Engineering artificial photosynthetic systems to afford solar fuel production can also be accomplished with a
the efficient and economical conversion of abundant photoelectrochemical (PEC) device, which contains
solar energy into chemical fuel on a scale in accordance a direct-semiconductor/liquid interface5. However, to
with global energy demand (tens of terawatts) is an enable practical PEC solar fuel production, low-cost,
important step towards the realization of a sustainable robust and high-performance semiconductor materials
carbon-neutral society 1. At the heart of any system for are needed. The past 5years have seen a surge in the
harvesting solar energy is the semiconductor material, development of promising semiconductor materials,
which transforms absorbed solar photons into excited owing, in part, to the rational engineering of classic
electronic states (both electrons and holes). Oppositely semiconducting materials.
charged excited states can be separated to generate an In this Review, we aim to give readers a brief overview
electronic potential (that is, a photopotential) suita- of the basics of PEC technology while focusing on the
ble to drive the electrochemical production of simple light-harvesting materials. We present a critical analy-
fuels, such as molecular hydrogen from the reduction of sis of both the advantageous and detrimental aspects of
1
Institut des Sciences et water or methanol from the reduction of carbon dioxide. classic and recently developed semiconductors for direct
Ingnierie Chimiques, cole Hydrogen production can be accomplished in a straight- solar fuel production to give insight into the potential for
Polytechnique Fdrale de
forward manner using traditional photovoltaic devices viable performance in PEC cells. Advances with ternary
Lausanne (EPFL), Station 6,
1015 Lausanne, Switzerland. together with standard electrolysers (splitting water into oxide semiconductors beyond the well-known binary
2
Helmholtz-Zentrum Berlin fr H2 and O2). However, this route is not economically transition metal oxides are examined in the context of
Materialien und Energie competitive with traditional methods using fossil fuels their optoelectronic properties and intrinsic stability.
GmbH, Institute for Solar (for example, the steam reforming of natural gas or coal Tuning the optoelectronics by adding nitrogen yields
Fuels, Hahn-Meitner-Platz 1,
14109 Berlin, Germany.
gasification). Although cost estimates vary, hydrogen oxynitrides, a new and emerging class of PEC materi-
produced from fossil fuels is less expensive (by a fac- als. Replacing the oxygen atom with a chalcogen (S or
Correspondence to K.S.
kevin.sivula@epfl.ch tor of two to five) than using photovoltaics combined Se) can offer improved light absorption, and suitable
with electrolysis24. Thus, the development of alterna- charge transport can be achieved using transition-metal
Article number: 15010
doi:10.1038/natrevmats.2015.10 tive, economically competitive strategies to produce dichalcogenides and multinary compounds, such as
Published online 20 Jan 2016 solar fuels is of strong interest. Semiconductor-based CZTS (copper zinc tin sulfide, Cu2ZnSnS4) and CIGS
C
EQF R
0 EH+/H2
e H+
Eg,2
e 1.23 V e
EF
+1
EH2O/O2
Eg,1 H2O h+
O
A
+2 EQF
H+
O2
h+
E(VRHE)
VB WOC VB
Photocathode
Membrane
Photoanode
Nature Reviews | Materials
Figure 1 | Photoelectrochemical cell basics. a | Schematic of a wired-type tandem cell for water splitting with incident
solar illumination striking the photoanode and transmitting to the photocathode. b|Working principle of the tandem
cell for water splitting using a photoanode with bandgap energy Eg,1, and a photocathode with Eg,2 (where Eg,1 > Eg,2).
Briefly, on absorption of a solar photon, an electron (e) from the valence band (VB) is promoted to the conduction band
(CB) leaving the corresponding electron hole (h+). The electric field in the depletion layer physically separates these
charges and, in the photocathode, the electrons in the CB drift to the semiconductorliquid junction, increasing the
quasi-Fermi energy of the cathode, ECQF, to drive the reduction of H+ to H2 at a water reduction catalysis (WRC) site.
