water at its surface.
A dual band gap p/n-PEC (Figure 4b) utilizes both electron
and hole minority charge carriers for water splitting reactions
at their respective semiconductor/liquid interfaces. This is
in contrast to Figure 4d, which depicts two p-n PV cells
Solar Water Splitting CellsChemical Reviews, 2010, Vol. 110, No. 11 6451
connected in series and coated with a metal cathode and
anode. In this configuration, the majority carriers are injected
from the PV cells into the metal cathode and anode to carry
out water reduction and oxidation, respectively.14bPV cells
integrated directly into water splitting cathodes and anodes
are termed “buried” photoelectrochemical junctions, because
no direct liquid/semiconductor junction is formed. Devices
containing buried junctions can be classified as photovoltaic
photoelectrolysis cells (PV-PEC), and are essentially identical
to configuring multiple PV cells wired directly to an
electrolyzer.22The semiconductor surfaces in these configu-
rations can be completely protected from the aqueous
solution using evaporated metal electrodes or thin oxide
layers. The junctions that provide the driving force for
splitting water are therefore buried, and are not directly
related to the difference between the redox potentials of the
H+/H2 or the O2/H2O redox couples or the conduction/
valence band-edge positions, respectively, of the semicon-
ductor photoelectrode.23This type of system encompasses
many reported PEC cell configurations that are composed
of multi band gap, buried p-n junction photoelectrodes, and
does not show semiconductor/liquid junction device proper-
ties that have been described in the next section.24
3. Semiconductor Photoelectrochemistry
To design an efficient solar water splitting cell, it is critical
to understand the fundamental device physics of semicon-
ductors, the thermodynamic and kinetic parameters of
semiconductor-liquid contacts, and the function of surface
electrocatalysts. Several introductory books and reviews that
address semiconductor device physics are suitable for chem-
ists and materials scientists.25Several reviews also address
the photovoltaic performance of semiconductors in contact
with fast, reversible redox couples (i.e., so-called regenerative
cells).25c,26This section aims to present a concise view of
photoelectrochemical semiconductor/liquid contacts, utilizing
the well-developed concepts presented in these earlier books
and reviews. The physics of the semiconductor/liquid contact
under equilibrium conditions will be discussed in section 3.1,
while efforts to optimize the interfacial kinetics of photo-
electrolysis, such the functionalization of the semiconductor
surface with dipoles, are presented in section 3.2. Finally an
analysis of the photogenerated charge carrier pathways and
their effect on the obtainable voltage of an illuminated
semiconductor is presented in section 3.3.
3.1. Physics of Semiconductor/Liquid Contacts
When a semiconductor is brought into contact with a liquid
that contains a redox couple (consisting of the acceptor, A,
and the donor, A-) having electrochemical potential -qE°(A/
A-) where E° is the Nernst potential of the redox pair (A/
A-), electrons will flow between the semiconductor and the
solution until equilibrium is established (Figure 5). Charge
transfer results in an interfacial electric field whose electro-
static potential balances the initial difference in electrochemi-
cal potentials between the solution and semiconductor. After
equilibration, the electrochemical potential (Fermi level) is
the same everywhere. For a photoelectrochemical water
splitting device, the redox couples of interest are the H+/H2
couple for a p-type semiconductor photocathode, and the O2/
H2O couple for an n-type semiconductor photoanode.
For a typical n-type semiconductor photoanode in equi-
librium with a redox species in solution (e.g., O2/H2O), the
electrode will have an excess positive charge, arising from
the ionized dopant atoms in the semiconductor, and the
solution will have an excess negative charge. The positive
Figure 4. Energy diagrams for (a) a single band gap photoanode (n-SC ) n-type semiconductor) PEC with metal cathode back contact;
(b) a dual band gap p/n-PEC configuration with n-type and p-type photoelectrodes electrically connected in series; (c) n-type photoelectrode
in series with an integrated p-n PV cell to provide additional bias and connected to a metal cathode for hydrogen evolution; (d) two p-n
PV cells connected in series and integrated into a metal cathode and anode for water oxidation and reduction.
6452 Chemical Reviews, 2010, Vol. 110, No. 11Walter et al.
charge is spread out over the depletion width, W, in the
semiconductor, whereas the negative charge is spread over
a much narrower region (the Helmholtz layer) in solution,
close to the electrode. An N-type semiconductor is tradition-
ally used as a photoanode because the electric field that is
developed in equilibrium with a redox couple results in band
bending, due to the drop in the electric field strength in
the solid, that directs photogenerated free minority charge
carriers (holes, for n-type semiconductors) to move into the
solution.26aP-type semiconducting electrodes behave in an
analogous manner, except that the ionized dopants are
negatively charged and the solution is positively charged.
Therefore, p-type semiconductors favor electron flow into
the positively charged acceptor species at the interface.
The electric field strength and, hence, the potential energy
barrier in the semiconductor, depend on the initial energy
difference between the Fermi level of the semiconductor and
the value of -qE(A/A-). Because the initial difference in
electrochemical potentials is on the order of 1 eV, and the
depletion width in the semiconductor is typically on the order
of hundreds of nanometers, the electric field in the semi-
conductor can be as large as 105V cm-1.26aCharge carriers
(electron-hole pairs) generated by the absorption of light are
of the relatively large mobilities of charge carriers (10-1000
cm2V-1s-1) in crystalline inorganic semiconductors.
3.2. Optimization of Interfacial Energetics for
The theoretical limit on the efficiency of a photoelectrode
is determined by the energy that can be extracted from the
photogenerated electron-hole pairs. In principle, the maxi-
mum internal energy that can be extracted from these pairs
for an n-type photoanode is given by the difference between
Ecband -qE(A/A-), that is, the barrier height (qφbin Figure
5b). To optimize performance of a semiconductor/liquid
junction, this difference should be as large as possible.
In a regenerative photoelectrochemical cell, the efficiency
of energy extraction can be optimized for a given semicon-
ductor by varying the electrochemical potential of the redox
system. For example, a series of studies has shown that, for
well-defined semiconductor surfaces, such as Si and InP in
nonaqueous solvents, systematic tuning of the redox potential
provides a method for chemically controlling, and optimizing,
the interfacial barrier height.27
In a photoelectrolysis cell, however, the redox couple is
fixed by the desired chemistry, i.e. water oxidation and
reduction. Therefore, the interface energetics cannot be tuned
by simply selecting the appropriate redox couple. However,
the same fundamental limits to energy extraction apply in a
photoelectrolysis cell as in a regenerative photovoltaic cell.
In principle, in water splitting devices several routes could
allow for tuning the relative semiconductor-solution ener-
getics. For example, the redox potentials for water oxidation
(and reduction) are pH dependent, eq 4.
This behavior suggests that the relative energetics of the
semiconductor-solution interface can be tuned by simply
adjusting the pH of the solution. Unfortunately, this tactic
has been largely unsuccessful. The surfaces of semiconductor
electrodes in contact with water are typically covered with
hydroxyl groups. As the pH changes, these surface groups
protonate/deprotonate, resulting in a pH-dependent surface
dipole.28This surface dipole generates a potential drop that
shifts the bands in unison with the shifting water reduction
and oxidation potentials (i.e., ∼60 mV per pH unit). Because
both the oxide band edge and the redox couple are pH
sensitive, the relative interfacial energetics remain pH
independent. One a strategy to circumvent this problem is
to functionalize the semiconductor surface with a pH
insensitive organic group, to thereby produce a surface dipole
that is no longer pH dependent (such as methylated silicon).29
It has yet to be shown, however, that such a strategy is
suitable for photoelectrosynthetic cells that must also incor-
Another promising approach is to operate at a set pH, and
to directly tune the band edge positions relative to the water
oxidation or reduction potentials via the incorporation of
fixed dipoles or charges on the semiconductor surface (Figure
6).26dThese dipoles/charges must be held very close to the
surface of the semiconductor, to avoid screening by the
electrolyte solution. Turner and co-workers demonstrated that
Figure 5. The band energetics of a semiconductor/liquid contact are shown in three cases: (A) before equilibration between the two
phases; (B) after equilibration, but in the dark; and (C) in quasi-static equilibrium under steady state illumination (discussed in section 3.3).
In panel B, qφbis known as the barrier height, and its magnitude determines the theoretical maximum energy that can be extracted from
a separated electron-hole pair at the semiconductor/liquid junction. In panel C, where steady-state illumination yields nonequilibrium
electron and hole populations, EF,nis the electron quasi-Fermi level and EF,pis the hole quasi-Fermi level. The voltage generated by the
junction under illumination is given by the difference between EF,nand -qE(A/A-).
2H2O + 4(h+) f O2+ 4H+
E°(O2/H2O) ) 1.23 V - 0.059 V × pH (vs NHE)
Solar Water Splitting Cells Chemical Reviews, 2010, Vol. 110, No. 11 6453
the chemical functionalization of GaAs photoelectrodes with
porphyrins having varying positive charge leads to systematic
shifts in the band-edge positions.30Other results have shown
that absorbed ions can also modify the band-edge positions
of a semiconductor photoelectrode.31For example, Mallouk
and co-workers showed that F doping and surface adsorption
on TiO2leads to large changes in the band-edge positions.32
Recent experiments by McFarland and co-workers have
shown that incorporation of F into Fe2O3 increases the
observed photovoltage for water oxidation, also presumably
due to the surface dipole and associated band-edge shift.33
3.3. Energetics of Semiconductor/Liquid
Junctions under Illumination
The free energy that is produced by a semiconductor/liquid
contact does not in practice reach the theoretical energy limit
dictated by the interfacial energetics discussed above. Instead,
the actual free energy obtained depends on the kinetics of
the charge carriers in the photostationary state that is
produced as a result of illumination of the solid/liquid
interface.27aIn the solid-state physics formalism, the free
energy generated by the semiconductor is given by the
difference between the hole and electron quasi-Fermi levels
under illumination, that is, by the free energy difference
between the majority carriers and the photoexcited minority
carriers (Figure 5c). The quasi-Fermi level is simply a
description of the electrochemical potential of one carrier
type at a time (i.e., either electrons or holes) under nonequi-
librium (e.g., illuminated) conditions, using Fermi-Dirac
statistics to describe separately the populations of electron
and of holes.34The prefix quasi- refers to the fact that
thermalization of the excited carriers is a fast process, leaving
the collection of holes (and electrons) each in quasi-thermal
equilibrium under steady-state illumination. True equilibrium
can be reached only in the dark, through slower conduction-
band to valence-band recombination processes, at which
point a single Fermi level describes the statistical distribution
of all carriers. The degree of splitting between the electron
and hole quasi-Fermi levels under no net current flow is
referred to as the open-circuit voltage (Voc) and can be
measured experimentally in suitable systems.34a-c
The magnitude of the photovoltage that is generated by a
semiconductor/liquid junction determines the photoelectro-
chemical reactions that can be driven by that system. Even
if the appropriate band-edge positions of the semiconductor
straddle the water oxidation and reduction levels, water
splitting is not possible unless the photovoltage exceeds 1.23
V.11,35This can be explained in a kinetic framework by the
principle of microscopic reversibility, in that the rates of
the forward, fuel-forming reactions must exceed the sum of
the reverse, fuel-consuming, rates.35a,36Understanding and
controlling the relevant kinetic processes that affect the
photovoltage is essential in the design of efficient semicon-
ductor photoelectrodes. These kinetic processes govern the
respective electron and hole concentrations at the interface
under quasi-equilibrium conditions, and can be broken down
into five different categories (Figure 7). The photogenerated
charge carriers can (1) recombine in the bulk of the solid
(Jbr), (2) recombine in the depletion region (Jdr), (3) tunnel
through the electric potential barrier near the surface (Jt),
(4) thermally surmount the interfacial potential barrier
(thermionic emission) (Jet), or (5) recombine at defects (trap
states) at the semiconductor/liquid interface (Jss).18In some
cases, for example for well-surface-passivated silicon single
crystals, all of the mechanisms except the fundamentally
limiting bulk recombination (Jbr) can be suppressed.27a,37Such
devices generate the maximum possible voltage in the
photostationary state. In other cases, surface recombination
Figure 6. The presence of an interfacial dipole layer in panels B and C can significantly affect the electrostatic barrier (qφb) and hence the
photoelectrode performance. For simplicity, the band diagrams are shown at the flat-band condition in the absence of band bending. Evac
is the energy of an isolated electron in vacuum and Emaxrepresents the theoretical maximum energy that can be extracted from an electron-hole
Figure 7. Recombination pathways for photoexcited carriers in a
semiconductor photoelectrochemical cell held at open circuit can
be broken down into at least five different categories, represented
by the thin arrows in the diagram. The electron-hole pairs can
recombine through a current density associated with radiative or
nonradiative recombination in the bulk of the semiconductor (Jbr),
depletion-region recombination (Jdr), surface recombination due to
defects (Jss), tunneling current (Jt), and electron-transfer current
associated with majority carriers traversing the interfacial barrier
(Jet). Electron collection by the back contact and hole collection
by the redox couple (e.g., oxidation of water to O2) are processes
that contribute positively to device efficiency, and these are depicted
by thick black arrows.26aReproduced with permission from ref 26a.
Copyright 2005 American Chemical Society.
6454 Chemical Reviews, 2010, Vol. 110, No. 11Walter et al.
(Jss) or charge transfer across the interface (Jet) dominates
the recombination processes.26d
The analytical expression for the photovoltage (Voc)
generated at a semiconductor/liquid junction is given by the
ideal diode equation solved for the situation at zero net
where n is the diode quality factor, kB(m2kg s-2K-1) is
Boltzmann’s constant, T (in K) is the temperature, q (C) is
the charge on an electron, Jph(A m-2) is the photocurrent
density, Jsis the saturation current density, which is related
to the sum of the recombination pathways outlined above,
and γ is the ratio of the actual junction area to the geometric
surface of the electrode (i.e., the roughness factor).38As will
be discussed in more detail below, this relationship implies
a decrease in Vocat room temperature of g59 mV for each
10-fold increase in γ. Section 8 provides further discussion
about the effects of surface roughness on the photoelectrode
performance. This equation also shows that the photovoltage
is dependent on the illumination intensity, with a higher
photon flux leading to more minority carriers and hence to
a larger splitting of the quasi-Fermi levels.39
When any of the recombination currents 2-5 contribute
significantly to Js, these rate processes limit the photovoltage
of the device, and thus limit the ability to drive photoelec-
trochemical reactions (Figure 7). This point is critically
important, in that it is often remarked that if a semiconductor
meets the thermodynamic requirements regarding the position
of the valence and conduction bands with regard to the redox
potentials of HER and OER, this semiconductor should drive
the unassisted photoelectrolysis of water. A semiconductor
can only drive the photoelectrolysis of water if recombination
is sufficiently suppressed such that the quasi-Fermi level
splitting and, hence, the photovoltage, exceeds 1.23 V under
illumination. Note also that, in optimized semiconductors
(e.g., single crystal silicon or GaAs), the maximum attainable
photovoltages are typically ∼0.4 V smaller than the band
gap of the semiconductor.26a
A key challenge in the development of water splitting
devices is to fabricate semiconductor photoelectrodes for
water oxidation and/or reduction that have bulk and surface
properties such that the generated photovoltage reaches the
fundamental bulk-recombination limit. The kinetics of charge
transfer from a bare semiconductor surface to a redox species
in solution depends both on the number of electrons (or
holes) at the semiconductor surface and on the energetics of
the semiconductor band edges.
Consider the well-studied case of an n-type semiconductor
in the dark, in contact with an electrolyte that contains a
redox couple and is under external potential control.27cThe
electron concentration at the surface of an n-type semicon-
ductor, ns, is determined through a Boltzmann-type relation-
ship by the difference between the potential applied to the
electrode, E, and the flat-band potential, Efb:
where Ndis the concentration of donor atoms. At forward
bias, ns increases exponentially with the application of a
negative potential, resulting in a net current across the
semiconductor/liquid interface. The net flux of electrons from
the conduction band to acceptors dissolved in solution is
given by the rate law:
where J is the current density (A cm-2), ketis the electron-
transfer rate constant (cm4s-1), and [A] is the acceptor
concentration (cm-3). Unlike the case for metallic electrodes,
the surface electron concentration is explicit in the rate law
for electron transfer at semiconductor electrodes. Hence,
application of a potential to an ideally behaving semiconduc-
tor electrode interface effects a change in the observed current
density (i.e., the charge-transfer rate) by changing the electron
concentration at the surface of the solid, as opposed to
changing the rate constant, or the energetics, of the interfacial
charge-transfer process. In the dark, few minority carrier
holes are present, so their contribution to the rate can be
neglected. This majority-carrier electron-transfer process is
a source of detrimental recombination current, and in
practical devices it is minimized by introduction of a large
electrostatic barrier at the semiconductor surface.25b
The kinetics of minority carrier charge transfer under
illuminationsi.e. useful currentscan be analyzed using a
similar kinetics framework.35aThe surface minority-hole
concentration for an n-type semiconductor is described by
the energy difference between the hole quasi-Fermi level and
the valence band edge.40Like the case in the dark, the rate
of forward hole transfer, from the n-type semiconductor to
the solution, is related to the concentration of holes at the
semiconductor surface. Furthermore, the driving force for
hole transfer to the solution is dictated by the energy
difference between the valence band edge and the redox level
of the species in solution (for water splitting, this is H2O or
OH-, depending on the pH) and is independent of the light
intensity or potential applied to the electrode, provided that
the semiconductor remains in depletion. If the potential of
the valence band edge is sufficiently positive of the formal
potential of redox couple to give rise to a large electron-
transfer rate constant, this condition can be considered a
“built-in” overpotential for driving a given redox reaction.26c
4. Photocathodes for Hydrogen Evolution
Photocathodes used for a water splitting cell need to supply
sufficient cathodic current to reduce protons to H2and must
be stable in aqueous environments. In addition, to success-
fully reduce protons to H2, the potential of the conduction
band edge of the photocathode must be more negative than
the hydrogen redox potential. The mechanism for the HER
is pH dependent: at low pH, the HER proceeds primarily by
the reduction of protons, whereas at high pH, water is
primarily reduced to produce hydroxide ions.41The overall
reactions are shown in eq 8.
