This is the pre-peer reviewed version of the following article: ChemSusChem 2017, 10, 4552 –
DOI: 10.1002/cssc.201701538
Full Papers
4560, which has been published in final form at https://onlinelibrary.wiley.com/doi/full/
10.1002/cssc.201701538. This article may be used for non-commercial purposes in accordance
with Wiley Terms and Conditions for Use of Self-Archived Versions.

Level Alignment as Descriptor for Semiconductor/Catalyst
Systems in Water Splitting: The Case of Hematite/Cobalt
Hexacyanoferrate Photoanodes
Franziska Simone Hegner,[a] Drialys Cardenas-Morcoso,[b] Sixto Gimÿnez,*[b] NÇria LÛpez,*[a]
and Jose Ramon Galan-Mascaros*[a, c]
The realization of artificial photosynthesis may depend on the
efficient integration of photoactive semiconductors and catalysts to promote photoelectrochemical water splitting. Many
efforts are currently devoted to the processing of multicomponent anodes and cathodes in the search for appropriate synergy between light absorbers and active catalysts. No single material appears to combine both features. Many experimental
parameters are key to achieve the needed synergy between
both systems, without clear protocols for success. Herein, we
show how computational chemistry can shed some light on
this cumbersome problem. DFT calculations are useful to predict adequate energy-level alignment for thermodynamically
favored hole transfer. As proof of concept, we experimentally

confirmed the limited performance enhancement in hematite
photoanodes decorated with cobalt hexacyanoferrate as a
competent water-oxidation catalyst. Computational methods
describe the misalignment of their energy levels, which is the
origin of this mismatch. Photoelectrochemical studies indicate
that the catalyst exclusively shifts the hematite surface state to
lower potentials, which therefore reduces the onset for water
oxidation. Although kinetics will still depend on interface architecture, our simple theoretical approach may identify and predict plausible semiconductor/catalyst combinations, which will
speed up experimental work towards promising photoelectrocatalytic systems.

Introduction
The production of solar fuels consists of harvesting sunlight as
an energy source to transform a substrate into an energy-rich
chemical through reduction. The reducing equivalents (electrons) are extracted from the complementary semireaction, in
which another substrate (ideally water) gets oxidized. Hence,
the efficiency and kinetics of the overall process undoubtedly
depends on the latter, although the oxidation products typically have no commercial value. Without a fast and robust oxidation process, solar fuels will never reach market interests.
This is the reason why water-oxidation catalysis has become
such a hot topic, as it is generally considered the major bottle[a] F. S. Hegner, Dr. N. LÛpez, Dr. J. R. Galan-Mascaros
Institute of Chemical Research of Catalonia (ICIQ)
Av. Paisos Catalans, 16 Tarragona, 43007 (Spain)
E-mail: nlopez@iciq.es
jrgalan@iciq.es
[b] D. Cardenas-Morcoso, Dr. S. Gimÿnez
Institute of Advanced Materials (INAM)
Universitat Jaume I
Castellon, 12006 (Spain)
E-mail: sjulia@uji.es
[c] Dr. J. R. Galan-Mascaros
ICREA
Pg. LluÌs Companys, 23. Barcelona, 08010 (Spain)
Supporting Information and the ORCID identification number(s) for the
author(s) of this article can be found under https://doi.org/10.1002/
cssc.201701538.
This publication is part of a Special Issue on the topic of Artificial Photosynthesis for Sustainable Fuels. To view the complete issue, visit:
http://dx.doi.org/10.1002/cssc.v10.22.

ChemSusChem 2017, 10, 4552 – 4560

neck towards the realization of artificial photosynthesis.
Beyond electric potential requirements, water oxidation is a
very slow process that includes a four-electron transfer and the
chemically challenging formation of an oxygen–oxygen bond.
Several photoactive semiconductors possess appropriately
aligned valence-band levels and a suitable band gap to drive
water oxidation in their photoexcited state.[1] In this context, ntype metal-oxide semiconductor materials (e.g., TiO2,[2] Fe2O3,[3]
WO3,[4] BiVO4,[5] etc.) have been extensively studied as promising candidates to develop this technology, as they present relatively good stability under operation in harsh environments.
From this family, a-Fe2O3 (hematite) is particularly appealing,
not only because its favorable band gap of 1.9–2.2 eV allows
for light absorption in the visible region but mainly because
“rust” is one of the most abundant and cheap materials on
earth.[3, 6–9]
However, its application in photoelectrochemical water oxidation is severely limited owing to its unfavorable conduction
band-edge level as well as its short carrier lifetimes and slow
oxygen evolution kinetics.[8] Recombination in the bulk as a
result of small polaron trapping impedes hole transport to the
semiconductor–electrolyte interface,[10–12] whereas surface-state
trapping causes electron–hole recombination before the oxidation reaction is able to proceed.[13–20]
To overcome the latter, that is, to enhance water-oxidation
kinetics with respect to surface recombination, a common
strategy is to modify the semiconductor surface with a wateroxidation catalyst (WOC).[9] An efficient hole-transfer catalyst