Analogously, in the photoanode, electron holes in the VB drift to the semiconductorliquid junction, increasing the
photoanodes quasi-Fermi energy, EAQF, sufficiently to surmount the overpotential for oxidation (O) and oxidize water to
O2 at a water oxidation catalysis (WOC) site. Photogenerated electrons in the CB of the photoanode travel through the
external circuit to recombine with the holes in the VB of the photocathode. EF, Fermi energy; E(VRHE), electronic potential
with respect to the reversible hydrogen electrode; R, overpotential for reduction.
(copper indium gallium selenide, CuIn xGa(1x)Se2). electrolysis, in a PEC cell, at least one of the electrodes
These materials, as well as the traditional high-perfor- is a semiconductor, which absorbs solar photons with
mance semiconductors used in photovoltaic devices, are energy greater than the semiconductor bandgap energy,
typically not stable in a direct-semiconductor/electrolyte Eg. The photopotential generated by the subsequent sep-
junction configuration but can be encapsulated and opti- aration of the photogenerated charges can directly drive
mized for high-performance PEC application. The recent an electrochemical solar-fuel-production reaction.
emergence of carbon- and carbon-nitride-based semi- Although it is possible to envisage a PEC cell with
conductors for the direct PEC splitting of water is also one semiconductor photoelectrode and one standard
discussed. In the final section, based on the development metallic (dark) electrode, identifying a single semicon-
of the above-mentioned materials, we present a realistic ductor material that can both develop sufficient pho-
outlook on device performance in terms of the goal of topotential for solar fuel production (approximately
cost-effective PEC solar-to-chemical energy conversion. 1.51.8 V for water splitting) and also harvest a signif-
icant portion of the solar spectrum is an ongoing and
Solar-to-chemical using semiconductors difficult challenge. However, because semiconductors
PEC solar fuel production conceivably offers a more can also transmit photons with energy less than their
straightforward route towards renewable solar-to- Eg, a PEC cell can be constructed with semiconductors
chemical energy conversion compared with conven- used for both electrodes (that is, an n-type photoanode
tional photovoltaic technology coupled with standard and a p-type photocathode) in a tandem configuration
devices that produce electrochemical fuel (for example, to increase both the photopotential and the amount of
electrolysers)6. This is partly because of the simplicity of light harvested by the device7.
the PEC device, which requires only a few elements to Many distinct tandem cell configurations are possi-
integrate the light-absorbing and electrocatalytic func- ble and have been discussed recently8. A schematic is
tions. Similar to a traditional electrochemical device, presented in FIG.1a of a common and representative
a PEC cell requires two electrodes (an anode for the example, in which a large-bandgap photoanode absorbs
oxidation reaction and a cathode for the reduction the high-energy portion of the incoming solar irradia-
reaction) and a membrane to separate the products tion and transmits the remainder to the photocathode.