As has been summarized earlier, a semiconductor brought
into contact with a liquid electrolyte phase will experience
Fermi level equilibration with the electrochemical potential
(Eredox) of the liquid by transferring charge across the
interface. For a p-type semiconductor, the bands bend in such
a way that photogenerated electrons are driven toward the
interface, while holes are swept into the bulk of the solid.
Photoexcitation thus injects electrons from the solid into
solution. This cathodic current may, to some extent, protect
the surface of the semiconductor from oxidation. For this
Voc) (nkBT/q) ln(Jph/γJs) (5)
J(E) ) -qket[A]ns
2H++ 2e-f H2
2H2O + 2e-f H2+ 2OH-
Solar Water Splitting CellsChemical Reviews, 2010, Vol. 110, No. 11 6455
reason, p-type semiconductors can be expected to be more
stable than their n-type counterparts.
Many p-type semiconductors have been investigated as
electrochemical photocathodes for the HER, but the scope
of this review will be limited to those that have been most
thoroughly investigated or that have received attention
relatively recently. Work has also been performed on
regenerative photoelectrochemical systems and on splitting
HBr or HI (which require a smaller overall potential
difference to drive the decomposition reaction), which
provides useful insights for water splitting applications.42
4.1. Photocathode Materials
GaP has an indirect band gap of 2.26 eV, with band edges
that straddle the hydrogen reduction potential. The n-type
form of this material is unstable in aqueous solution, but
p-GaP is stable for extended periods of time under reducing
conditions.43Memming demonstrated the ability of GaP to
produce H2positive of the H+/H2thermodynamic potential.
One of the drawbacks of GaP is that it has small minority-
carrier diffusion lengths relative to the absorption depth of
visible light in the solid.44Recent work using a regenerative
ferrocenium/ferrocene redox system has demonstrated that
structuring this material, by etching macropores, increases
the photocurrent of n-GaP, and similar strategies should be
applicable to p-type materials.45
InP has a band gap of 1.35 eV, which makes it a good
solar absorber; however the scarcity and high demand of
indium limit the commercial viability of the material, at
least in traditional wafer format.46Using Ru catalyst islands
on the surface of oxidized InP, Heller et al. were able to
achieve 12% solar to chemical conversion for the produc-
tion of H2.47By electrodepositing Rh and Re on p-InP
electrodes, efficiencies of 13.3% and 11.4% were obtained,
respectively (efficiencies were calculated as the ratio of
energy that could be produced in an ideal fuel cell using the
photoelectrochemically produced H2and the solar irradiance
striking the photocathode).
Turner and other researchers have systematically studied
the properties of GaInP2, whose band gap of 1.83 eV is large
enough to split water. Although the potential of the conduc-
tion band edge of GaInP2is more negative than the hydrogen
reduction potential, the valence band potential is negative
of the oxygen potential, so an additional bias is needed to
effect the overall decomposition of water.48Some success
was achieved in shifting the VB edge more positively by
modifying the surface with quinolinol groups.31aMeasure-
ments in a variety of different electrolyte pHs has indicated
that p-GaInP2is most stable under acidic conditions.15c
With a band gap of 1.12 eV, p-Si is a desirable p-type
small band gap absorber for possible use in dual band gap
p/n-PEC water splitting configurations.29Several groups have
demonstrated that planar p-Si photocathodes, combined with
a variety of metal catalysts, can be used to reduce the voltage
required to electrochemically produce H2.49Photon to
hydrogen conversion efficiencies as high as 6% (under low-
level monochromatic 633 nm illumination) have been
reported for p-Si decorated with Pt nanoparticles.49bSi is
stable in acidic conditions, but surface oxidation can occur
over extended periods of time. Passivation of the Si surface
by covalent attachment of methyl groups has been shown to
improve the stability of p-Si photocathodes.29,50
Currently, II-VI semiconductors such as CdTe and
CdIn1-xGaxSe2(CIGS) dominate the thin film photovoltaics
market. These materials have band gaps that can be
controlled by modifying their composition and processing.
Several different groups have investigated the properties of
II-VI semiconductors, and have used them to split HI and
HBr with relatively high efficiencies. Bard and co-workers
made single crystals of the group VI selenides and used them
to split HI in water, using methyl viologen as a redox shuttle
for the hydrogen evolution process.42,51Parkinson et al. used
molybdenum and tungsten diselenide anodes, along with InP
cathodes coated with noble metal catalysts, to split HBr with
an efficiency of 7.8% (based on the amount of free energy
stored).14cOther work showed that during HI decomposition,
H2bubbles formed at the crystal steps of WSe2, while I2
was produced over the flat surfaces of the crystal.52The
Wrighton group tested p-WS2for a variety of PEC reduction
reactions, and achieved 6-7% efficiency, and very high
open-circuit voltages for hydrogen production, through the
use of platinum catalysts.53More recently, several groups
have reported new ways to create thin film photocathodes
by controlling the stoichiometry of copper chalcopyrites.
CuInxGayS2materials are interesting because they have direct
band gaps that can be adjusted by modifying the In:Ga ratio
(ranging between 1.52 to 2.5 eV).54Work by Fernandez et
al. demonstrated that p-type films of CuIn1-xGaxS2, fabricated
by sputtering the materials onto Mo-coated glass and then
sulfurizing the electrodes, have the appropriate energetics
to split water at moderate pH (5-7). Miller et al. made
CuGaSe2(Eg) 1.65 eV) films that had nonideal band-edge
alignment for photoassisted water splitting, but nevertheless
exhibited very high photocurrents.24cWork by Valderrama
showed that the onset for cathodic current was negative of
the H2reduction potential and that Vfbwas 0.168 V vs SCE
(∼0.4 V vs NHE, more positive than H+/H2).55Cuprous
oxide has a direct band gap of 2.0 eV, and can be fabricated
by several different methods including thermal oxidation,
sputtering, and electrodeposition.56Although Cu2O has been
used as both a photoanode and a photocathode, the VB edge
of Cu2O is located at ∼0.9 V vs NHE,15amaking it ineffective
for oxygen evolution.57Additionally, Cu2O is unstable with
respect to photocorrosion in aqueous conditions, and has a
relatively low electrical conductivity. The favorable band gap
of Cu2O thus makes it a more likely candidate for use in a
regenerative photoelectochemical cell.57,58
4.2. Effects of Catalyst Particles on
Although several semiconductors have band-edge positions
that are appropriate for the photoelectrochemical reduction
of water, the kinetics of HER on the bare semiconductor
surface generally limit the efficiency of this reaction.41
Overcoming this kinetic limitation requires a stronger driving
force, i.e. an overpotential, to drive the desired chemical
reaction. In turn, the overpotential lowers the usable voltage
output, and hence lowers the efficiency of the photocathode.59
Addition of a catalyst to the surface (often in the form of a
nanoparticulate metal film) can improve the kinetics of the
reaction.60Additionally, if the metal is deposited such that
the metal particles are discontinuous and smaller than the
wavelength of incident photons, the metal film will ef-
fectively be optically “transparent” and not significantly
affect the light absorption properties of the semiconductor.61
To date, the most efficient system based on a p-type
semiconductor has been achieved using p-InP decorated with
6456 Chemical Reviews, 2010, Vol. 110, No. 11Walter et al.
Pt catalyst islands, yielding a 13.3% conversion efficiency
Semiconductors that are coated with discontinuous films
of metal nanoparticles behave essentially like the semicon-
ductor alone in contact with a redox couple, albeit with the
catalytic activity of the metal coating. This result is different
from the behavior of a continuous metal film on a semicon-
ductor surface, which forms a Schottky barrier whose height
is determined by the metal/semiconductor interface and
whose energetics are independent of the solution. Two
approaches have been advanced to explain this phenomenon.
The first, developed by Heller, is based on the fact that
alloying the metal with H2can change the work function of
the metal and hence increase the semiconductor/metal barrier
height.62Because the catalyst material has equilibrated its
Fermi level with the H+/H2redox potential, the driving force
for the reduction reaction will only depend on the barrier
height of the semiconductor/metal junction.8cThe need for
nonuniform metal films can be explained by the fact that a
continuous film would limit the amount of light absorbed
by the semiconductor, and would thus reduce the photocur-
rent from the system.
Another approach to explaining the hybrid behavior of
nanoscale metal particles on semiconductors is to treat the
semiconductor/liquid interface as a junction with an inho-
mogeneous barrier height. It is well-known that semiconduc-
tor-metal Schottky junctions exhibit “Fermi level pinning,”
in which the barrier height (φb) of the junction is much lower
than the theoretical difference in work functions between
the two materials. However, if the diameters of the metal
particles are comparable to the depletion width of the
semiconductor, under certain conditions, the high barrier
height liquid junction can dominate the band bending in the
semiconductor. The predicted size dependence on the effec-
tive barrier height has been observed for n-Si/Ni/electrolyte
systems.63In the regime in which the scale of the metal
particle is comparable to, or smaller than, the depletion width
of the semiconductor, the current density depends strongly
on the band bending of the semiconductor as well as on the
spatial dependence of the barrier height.63If the metal
nanoparticles are small and well dispersed on the semicon-
ductor surface, the effective barrier height of the system will
be controlled by the semiconductor/electrolyte interface, so
the effect of the metal on the energetics of the system will
be “pinched off” (Figure 8).
Both the H2 metal-alloy and “pinch-off” theories offer
reasonable explanations for the behavior of semiconductor/
metal electrolyte systems with metal island nanoparticles.
Work by Szklarczyk et al. and by Rossi et al. provides
experimental evidence that supports the “pinch-off” theory.21,63
Szklarczyk’s work systematically investigated photoelectro-
catalysis for a variety of different metals on p-Si, and found
that the photoevolution of H2 for a variety of metal
nanoparticle catalysts was directly related to the metal’s
exchange current density for H2evolution in the dark.21
Theoretically, the pinch-off effect can be exploited to
promote the desired catalytic reaction at the semiconductor
surface. A metal has a continuum of energetic states, so a
higher density of majority carriers exists in the metal than
in the semiconductor. This should enable the facile transport
of minority carriers into the metal, where the electrons can
more easily be used in the redox reaction. Decoration of the
semiconductor surface with pinched-off catalysts will also
protect the semiconductor from corrosion, because the surface
concentration of minority carriers will be decreased in the
semiconductor if the carriers are preferentially transported
through the catalyst islands. It is well documented that the
addition of Pt nanoparticles to a p-type semiconductor surface
significantly lowers the overpotential required to evolve H2,
but these systems are often limited by low photovoltages.49b
A more detailed understanding of the interface of semicon-
ductor/metal/aqueous systems may help to improve the
efficiencies of photocathode devices by exploiting this
While several promising candidate materials exist for the
photocathode of a solar water splitting device, more work is
needed to simultaneously optimize both the photoconversion
efficiency of the semiconductor and the rate of catalysis of
the HER. This issue will be discussed in greater detail in
5. Photoanodes for Water Splitting
An oxygen evolving photoanode material must be an
n-type semiconductor, such that the electric field generated
by band bending drives holes toward the surface. The
material must have a band gap and band-edge positions that
are suitable for use in a single or multiple band gap system,
as well as electrical properties such as doping and resistivity
that allow for efficient collection of charge carriers. Ad-
ditionally, the material needs to be stable under water
oxidization conditions, and if the interfacial kinetics of the
OER are determined to be rate limiting, an oxygen evolution
catalyst must be placed on the electrode surface.
Due to the requirement of stability under oxidizing
conditions, most of the photoanode materials that have been
investigated are metal oxides or metal oxide anions (oxo-
metalates), in pure, mixed, or doped forms. A general trend
in the electronic structures of these oxides and oxometalates
is that the valence band (VB) consists of O 2p orbitals, and
the conduction band (CB) is formed by the valence orbitals
Figure 8. Plot of the valence band edge as a function of position
in a p-type semiconductor behind a mixed barrier height contact,
where ΦE is the barrier height between the semiconductor and
solution and ΦMis the barrier height between the semiconductor
and the metal. ∆ ≡ ΦE- ΦM(the difference in barrier height
between the two contacting phases), and W is the depletion width
behind an unperturbed region of the liquid contact. A particle with
a sufficiently small radius will be pinched off, meaning that the
effective barrier height (Φeff) across a mixed contact junction will
be much larger than that of the low barrier height metal contact.
The extent to which pinch off occurs in a given system depends
on ∆, the band bending, and the depletion width in the semiconduc-
Solar Water Splitting Cells Chemical Reviews, 2010, Vol. 110, No. 11 6457
of one or more metals. An implication of this trend is that,
especially in ionic crystals, the potential of the VB edge stays
relatively unchanged at 3.0 ( 0.5 V vs NHE for most metal
oxides and oxometalates including TiO2, SrTiO3, WO3, Fe2O3
and ZnO (Figure 2).64Metal ions, either as bulk matrix or
dopant species, serve to tune the CB position, and hence the
Upon photoexcitation and charge separation of an n-type
semiconductor, minority carriers (holes) in the VB diffuse
to the semiconductor-electrolyte interface to oxidize water.
The disparity between the oxygen-centered VB at ∼3.0 V
and the OER potential at 1.23 V vs NHE presents a major
challenge for the development of high-performance photo-
anode materials. Thus much of the excess ∼1.77 eV absorbed
by the oxide is wasted by thermal relaxation. Unfortunately,
few semiconductors satisfy both requirements of electronic
structure and stability for photoanodes; thus most examples
of functioning photoanodes convert sunlight to O2 at
relatively low efficiencies.
In the following sections, we discuss first several classes
of photoanode materials that have been shown to produce
oxygen under illumination. Second, we outline several efforts
at effecting full water splitting using a single oxide photo-
electrode, and describe the corresponding limitations. Third,
we describe strategies that have been used to increase the
sensitivity of oxides to visible light. Finally, we discuss
general factors that have led to improved photoanode
performance, implying appropriate strategies for moving
5.1. Transition Metal Oxides as Photoanodes
Several classes of transition metal oxide materials have
been shown to satisfy some of the requirements for efficient
photoelectrochemical oxygen evolution. The redox-active
metal ions in metal oxide photoanodes can include early
transition metals, e.g., Ce(IV), Ti(IV), Zr(IV), Nb(V), Ta(V),
as well as d10configuration ions, e.g., Zn(II), Ga(III), Ge(IV),
Sn(IV) and Sb(V). Additionally, group 1 through 3 metal
ions can be added as inert components to help produce
specific crystal structures. TiO2has been extensively studied
following the experiment by Fujishima and Honda,7and has
also played a central role in dye-sensitized solar cells
(DSSCs).65However, the potential of the CB edge of TiO2
lies slightly positive of the HER potential and therefore
electrons in the CB do not effect the net reduction of water
into H2,65,66unless the photoelectrode is operated under non-
standard-state conditions (e.g., with a gradient in pH between
the photoanode and cathode).7,67This limitation can be
overcome by use of the titanates SrTiO3and BaTiO3, for
which the addition of Sr2+and Ba2+cations results in the
perovskite structure, and moves the potential of the CB edge
more negative than NHE. When loaded with appropriate H2
evolution cocatalysts such as Rh or Pt, SrTiO3 fully
decomposes water, using near-UV photons, with internal
quantum yields that approach unity.
As an isoelectronic compound to SrTiO3, NaTaO3also
adopts the perovskite structure. Partly because of its sizable
band gap (4.0 eV), and therefore its greater driving force
for H2evolution, NaTaO3exhibits a relatively high activity
for water splitting, albeit with deep-UV light. In particular,
by doping with 2% La, the surface area of the NaTaO3crystal
can be expanded by a factor of 8, with the emergence of
terrace-shaped dislocations that spatially separate the active
sites for H2(cocatalyzed by NiO) and O2evolution.68As a
result, this system has yielded an O2evolution rate of 9.7
mmol/h in pure water, when illuminated with a 400 W Hg
The stoichiometric addition of more highly charged metal
counterions into titanates, niobates and tantalates tends to
produce modified perovskite structures that have intervening
O2-layers. Representative photoanodes of this type include
La2Ti2O7, K2La2Ti3O10, Ba5Nb4O15, Sr2Ta2O7and Ba5Ta4O15.
Nonbridging O atoms in these structures can be viewed as
terminal oxo MdO groups, and are believed to form the
catalytic center for O2evolution.69Doping La2Ti2O7with
Ba increases the surface concentration of hydroxyl groups,
and greatly enhances the rate of water photoelectrolysis.
Crystals with pyrochlore-type structures, including Gd2Ti2O7,
Y2Ti2O7 and Cs2Nb4O11, have been reported to show
considerable water splitting activity. By comparison, oxides
with d10-configuration ions generally yield far inferior
performance for water splitting, due to the higher electrone-
gativity of these elements. However, Zn-doped Ga2O3can
decompose water with a Ni cocatalyst at a rate analogous to
that of the most efficient Ti-, Nb-, or Ta-based photocatalysts.
One explanation for this behavior is that the introduction of
Zn2+produces p-doping, and thus increases the hole mobility
in the VB.70
5.2. Photoanode-Based Unassisted Water
Single band gap semiconductor cells that drive the
unassisted photoelectrolysis of water are typically composed
of wide band gap n-type semiconductors that can generate
the necessary photovoltages needed to drive both the OER
and the HER. Such electrodes are presently limited to
materials that have band gaps greater than ∼3.0 eV, which
sets an upper limit on the attainable solar conversion
efficiency at ∼2%.