4552

⌫ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
on top of a photoactive semiconductor is thought to overcome the kinetic barriers of the sluggish water-oxidation reaction, which boosts the performance of the photoanode.[21, 22]
Consequently, the deposition of a stable and cost-effective
WOC on a photoactive semiconductor material is crucial to
achieve the targeted technoeconomical requirements.
Notwithstanding, the simple combination of such two systems does not guarantee success. Indeed, detailed studies on
the mechanism of WOCs on iron-oxide photoanodes have
pointed out that most “catalysts”, although enhancing photoelectrochemical behavior, do not act as genuine catalysts, that
is, they do not provide an effective hole-transfer pathway to
increase the rate of water oxidation by lowering the activation
barrier of the reaction. In many cases, the formation of the
semiconductor/catalyst interface “only” increases the lifetime
of surface recombination by acting as a capacitive layer or by
passivating surface states.[20, 23–25] A thick catalyst layer may
even counteract photoelectrochemical activity by blocking
light absorption or by inhibiting charge transport and ion diffusion from the electrolyte.[26, 27] Moreover, an improvement in
the photoelectrochemical (PEC) performance was also observed for noncatalytic overlayers, such as Al2O3 and Ga2O3,
which either passivate surface states, as in the case of
Ga2O3,[28–30] or change the energetic levels favorably, as in the
case of Al2O3.[31]
The extent to which any semiconductor/catalyst combination has a chance of success can be predicted by computational methods, which in turn can save much time and efforts towards appropriate interface constructions.
We recently reported that cobalt hexacyanoferrate (CoFePB), a competitive WOC, acted preferentially as a genuine catalyst on top of photoactive BiVO4 thin films.[32] Through DFT calculations, we assigned this behavior to favorable alignment of
the energy levels, allowing CoFe-PB to participate as a viable
bridge between the semiconductor and the water molecules.
This synergy was experimentally assessed with CoFe-PB-decorated photoanodes, which exhibited a + 0.8 V gain in photocurrent and a hole-extraction efficiency above 80 %.[32]
Herein, we report the DFT analysis and experimental validation of a similar system, the hematite/CoFe-PB (Figure 1) composite photoanode. Computational models indicate that there

Figure 1. Representation of the crystal structures of a) the trigonal-hexago¯
nal unit cell (space group R36), b) the predominant (0 0 1) surface facet (top
¯
view) of hematite (a-Fe2O3), and c) the cubic unit cell (space group F43m) of
CoFe-PB (KCo[Fe(CN)6]).

ChemSusChem 2017, 10, 4552 – 4560

www.chemsuschem.org

is an intrinsic mismatch between these two materials, which
implies that photoelectrocatalytic synergy should not occur.
Indeed, our experiments are in excellent agreement with these
predictions. CoFe-PB-decorated hematite electrodes show only
a small improvement in photoelectrocatalytic water-oxidation
activity that cannot be assigned to genuine catalytic behavior
of CoFe-PB. Most probably, this is due to a surface-state shift
that also shifts the water-oxidation onset to a lower potential.
Such robust computational analysis will be very useful for any
other semiconductor/catalyst ensemble by allowing for prior
assessment of the validity of a proposed interface combination. These theoretical analyses will facilitate, and even push,
future experimental research in the right direction towards
final optimization of interfaces in photoelectrodes.