of the two half-cell reactions. However, in contrast to The two photoelectrodes are electrically connected by
Reference
electrode
Counter
electrode
Photoanode Photocathode
100
Illuminated E3app Light on
Dark E2app
JOP
E1app
Current density ( J)
J(Eapp)
IPCE or APCE (%)
0
0
Potential, E Photon wavelength, Time
For PEC devices, different performance metrics are used to assess which photon energies contribute to the solar photocurrent. These
solar-to-fuel energy conversion. The most common method (panel a in metrics are measured at a fixed applied potential, E1app (panel b in the
the figure) is to measure the photocurrent density, Jph, produced by a figure) by varying the wavelength, , of the monochromatic incident
photoelectrode in an electrolyte as a function of the applied potential, illumination. The photocurrent density Jph is measured at each wave-
E, using a potentiostat. The potential E is reported relative to a reference length, and from the respective incident power density, P, the IPCE
electrode, and the current flows between the photoelectrode and the is calculated as follows:
counter-electrode during testing. The onset potential, Eon, for the hc
photocurrent can be defined from the plot, and Jph at a particular applied IPCE = Jph, P e (2)
potential (Eapp) can be used to calculate the applied-bias photon-to-current
efficiency (ABPCE)163. Assuming PEC water splitting and a photoanode for where hc/e is equal to 1,240 Vnm. The APCE can be determined by
water oxidation simply dividing the IPCE by the fraction of photons absorbed by the
photoelectrode at each wavelength. The IPCE plot shows how the IPCE
Jph (EH2O/O2 Eapp)
ABPCE = (1) can vary with Eapp. The onset wavelength, on, is closely related to the
Pin
semiconductor bandgap, and we note that the IPCE multiplied by the
where EH2O/O2 is the standard potential for water oxidation (+1.23 V versus spectral distribution of standard solar illumination (AM 1.5G,
the reversible hydrogen electrode, VRHE) and Pin is the incident 100 mW cm2) can be integrated to give an estimate of the expected Jph
illumination power density. Note that the chemical products are not under solar illumination.
directly measured. When the photoelectrode is at a specific Eapp, called Reliable measurements of the true solar-to-chemical conversion
the flat-band potential (EFB), the applied potential cancels the efficiency cannot be made in a three-electrode configuration, but instead
band-bending in the space-charge layer. The bands of the semiconductor require a two-electrode configuration, as shown in panel c of the figure
thus become flat, indicating that there is no longer an electric field (note that for every H2 molecule, two electrons flow in the external
present to assist with the separation of the photogenerated electrons circuit). Measuring the operating photocurrent density, Jop, with an
and holes. Determining EFB is useful for estimating the positions of the ammeter as a function of time gives an indication of the stability of the
conduction and valence bands with respect to the redox energy levels device. The integrated Jop gives the total charge passed, which can be
in the electrolyte, which, in turn, allows estimation of the maximum compared with the total amount of gas evolved (accounting for the
photovoltage and photocurrent onset potential (Eon) that can be reaction stoichiometry) to give the faradaic efficiency, F. For the
produced by the junction. Ideally, EFB = Eon; however, the kinetic over- water-splitting reaction, the overall solar-to-hydrogen conversion
potentials typically make this a poor estimation. Thus, EFB is usually efficiency (STH) is calculated as
measured using other techniques such as electrochemical impedance
F (3)
spectroscopy164 or electroreflectance. STH = Jop (1.23 V)
Pin
Quantum-efficiency metrics, such as the incident (or absorbed)
photon-to-current efficiency, IPCE (or APCE), are useful for determining where 1.23 V corresponds to the Gibbs free energy (G) of the reaction.
0 Fe 3d EH+/H2
W 5d W 5d
3.2 eV
2.8 eV
2.0 eV
2.4 eV
2.1 eV
2.0 eV
1.9 eV
1.8 eV
2.0 eV
1.5 eV
2.4 eV
2.0 eV
+1
EH2O/O2
Cu 3d
Fe 3d
Bi 6s Cu3d
+2
Fe 3d N 2p N 2p
Fe 3d
Bi 6s Cu 3d O 2p Valence
+3 O 2p O 2p O 2p band
O 2p O 2p O 2p O 2p O 2p O 2p O 2p O 2p
TiO2 WO3 Fe2O3 BiVO4 CuWO4 ZnFe2O4 CaFe2O4 CuBi2O4 CuNbO4 CuFeO2 TaON LaTiO2N
Figure 2 | Bandgap structure and selected crystal structures of oxide and oxynitride semiconductors for PEC
applications. a|Contribution of metal cation and oxygen anion states to the conduction and valence bands. The bandgap
energy (red for n-type, black for p-type) is shown with respect to the reversible hydrogen electrode and the water redox
energy levels (assuming Nernstian behaviour for the band-edge energies with respect to electrolyte pH). b|Crystal
structure of the monoclinic scheelite polymorph of BiVO4. Here, the blue Bi3+ cations are coordinated to the red oxygen
anions of the VO43 tetrahedra (in a pentagonal bipyramidal structure). c|Delafossite crystal structure of CuFeO2 showing
the light blue Cu+ cations forming layers between face-sharing FeO6 octahedra. d|LaTiO2N adopts a triclinic perovskite
structure where the large La cations (green) occupy the A site between two distinct types of Ti(O4N2) octahedra.