Several metal oxides with n-type semiconducting proper-
ties, such as SrTiO3and KTaO3, can drive the unassisted
photoelectrolysis of water under solar illumination; however,
solar-to-chemical conversion efficiencies are <1%.71Waki
et al. demonstrated that GaN (Eg∼ 3.4 eV) could split water
with UV illumination, but both the lack of stability of the
electrode in aqueous solution and the excess current gener-
ated from photocorrosion limit the practicality of using this
material.72?-Ge3N4 is a rare example of a non-oxide
photocatalyst that stably effects overall water splitting.73
5.3. Photoanodes with a Response to Visible
Several strategies can be used to decrease the band gap
of oxides and oxometalates that have d0or d10configurations.
Metal ions that have a strong polarizing capability result in
MsO bonds that have substantial covalent character, and
thus produce an oxygen-to-metal charge transfer absorption
in the visible region.74Moreover, the potential of the CB
edge can be made more positive through the incorporation
of transition metals, whereas introduction of N3-and S2-
anions shifts the VB edge to more negative potentials.75
The high nuclear charge of Mo6+and W6+enables these
ions to form covalent MsOsM networks that are respon-
sible for the yellow color of the oxides and of the polyoxo-
metalate clusters of these metals. The 2.7 eV band gap of
WO3results in a CB edge potential slightly positive of NHE
(Figure 2); nonetheless, WO3can oxidize water to O2as a
6458 Chemical Reviews, 2010, Vol. 110, No. 11Walter et al.
photoanode76or with sacrificial oxidants, without the need
for an additional OER cocatalyst. Computational studies
suggest that, under working conditions, the WO3surface is
completely covered by oxygen atoms, with water oxidation
taking place via hydroperoxide and hydroxyl intermediates.77
The soft and polarizable d10and d10s2outer shells of Cu+,
Ag+, Cd2+, Hg2+, In3+, Tl3+, Sn2+, Pb2+and Bi3+have a
propensity for mixing into the oxygen-centered VB, and the
oxides of these metals appear yellow, red or brown. In2O3
is well-known to be an efficient photoanode material,
although it is only responsive to near-UV light.78Litharge
(R-PbO) has a band gap of 1.9 eV, and with the ferri-/
ferrocyanide redox couple gives rise to photocurrent over a
wide range of the visible spectrum (400 nm < λ < 650 nm).79
However, during water splitting, PbO is converted into
metallic PbO2, and no O2is evolved.80
Recently, many colored composite oxides composed of
soft metal ions (e.g., Bi3+, Pb2+) have been synthesized.
Monoclinic BiVO4has a band gap of 2.4 eV, resulting from
a high-lying VB composed of the Bi 6s and O 2p atomic
orbitals.81BiVO4catalyzes O2evolution driven either by
visible photons or by the oxidant AgNO3, as is the case for
the isoelectronic material PbMo1-xCrxO4(Eg) 2.3 eV).82
For oxides in the series InVO4(Eg) 1.9 eV), InNbO4(2.5
eV) and InTaO4(2.6 eV), potentials of the VB edges are
located at ∼1.8 V vs NHE. The band gaps of these oxides
also straddle the HER and OER potentials.83Although H2is
generated from the photolysis of pure water with a NiO
cocatalyst, conflicting results regarding O2evolution with
these oxides have been published by the same authors.83,84
BiYWO6 (Eg ) 2.7 eV) also effects the stoichiometric
splitting of water under visible light.85PbBi2Nb2O9possesses
a sandwiched perovskite structure with alternating (Bi2O2)2+
and (PbNb2O7)2-layers, and displays a visible-light response
for either the HER or OER in the presence of appropriate
Group 7-10 transition metal oxides have been more
widely utilized as OER cocatalysts than as bulk photoanode
materials, because of the scarcity of the metals and/or the
tendency to form partially filled d shells. The VB-to-CB
transitions of these compounds are localized d-d transitions.
In addition, the unpaired d electrons create low-lying excited
states that have energy levels between the VB and CB. The
orbitals provide an energy ladder for recombination of charge
carriers by nonradiative relaxation processes. For hematite
(R-Fe2O3), with a band gap of 2.2 eV, optical measurements
have revealed that, below the main absorption edge, two
weak peaks are present, at 1.4 and 2.0 eV, that can be
attributed to crystal field transitions.87Consequently, the
charge recombination rate in hematite is intrinsically high,
and the hole diffusion length is only 2-4 nm,88severely
impairing the performance of this material as a photoanode.
Substantial effort has thus been devoted to the fabrication
of thin layers89and nanoparticles90of hematite, to accelerate
charge collection. Gra ¨tzel et al. have reported that silicon
doping of hematite promotes both its electrical conductivity
and the development of dendritic nanostructures.91These
researchers have thus far attained a short-circuit current
density of >3 mA cm-2under AM 1.5G sunlight.92Recent
work from the Gra ¨tzel group (Sivula et al.) has also shown
good photoactivity for mesoporous films that were sintered
on SnO2glass electrodes at high temperatures (∼800 °C),
demonstrating the ability to form efficient hematite photo-
anodes from colloidal suspensions.93Mixed oxide semicon-
ductors in the Ti-Fe-O system have also been synthesized
and characterized.94TiO2codoped with Cr3+/Sb5+as well
as with Rh3+/Sb5+exhibited d-d type band gaps of 2.1-2.2
eV, and showed activity for O2evolution in the presence of
AgNO3.82,95The Sb5+codopant maintained charge balance
and the oxidation states of Cr3+and Rh3+in the crystal lattice.
Chalcogenide and pnictide semiconductors have been
intensively studied with respect to their physical parameters,
but not with respect to their interfacial properties in contact
with aqueous electrolytes. These anions are softer bases than
O2-, and thus give rise to more negative VB potentials than
those of metal oxides. GaN and ZnO are isoelectronic and
possess essentially identical wurtzite structures, as well as
very similar direct band gaps (3.3-3.4 eV). In contrast, in
solid solutions of GaN:ZnO, the two relatively soft ions, Zn2+
and N3-interact to generate a new absorption band in the
400-500 nm region.96Domen et al. reported that these
compounds, along with Rh2-xCrxO3, are active and stable
photocatalysts for the decomposition of water at pH ) 4.5.97
Photoelectrochemical water oxidation has also been dem-
onstrated using GaN:ZnO and IrO2at a potential of ∼0.95
V vs RHE.98(Oxy)nitrides of Ti and Ta offer further
reductions in the band gaps, to values of 2.0-2.5 eV, and
rely partially on AgNO3 as a sacrificial reagent for O2
evolution.99In situ IR characterization has indicated that the
surface of TaON is covered by an overlayer of Ta2O5, which
serves as the OER active site and protects the buried N3-
anions from oxidation.100CdS features an attractive, direct
band gap of 2.4 eV, and can thermodynamically drive full
water splitting. However, in practice, elemental sulfur or
polysulfide are usually the oxidation products.101By applying
a Nafion film that adsorbs a Ru complex as the OER
cocatalyst, photoinduced holes have been rapidly harvested
at the CdS surface and passed on for O2evolution.102The
potentials of the VB edge of GaP (Eg) 2.3 eV) is almost
equal to the standard OER potential, but GaP can neverthe-
less function as a photoanode for water oxidation in neutral
or basic water. To eliminate photocorrosion, GaP needs to
be covered by a protective layer, such as Sn-doped In2O3
(ITO), before attachment of an OER cocatalyst such as
5.4. Factors Leading to Improved Photoanode
The development of stable photoanode materials that can
absorb visible light is a key challenge for effecting the
oxidation of water at semiconductor surfaces. An ideal light
absorber would behave as a low-pass color filter, whose
extinction coefficient approaches infinity above the band gap
of the absorber. Such materials are characterized by their
pure, rich color, and include materials that are widely used
as pigments, including BiVO4, better known as bismuth
yellow,and CdS1-xSex, known as cadmium red.104An indirect
band gap results in weaker absorption at longer wavelengths
and produces a dull, unsaturated color. In a direct band gap
semiconductor, the absorption of incident photons is complete
within a surface depth of 100-1000 nm, so charges need
only to travel a comparatively short distance to reach the
solution. As a result, semiconductors with direct band gaps
can be made into nanoparticles or thin films as a path to
obtain efficient devices. Unfortunately, it is not easy to
predict based on crystal or putative electronic structure
whether a semiconductor’s band gap will be direct or indirect.
For example, although GaN, GaP and GaAs all exist in the
Solar Water Splitting Cells Chemical Reviews, 2010, Vol. 110, No. 11 6459
zincblende structure, GaP has an indirect band gap, whereas
GaAs and GaN exhibit direct band gaps.
Crystal structures and crystal phases cast profound influ-
ence on the properties of oxide semiconductors. For VBs
that consist of mostly O 2p orbitals, the electronic coupling
between adjacent orbitals is modulated by the dihedral angle,
with stronger O 2p-2p couplings contributing to more
positive VB edge potentials as well as larger hole mobilities.
Accordingly, a more delocalized or “thicker” VB, centered
on O 2p orbitals, is facilitated by the 180° Ta-OsTa angle
in KTaO3, compared to the 143° angle in LiTaO3.105The
cubic perovskite- or ReO3-type structure is therefore arguably
optimized for the purpose of acting as a photoanode material.
Indeed, almost all of the highest-performance photoanodes
based on Ti, Nb or Ta have a perovskite, or layered
perovskite, structure. Isoelectronic materials adopt the same
crystal structure, such as the aforementioned pairs SrTiO3/
KTaO3, PbMo1-xCrxO4/BiVO4, and ZnO/GaN. Isoelectronic
substitution or mixing could thus be a helpful approach to
fine-tune the electronic structure of a known material.
Introduction of inert, lattice-mismatching atoms, as in
NaTaO3:La and Fe2O3:Si, creates crystal defects, and frus-
trates the growth of large single crystallites. Therefore,
various nanostructures can be readily prepared to obtain a
high surface area and/or small crystal domains, without
special dispersion techniques, possibly allowing for more
efficient collection of charge carriers in spite of low
mobilities. Additionally, doping that maintains the crystal
lattice produces free charge carriers, and thus dramatically
increases the electrical conductivity, as in the case of Ga2O3:
Zn. As an even simpler doping strategy, oxygen vacancy
sites as n-type dopants can be produced at higher tempera-
tures and lower O2partial pressures. A third type of doping
can be considered alloying with an element from the same
group. More importantly, like the formation of eutectic
mixtures, nonlinear results are often obtained. For instance,
doping GaP with 0.5-1% N shrinks its 2.26 eV indirect
band gap into a ∼2.1 eV direct band gap,106and alloying
colorless PbMoO4with yellow PbCrO4yields orange or red
From a materials perspective, a huge range of candidate
oxides and structured wide band gap materials may have
many of the characteristics desired for efficient photoanodic
water oxidation. A recent review mentions many of the 130
candidate metal oxides that have been studied for electro-
chemical water splitting.107To sort through these candidates,
and seek out new materials, combinatorial methods have been
developed to quickly screen large numbers of material
combinations.108Metal oxide material combinations can be
deposited onto electrodes, and automated systems have been
used to test for photoactivity under illumination. For example,
a metal oxide testing kit has been developed by Parkinson
et al. at the University of Wyoming under the name SHArK
(Solar Hydrogen Activity Research Kit), to seek out new
photoelectrolysis candidates.108a,109The kits are being utilized
as a research and educational platform for use in high schools
and undergraduate/graduate institutions. While such combi-
natorial efforts hold promise for discovery of new photoan-
ode candidate materials, they should not be considered a
substitute for targeted study of known systems or their
6. Dual Band Gap Solar Water Splitting Cells
Honda and Fujishima’s photoelectrolysis report suggested
that the efficiency of the water decomposition process could
be increased by coupling a p-type photocathode to the TiO2
photoanode.7This type of water splitting cell provides an
increase in light absorption through the use of a smaller band
gap semiconductor, and more importantly, generates enough
photovoltage to drive the hydrogen evolution reaction (p/n-
PEC, Figure 4b).14cThe use of TiO2as a photoanode material
has significant limitations with respect to the amount of
photocurrent that can be generated under white light il-
lumination. In addition, with two semiconductor-liquid
junctions present, two surface electrocatalysts are required,
to reduce the overpotential for each electrode. This constraint
is especially relevant at the photoanode, at which the
overpotential losses often exceed 0.5 V. The maximum
system external quantum efficiency is limited to 50%,
because absorption of two photons, one photon per semi-
conductor material, is required to generate one electron that
is capable of fuel production. From a systems perspective,
in this arrangement, four photons are required to produce
one molecule of H2, and eight photons are required to
produce one molecule of O2.
6.1. p/n-Photoelectrolysis Cells
Yoneyama et al. constructed a photoelectrolysis cell using
an n-type TiO2photoanode and a p-GaP photocathode p/n-
PEC configuration. Open-circuit voltages were, however, less
than the 1.23 V needed to successfully demonstrate unas-
sisted photoelectrochemical water splitting.110Stability issues
with the p-GaP electrode precluded sufficient evaluation of
this cell. Nozik reported unassisted solar water splitting using
a back-to-back photoelectrode configuration, again using
p-type GaP photocathodes and TiO2photoanodes.14a,bThe
dual p/n-PEC cell geometry was termed a “photoelectro-
chemical diode” and represented a p/n-PEC that contained
two semiconductor/liquid junctions (Figure 4b). These
devices operated at less than 1% efficiency, and encountered
stability issues due to the degradation of the p-GaP photo-
cathodes used to drive the hydrogen evolution reaction.
Ohashi et al. also evaluated the n-TiO2/p-GaP dual band gap
system, and calculated a solar to hydrogen conversion
efficiency (STH) of ∼0.1% under 50 mW cm-2of illumina-
tion.111This report also evaluated several p/n-PEC configura-
tions of p-CdTe/n-TiO2(STH of 0.04%), p-GaP/n-SrTiO3
(STH ) 0.7%), and p-CdTe/n-SrTiO3(STH ) 0.2%). It is
important to note that, with a band gap of 3.4 eV, SrTiO3
can split water by itself, albeit with lower overall efficiencies
than might be obtainable in a p/n-PEC.112
Mettee and Calvin reported a dual band gap p/n-PEC
“heterotype” cell that used n-Fe2O3 with RuO2 surface
catalysts, and p-GaP with Pt islands, to drive the photoelec-
trolysis of water.113Due to their electrode setup and
configuration, the cells were illuminated side-by-side and
were wired together externally, thereby halving the photo-
current density compared to a “stacked” cell design. A close
comparison was made between this dual band gap cell and
the two-step light reactions of natural photosynthesis, to
oxidize water and reduce NAD to NADH2. The dual n-Fe2O3/
p-GaP cell was run in neutral pH (Na2SO4) and in seawater,
exhibiting solar-to-hydrogen efficiencies of 0.01% and 0.1%,
respectively. The authors noted the poor electrical perfor-
6460 Chemical Reviews, 2010, Vol. 110, No. 11Walter et al.
mance and quantum efficiency of the n-Fe2O3as significantly
limiting the overall efficiency of the photoelectrolysis
An efficient p/n-PEC for the unassisted photoelectrolysis
of water was a configuration developed by Kainthla et al.
that consisted of a photocathode composed of p-type InP
coated with Pt islands, and an n-type GaAs electrode coated
with a thin protective MnO2layer.20,114The photoelectrolysis
cell operated at 5 mA cm-2with an overall solar-to-chemical
conversion efficiency of 8.2%. Testing of the individual
photoelectrodes in separate three-electrode cells indicated
short-circuit photocurrent densities over 18 mA cm-2for the
MnO2-protected GaAs electrode and almost 30 mA cm-2
for the Pt decorated p-InP photocathode. Although the two
curves were not overlaid in the report, it is clear that the
operating current density of a “stacked” configuration would
have been near 10 mA cm-2, but the cells were illuminated
side-by-side, halving the current density. Although the
efficiency is impressive, the materials used in this deviced
incorporated expensive, non-earth abundant materials (In),
with surface bound Pt electrocatalysts, suggesting expensive
Akikusa and Khan reported a water splitting cell that used
two wide band gap materials, p-SiC (Eg ∼ 2.9 eV) and
n-TiO2 (Eg ∼ 3.0 eV).115Although n-SiC is prone to
photocorrosion in electrolyte solutions,116p-SiC is cathodi-
cally protected under illumination, and exhibits a remarkably
negative conduction band-edge potential (-1.3 V vs NHE)
for reducing protons to H2. A photoelectrode composed of
p-SiC coated with Pt islands and a TiO2 photoelectrode
illuminated side-by-side split water with an open-circuit
voltage of 1.25 V and an operating current density of 0.05
mA cm-2, for an overall solar-to-hydrogen efficiency of
0.06%. The use of two such wide band gap semiconductors
allows for the generation of sufficient photovoltages; how-
ever, it limits the available operating photocurrent to the
maximum output of the higher band gap photoelectrode.
A homotype dual band gap configuration incorporating
both p-type/n-type Fe2O3photoelectrodes was reported by
Kahn and Ingler, in which the two photoelectrodes were used
in tandem for the unassisted photoelectrolysis of water.117
Unfortunately, the poor photoresponse and carrier transport
of p-type Fe2O3severely limited the efficiency of these cells,
resulting in a photocurrent density of 0.1 mA cm-2under
100 mW cm-2illumination (solar-to-hydrogen efficiency of
0.11%). Other devices that used a dual band gap configu-
ration included a two electrode p/n-PEC made from an
n-type, nanostructured WO3photoanode and a p-type GaInP2
photocathode.118The structured device could not split water
with light intensities <1 W cm-2, due to a Fermi level
mismatch of the two semiconductors and low absorption of
visible light by the WO3. At higher light intensities, a
photocurrent of 20 µA cm-2was observed.