Results and Discussion
For efficient hole-transfer catalysis to occur, the catalyst needs
to have its energy levels correctly aligned with the photoactive
material, as schematized in Figure 2 a for the case of the synergetic BiVO4/CoFe-PB system.[32] In particular, this indicates that
once the electron–hole pair is formed, the hole needs to be
transferred to the catalyst; therefore, the corresponding levels
in the catalyst need to be higher in energy than the valence
band (VB) of the photoanode. This energy-alignment requisite
is at the core of the process. If this thermodynamic condition
is not fulfilled, as shown in Figure 2 b, the “catalyst” will not
exert efficient hole transfer.
Hybrid density functional theory (DFT) was used to calculate
the electronic structures of a-Fe2O3, CoFe-PB (KCo[Fe(CN)6]),
and a solvated water molecule, as further described in the Experimental Section. Figure 2 c shows the aligned densities of
states (DOS) of the photoelectrochemical system: the hematite
photoanode (left), the catalyst (middle), and water (right). To
avoid confusion, it has to be noted that the calculated HOMO
must not be mistaken with the electrochemical H2O redox potential, although a linear relationship may exist.[33–35]
The calculated band gap of 1.95 eV for a-Fe2O3 matches well
with its experimental value of 1.9–2.1 eV (Figure S6 in the Supporting Information).[6, 7] This shows that the hybrid functional
HSE03-13 %, which includes 13 % of exact exchange and which
was previously optimized to match the electronic properties of
CoFe-PB,[32] is also adequate to describe the electronic structure of a-Fe2O3 and follows previous indications by Pozun and
Henkelman.[36] The VB edge of hematite consists of both O 2p
and Fe 3d t2g antibonding orbitals, which hybridize as a result
of longitudinal lattice distortion along the main axis of the
hexagonal unit cell (Figure 1). Hybridization of the strongly correlated Fe 3d levels and the bandlike O 2p levels in the VB was
also found by photoemission studies and allows ligand-tometal charge transfer by photoexcitation.[37, 38] The distortion of
the FeO6 octahedra in the lattice also stabilizes the empty
Fe dz2 orbitals along the principal axis, which thus forms the
conduction band (CB) minimum (together with a small Fe 4s
contribution).
If a photon is absorbed, an electron of the O 2p VB will be
excited to the Fe CB by a charge-transfer transition, as was

4553

⌫ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers

Figure 2. Energy-level alignment of a) a synergetic photoanode/catalyst interface: BiVO4/CoFe-PB and b) a noncatalytic photoanode/catalyst combination:
Fe2O3/CoFe-PB. c) Densities of states of a-Fe2O3 (left), KCoFe[(CN)6] (middle), and a solvated H2O molecule (right) aligned by their O 2s bands. Filled electronic
states are represented by filled areas. The CoFe-PB valence band edge is set as the zero energy level, that is, the given electronic levels do not coincide with
electrochemical potentials, which are potential differences measured against a reference electrode.

seen in several X-ray photoemission and absorption studies,
which hence leaves a hole in the VB that is then available for
water oxidation.[37, 39–41] The highest electronic state, that is, the
highest occupied molecular orbital (HOMO), of water lies at
the same energy (within the computational error of ⌃ 0.1 eV)
and, thus, is readily available to accept the hole at the Fe2O3
VB. From a thermodynamic viewpoint, disregarding kinetic effects and surface requirements, the hematite VB edge and the
H2O HOMO levels match adequately to induce electronic overlap, which is a necessary requirement for efficient electron
(hole) transfer to take place. Nonetheless, small polaron effects
and surface recombination prevent the efficient use of the
hole.[10–20]
The CoFe-PB catalyst, however, has its filled Co d t2g states,
responsible for water oxidation,[32, 42] slightly below the Fe2O3
VB edge (Figures 2 b, c). Therefore, there is no thermodynamic
driving force to favor hole transfer from Fe2O3 to CoFe-PB and,
subsequently, to H2O. This, in turn, does not mean that charge
transfer to the catalyst will not occur. The Fe t2g states, which
lie approximately 1 V above the Fe2O3 VB, do not directly participate in the water-oxidation catalytic reaction, nor are they
involved in the formation of the CoFe-PB/Fe2O3 interface. The
strong CN binding from cyanide’s C site makes the Fe centers
inaccessible to coordination. Moreover, their electronic levels
are far above the water HOMO, which excludes electronic overlap, required for water oxidation to occur. At finite temperatures and in an electrochemical environment, the CoFe-PB
electronic levels, which lie close to the Fe2O3 VB, may accept/
donate holes from/to Fe2O3/H2O. All the same, the probability
ChemSusChem 2017, 10, 4552 – 4560