materials (for example, IIIV semiconductors, such as are termed photovoltaic-biased electrosynthetic cells
GaAs and GaInP)19. However, because these materials or integrated photovoltaicelectrolysis cells (REF.24).
corrode or oxidize when they are in direct contact with Although the development of these devices drives inno-
an aqueous electrode, greater complexity is required. vation in PEC-cell design, the complexity and cost of
Typically, thin and conformal overlayers of stable these devices25 make scaling up to the required mag-
oxides such as TiO2 are deposited by atomic layer depo- nitude very difficult. As a result, achieving solar fuel
sition to bury the junction and protect the unstable sem- production at a cost significantly less than (or even
iconductor. Alternatively, the semiconductor junction equivalent to) photovoltaics combined with electrolysis
can be completely removed from the harsh conditions will be challenging 26. Indeed, an analogy with technol-
of direct contact with the aqueous electrolyte and placed ogy for solar electricity production can be drawn here:
outside the liquid cell2023. However, as these devices do the most prolific photovoltaic technology is not based
not derive an electronhole charge-separating junction on the devices with the highest performance, as these
from the direct semiconductorliquid contact, they are far too expensive. Thus, a clear challenge on the way
Potential (VRHE)
0 EH+/H2
1.8 eV
1.7 eV
1.5 eV
1.0 eV
1.5 eV
1.2 eV
2.4 eV
3.2 eV
+1 EH2O/O2
+2
CuInxGasyS2 CuInSe2 CZTS WSe2 CdS TiO2
N N N N N N NH N O
N N N N N N N
N HN OH
N N N N
N N N N N N
O N O
N N N N N N N
O
P
O OH
b
2
1 LUMO
Potential (VRHE, pH 7)
0 EH+/H2
2.7 eV
1.7 eV
1.8 eV
3.2 eV
+1
EH2O/O2
+2
HOMO
Figure 4 | Emerging carbon-based semiconductors for photoelectrochemical devices. a|Chemical Nature Reviews | Materials
structures of
graphitic carbon nitride (g-C3N4); 5-(4-carboxyphenyl)-10,15,20-tris(2,4,6-trimethylphenyl)porphyrin (TMP); and
N,N-bis(phosphonomethyl)-3,4,9,10-perylenediimide (PMPDI). b|Approximate highest occupied molecular orbital
(HOMO) and lowest unoccupied molecular orbital (LUMO) locations of the semiconductors in water at pH 7. RHE,
reversible hydrogen electrode.