6.2. Photoanode-Photovoltaic Cells
An alternative approach to providing additional voltage
to a single band gap water splitting device involves integra-
tion of a p-n photovoltaic (PV) cell in series with the PEC
photoanode material (Figure 4c). Morisaki et al. reported a
hybrid TiO2 photoanode coupled with a p-n junction Si
photovoltaic device, by depositing the TiO2photoelectrode
on top of the PV cell, with the latter protected from the
solution by a Pt metal electrode. Light that was not absorbed
by TiO2could be captured by the p-n Si photovoltaic layer,
and thus provided the additional bias needed for water
splitting.19bIn a single band gap PEC + (PV cell) device, a
dark anode or cathode with catalytic properties can be used,
or surface-bound catalysts can help drive the hydrogen/
oxygen evolution reactions.
An efficient single band gap n-type photoanode PV-
coupled device has been reported by Gra ¨tzel and Augustyn-
ski, who reported a structured WO3photoanode in series with
a dye-sensitized TiO2solar cell (DSSC).65Some features of
this cell include the use of a DSSC module to provide the
necessary bias to drive the hydrogen evolution reaction at a
platinum cathode in series with a WO3photoanode. Several
patents related to the use of an Fe2O3photoelectrode coupled
to a DSSC report efficiencies as high as 2.2%. This work is
based upon success with Si-doped nanostructured Fe2O3
electrodes deposited on Sn-doped In2O3, which have shown
incident-photon-to-current efficiency (IPCE) values as high
as 42% (at 370 nm) and photocurrent densities (Jsc) of 2.2
mA cm-2under 1 Sun illumination.91Subsequent work by
Brillet et al. used an Fe2O3photoanode that was connected
in series to two DSSCs, one of which used a squaraine
sensitized cell, followed by an N749 (black dye) sensitized
DSSC to form a trilevel tandem configuration.119The
reported operating photocurrent densities (Jsc) were 0.94 mA
cm-2with an overall STH of 1.36%.
Miller et al. reported a single band gap device that used a
PV layer to provide an additional bias with good stability.
The system consisted of a WO3semiconductor photoanode
coupled to a triple-junction amorphous photovoltaic stack.19a,120
This configuration utilized a photoactive nanostructured thin
film of either WO3, Fe2O3, or TiO2that had been deposited
using low temperature (<300 °C) reactive sputtering.19bThe
lower temperature deposition of the WO3 films is an
important consideration for both the cost effectiveness of
developing this system and for the materials chosen to
construct the photoelectrodes. These metal oxide films were
sputtered onto an a-Si:Ge tandem PV cell that was electrically
connected to a hydrogen evolution catalyst electrode, such
as Pt or RuO2, and ohmically connected to the front metal
oxide photoanode. This configuration provides another
example of complementary absorbing layers in which the
bias provided by the tandem PV cell is buried in a solid-
state junction, and utilizes photons with energies that are
not absorbed by the metal oxide film. These devices showed
operating photocurrent densities of <1 mA cm-2when Fe2O3
and TiO2were used as photoanodes. Devices that used WO3
exhibited operating current densities of 0.45 mA cm-2, with
solar-to-hydrogen efficiencies of 0.7% while operating for
10 h in acidic media under AM 1.5 illumination. A similar
WO3photoanode PEC-PV based device was later reported
using a triple junction amorphous-Si tandem PV cell to
increase the bias voltage, resulting in a solar-to-hydrogen
efficiency of 0.6% with generated current densities of 3 mA
cm-2.19aMiller et al. also have reported the use of a triple
junction a-Si solar cell that was coated with NiFeyOxand
CoMo catalysts to yield 7.8% efficiency.121This design is
an example of a photovoltaic-photoelectrolysis cell with
catalyst-coated electrodes that are essentially wired to an a-Si
solar cell. Several other configurations similar to this
particular implementation are presented in section 6.4.
6.3. Photocathode-Photovoltaic Cells
A very efficient photoelectrolysis cell that utilizes a direct
semiconductor/liquid electrolyte junction is the 12.4% solar
Solar Water Splitting Cells Chemical Reviews, 2010, Vol. 110, No. 11 6461
water splitting configuration based on a p-GaInP2 photo-
cathode connected in series to a p-n GaAs junction
photovoltaic layer, reported by Khaselev and Turner.48bThis
device was modeled after a solid-state tandem cell and used
complementary absorbing layers, with the p-GaInP2photo-
cathode having a band gap of 1.83 eV to absorb the visible
part of the spectrum, while the p-n GaAs, with a band gap
of 1.42 eV, absorbed the near-infrared part of the spectrum
that was not absorbed by the p-GaInP2layer. The p-GaInP2
layers were grown using metal-organic chemical-vapor
deposition (MOCVD) and were coated with a Pt catalyst.
This device can be classified as a dual band gap configura-
tion, with the p-GaInP2semiconductor/liquid junction car-
rying out hydrogen evolution with photogenerated minority
charge carriers, while the GaAs PV cell layer provided the
additional bias needed to drive the oxygen evolution reaction.
Although p-GaInP2has an ideal band gap of 1.8 to 1.9 eV,
its band edges do not straddle the HER and OER reactions,
requiring a 0.3 V bias to drive the hydrogen/oxygen evolution
reactions.14bThe strategy of using a single semiconductor/
liquid junction coupled to a complementary absorbing
photovoltaic layer utilizes an unsymmetrical charge-carrier
injection configuration. Photogenerated minority carriers
(electrons) in the p-GaInP2layer effect water reduction while
majority carriers (holes) generated from the GaAs PV cell
effect water oxidation at a Pt electrode. The cell produced
120 mA cm-2at short circuit under 1190 mW cm-2white
light illumination, and showed limited long-term stability.
The limitations of this configuration include a short photo-
cathode lifetime, in addition to pitting of the GaInP2electrode
surface under focused illumination.15cThis device configu-
ration also has limited large-scale feasibility due to the
expensive vapor-phase epitaxy processing that is required
to grow the GaAs p-n junction as well as to produce the
6.4. Photovoltaic-Photoelectrolysis Cells
As was mentioned in section 2.3, a PV-PEC configuration
is essentially a PV cell that is connected directly to an
electrolyzer, with the major difference being that the metal
catalyst is deposited directly onto shielded, photoactive
semiconductor materials.121These semiconductor/metal/liquid
junctions can, in principle, operate at any pH, because the
semiconducting layers are completely protected from the
acidic/alkaline solutions used for photoelectrochemical water
splitting, thereby avoiding changes in the semiconductor/
liquid junction properties and possible photoelectrode stabil-
ity issues. The cost of a buried junction configuration,
however, is likely to be greater than that of a single
photoelectrode, due to the necessity of creating junctions
within the electrode, to create the needed photovoltage.
Several device structures are presented here to illustrate the
efficiency and stability properties of buried junction photo-
Khaselev, Bansal, and Turner demonstrated a 16% efficient
n/p-GaInP2/GaAs(Pt) and 7.8% triple-junction p-i-n a-
Si(Pt) photovoltaic/electrolysis cell configurations.122The n/p-
GaInP2/GaAs(Pt) design is similar to the earlier reported
p-GaInP2/p-n GaAs photoelectrolysis cell with the addition
of an additional n-GaInP2to form two PV cells in series.
The n/p-GaInP2/GaAs tandem cell exhibited a Voc) 2.32 V
under 100 mW cm-2of illumination, Jsc) 13.4 mA cm-2,
and ff ) 0.62. The cell configurations required a Pt
electrocatalyst surface for both the HER and OER reactions.
The cells demonstrated highly efficient solar water splitting
cells that incorporate an electrolyzer directly onto the surface
of the PV cells and were found to be more efficient than
separate electrolyzers coupled to PV cells.
Yamane et al. have recently reported a composite semi-
conductor electrode for water splitting that used an n-Si/p-
CuI/ITO/n-i-p a-Si/n-p GaP/ITO/RuO2configuration. With
this cell, they observed a 2.2 V shift from the anodic current
that was observed when only a RuO2electrode was used to
oxidize water.123This cell was developed around a photo-
voltaic structure that used p-CuI/n-Si to generate a high open-
circuit photovoltage, with a sputtered ITO ohmic top contact.
The RuO2catalyst layer was prepared on ITO using either
an electrodeposition step or a chemical deposition technique,
with the latter exhibiting smaller current density under water
oxidation conditions. The overall chemical conversion ef-
ficiency of this cell was 2.3%, while generating a photocur-
rent density of 1.88 mA cm-2. This device demonstrates an
effective method of integrating photovoltaic junctions within
a photoelectrode that is coated with highly active electro-
catalysts. For the unassisted photoelectrolysis of water, it
was necessary to generate over 2.0 V of potential to drive
the reactions at both electrodes. A platinum counter electrode
was required for hydrogen evolution, and RuO2was depos-
ited onto ITO to effect oxygen evolution.
Yamada and co-workers demonstrated a 2.5% STH
efficient amorphous-Si triple-junction (PV) device that
integrated hydrogen and oxygen evolution catalysts coated
on the electrode surface.124A Co-Mo alloy was used as the
HER catalyst, while Fe-Ni-O was used for the oxygen
evolving catalyst. The novelty of this device was that the
entire photovoltaic electrolysis cell was built on an ITO
electrode and was protected from the solution using epoxy,
allowing the entire device to be submerged for testing. The
device was operationally stable for 18 h. This device can
also be categorized as a buried junction, due to the buried
triple junction a-Si photovoltaic stack that was protected from
solution by the catalyst layer.
A unique configuration developed by Licht et al. placed
an AlGaAs/Si photovoltaic stack on top of a RuO2/Pt dual
electrode that was submersed in the electrolyte solution.24a
This design was essentially a photovoltaic cell connected
directly to the electrocatalyst surface components (RuO2and
Pt). However, the system avoided problems of light attenu-
ation due to bubble formation on the photoelectrode surface
and avoided any possible photoelectrode degradation due to
splitting water at the photoelectrode surface. Impressive solar-
to-hydrogen efficiencies of 18% were obtained with good
operational stability over >12 h. The cost limitations for this
cell are significant when considering the large area of the
Pt/RuO2dual electrode, in addition to the MOCVD-grown
AlGaAs/Si photovoltaic junctions needed to provide a large
The best known industrial attempt at a photoelectrochemi-
cal system for hydrogen production and storage was devel-
oped by Texas Instruments in 1981, and used HBr as the
medium, evolving H2 at a photocathode and Br2 at a
photoanode.125Researchers developed pairs of silicon p-n
junction particles that were wired in series to generate enough
photovoltage to oxidize Br-to Br2and to reduce protons to
evolve H2. The H2and Br2were stored (hydrogen in a metal
hydride and Br2in solution), and used when needed in a
fuel cell to generate electricity. The efficiency of the H2-
6462 Chemical Reviews, 2010, Vol. 110, No. 11Walter et al.
and Br2-generating solar panels was 8.6%, with all reactants
and products stored and continually reused in a closed
A summary of the various multiple band gap water
splitting cell configurations reviewed, photoelectrodes used,
and their efficiencies (organized by cell type) is presented
in Table 1.
7. Effects of Surface-Attached Catalysts on
Photoelectrodes for Water Splitting
Typically, electrolysis requires application of greater than
1.23 V between the anode and the cathode because of the
kinetic barriers that are commonly encountered in performing
multistep, multielectron reactions. For example, hydrogen
evolution from acidic water requires combining two protons
and two electrons to make a chemical bond. On most bare
semiconductor surfaces, the formation, or reductive desorp-
tion of intermediate hydride species presents a large energy
barrier to hydrogen production.
The catalytic behavior of a given material (in the dark)
can be quantified by the current density that can be passed
at a given overpotential. The overpotential is simply the
voltage applied to the electrode relative to the redox potential
of the relevant couple (e.g., H+/H2) in the electrolyte of
interest. The overpotential is necessary to drive the kinetically
rate-limiting step of the multistep oxidation or reduction
reaction. In the case of photoelectrodes, the electrocatalytic
behavior is convoluted with the device properties of the
semiconductor/liquid contact, and will affect the overall
performance of a photoelectrolysis cell (Figure 9).
A common, and often necessary, strategy to improve the
performance of photoelectrochemical devices is to add
catalytic units to the surface of the semiconductor. Typically
these electrocatalysts are deposited as thin layers or as
nanoparticles so as to avoid excessive light absorption or
reflection, preserve desired interfacial energetics (e.g., via
“pinch off,” see section 4.2, Figure 8), and improve the
kinetics of the desired reactions.
Aside from improving the photoelectrode kinetics, the
addition of catalyst particles fundamentally changes the
energetics of the electron transfer process at the semiconduc-
tor surface. As discussed in section 3.1, the excited minority
carriers in a pristine semiconductor/liquid junction thermalize
in the semiconductor to the band edge level (Figure 5).
However, when a metallic catalyst particle is on the
semiconductor surface, the minority carriers that participate
in the redox reaction come from the catalyst particle, whose
Fermi energy is in equilibrium with the minority carrier
quasi-Fermi level in the semiconductor, as in a traditional
metal-semiconductor Schottky contact.34a,bTherefore, the
addition of catalyst particles to a bare semiconductor surface
can actually lead to a loss in driving force, and thus produces
a slower electron-transfer rate constant for the reaction.
However, this loss in driving force is often offset by
drastically increased catalytic turnover rates, thereby increas-
ing the overall device efficiency.
Figure 9 depicts the effects of surface catalysts on a
photoelectrode. For a generic “good” catalyst, ∼100 mV is
needed to drive a current density of ∼10 mA cm-2, whereas
for a “bad” catalyst ∼600 mV is needed to produce the same
current density. In principle, the open-circuit voltage (Voc)
produced by use of either catalyst should be the same,
because the catalyst has no effect on the thermodynamics
of the reaction. However, the fill factor and, hence, overall
photoelectrode efficiency are drastically reduced by poor
7.1. Electrocatalysis for Solar Water Splitting
For use in a water splitting device, a good HER or OER
catalyst must satisfy two basic requirements. First, the
catalyst must be highly active toward its respective reaction,
meaning it must be capable of producing, at a minimum
overpotential, large quantities of hydrogen or oxygen as
quickly as the absorber can supply electrons or holes to the
catalyst. Second, an effective catalyst must be robust enough
to maintain its efficiency over time scales relevant to
commercial use. Catalysts described in the industrial elec-
trolysis literature largely meet these requirements, although
continued research in the area demonstrates that improve-
ments would still be beneficial.
A Tafel relationship describes well the catalytic perfor-
mance of electrodes for the hydrogen or oxygen evolution
half-reactions (eq 9),
where η is the overpotential, I is the observed current, and
I0is the exchange current. The Tafel slope, b, is a measure
of the potential increase required to increase the resulting
Table 1. Mutiple Band Gap Solar Water Splitting Cells
materials and configurations
η ) (STH) %
Figure 9. Schematic showing the qualitative effect of surface
catalysts on photoanode performance. Dotted curves show the
current density vs overpotential relationship for generic “good” and
“bad” catalysts. The solid curves show the performance of a
hypothetical oxygen-evolving semiconductor photoanode integrated
with either a “bad” or “good” catalyst.
η ) blog(I/I0) (9)
Solar Water Splitting Cells Chemical Reviews, 2010, Vol. 110, No. 11 6463
current 1 order of magnitude, usually reported in mV/decade.
The exchange current corresponds to the intercept at η ) 0,
extrapolated from a linear portion of a plot of η versus log(I)
(i.e., a Tafel plot). This reaction rate under dynamic
equilibrium is also described in terms of geometric area as
the exchange current density, J0(e.g., A cm-2). Qualitatively,
J0can be thought of as a measure of how vigorously the
forward and reverse reactions occur during dynamic equi-
librium, while b is a measure of how efficiently the electrode
can respond to an applied potential to produce current.
Though researchers often invoke J0values alone to describe
the catalytic performance of an electrode, the Tafel slope is
also an important measure of electrode performance, because
it accounts for changes in mechanism at different overpo-
tentials. Taken together, the Tafel slope and the exchange
current density can often determine whether improved
performance of an electrode after some modification is due
to electronic, geometric (surface-area), or combined effects.
For example, Figure 10(A) compares two hypothetical
catalysts with comparable active surface areas (correlated
with J0) and different electronic activities (correlated with
b), whereas Figure 10(B) compares two catalysts with
comparable electronic activity but with different active
7.2. Mechanismand Theory of the Hydrogen
Many materials that are potentially useful as photocathodes
for the HER do not have surfaces that are sufficiently
electrocatalytic to support light-driven H2evolution without
a large additional electrical bias. The construction of an
efficient device for splitting water with sunlight thus requires
the attachment of a more active HER catalyst to the absorber
surface. Much of the work on candidate catalysts has come
from academic and industrial efforts that have been focused
on the development of cheap, efficient water electrolysis
systems. An overview of HER catalysis by Trasatti contains
an exhaustive description of the available heterogeneous
The HER is one of the most well-studied electrochemical
reactions. It is understood to proceed by one of two
mechanisms,41each consisting of two primary steps, as
outlined in eqs 10a, 10b, and 10c. The “discharge” step, eq
10a, proceeds in all cases, while only one of 10b or 10c
typically predominates to complete the reaction. The asterisk
(*) represents a binding site at the electrode surface, while
A-refers to the conjugate base of the reduced (acidic) proton,
namely, H2O in acidic media and OH-under alkaline
Exchange current densities for the HER on pure metals
in acid have been measured by numerous researchers.126
Plotting these values versus the metal-hydrogen bond
strength gives rise to a characteristic pattern, known as the
“volcano” relation (Figure 11), where the HER activity
increases to a peak value obtained at intermediate bond
strengths and then decreases at higher bond strengths.
In highly alkaline media, comparable collected values of
the exchange current density are not readily available; rather,
the overpotentials required to achieve a set current density
were measured for the transition metals by Miles.127A plot
of such data against periodic group gives a similar volcano
Figure 10. Polarization curves and Tafel plots (inset) illustrating the electronic and geometrical effects of catalysts. (A) Dashed blue: J0
) 10-5A cm-2, b ) 40 mV/decade. Solid black: J0) 10-5A cm-2, b ) 120 mV/decade. (B) Dashed blue: J0) 3 × 10-3A cm-2, b )
120 mV/decade. Solid black: J0) 3 × 10-5A cm-2, b ) 120 mV/decade.