www.chemsuschem.org

of CoFe-PB to have a true catalytic function, providing a faster
hole-transfer pathway, on top of a hematite photoanode is
negligible owing to level-alignment considerations. According
to the DOS levels (Figure 2 c), it is clear that there is an intrinsic
mismatch, as the Co t2g hole-acceptor level of CoFe-PB is at
lower energy than the valence band of hematite. This indicates
that hole transfer is neutral or slightly uphill and, thus, does
not favor catalysis. Nevertheless, other beneficial effects of a
CoFe-PB “catalyst” on hematite cannot be excluded. Tunneling[43] and hopping between sub-band-edge states within the
barrier layer[44] have also been reported to describe charge
transfer between photoactive and catalytic materials if the
energy levels are not correctly aligned.
We also performed experiments to assess the validity of
such a prediction. Nanostructured Zr-doped hematite films
(see the Supporting Information for more details) were modified with the CoFe-PB catalyst by a sequential coating method
at room temperature, as previously described.[32] The hematite
electrode was submerged first in a ferricyanide solution and
then in a cobalt chloride solution (both at neutral pH), which
promoted the growth of the insoluble cobalt hexacyanoferrate.
For our experiments, we generally used between three and six
dipping cycles. Increasing the number of cycles did not significantly alter the photoelectrochemical performance of the photoanodes (as shown in Figure S3).
The electrodes were characterized by scanning electron microscopy (SEM) (Figure S4) and high-resolution transmission
electron microscopy (HRTEM, Figure 3), both combined with
energy-dispersive X-ray (EDX) spectroscopy. A total Fe/O ratio

4554

⌫ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
of 32:68 is found, which is in good agreement with the Fe2O3
stoichiometry. More consistent evidence for the formation of
Fe2O3 is provided by X-ray photoelectron spectroscopy (XPS)
analysis (see Figure S5 and Table S2). The amount of the CoFe-

Figure 3. a) Electron micrographs obtained by HRTEM of a scratched CoFePB/a-Fe2O3 electrode with b, c) magnifications of the CoFe-PB crystallites
found on the surface.

PB catalyst is less than 1 % and cannot be detected under SEM
conditions. With HRTEM, however, cubic CoFe-PB crystallites
can clearly be found on the Fe2O3 surface (Figure 3 b, c). Their
Co/Fe ratios are found to be between 50:50 and 60:40 by EDX
spectroscopy, and they all lie in the stoichiometric possible
range of 1:1 (KCo[Fe(CN)6]) and 3:2 (Co3[Fe(CN)6]2).[32, 45, 46] The
presence of CoFe-PB on the surface is further confirmed by
XPS, as described in the Supporting Information. Catalyst deposition does not show a significant influence on the optical
UV/Vis absorption or band gap of the hematite semiconductor
(Figure S6).
The photoelectrochemical behavior of the photoanodes was
examined by cyclic voltammetry (CV) under chopped and constant illumination (Figure 4) in neutral (pH 7) potassium phosphate (KPi) buffer (0.1 m) solution. In the anodic scans at
50 mV sˇ1 (Figure 4 a, c) an apparent cathodic shift in the photocurrent onset potential of 0.3 V can be seen. Nevertheless,
scanning in the cathodic direction (Figure 4 b) reveals a much
smaller improvement in the photocurrent, which indicates that
the apparent shift in the water-oxidation onset in the anodic
direction is the result of the capacitive effect of the CoFe-PB
layer, as the difference in anodic and cathodic scans for bare
Fe2O3 is negligible. Indeed, scanning the CoFe-PB/a-Fe2O3 electrode at lower scan rates (Figures 4 c and S7) significantly de-

Figure 4. The j–V curves of CoFe-PB/Fe2O3 (red) and bare Fe2O3 (black) in 0.1 m KPi buffer solution at pH 7. a) Anodic and b) cathodic CV scans under chopped
light (1 sun) at a scan rate of 50 mV sˇ1. c) Anodic CV curve in the dark (dotted thin lines) and under light (thick solid lines) collected at scan rates of 50 (thick
solid lines) and 2 mV sˇ1 (blue dashed line). d) Steady-state j–V curve extracted from the electrochemical impedance analysis data.