structure (as shown in FIG.3c) have recently attracted Silicon and IIIV semiconductors. The outstanding per-
widespread attention in the materials science commu- formance of the classic semiconductors used in photo-
nity owing to their ability to be exfoliated into atomically voltaic energy conversion has not been ignored by the
thin 2D sheets, which results in high extinction coeffi- PEC community. Although silicon is unstable under
cients and allows the bandgap to be tuned according to direct PEC water oxidation or reduction reactions, pho-
the number of atomic layers126,127. In particular, WSe2 has toanode or photocathode operation can be made possi-
been identified as a promising material for PEC appli- ble with conformal protecting layers of TiO2 deposited
cations. Although not stable for water oxidation, initial by atomic layer deposition. The Si and TiO2 band align-
efforts with non-exfoliated (bulk) WSe2 showed its high ment is suitable for conduction-band electron transfer
performance as an n-type photoanode for PEC oxidation from the Si to the TiO2, and subsequent water reduction
reactions128,129. More recently, Lewis and co-workers130 occurs with high photocurrent densities (16 mA cm2 at
reported PEC water reduction with p-type (Nb-doped) +0.2 VRHE)132. Misalignment of the valence-band levels,
WSe2 bulk single crystals as photocathodes. Although however, typically requires that ultrathin tunnel junc-
bare WSe2 electrodes exhibited negligible photocurrents, tions be used133. Recently, a breakthrough in the prepa-
significant hydrogen evolution was observed from a cat- ration of leaky TiO2, in which hole transport can occur
alyst (PtRu)-coated WSe2 electrode in acidic electrolyte by hopping through mid-gap states, has enabled the use
with a (1-sun) photocurrent density of 24.5 mA cm2 at of n-type silicon (n-Si) and IIIV semiconductors as
0 VRHE. As WSe2 can be exfoliated in 2D nanosheets and stable photoanodes for water oxidation134. Multijunction
solution-processed into thin films, it could be possible amorphous silicon (a-Si), which has a tunable bandgap,
to produce ultracheap roll-to-roll fabricated photoelec- was investigated in early solar fuel research on photo-
trodes131. In addition, as a critical advantage over other voltaic-biased electrosynthetic cells25 and has recently
chalcogenide-based materials, no charge extraction or also been used as a photocathode with a protective layer
protection layers are needed on WSe2. Further develop- created by atomic layer deposition135.
ment of this family of semiconductors is warranted in The IIIV class of semiconductors (GaP (REFS136,137),
light of these recent encouraging results. GaInP2 (REFS138,139) and InP (REFS140,141)) also offers
CuWO4 2.02.2 0.3 ~0.2 mA cm2 at 1.23 VRHE under 1-sun Indirect bandgap and high resistivity Unlikely 7072
illumination and IPCEs over 20%
ZnFe2O4 2.0 0.40.6 0.25 mA cm2 at 1.23 VRHE under 1-sun Poor minority carrier LD; bulk and Unlikely 77,80,165
illumination and IPCEs up to 15% surface recombination
TaOxNy 2.12.5 0.1 5 mA cm2 at +1.23 VRHE under 1-sun Low stability (can be addressed with Potentially yes 105,106
illumination and IPCEs up to 76% co-catalyst coating in the short term)
LaTiO2N 2.1 0.1 3.5 mA cm2 at +1.23 VRHE under 1-sun Faradaic efficiency for water Potentially yes 109111
illumination and (sacrificial) IPCEs oxidation of 76%; long-term stability
up to 40% unknown
BaTaO2N 1.8 0.2 4.2 mA cm2 at 1.2 VRHE under 1-sun Long-term stability unknown Potentially yes 114
illumination and IPCEs up to 25%,
ABPCE of 0.7% at 1.0 VRHE
g-C3N4 2.32.7 0.73 Small anodic photocurrents of Poor charge separation; long-term Unlikely 149,150
~5 A cm2 stability unknown
TMP 1.9 NA Small anodic photocurrents of Requires TiO2 scaffold; stable Unlikely 153
~100 A cm2 and an IPCE of 0.028% photocurrent not observed
PMPDI 1.75 0.3 150 A cm2 at +1.6 VRHE IPCEs of ~1% Stable photocurrent not observed Unlikely 154
ABPCE, applied-bias photon-to-current efficiency; EFB(VRHE), flat-band potential with respect to the reversible hydrogen electrode; Eg, bandgap energy; IPCE,
incident photon-to-current efficiency; LD, diffusion length; NA, not applicable; PMPDI, N,N-bis(phosphonomethyl)-3,4,9,10-perylenediimide; STH,
solar-to-hydrogen; TMP, 5-(4-carboxyphenyl)-10,15,20-tris(2,4,6-trimethylphenyl)porphyrin. *For the bare material in contact with aqueous electrolyte.
ERRATUM