Figure 11. The volcano relation for pure metals in acidic solution
(Trasatti).126The noble metals Pt and Pd demonstrate exceptionally
high activity, with Ni as the most active non-precious metal.
Reproduced with permission from ref 126. Copyright 1972 Elsevier.
HA + e-* f H•* + A-
HA + H•* + e-* f H2+ A-
2H•* f H2
6464 Chemical Reviews, 2010, Vol. 110, No. 11Walter et al.
relation, with Pt, Pd, and Ni among the most active pure
metals (Figure 12). This behavior is consistent with the notion
that the predominant HER mechanism for a given metal is
the same in acidic and basic media.
Based on the observed volcano relation, the catalytic
activity toward the HER is understood to arise from the
strength of the interaction between the catalyst surface and
adsorbed hydrogen. In the 1950s Gerischer and Parsons
independently introduced the first frameworks for under-
standing the nature of hydrogen catalysis on transition metals
on the basis of hydrogen adsorption energies.128Both reached
the conclusion that the ideal HER electrocatalyst is the one
with negligible hydrogen adsorption free energy. The fol-
lowing is a contracted form of Parsons’ discussion of why
this is the case.
Considering only the discharge step, which all mechanisms
share in common, Parsons showed that the exchange current
can be described by eq 11:
where aH+ is the activity of protons, θ* is the fractional
equilibrium coverage of hydrogen on the metal, Eais the
activation energy for the reaction step, and R is the symmetry
factor for the electrochemical energy barrier. The quantity
θ* is determined by the free energy of adsorption. In case
of R ) 0.5, the maximum rate is attained when both θ* and
(1 - θ*) are equal to 0.5, corresponding to coverage of half
of the available metal sites by reduced hydrogen and an
adsorption free energy of zero, that is, equal stability for
both adsorbed and absent hydrogen at adsorption sites.
Parsons showed that all three distinct HER steps obey a
similar relation, so that, regardless of the predominant
mechanism or rate-limiting step, the maximum exchange
current for the reaction is attained when the free energy of
hydrogen adsorption is zero, or nearly zero. Extensive
experimental work has indeed shown that protons converted
to hydrogen on platinum surfaces bind with small adsorption
Unfortunately it is very difficult to use any fundamental
parameters of a material to predict hydrogen adsorption
energies or resulting HER activities. Recently, though,
Norskøv and co-workers have used density functional theory
to build a predictive model of HER activity on the basis of
calculated adsorption energies.130This model was able to
reproduce, with some accuracy, the volcano curve for metal
catalysts. Furthermore, the calculations predicted high cata-
lytic activity for certain metal composites and reasonably
high activity for edge sites on lamellar MoS2. Both predic-
tions have been supported by subsequent experiments.131
7.3. Catalyst Materials for Hydrogen Evolution
Aside from pure metal catalysts, electrochemists have
studied myriad other materials for efficient, low-cost hydro-
gen evolution.132These can be divided into two broad
categories: metal composites/alloys and compounds that
incorporate nonmetallic elements. A few examples are
highlighted below, but the reader is directed to the available
review literature for a more exhaustive analysis.58Several
selected catalyst Tafel plots are collected and presented in
Nickel metal has demonstrated high performance when
subjected to treatments to increase its active surface area.138
Accordingly, the majority of examples from the literature
of mixed-metal catalysts involve additions of another metal,
or metals, to nickel. Of particular importance are the binary
mixtures of Ni-Mo134,139and Ni-Co,134,138aas well as the
ternary systems Ni-Mo-Cd133,140and Ni-Mo-Fe.141Elec-
trochemical studies of such compounds have indicated an
enhancement of catalytic activity through geometric, elec-
tronic, and mixed (so-called synergistic) effects. Nickel has
also been mixed with lanthanum to form bulk intermetallic
La5Ni, which is understood to behave as a high surface-area
nickel catalyst.142In many cases, these materials have been
shown to exhibit activity comparable to the noble metals they
are intended to replace, with low Tafel slopes extending to
Figure 12. The volcano relation (in 30 wt % base solution) based
on potentials required to attain a current density of 2 mA cm-2,as
measured by Miles.127Reproduced with permission from ref 127.
Copyright 1975 Elsevier.
1-R(θ*)R(1 - θ*)1-Re-Ea/kBT
Figure 13. Tafel plots for the following HER catalysts, reproduced
from the associated references: bright Ni, Ni-Mo-Cd;133bright
Ni-Mo;132cNi-Fe, Ni-Co, Ni-Mo;134bright WC;135RuO2-
IrO2;136bright Pt.137“Bright” refers to electrode surfaces that have
been polished such that the electrochemically active area equals
the projected area, whereas electrodes not noted as “bright” have
not been corrected for electrochemically active surface area. Solid
red curves were collected in basic electrolyte, whereas dashed blue
curves were collected in acidic electrolyte. Plots of bright Ni and
Ni-Mo-Cd were reproduced with permission from ref 133.
Copyright 1985 the Royal Society of Chemistry. Bright Ni-Mo,
bright WC, and RuO2-IrO2data was reproduced with permission
from refs 132c, 135, and 136, respectively. Copyright 1998, 1987,
and 1994 Elsevier. Ni-Fe, Ni-Co, and Ni-Mo data was repro-
duced with permission from ref 134. Copyright 1993 Springer
Science + Business Media. Bright Pt data was reproduced with
permission from ref 137. Copyright 1957 American Chemical
Solar Water Splitting Cells Chemical Reviews, 2010, Vol. 110, No. 11 6465
quite high current densities. For example, high surface area
Ni-Mo electrodes have been reported to require less than
100 mV of overpotential to achieve a current density of 1 A
cm-2in alkaline media for thousands of hours.139c
Many mixtures of metals with nonmetal components have
been employed in search of a high-activity, low-cost catalyst.
Trasatti has studied the electrochemistry and HER catalytic
properties of a set of metal oxides, mostly based on RuO2,
which show high catalytic activity.136,143Among the oxides,
the ternary mixture SrxNbO3-δis particularly promising due
to its low-cost components and high apparent stability.144
Sulfides have also been pursued, especially nickel sulfide,
which appears to exhibit higher activity and stability under
alkaline conditions than pure nickel.145Tungsten carbide has
been shown to be an active catalyst for the HER,146likely
due to inhibition by carbon of the formation of a deactivating
surface oxide.135,147Silicotungstates that have been deposited
onto conductive substrates have been reported to give
extremely high exchange current densities,148although
control experiments indicate that this activity is likely due
to platinum impurities that were stripped from the counter
electrode and plated onto the working electrode.149
Platinum and other noble metals have been used exten-
sively as catalysts in proof-of-concept photoelectrochemical
HER systems, allowing attainment of ∼13% solar conversion
efficiency.47In spite of the expansive electrolysis literature,
however, only a few examples of working photoelectro-
chemical HER photocathodes have employed catalysts other
than noble metals. Perhaps most notable among these is the
integrated device developed by Rocheleau et al., who used
a sputtered cobalt-molybdenum alloy cathode in strong
alkali.121Several cases have also emerged in which other
materials perform comparably to or better than platinum in
systems that utilize sacrificial agents, rather than a coupled
anode, for the necessary reductive equivalent. Kakuta et al.,
for example, reported that coprecipitated ZnS/CdS produced
hydrogen with comparable performance to Pt/CdS when
illuminated in the presence of a sacrificial reductant.150More
recently, Zong et. al reported a MoS2cocatalyst that, along
with CdS, produced hydrogen more efficiently than CdS
coated with a platinum cocatalyst, under illumination and
in the presence of a sacrificial reductant.151
7.4. Stability of Catalysts for Hydrogen Evolution
Aside from catalytic activity, stability and long-term
performance have been primary concerns for the develop-
ment of HER electrocatalysts.152In general, three broad
sources contribute to the degradation of electrode perfor-
mance over time. The first is corrosion, which is a concern
for any material, often taking its toll on electrode activity
only over long periods. The second is catalyst poisoning by
solution impurities. Only very small amounts of deleterious
material are required to significantly deactivate an electrode.
Finally, electrode performance can be diminished due to
changes in the electrode composition or morphology. These
effects can occur over long or short time scales, as with
hydrogen absorption in Ni, which begins immediately and
progresses as the electrode continues to be used.153All three
of these degradation mechanisms are intimately dependent
on the exact conditions under which the electrode is
fashioned and operated, including the electrolyte composition
and pH, the cell housing, and the specific operating condi-
tions of temperature, potential, and current density. Though
general steps can be taken to eliminate rapid catalyst
deterioration, procedures to fully minimize these deleterious
effects must be developed individually for each set of
material systems and operating conditions.
7.5. Materials and Mechanismfor the Oxygen
The exchange current densities for water oxidation on the
best known catalysts, such as RuO2, are on the order of Jo
∼1 × 10-5to 1 × 10-6A cm-2.154This vast difference in
activity relative to the 10-3A cm-2exchange current density
of Pt or Pd for H2production in acid is generally attributed
to the more complex intermediate structure and kinetics of
an oxide phase, as opposed to a metal surface, for the
oxidation of water. As a result, electrodes for the oxidation
of water generally must operate at relatively high overpo-
tentials. The activities of such catalysts are thus not well
described by the exchange current densities of the materials,
but are better understood in terms of the corresponding Tafel
slopes and η-J behavior (Figure 14). The most active known
catalysts for water oxidation that are stable enough for
laboratory testing are RuO2 under acidic conditions, and
NiCo2O4or doped lanthanum oxides in alkaline media. Tafel
slopes range from 30 mV/decade to 120 mV/decade, while
40 mV/decade is the most common. In general, the Tafel
slopes of most catalysts drastically increase at higher current
densities, due to several factors, including changes in
mechanism, uncompensated resistance from bubble forma-
tion, and degradation of the catalyst.154Note the high
overpotentials as compared to those needed to drive the
reduction of water to hydrogen. To pass 1 A cm-2with a
highly structured NiCo2O4 catalyst, one would need ap-
proximately 300 mV, a large overpotential compared to ∼100
mV required for structured Ni-Mo alloy to produce H2at
the same current density.
The apparent activities of all water splitting catalysts are
greatly affected by catalyst preparation, as seen by the great
Figure 14. Collected Tafel plots for OER catalysts, adapted from
IrO2;160RuO2 (both “compact” and not);161La0.6Sr0.4CoO3;162
NiOX;163NiLa2O4.164,165Solid lines represent alkaline conditions,
and dashed lines are under acidic conditions. Current densities are
in terms of independently measured electrochemically active surface
area. The dot-dashed line, however, shows NiCo2O4with respect
to projected, rather than active surface area, demonstrating that the
apparent activity of a catalyst with a high roughness factor increases
dramatically if one does not account for the catalytic activity with
respect to the active surface area. Data reproduced with permission
from ref 165 and from S. Trasatti. Copyright 1981 Elsevier. [Plots
of PtO2, MnOx, Co3O4, SrFeO3, IrO2, RuO2 (both forms) and
NiLa2O4were reproduced with permission from refs 155, 156, 157,
159, 160, 161, and 165, respectively. Copyright 1976, 1978 (2),
1979 (2), and 1981 (2) Elsevier. Plots of NiCo2O4, La0.6Sr0.4CoO3,
and NiOxwere reproduced with permission from refs 158, 162,
and 163, respectively. Copyright 1979, 1984, and 1978 the Royal
Society of Chemistry.]
6466 Chemical Reviews, 2010, Vol. 110, No. 11Walter et al.
disparity in activity between two differently prepared RuO2
catalysts (Figure 14). These differences can be attributed to the
method of synthesis and to the conditioning prior to measure-
ment, as well as to the different activities of materials that have
widely varying crystallinity. Possible differences in roughness
factors and porosities also can contribute to discrepancies in
performance of electrodes of the same material. High surface
area, Teflon-bonded NiCo2O4electrodes, for example, exhibit
drastic differences in activity when the projected area is
considered relative to when the electrochemically active surface
area of the electrode is considered.158
A general mechanism for the OER on metal oxides in
acidic or alkaline solutions has been proposed on the basis
of experimentally measured Tafel slopes.166A general
mechanism for the OER on metal oxides in acidic solution
is given by eqs 12a, 12b, 12c, and 12d.
The primary discharge of either water (in acid) or hydroxide
(in base) to oxidize a surface-active site (asterisk) is the first
step, eq 12a corresponding to a Tafel slope of 120 mV/decade.
The result is an unstable intermediate species, designated in
square brackets in eqs 12a and 12b. After further chemical
conversion to a more stable surface species (eq 12b; 60 mV/
Oxygen is finally liberated from the reaction of two highly
oxidized surface sites (eq 12d). This mechanism makes no
distinction between the oxygen intermediates that arise from
the oxide lattice or from water species. Indeed, the oxide lattice
has been shown to participate in the oxidation, with isotopically
labeled RuO2and NiCo2O4surfaces evolving various isotopo-
logues of O2.167
7.6. Theory for the Activity of Oxygen Evolution
The activity of oxidation catalysts can be generally under-
stood by the ability of the surface oxide to transition between
Rasiyah and Tseung proposed that O2evolution follows a metal
oxide redox transition, and hypothesized that catalysts that
undergo such transitions close to the reversible potential for
oxygen evolution should possess the highest activity.168These
two researchers demonstrated a linear correlation between
the minimum potential required for oxygen evolution and
their lower oxide/higher oxide redox potentials. Trasatti
departed from this theory, relating the catalytic activity to
the metal-oxygen bond strength on the surface of the
oxides.169A “volcano” plot, analogous to that observed for
the HER, consequently results from correlating the overpo-
tential at a fixed current density to the enthalpy of a lower-
to-higher oxide transition (Figure 15). Left of the apex of
the volcano plot, the required overpotential increases as the
strength of the oxide-intermediate interaction increases.
Materials to the right of the apex are easily oxidized, and
thus they have a high coverage of the absorbed intermediate,
increasing the overpotential for oxygen evolution. Those
oxides near the apex of the curve (RuO2, IrO2) possess
optimized bond strengths for the catalysis.
Neither Trasatti nor Tseung relate their observations to
any rate-determining step in a mechanism for water oxida-
tion. Tamura and co-workers have correlated experimental
activation energies with those calculated from consideration
of the ligand-field stabilization energy of complex intermedi-
ates formed on the surface during O2 evolution.170More
recently, Nørskov et al. implemented density functional
theory calculations to model the energetics of the OER on
rutile-type oxides.171A rudimentary volcano curve was
derived, relating the oxygen-evolving activity to the binding
energy of various oxygen species on the active oxide surface.
7.7. Catalyst Materials for Oxygen Evolution
A major technological advance in oxygen evolution
electrocatalysis came with H. Beer’s 1965 patent on the
dimensionally stable anode, abbreviated DSA.172These
electrodes are highly active for electrocatalytic oxidation
reactions and also are resistant to chemical and electrochemi-
cal degradation. DSAs generally consist of an active com-
ponent metal oxide, usually RuO2 and/or IrO2, thermally
decomposed on an inert support, such as Ti. To date,
electrodes for water oxidation in acid have not strayed far
from this composition: precious metal oxides (RuO2 and
IrO2) stabilized by inert, inexpensive metal oxides (e.g., TiO2,
SnO2, Ta2O5, ZrO2) for optimal performance and stability.173
However, due to the expense of precious metals, nickel
operated in hot alkaline solution remains the anode of choice
in commercial electrolyzers.
Most interest in conductive metal oxides for oxygen
evolution has focused on four classes of crystal structures:
dioxides, spinels, perovskites, and pyrochlores.174Typical
catalysts of the first class include RuO2and IrO2, whereas
members of the second class include Co3O4and NiCo2O4.
The third class includes doped lanthanum oxides, such as
NiLa2O4 and LaCoO3, and the fourth class consists of
materials such as Pb2Ru(Ir)2O7. For more information, the
reader is encouraged to consult the extensive review
With respect to OER catalysts in photoelectrolysis systems,
it is possible that metal oxide photoanodes can provide
enough overpotential to drive the oxidation of water without
a catalyst, because the oxide valence band edge potentials
are generally very positive of the O2/H2O equilibrium
*OH + H2O f [HO*OH] + H++ e-
[HO*OH] f HO*OH(12b)
HO*OH f O*OH + e-+ H+
2O*OH f 2*OH + O2
Figure 15. OER volcano plot from Trasatti.169Closed circles: in
acid. Open circles: in alkaline media. Data reproduced with
permission from ref 169. Copyright 1980 Elsevier.
Solar Water Splitting Cells Chemical Reviews, 2010, Vol. 110, No. 11 6467
potential. For example, under AM 1 illumination, WO3is
able to oxidize water at >0.6 V negative of the thermody-
namic potential.76However, most metal oxide semiconduc-
tors, including WO3, possess wide band gaps, and if a lower
band gap oxide that absorbed in the visible and had a less
positive valence band potential was used as a photoanode, a
catalyst would probably be useful to provide more favorable
kinetics for water oxidation.
In practice, the use of oxidation catalysts in existing
devices that use light to split water has been largely limited
to Pt, either physically deposited on a photoanode surface
or used as a “dark” counter electrode, so that the OER can
occur separately from the photocathode surface. The choice
of Pt as an OER catalyst is not optimal, as a large applied
overpotential is required to drive the oxidation of water at
∼10 mA cm-2operating current density.176
Although the use of oxidation catalysts has been limited
in devices fabricated to date, Gra ¨tzel and co-workers have
observed improvements in the photocurrent-potential re-
sponse of Fe2O3after it had been treated with Co(NO3)2, a
precursor for Co3O4.91By deposition of approximately one
monolayer of cobalt catalyst on the surface, an 80 mV
increase in Vocwas observed, accompanied by an increase
in maximum observed internal quantum yield from ∼37%
to ∼45%. Pyrolysis of the deposited precursor onto the
surface resulted in a decrease in activity, attributed to the
aggregation of catalyst particles. Zhong and Gamelin ob-
served a similar increase in the rate of photoelectrochemical
O2evolution177upon electrodeposition in pH 7 phosphate
buffer of an amorphous cobalt oxide (CoPi) catalyst.178
Additionally, Choi and co-workers observed increased anodic
photocurrent from illuminated ZnO electrodes upon photo-
deposition of a CoPicatalyst at neutral pH.179More informa-
tion on this cobalt-based water oxidation catalyst is available
in the recent literature.180
7.8. Stability of Catalysts for Oxygen Evolution
Although it is important to consider catalyst performance,
the long-term stability of the conductive metal oxides is also
a relevant parameter. The tendency toward anodic dissolution
of the metal oxides can be predicted based on thermodynam-
ics.181Precious metal dioxides, such as RuO2and IrO2, are
more stable in acid, while anodic dissolution occurs in base.
Hence, spinels and perovskites are recommended for use in
base.174Although RuO2is widely considered the most active
material for water oxidation, this compound is generally
unstable at high overpotentials. RuO2readily dissolves as
ruthenate in basic solutions, or even as volatile RuO4if the
potentials for the oxidation of surface sites to Ru(VI) or
Ru(VIII), respectively, are reached. RuO4has been shown
to form simultaneously with O2evolution on Ru metal at
moderate overpotentials (∼200 mV) in acidic media.182To
minimize anodic corrosion, ruthenium oxide can be stabilized
by the addition of an inert metal oxide such as SnO2or TiO2,
as in a DSA, or by mixing with IrO2.183
The processing temperature and method of preparation of
the material also play major roles in the stability of oxides.
Materials that are prepared at higher temperatures, and that
have high crystallinity, generally show more stability than
amorphous, “hydrous” films that are formed at lower
temperatures.184However, lower overall catalytic activity,
resulting from a lower number of active sites, often ac-
companies an increase in crystallinity. The optimal trade-
off between activity and stability for RuO2thus appears to
occur for catalysts synthesized at intermediate temperatures
of ∼450 °C.185
7.9. Applications of Electrocatalysts to Solar
There is no shortage of HER and OER catalyst options
available at present. However, the particular concerns
involved with the attachment of catalysts directly to semi-
conductor surfaces place significant constraints on which of
the known systems can, or should, be utilized. The following
section outlines some of the concerns that are unique to
Several considerations make the development of catalyst
materials that are useful in systems for splitting water, and
coupling those catalysts to light-absorbers, different from the
use of such catalysts in standard electrolysis. One of the most
important differences is that light-coupled electrolysis re-
quires absorbers that have large areas, to maximize the
capture of the solar flux. If the catalyst is to be deposited
directly on the absorber surface, this drastically reduces the
requirements for current production per unit geometric area.
Consequently, commercial electrolysis systems run at current
densities as high as 1 A cm-2, whereas a light-coupled HER
apparatus only need run at 10-20 mA cm-2, about a
hundred-fold smaller flux.
Concomitant with reduced material performance require-
ments, however, come more stringent requirements on the
cost of catalyst materials, because comparatively more mass
of catalyst will likely be required to cover the relevant area.
Many of the problems with the stability of conventional
metal, metal composite, and metal oxide electrodes can be
attributed to the requirement of running electrolyzers at high
current densities, at elevated temperatures, and in highly
caustic environments. None of these are required of semi-
conductor-coupled systems. Thus, catalysts in these cases
can possibly be expected to be more robust.
For electrolyzers, Tafel slopes are preferred relative to
exchange current densities as metrics for evaluating the
activity of dark electrolysis catalysts, due to the Tafel slope’s
ability to distinguish between electronic and geometric
(surface-area) effects on apparent catalyst performance.
Indeed, this distinction is particularly important for elec-
trolysis systems because, at the overpotentials required to
drive their characteristically high current densities, the
benefits of extremely high surface areas tend to erode due
to the generally high Tafel slopes (∼120 mV/decade) of
geometrically enhanced materials. At the comparatively low
current densities required by light-coupled water splitting,
however, there is less incentive to focus on electronic
performance gains relative to geometric gains. Figure 16
shows that, for current densities of tens of mA cm-2, a
hypothetical water splitting catalyst that has relatively high
surface area (high observed J0), but has a low fundamental
electronic activity (high Tafel slope), outperforms a catalyst
that has a lower surface area but a higher electronic activity.
Another way to describe this behavior is to evaluate the
photocurrent density required to match the incoming solar
flux in a light-coupled cell, based on the geometrical (i.e.,
projected) area of the cell. A highly structured catalyst or
absorber substrate could possess an active surface area that
is, conservatively, ten times greater than its projected area.
This means that the coupled catalysts would only need to
produce a few mA cm-2of electrochemically active catalyst
area. As a result, high surface-area forms of cheap, abundant
6468 Chemical Reviews, 2010, Vol. 110, No. 11Walter et al.
materials, e.g. pure nickel, could be sufficient to provide the
catalysis necessary for these types of systems.
Photoelectrolysis catalysts, however, must not obscure a
significant fraction of light incident on the surface of a device.
Transition metal or conductive metal oxide catalysts will
often absorb or reflect some of the light, decreasing the
resulting efficiency. A thick, continuous catalyst layer with
an extremely high surface area like that found in industrial
electrolyzers, while perhaps sufficiently catalytic, is useless
in practice for a semiconductor-coupled system, because the
metal overlayer will absorb or reflect nearly all of the incident
light. Concerns over both catalytic activity and light absorp-
tion/reflection can perhaps be mitigated together either by
the development of catalysts that are transparent (e.g., a
transparent conducting oxide) or by moving toward systems
in which both the absorber and the catalyst are micro- or
nanostructured, so as to produce a high surface area for both.
Another option is to actually take advantage of the optical
properties of an absorber or a catalyst material so as to
enhance rather than diminish light absorption.186
Another consideration that is unique to light-coupled water
splitting systems is the set of restrictions on the deposition
of catalysts that is imposed by the nature of the absorber
material. In particular, high-temperature and/or metallurgical
preparation methods are not likely to be useful, due to the
likelihood of undesired reactions with the absorber medium
under such conditions, e.g. silicide formation for metallic
catalysts on Si semiconductors at high temperatures. Instead,
vacuum (evaporation, sputtering) and solution-phase (electro-
and electroless) deposition processes are likely necessary to
protect the integrity of the absorber. This constraint renders
many of the most active and robust dark HER and OER
catalysts inaccessible, unless alternative processing methods
can be developed to create materials that have comparable
catalytic activity and stability.
A third key consideration that is unique to light-coupled
water splitting systems is the necessity of intimately contact-
ing the semiconductor with the catalyst material. In the case
of semiconductor/liquid junctions, for example, it will be
important to ensure that highly rectifying or appropriately
“pinched off” contacts are made between the semiconductor
and the metal. Furthermore, interfacial energy states created
at metal-semiconductor contacts could enhance interfacial
recombination losses, giving rise to decreased voltages and
resulting in losses in efficiency.
8. Micro- and Nanostructural Effects on the
Efficiency of Photoelectrodes
Recently much interest has been directed toward the utiliza-
tion of micro- and nanostructured electrodes for solar energy
conversion, in the form of either photovoltaics or direct
associated with a structured electrode, compared to a planar
system, is the decoupling of the directions of light absorption
and charge-carrier collection.38,188The distance that a minority
carrier can diffuse before recombining is termed the diffusion
length (LD) and is defined as eq 13:
where τ is the minority-carrier lifetime and D is the minority
carrier diffusion coefficient. The diffusion coefficient is related
to the minority-carrier mobility, µ (m2V-1s-1), by the Einstein
relation, eq 14:
In a traditional, planar solar cell, the direction of light
absorption is the same as that of charge-carrier collection.
Thus, to build an efficient cell, the absorber must be thick
enough to absorb all the light, but also must be of sufficient
electronic quality (i.e., purity and crystallinity) such that the
excited minority carriers that are photogenerated deep within
the sample are able to diffuse to the surface, where they can
be collected. This constraint requires that LDg 1/R (i.e., the
absorption length), where R is the absorption coefficient of
the semiconductor near the band gap energy. To achieve
sufficient diffusion lengths in a planar geometry, high purity
semiconductors with few defects that act as recombination
sites generally must be used.
The diffusion length requirement can be decoupled from
the absorption length if nonplanar geometries, such as a
semiconductor rod array, are implemented (Figure 17). As
has been shown by device physics modeling, and demon-
strated experimentally in several different configurations,38,188
high surface area semiconductor structures reduce the
distance that minority carriers must travel, and hence enable
near-unity collection efficiencies despite short minority
carrier diffusion lengths.
hypothetical HER catalysts. Dotted blue: J0) 10-6A cm-2, b )
40 mV/decade. Solid black: J0) 10-3A cm-2, b ) 120 mV/decade.
Note the crossover point at ∼0.4 V and ∼30 mA cm-2, where the
high-b catalyst overtakes the high-J0one in HER performance.
Polarization curve and Tafel plot (inset) of two
Figure 17. In a planar device (A), photogenerated carriers must
coefficient), before collection. In a rod-array cell (B), carriers must
only reach the rod surface before recombination. LDis the diffusion
length of the photogenerated minority carrier (open circle).
Solar Water Splitting Cells Chemical Reviews, 2010, Vol. 110, No. 11 6469
Increasing the junction area of a semiconductor photoelec-
trode or of a photovoltaic device via nano- or microstructuring
has also been shown to reduce the Vocand in this respect is
detrimental to device performance.38,188dThis behavior is
expected from a fundamental analysis of the dependence of Voc
on the dark and light currents, as presented in section 3 (eq 6).
The phenomenon of decreased Vocupon increased junction area
the photogenerated charge carriers are diluted over a large
junction area.34This situation holds even in the ideal case
when surface recombination is negligible and recombination
in the bulk (i.e., quasi-neutral) region dominates the system
performance. The photovoltage is predicted to decrease by
g60 mV per order of magnitude increase in junction (i.e.,
solution-semiconductor contact) area.
Practically, this effect means that, to achieve the highest
performance from a rod-array electrode, the junction area
should be enhanced enough to collect all the carriers (i.e.,
radius ) LD) but not more. This analysis also suggests that
highly nanostructured semiconductor electrodes will suffer
from a loss in Vocunless their geometries also significantly
enhance light absorption, thereby offsetting the loss from
Finally, structured semiconductor surfaces should reduce
electrocatalysis losses in the form of overpotentials, because
of the lower current flux per real area of the electrode. This
effect, in principle, might allow for earth-abundant catalysts,
with lower activities, and finely spread over a structured
electrode, to replace highly active precious metal catalysts.
Numerous combinations of semiconductor materials, elec-
trocatalysts, and cell configurations are available for photo-
electrolysis research. As solar fuels research expands,
standardizing both research methodologies and characteriza-
tion techniques becomes paramount for accurate reporting,
and ultimately helps to move the field forward into new areas
of development and discovery.16
In contrast to the use of a single band gap configuration
(S2) to split water, the use of a dual band gap (D4) water
splitting cell configuration, where the electric field is
generated at a semiconductor liquid junction or through a
“buried junction”, appears to be the most efficient and robust
use of complementary light absorbing materials.10,11Multiple
junction configurations require innovative contacts and new
electrocatalysts. Semiconducting materials continue to be
developed, and many of their electrochemical properties
remain to be fully understood and optimized.
Recent efforts toward the development of efficient photo-
electrolysis devices have centered largely on advances in
controlling the size and shape of micro- and nanoscale features
of semiconductors and catalysts. The use of structured photo-
electrodes lowers material purity constraints by controlling the
directionality of charge movement and by controlling the light
absorption pathways in the semiconductor. High aspect ratio
photoelectrode surfaces also might allow for the use of less
expensive catalysts with lowered catalytic activities.
The search for earth-abundant materials that can be used
in solar water splitting cells remains an important goal for
affordable and environmentally benign methods for energy
conversion and storage. Photoelectrode stability continues
to be a major challenge for the development of efficient
photocathodes and photoanodes. Nature uses a continually
renewed dual band gap photosystem to capture light and store
the energy in simple sugar molecules. A similar photoelec-
trosynthetic strategy can be used to decompose water using
two semiconductors and store the energy in the simplest
chemical bond, H2.
concentration of species A
Air Mass standardized solar irradiance
minority carrier diffusion coefficient
PEC classification involving two semiconduc-
tors requiring four photons to produce one
dimensionally stable anode
dye-sensitized solar cell
potential (e.g., volts)
signed electron charge
standard potential for the reduction of A
formal potential for the reduction of A to A-
conduction band edge energy
flat band potential
quasi-Fermi level for electrons
quasi-Fermi level for holes
band gap energy
theoretical maximum potential extractable from
valence band edge energy
hydrogen evolution reaction
current (e.g., amperes)
incident photon to current efficiency
tin doped indium oxide
current density (e.g., A m-2)
exchange current density
bulk recombination current density
depletion region recombination current den-
current density of interfacial charge transfer
absorbed photon flux, see eq 1
current at maximum power output
operating current density
saturation current density
short-circuit current density
surface recombination current density
tunneling current density
electron-transfer rate constant (e.g., cm4s-1)
metal-organic chemical-vapor deposition
number of electrons transferred in an electro-
normal hydrogen electrode
electron concentration at the surface of a
nsat zero applied potential
6470 Chemical Reviews, 2010, Vol. 110, No. 11Walter et al.
oxygen evolution reaction
tandem p/n junction photoelectrochemical cell
incoming irradiance (power flux)
power at maximum power point
unsigned electronic charge
electrochemical (in eV) of the standard reduc-
tion potential of A to A-
reversible hydrogen electrode
PEC classification involving one semiconduc-
tor requiring two photons to produce one
solar-to-hydrogen conversion efficiency
voltage at maximum power output
free energy change
electrochemical potential difference under stan-
solar energy conversion effiency; overpoten-
fractional surface coverage
wavelength corresponding to the band gap
minority carrier mobility
excess chemical potential generated by a PEC
cell, see eq 1
minority carrier lifetime
This work was supported by a National Science Founda-
tion (NSF) Center for Chemical Innovation (CCI) Powering
the Planet, Grants (CHE-0802907, CHE-0947829) and (NSF-
ACCF) support for MGW (CHE-0937048). The authors
would also like to thank Dr. Nick Strandwitz and Dr. Bruce
Brunschwig for help reviewing the manuscript.
(1) Basic Research Needs for Solar Energy Utilization; Lewis, N. S.,
Crabtree, G., Eds.; Office of Science, U. S. Department of Energy:
Washington, DC, 2005.
(2) Lewis, N. S. Science 2007, 315, 798.
(3) Bard, A. J.; Fox, M. A. Acc. Chem. Res. 1995, 28, 141.
(4) Lewis, N. S. Nature 2001, 414, 589.
(5) Turner, J. A. Science 1999, 285, 1493.
(6) Heller, A. Acc. Chem. Res. 1981, 14, 154.
(7) Fujishima, A.; Honda, K. Nature 1972, 238, 37.
(8) (a) Bak, T.; Nowotny, J.; Rekas, M.; Sorrell, C. C. Int. J. Hydrogen
Energy 2002, 27, 991. (b) Currao, A. Chimia 2007, 61, 815. (c)
Heller, A. Science 1984, 223, 1141.
(9) Developments using sacrificial reagents will not be covered in this
review, except where promising results with a semiconductor material
have only be achieved in this mode.
(10) Bolton, J. R.; Strickler, S. J.; Connolly, J. S. Nature 1985, 316, 495.
(11) Weber, M. F.; Dignam, M. J. J. Electrochem. Soc. 1984, 131, 1258.
(12) Werner, J. H.; Kolodinski, S.; Queisser, H. J. Phys. ReV. Lett. 1994,
(13) The Schockley-Quiesser limit for maximum solar conversion
efficiency is 32% for a 1.4 eV band gap material.
(14) (a) Nozik, A. J. Appl. Phys. Lett. 1976, 29, 150. (b) Nozik, A. J.
Appl. Phys. Lett. 1977, 30, 567. (c) Fornarini, L.; Nozik, A. J.;
Parkinson, B. A. J. Phys. Chem. 1984, 88, 3238.
(15) (a) Gerischer, H. J. Electroanal. Chem. 1977, 82, 133. (b) Bard, A. J.;
Wrighton, M. S. J. Electrochem. Soc. 1977, 124, 1706. (c) Khaselev,
O.; Turner, J. A. J. Electrochem. Soc. 1998, 145, 3335.
(16) Chen, Z.; Jaramillo, T. F.; Deutsch, T. G.; Kleiman-Shwarsctein, A.;
Forman, A. J.; Gaillard, N.; Garland, R.; Takanabe, K.; Heske, C.;
Sunkara, M.; McFarland, E. W.; Domen, K.; Miller, E. L.; Turner,
J. A.; Dinh, H. N. J. Mater. Res. 2010, 25, 3.
(17) (a) Varghese, O. K.; Grimes, C. A. Sol. Energy Mater. Sol. Cells
2008, 92, 374. (b) Parkinson, B. Acc. Chem. Res. 1984, 17, 431.
(18) Lewis, N. S. Acc. Chem. Res. 1990, 23, 176.
(19) (a) Miller, E. L.; Marsen, B.; Paluselli, D.; Rocheleau, R. Electrochem.
Solid-State Lett. 2005, 8, A247. (b) Miller, E. L.; Paluselli, D.;
Marsen, B.; Rocheleau, R. E. Sol. Energy Mater. Sol. Cells 2005,
(20) Kainthla, R. C.; Khan, S. U. M.; Bockris, J. O. M. Int. J. Hydrogen
Energy 1987, 12, 381.
(21) Szklarczyk, M.; Bockris, J. O. M. J. Phys. Chem. 1984, 88, 1808.
(22) (a) Jahagirdar, A. H.; Dhere, N. G. Sol. Energy Mater. Sol. Cells
2007, 91, 1488. (b) Shukla, P. K.; Karn, R. K.; Singh, A. K.;
Srivastava, O. N. Int. J. Hydrogen Energy 2001, 27, 135.
(23) Bansal, A.; Tan, M. X.; Tufts, B. J.; Lewis, N. S. J. Phys. Chem.
1993, 97, 7309.
(24) (a) Licht, S.; Wang, B.; Mukerji, S.; Soga, T.; Umeno, M.; Tributsch,
H. J. Phys. Chem. B 2000, 104, 8920. (b) Park, J. H.; Bard, A. J.
Electrochem. Solid-State Lett. 2005, 8, G371. (c) Marsen, B.; Cole,
B.; Miller, E. L. Sol. Energy Mater. Sol. Cells 2008, 92, 1054.
(25) (a) Pierret, R. F. AdVanced Semiconductor Fundamentals; Prentice
Hall: 2003. (b) Sze, S. M.; Ng, K. K. Physics of Semiconductor
DeVices; John Wiley & Sons, Inc.: Hoboken, NJ, 2007. (c) Tan,
M. X.; Laibinis, P. E.; Nguyen, S. T.; Kesselman, J. M.; Stanton,
C. E.; Lewis, N. S. Prog. Inorg. Chem. 1994, 41, 21.
(26) (a) Lewis, N. S. Inorg. Chem. 2005, 44, 6900. (b) Memming, R.
Electrochim. Acta 1980, 25, 77. (c) Nozik, A. J.; Memming, R. J.
Phys. Chem. 1996, 100, 13061. (d) Gerischer, H. J. Electroanal.
Chem. 1983, 150, 553.
(27) (a) Lewis, N. S. J. Electrochem. Soc. 1984, 131, 2496. (b) Fajardo,
A. M.; Lewis, N. S. Science 1996, 274, 969. (c) Fajardo, A. M.;
Lewis, N. S. J. Phys. Chem. B 1997, 101, 11136. (d) Pomykal, K. E.;
Lewis, N. S. J. Phys. Chem. B 1997, 101, 2476.
(28) (a) Hamann, T. W.; Gstrein, F.; Brunschwig, B. S.; Lewis, N. S.
Chem. Phys. 2006, 326, 15. (b) Gerischer, H. Electrochim. Acta 1989,
(29) Hamann, T. W.; Lewis, N. S. J. Phys. Chem. B 2006, 110, 22291.
(30) Hilal, H. S.; Turner, J. A. Electrochim. Acta 2006, 51, 6487.
(31) (a) Kocha, S. S.; Turner, J. A. J. Electrochem. Soc. 1995, 142, 2625.
(b) Turner, J. A.; Parkinson, B. A. J. Electroanal. Chem. 1983, 150,
611. (c) Lyon, L. A.; Hupp, J. T. J. Phys. Chem. 1995, 99, 15718.
(32) (a) Wang, C. M.; Mallouk, T. E. J. Phys. Chem. 1990, 94, 4276. (b)
Wang, C. M.; Mallouk, T. E. J. Phys. Chem. 1990, 94, 423.
(33) Hu, Y. S.; Kleiman-Shwarsctein, A.; Stucky, G. D.; McFarland, E. W.
Chem. Commun. 2009, 2652.
(34) (a) Tan, M. X.; Kenyon, C. N.; Kruger, O.; Lewis, N. S. J. Phys.
Chem. B 1997, 101, 2830. (b) Kruger, O.; Kenyon, C. N.; Tan, M. X.;
Lewis, N. S. J. Phys. Chem. B 1997, 101, 2840. (c) Kenyon, C. N.;
Tan, M. X.; Kruger, O.; Lewis, N. S. J. Phys. Chem. B 1997, 101,
2850. (d) Tan, M. X.; Kenyon, C. N.; Lewis, N. S. J. Phys. Chem.
1994, 98, 4959.
(35) (a) Kumar, A.; Santangelo, P. G.; Lewis, N. S. J. Phys. Chem. 1992,
96, 834. (b) Gregg, B. A.; Nozik, A. J. J. Phys. Chem. 1993, 97,
(36) Shreve, G. A.; Lewis, N. S. J. Electrochem. Soc. 1995, 142, 112.
(37) Rosenbluth, M. L.; Lieber, C. M.; Lewis, N. S. Appl. Phys. Lett.
1984, 45, 423.
(38) Kayes, B. M.; Atwater, H. A.; Lewis, N. S. J. Appl. Phys. 2005, 97,
(39) Fahrenbruch, A. L.; Bube, R. H. Fundamentals of Solar Cells;
Academic Press: New York, 1983.
(40) Memming, R. In Semiconductor Electrochemistry; Wiley VCH Verlag
GmbH: Weinheim, 2001.
(41) Conway, B. E.; Salomon, M. Electrochim. Acta 1964, 9, 1599.
(42) Levy-Clement, C.; Heller, A.; Bonner, W. A.; Parkinson, B. A. J.
Electrochem. Soc. 1982, 129, 1701.
(43) Memming, R.; Schwandt, G. Electrochim. Acta 1968, 13, 1299.
(44) Aspnes, D. E.; Studna, A. A. Phys. ReV. B 1983, 27, 985.
(45) Price, M. J.; Maldonado, S. J. Phys. Chem. C 2009, 113, 11988.
(46) USGS. Mineral Commodity Summaries 2010; U.S. Geological
(47) Aharon-Shalom, E.; Heller, A. J. Electrochem. Soc. 1982, 129, 2865.
Solar Water Splitting Cells Chemical Reviews, 2010, Vol. 110, No. 11 6471
(48) (a) Kocha, S. S.; Turner, J. A.; Nozik, A. J. J. Electroanal. Chem.
1994, 367, 27. (b) Khaselev, O.; Turner, J. A. Science 1998, 280,
(49) (a) Nakato, Y.; Yano, H.; Nishiura, S.; Ueda, T.; Tsubomura, H. J.
Electroanal. Chem. 1987, 228, 97. (b) Dominey, R. N.; Lewis, N. S.;
Bruce, J. A.; Bookbinder, D. C.; Wrighton, M. S. J. Am. Chem. Soc.
1982, 104, 467.
(50) Bookbinder, D. C.; Bruce, J. A.; Dominey, R. N.; Lewis, N. S.;
Wrighton, M. S. Proc. Natl. Acad. Sci. U.S.A. 1980, 77, 6280.
(51) Fan, F.-R. F.; White, H. S.; Wheeler, B. L.; Bard, A. J. J. Am. Chem.
Soc. 1980, 102, 5142.
(52) Bicelli, L. P.; Razzini, G. Surf. Tech. 1983, 20, 393.
(53) Baglio, J. A.; Calabrese, G. S.; Harrison, D. J.; Kamieniecki, E.;
Ricco, A. J.; Wrighton, M. S.; Zoski, G. D. J. Am. Chem. Soc. 1983,
(54) Fernandez, A. M.; Dheree, N.; Turner, J. A.; Martinez, A. M.; Arriaga,
L. G.; Cano, U. Sol. Energy Mater. Sol. Cells 2005, 85, 251.
(55) Valderrama, R. C.; Sebastian, P. J.; Enriquez, J. P.; Gamboa, S. A.
Sol. Energy Mater. Sol. Cells 2005, 88, 145.
(56) Siripala, W.; Ivanovskaya, A.; Jaramillo, T. F.; Baeck, S. H.;
McFarland, E. W. Sol. Energy Mater. Sol. Cells 2003, 77, 229.
(57) de Jongh, P. E.; Vanmaekelbergh, D.; Kelly, J. J. J. Electrochem.
Soc. 2000, 147, 486.
(58) Trasatti, S. AdV. Electrochem. Sci. Eng. 1992, 2, 1.
(59) Bookbinder, D. C.; Lewis, N. S.; Bradley, M. G.; Bocarsly, A. B.;
Wrighton, M. S. J. Am. Chem. Soc. 1979, 101, 7721.
(60) Bocarsly, A. B.; Bookbinder, D. C.; Dominey, R. N.; Lewis, N. S.;
Wrighton, M. S. J. Am. Chem. Soc. 1980, 102, 3683.
(61) Heller, A.; Aspnes, D. E.; Porter, J. D.; Sheng, T. T.; Vadimsky,
R. G. J. Phys. Chem. 1985, 89, 4444.
(62) Heller, A.; Aharonshalom, E.; Bonner, W. A.; Miller, B. J. Am. Chem.
Soc. 1982, 104, 6942.
(63) Rossi, R. C.; Lewis, N. S. J. Phys. Chem. B 2001, 105, 12303.
(64) (a) Scaife, D. E. Sol. energy 1980, 25, 41. (b) Matsumoto, Y. J. Solid
State Chem. 1996, 126, 227.
(65) Gra ¨tzel, M. Nature 2001, 414, 338.
(66) Khan, S. U. M.; Al-Shahry, M.; Ingler, W. B. Science 2002, 297,
(67) Nozik, A. J. Nature 1975, 257, 383.
(68) (a) Kato, H.; Kudo, A. Catal. Today 2003, 78, 561. (b) Kato, H.;
Asakura, K.; Kudo, A. J. Am. Chem. Soc. 2003, 125, 3082.
(69) Kim, J.; Hwang, D. W.; Kim, H. G.; Bae, S. W.; Lee, J. S.; Li, W.;
Oh, S. H. Top. Catal. 2005, 35, 295.
(70) Sakata, Y.; Matsuda, Y.; Yanagida, T.; Hirata, K.; Imamura, H.;
Teramura, K. Catal. Lett. 2008, 125, 22.
(71) (a) Morisaki, H.; Watanabe, T.; Iwase, M.; Yazawa, K. Appl. Phys.
Lett. 1976, 29, 338. (b) Wagner, F. T.; Somorjai, G. A. J. Am. Chem.
Soc. 2002, 102, 5494. (c) Paulauskas, I. E.; Katz, J. E.; Jellison, G. E.;
Lewis, N. S.; Boatner, L. A.; Brown, G. M. J. Electrochem. Soc.
2009, 156, B580. (d) Mavroides, J. G.; Tchernev, D. I.; Kafalas, J. A.;
Kolesar, D. F. Mater. Res. Bull. 1975, 10, 1023.
(72) Waki, I.; Cohen, D.; Lal, R.; Mishra, U.; DenBaars, S. P.; Nakamura,
S. Appl. Phys. Lett. 2007, 91, 093519/1.
(73) Sato, J.; Saito, N.; Yamada, Y.; Maeda, K.; Takata, T.; Kondo, J. N.;
Hara, M.; Kobayashi, H.; Domen, K.; Inoue, Y. J. Am. Chem. Soc.
2005, 127, 4150.
(74) Woodward, P. M.; Mizoguch, H.; Kim, Y.-I.; Stoltzfus, M. W. In
Metal Oxides: Chemistry and Applications; Fierro, J. L. G., Ed.;
Taylor & Francis: Boca Raton, 2006.
(75) Kudo, A.; Miseki, Y. Chem. Soc. ReV. 2009, 38, 253.
(76) Hodes, G.; Cahen, D.; Manassen, J. Nature 1976, 260, 312.
(77) Valdes, A.; Kroes, G. J. J. Chem. Phys. 2009, 130, 114701.
(78) (a) Liou, F. T.; Yang, C. Y.; Hakim, K.; Levine, S. N. J. Appl.
Electrochem. 1983, 13, 377. (b) Schumacher, L. C.; Mamicheafara,
S.; Leibovitch, M.; Dignam, M. J. J. Electrochem. Soc. 1988, 135,
(79) Veluchamy, P.; Minoura, H. Appl. Phys. Lett. 1994, 65, 2431.
(80) Hardee, K. L.; Bard, A. J. J. Electrochem. Soc. 1977, 124, 215.
(81) Kudo, A.; Omori, K.; Kato, H. J. Am. Chem. Soc. 1999, 121, 11459.
(82) Shimodaira, Y.; Kato, H.; Kobayashi, H.; Kudo, A. Bull. Chem. Soc.
Jpn. 2007, 80, 885.
(83) Ye, J.; Zou, Z.; Arakawa, H.; Oshikiri, M.; Shimoda, M.; Matsushita,
A.; Shishido, T. J. Photochem. Photobiol., A 2002, 79.
(84) Zou, Z. G.; Ye, J. H.; Sayama, K.; Arakawa, H. Nature 2001, 414,
(85) Liu, H.; Yuan, J.; Shangguan, W. F.; Teraoka, Y. J. Phys. Chem. C
2008, 112, 8521.
(86) Kim, H. G.; Hwang, D. W.; Lee, J. S. J. Am. Chem. Soc. 2004, 126,
(87) (a) Bailey, P. C. J. Appl. Phys. 1960, 31, S39. (b) Marusak, L. A.;
Messier, R.; White, W. B. J. Phys. Chem. Solids 1980, 41, 981.
(88) Kennedy, J. H.; Frese, K. W. J. Electrochem. Soc. 1978, 125, 709.
(89) Sartoretti, C. J.; Ulmann, M.; Alexander, B. D.; Augustynski, J.;
Weidenkaff, A. Chem. Phys. Lett. 2003, 376, 194.
(90) Sivula, K.; Formal, F. L.; Gra ¨tzel, M. Chem. Mater. 2009, 21, 2862.
(91) Kay, A.; Cesar, I.; Gra ¨tzel, M. J. Am. Chem. Soc. 2006, 128, 15714.
(92) Tilley, S. D.; Cornuz, M.; Sivula, K.; Gra ¨tzel, M. Angew. Chem.,
Int. Ed. 2010, 49, 6405.
(93) Sivula, K.; Zboril, R.; Le Formal, F.; Robert, R.; Weidenkaff, A.;
Tucek, J.; Frydrych, J.; Gra ¨tzel, M. J. Am. Chem. Soc. 2010, 132,
(94) (a) Mor, G. K.; Prakasam, H. E.; Varghese, O. K.; Shankar, K.;
Grimes, C. A. Nano Lett. 2007, 7, 2356. (b) Thimsen, E.; Biswas,
S.; Lo, C. S.; Biswas, P. J. Phys. Chem. C 2009, 113, 2014.
(95) Kato, H.; Kudo, A. J. Phys. Chem. B 2002, 106, 5029.
(96) Maeda, K.; Takata, T.; Hara, M.; Saito, N.; Inoue, Y.; Kobayashi,
H.; Domen, K. J. Am. Chem. Soc. 2005, 127, 8286.
(97) Maeda, K.; Teramura, K.; Lu, D. L.; Takata, T.; Saito, N.; Inoue,
Y.; Domen, K. Nature 2006, 440, 295.
(98) Hashiguchi, H.; Maeda, K.; Abe, R.; Ishikawa, A.; Kubota, J.; Domen,
K. Bull. Chem. Soc. Jpn. 2009, 82, 401.
(99) (a) Hitoki, G.; Takata, T.; Kondo, J. N.; Hara, M.; Kobayashi, H.;
Domen, K. Chem. Commun. 2002, 1698. (b) Liu, M. Y.; You, W. S.;
Lei, Z. B.; Zhou, G. H.; Yang, J. J.; Wu, G. P.; Ma, G. J.; Luan,
G. Y.; Takata, T.; Hara, M.; Domen, K.; Can, L. Chem. Commun.
(100) Nakamura, R.; Tanaka, T.; Nakato, Y. J. Phys. Chem. B 2005, 109,
(101) Ellis, A. B.; Kaiser, S. W.; Wrighton, M. S. J. Am. Chem. Soc. 1976,
(102) Kaneko, M.; Yao, G. J.; Kira, A. J. Chem. Soc., Chem. Commun.
(103) Yamane, S.; Kato, N.; Kojima, S.; Imanishi, A.; Ogawa, S.; Yoshida,
N.; Nonomura, S.; Nakato, Y. J. Phys. Chem. C 2009, 113, 14575.
(104) Buxbaum, G. Industrial inorganic pigments, 2nd ed.; Wiley-VCH:
Weinheim, New York, 1998.
(105) Kato, H.; Kudo, A. J. Phys. Chem. B 2001, 105, 4285.
(106) Shan, W.; Walukiewicz, W.; Yu, K. M.; Wu, J.; Ager, J. W., III;
Haller, E. E.; Xin, H. P.; Tu, C. W. Appl. Phys. Lett. 2000, 76, 3251.
(107) Osterloh, F. E. Chem. Mater. 2008, 20, 35.
(108) (a) Woodhouse, M.; Parkinson, B. A. Chem. Mater. 2008, 20, 2495.
(b) Katz, J. E.; Gingrich, T. R.; Santori, E. A.; Lewis, N. S. Energy
EnViron. Sci. 2009, 2, 103. (c) Lee, J.; Ye, H.; Pan, S.; Bard, A. J.
Anal. Chem. 2008, 80, 7445. (d) Arai, T.; Konishi, Y.; Iwasaki, Y.;
Sugihara, H.; Sayama, K. J. Comb. Chem. 2007, 9, 574. (e) Jaramillo,
T. F.; Baeck, S.-H.; Kleiman-Shwarsctein, A.; Choi, K.-S.; Stucky,
G. D.; McFarland, E. W. J. Comb. Chem. 2005, 7, 264.
(109) Woodhouse, M.; Parkinson, B. A. Chem. Soc. ReV. 2009, 38, 197.
(110) Yoneyama, H.; Sakamoto, H.; Tamura, H. Electrochim. Acta 1975,
(111) (a) Ohashi, K.; McCann, J.; Bockris, J. O. M. Nature 1977, 266,
610. (b) Ohashi, K.; McCann, J.; Bockris, J. O. M. Int. J. Hydrogen
Energy 1977, 1, 259.
(112) Wrighton, M. S.; Ellis, A. B.; Wolczanski, P. T.; Morse, D. L.;
Abrahamson, H. B.; Ginley, D. S. J. Am. Chem. Soc. 1976, 98, 2774.
(113) Mettee, H.; Otvos, J. W.; Calvin, M. Sol. Energy Mater. 1981, 4,
(114) Kainthla, R. C.; Zelenay, B.; Bockris, J. O. J. Electrochem. Soc. 1987,
(115) Akikusa, J.; Khan, S. U. M. Int. J. Hydrogen Energy 2002, 27, 863.
(116) Inoue, T.; Yamase, T. Chem. Lett. 1985, 869.
(117) Ingler, W. B.; Khan, S. U. M. Electrochem. Solid-State Lett. 2006,
(118) Wang, H. L.; Deutsch, T.; Turner, J. A. J. Electrochem. Soc. 2008,
(119) Brillet, J.; Cornuz, M.; Le Formal, F.; Yum, J. H.; Gratzel, M.; Sivula,
K. J. Mater. Res. 2010, 25, 17.
(120) Miller, E. L.; Marsen, B.; Cole, B.; Lum, M. Electrochem. Solid-
State Lett. 2006, 9, G248.
(121) Rocheleau, R. E.; Miller, E. L.; Misra, A. Energy Fuels 1998, 12, 3.
(122) Khaselev, O.; Bansal, A.; Turner, J. A. Int. J. Hydrogen Energy 2001,
(123) Yamane, S.; Kato, N.; Kojima, S.; Imanishi, A.; Ogawa, S.; Yoshida,
N.; Nonomura, S.; Nakato, Y. J. Phys. Chem. C 2009, 113, 14575.
(124) Yamada, Y.; Matsuki, N.; Ohmori, T.; Mametsuka, H.; Kondo, M.;
Matsuda, A.; Suzuki, E. Int. J. Hydrogen Energy 2003, 28, 1167.
(125) Johnson, E. L. Tech. Dig.sInt. Electron DeVices Meet. 1981, 1.
(126) Trasatti, S. J. Electroanal. Chem. 1972, 39, 163.
(127) Miles, M. H. J. Electroanal. Chem. 1975, 60, 89.
(128) (a) Gerischer, H. Bull. Soc. Chim. Belg. 1958, 67, 506. (b) Parsons,
R. Trans. Faraday Soc. 1958, 54, 1053.
(129) Conway, B. E.; Tilak, B. V. Electrochim. Acta 2002, 47, 3571.
(130) (a) Hinnemann, B.; Moses, P. G.; Bonde, J.; Jorgensen, K. P.; Nielsen,
J. H.; Horch, S.; Chorkendorff, I.; Norskov, J. K. J. Am. Chem. Soc.
6472 Chemical Reviews, 2010, Vol. 110, No. 11Walter et al.
2005, 127, 5308. (b) Greeley, J.; Norskov, J. K.; Kibler, L. A.; El- Download full-text
Aziz, A. M.; Kolb, D. M. ChemPhysChem 2006, 7, 1032.
(131) (a) Greeley, J.; Jaramillo, T. F.; Bonde, J.; Chorkendorff, I.; Norskov,
J. K. Nat. Mater. 2006, 5, 909. (b) Jaramillo, T. F.; Jorgensen, K. P.;
Bonde, J.; Nielsen, J. H.; Horch, S.; Chorkendorff, I. Science 2007,
(132) (a) Hu, W. K.; Cao, X. J.; Wang, F. P.; Zhang, Y. S. Int. J. Hydrogen
Energy 1997, 22, 621. (b) Kawashima, A.; Akiyama, E.; Habazaki,
H.; Hashimoto, K. Mater. Sci. Eng., A 1997, 226, 905. (c) de Chialvo,
M. R. G.; Chialvo, A. C. J. Electroanal. Chem. 1998, 448, 87. (d)
Hu, C. C.; Weng, C. Y. J. Appl. Electrochem. 2000, 30, 499. (e) Hu,
W. K. Int. J. Hydrogen Energy 2000, 25, 111. (f) Kaninski, M. P. M.;
Stojic, D. L.; Saponjic, D. P.; Potkonjak, N. I.; Miljanic, S. S. J.
Power Sources 2006, 157, 758. (g) Krstajic, N. V.; Jovic, V. D.;
Gajic-Krstajic, L.; Jovic, B. M.; Antozzi, A. L.; Martelli, G. N. Int.
J. Hydrogen Energy 2008, 33, 3676.
(133) Conway, B. E.; Bai, L. J. Chem. Soc., Faraday Trans. 1985, 81,
(134) Raj, I. A. J. Mater. Sci. 1993, 28, 4375.
(135) Tsirlina, G. A.; Petrii, O. A. Electrochim. Acta 1987, 32, 649.
(136) Kodintsev, I. M.; Trasatti, S. Electrochim. Acta 1994, 39, 1803.
(137) Bockris, J. O. M.; Ammar, I. A.; Huq, A. K. M. S. J. Phys. Chem.
1957, 61, 879.
(138) (a) Wendt, H.; Plzak, V. Electrochim. Acta 1983, 28, 27. (b) Endoh,
E.; Otouma, H.; Morimoto, T.; Oda, Y. Int. J. Hydrogen Energy 1987,
(139) (a) Raj, I. A.; Vasu, K. I. J. Appl. Electrochem. 1990, 20, 32. (b)
Brown, D. E.; Mahmood, M. N.; Man, M. C. M.; Turner, A. K.
Electrochim. Acta 1984, 29, 1551. (c) Brown, D. E.; Mahmood,
M. N.; Turner, A. K.; Hall, S. M.; Fogarty, P. O. Int. J. Hydrogen
Energy 1982, 7, 405.
(140) (a) Conway, B. E.; Angersteinkozlowska, H.; Sattar, M. A.; Tilak,
B. V. J. Electrochem. Soc. 1983, 130, 1825. (b) Stachurski, D.; Pouli,
G.; Pokrzyk, G.; Ripas, J. Low oVerVoltage hydrogen cathodes. U.S.
Patent 4,354,915, 1981.
(141) (a) Hu, W. K.; Zhang, Y. S.; Song, D. Y.; Zhou, Z. X.; Wang, Y.
Mater. Chem. Phys. 1995, 41, 141. (b) Jayalakshmi, M.; Kim, W. Y.;
Jung, K. D.; Joo, O. S. Int. J. Electrochem. Sci. 2008, 3, 908. (c)
Raj, I. A. Bull. Electrochem. Soc. 1999, 15, 519.
(142) (a) Hall, D. E.; Sarver, J. M.; Gothard, D. O. Int. J. Hydrogen Energy
1988, 13, 547. (b) Boonstra, A. H.; Bernards, T. M. N. J. Less
Common Met. 1990, 161, 355. (c) Wu, S. H.; Huang, W. P.; Yang,
X. L.; Xie, Q. X.; Wei, W.; Deng, Z. P. J. Rare Earths 1999, 17, 66.
(143) (a) Krstajic, N.; Trasatti, S. J. Appl. Electrochem. 1998, 28, 1291.
(b) Tavares, A. C.; Trasatti, S. Electrochim. Acta 2000, 45, 4195.
(c) Kircheva, N.; Guerrini, E.; Trasatti, S. Russ. J. Electrochem. 2004,
40, 1156. (d) Bianchi, I.; Guerrini, E.; Trasatti, S. Chem. Phys. 2005,
319, 192. (e) Popa, L.; Guerrini, E.; Trasatti, S. J. Appl. Electrochem.
2005, 35, 1213. (f) Fachinotti, E.; Guerrini, E.; Tavares, A. C.;
Trasatti, S. J. Electroanal. Chem. 2007, 600, 103.
(144) Manoharan, R.; Goodenough, J. B. J. Electrochem. Soc. 1990, 137,
(145) Nidola, A.; Schira, R. Int. J. Hydrogen Energy 1986, 11, 449.
(146) (a) Nikolov, I.; Petrov, K.; Vitanov, T.; Guschev, A. Int. J. Hydrogen
Energy 1983, 8, 437. (b) Struck, B. D.; Neumeister, H.; Naoumidis,
A. Int. J. Hydrogen Energy 1986, 11, 541. (c) Wiecek, B. Pol.
J. Chem. 1996, 70, 343. (d) Ma, C. A.; Gan, Y. P.; Chu, Y. Q.;
Huang, H.; Chen, D. H.; Zhou, B. X. Trans. Nonferrous Met. Soc.
China 2004, 14, 11.
(147) Tsirlina, G. A.; Petrii, O. A. Electrochim. Acta 1987, 32, 637.
(148) (a) Keita, B.; Nadjo, L. J. Electroanal. Chem. 1987, 227, 265. (b)
Keita, B.; Nadjo, L.; Haeussler, J. P. J. Electroanal. Chem. 1987,
230, 85. (c) Keita, B.; Nadjo, L.; Krier, G.; Muller, J. F. J.
Electroanal. Chem. 1987, 223, 287. (d) Keita, B.; Nadjo, L. J.
Electroanal. Chem. 1990, 287, 149.
(149) McKone, J.; Lewis, N. S.; California Institute of Technology:
Pasadena, CA. Unpublished work, 2009.
(150) Kakuta, N.; Park, K. H.; Finlayson, M. F.; Ueno, A.; Bard, A. J.;
Campion, A.; Fox, M. A.; Webber, S. E.; White, J. M. J. Phys. Chem.
1985, 89, 732.
(151) Zong, X.; Yan, H.; Wu, G.; Ma, G.; Wen, F.; Wang, L.; Li, C. J. Am.
Chem. Soc. 2008, 130, 7176.
(152) (a) Leroy, R. L.; Janjua, M. B. I.; Renaud, R.; Leuenberger, U. J.
Electrochem. Soc. 1979, 126, 1674. (b) Divisek, J.; Mergel, J.;
Schmitz, H. Int. J. Hydrogen Energy 1982, 7, 695. (c) Nidola, A.;
Schira, R. J. Electrochem. Soc. 1986, 133, 1653. (d) Nidola, A.;
Schira, R. J. Electrochem. Soc. 1986, 133, C123.
(153) (a) Huot, J. Y. J. Electrochem. Soc. 1989, 136, 1933. (b) Rommal,
H. E. G.; Moran, P. J. J. Electrochem. Soc. 1985, 132, 325.
(154) Trasatti, S.; Lodi, G. In Electrodes of ConductiVe Metallic Oxides;
Trasatti, S., Ed.; Elsevier: Amsterdam, 1981; Vol. B.
(155) Iwakura, C.; Fukuda, K.; Tamura, H. Electrochim. Acta 1976, 21,
(156) Morita, M.; Iwakura, C.; Tamura, H. Electrochim. Acta 1979, 24,
(157) Iwakura, C.; Honji, A.; Tamura, H. Electrochim. Acta 1981, 26, 1319.
(158) Jasem, S. M.; Tseung, A. C. C. J. Electrochem. Soc. 1979, 126, 1353.
(159) Matsumoto, Y.; Kurimoto, J.; Sato, E. J. Electroanal. Chem. 1979,
(160) Gottesfeld, S.; Srinivasan, S. J. Electroanal. Chem. 1978, 86, 89.
(161) Lodi, G.; Sivieri, E.; Debattisti, A.; Trasatti, S. J. Appl. Electrochem.
1978, 8, 135.
(162) Bockris, J. O.; Otagawa, T. J. Electrochem. Soc. 1984, 131, 290.
(163) Miles, M. H.; Huang, Y. H.; Srinivasan, S. J. Electrochem. Soc. 1978,
(164) Fiori, G.; Mari, C. M. Int. J. Hydrogen Energy 1982, 7, 489.
(165) Trasatti, S. Electrodes of ConductiVe Metallic Oxides; Elsevier:
Amsterdam, 1980 and1981.
(166) Trasatti, S. In Electrochemistry of NoVel Materials; Lipkowski, J.,
Ross, P. N., Eds.; VCH Publishers: New York, 1994.
(167) (a) Hibbert, D. B. J. Chem. Soc., Chem. Commun. 1980, 202. (b)
Wohlfahrt-Mehrens, M.; Heitbaum, J. J. Electroanal. Chem. 1987,
237, 251. (c) Willsau, J.; Wolter, O.; Heitbaum, J. J. Electroanal.
Chem. 1985, 195, 299. (d) Hibbert, D. B.; Churchill, C. R. J. Chem.
Soc., Faraday Trans. 1984, 80, 1965.
(168) Rasiyah, P.; Tseung, A. C. C. J. Electrochem. Soc. 1984, 131, 803.
(169) Trasatti, S. J. Electroanal. Chem. 1980, 111, 125.
(170) (a) Arikado, T.; Iwakura, C.; Tamura, H. Electrochim. Acta 1978,
23, 9. (b) Iwakura, C.; Tamura, H.; Inai, M. J. Electrochem. Soc.
Jap. 1980, 48, 229. (c) Iwakura, C.; Tamura, H.; Inai, M. J.
Electrochem. Soc. Jap. 1980, 48, 173.
(171) Rossmeisl, J.; Qu, Z. W.; Zhu, H.; Kroes, G. J.; Norskov, J. K. J.
Electroanal. Chem. 2007, 607, 83.
(172) (a) Beer, H. B. (Patent) England, 1965; Vol. 1 147-442. (b) Beer,
H. B. J. Electrochem. Soc. 1980, 127, 303C.
(173) Trasatti, S. Electrochim. Acta 2000, 45, 2377.
(174) Trasatti, S. In Interfacial Electrochemistry: Theory, Experiment, and
Applications; Wieckowski, A., Ed.; Marcel Dekker: New York, 1999.
(175) (a) Trasatti, S. In Electrochemical Hydrogen Technologies; Wendt,
H., Ed.; Elsevier: Amsterdam, 1990. (b) Matsumoto, Y.; Sato, E.
Mater. Chem. Phys. 1986, 14, 397.
(176) Bockris, J. O.; Huq, A. Proc. R. Soc. London A 1956, 237, 277.
(177) Zhong, D. K.; Gamelin, D. R. J. Am. Chem. Soc. 2010, 132, 4202.
(178) (a) Lutterman, D. A.; Surendranath, Y.; Nocera, D. G. J. Am. Chem.
Soc. 2009, 131, 3838. (b) Kanan, M. W.; Nocera, D. G. Science 2008,
(179) Steinmiller, E. M. P.; Choi, K. S. Proc. Natl. Acad. Sci. U.S.A. 2009,
(180) (a) McAlpin, J. G.; Surendranath, Y.; Dinca ˇ, M.; Stich, T. A.; Stoian,
S. A.; Casey, W. H.; Nocera, D. G.; Britt, R. D. J. Am. Chem. Soc.
2010, 132, 6882. (b) Surendranath, Y.; Dinca ˇ, M.; Nocera, D. G.
J. Am. Chem. Soc. 2009, 131, 2615. (c) Kanan, M. W.; Surendranath,
Y.; Nocera, D. G. Chem. Soc. ReV. 2009, 38, 109. (d) Cook, T. R.;
Dogutan, D. K.; Reece, S. Y.; Surendranath, Y.; Teets, T. S.; Nocera,
D. G. Chem. ReV. 2010, 110, doi: 10.1021/cr100246c.
(181) (a) Pourbaix, M. Atlas of Electrochemical Equilibria in Aqueous
Solutions; Pergamon Press: New York, 1966. (b) Tseung, A. C. C.;
Jasem, S. Electrochim. Acta 1977, 22, 31. (c) Barral, G.; Diard, J. P.;
Montella, C. Electrochim. Acta 1986, 31, 277.
(182) Kotz, R.; Stucki, S.; Scherson, D.; Kolb, D. M. J. Electroanal. Chem.
1984, 172, 211.
(183) Kotz, R.; Stucki, S. Electrochim. Acta 1986, 31, 1311.
(184) Trasatti, S. Electrochim. Acta 1984, 29, 1503.
(185) Iwakura, C.; Tamura, H. Int. J. Hydrogen Energy 1982, 7, 857.
(186) (a) Kelzenberg, M. D.; Boettcher, S. W.; Petykiewics, J. A.; Turner-
Evans, D. B.; Putnam, M. C.; Warren, E. L.; Spurgeon, J. M.; Briggs,
R. M.; Lewis, N. S.; Atwater, H. A. Nat. Mater. 2010, 9, 239. (b)
Heller, A. Pure Appl. Chem. 1986, 58, 1189.
(187) (a) Shankar, K.; Basham, J. I.; Allam, N. K.; Varghese, O. K.; Mor,
G. K.; Feng, X. J.; Paulose, M.; Seabold, J. A.; Choi, K. S.; Grimes,
C. A. J. Phys. Chem. C 2009, 113, 6327. (b) Goodey, A. P.; Eichfeld,
S. M.; Lew, K. K.; Redwing, J. M.; Mallouk, T. E. J. Am. Chem.
Soc. 2007, 129, 12344. (c) Park, J. H.; Kim, S.; Bard, A. J. Nano
Lett. 2006, 6, 24. (d) Beermann, N.; Vayssieres, L.; Lindquist, S. E.;
Hagfeldt, A. J. Electrochem. Soc. 2000, 147, 2456. (e) Yang, X.;
Wolcottt, A.; Wang, G.; Sobo, A.; Fitzmorris, R. C.; Qian, F.; Zhang,
J. Z.; Li, Y. Nano Lett. 2009, 9, 2331.
(188) (a) Spurgeon, J. M.; Atwater, H. A.; Lewis, N. S. J. Phys. Chem. C
2008, 112, 6186. (b) Erne, B. H.; Vanmaekelbergh, D.; Kelly, J. J.
AdV. Mater. 1995, 7, 739. (c) Erne, B. H.; Vanmaekelbergh, D.; Kelly,
J. J. J. Electrochem. Soc. 1996, 143, 305. (d) Maiolo, J. R.; Atwater,
H. A.; Lewis, N. S. J. Phys. Chem. C 2008, 112, 6194. (e) Maiolo,
J. R.; Kayes, B. M.; Filler, M. A.; Putnam, M. C.; Kelzenberg, M. D.;
Atwater, H. A.; Lewis, N. S. J. Am. Chem. Soc. 2007, 129, 12346.
Solar Water Splitting CellsChemical Reviews, 2010, Vol. 110, No. 11 6473