ChemSusChem 2017, 10, 4552 – 4560

www.chemsuschem.org

4555

⌫ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
creases the observed photocurrent, whereas the scan rate
barely affects the current density–voltage (j–V) behavior of
bare hematite (Figure S7). Moreover, from the steady-state j–V
curve (Figure 4 d), which can be extracted from electrochemical
impedance spectroscopy (EIS) analysis, as well as from scans at
very low scan rates, the “real” photocurrent onset potential
shift of 0.1 V can be determined. In here, the voltage needed
to attain a photocurrent of 0.01 mA cmˇ2 was taken as the
onset potential.
To gain mechanistic understanding of the effect of CoFe-PB
on hematite photoelectrodes, PEC experiments were performed in the presence of hydrogen peroxide (0.5 m), which
was found to be an optimal hole scavenger for hematite.[15]
Under these conditions it is assumed that all holes that reach
the surface will be rapidly injected into solution and no electron–hole recombination will take place at the semiconductor–
liquid interface.[15] Figure 5 a shows the photocurrents obtained
with and without a hole scavenger under illumination at pH 7,
and Figure 5 b shows the calculated charge-transfer efficiencies
as a function of applied potential (see Supporting Information
for calculation details).
The measured CV curves in hole scavenger solution do not
show significant differences, which implies that an equal
number of holes reaches the semiconductor/electrolyte interface for both bare and CoFe-PB-modified hematite; thus, their
charge-separation efficiencies are similar. This, in turn, implies
that charge mobilities in the bulk and band bending are not
affected by surface catalyst modification. Consequently, it can
be safely claimed that surface modification of Fe2O3 by CoFePB does not alter its bulk properties. The photocurrent onset
in hole scavenger, that is, without surface recombination, is
equal to the flat band potential (VFB) and is approximately
0.6 V versus reversible hydrogen electrode (RHE).[47]
The charge-transfer efficiency of CoFe-PB/Fe2O3 reaches its
maximum at approximately 1.35 V versus RHE, whereas for
bare Fe2O3 it is maximal at 1.57 V versus RHE. The magnitude
of 60 %, however, is equal for both bare and modified Fe2O3.
This indicates that the CoFe-PB catalyst does not improve
charge transfer to the electrolyte but that the catalyst shifts it

to a lower potential, which is in line with the observed cathodic shift in the onset potential.
EIS measurements were performed to further investigate the
effect of the catalyst layer on the photocurrent. The obtained
Nyquist plots systematically show a single arc in the dark and,
thus, were fit to a simple Randles’ circuit,[48] whereas two arcs
are obtained under illumination at high potential (see Figure S8). This is due to a different, indirect charge-transfer
mechanism taking place, in which holes are trapped in surface
states and are then transferred to the electrolyte.[17, 23, 49] It is apparent that the potential at which indirect charge transfer
starts to occur is lower for CoFe-PB/Fe2O3 electrodes than for
bare Fe2O3, as also described in the Supporting Information.
These EIS data can be fit to a previously established equivalent
circuit that accounts for indirect hole transfer via hematite surface states and that is shown in Figure 6 a.[17, 23, 47, 49] The equivalent circuit elements include the contact-dependent series resistance (RS); the trapping resistance (Rtrap) (Figure 6 b), which
describes trapping of charges into the surface state; the
charge-transfer resistance (RCT) from the surface state to the
electrolyte (Figure 6 b); the space-charge capacitance (CSC) of
the bulk semiconductor (Figure 6 c); and the surface-state capacitance (CSS) of the hematite surface traps (Figure 6 d).
The resistances Rtrap and RCT (Figure 6 a, b) are not influenced
by the amount of catalyst on the surface. The trapping of
holes into the surface state (described by Rtrap) is not affected
by the catalyst at all; this is in good agreement with the information provided by the experiments with the hole scavenger
(Figure 5), for which no significant changes in the bulk properties of Fe2O3 are identified. On the other hand, hole transfer
from the surface states occurs at lower potentials, in accordance with the cathodic shift of the onset potential (Figure 4)
and the calculated charge-transfer efficiency (Figure 5).
The bulk capacitance of hematite CSC (Figure 6 c), which results from band bending in the space-charge region, seems to
decrease slightly upon increasing the number of dipping
cycles during deposition. This, however, can be attributed to
having less Fe2O3 surface area exposed to the solution in the
CoFe-PB/Fe2O3 anode as more CoFe-PB is added. With Mott–
Schottky (MS) analysis (see the Supporting Information), the

Figure 5. a) Anodic CV curves of CoFe-PB/ Fe2O3 (red) and bare Fe2O3 (black) in 0.1 m KPi buffer (solid lines) and 0.5 m H2O2 (dashed lines) solutions (both at
pH 7) recorded at a slow scan rate of 2 mV sˇ1 to exclude catalytic, nonfaradaic current. b) Calculated charge-transfer efficiencies for both systems at pH 7.

ChemSusChem 2017, 10, 4552 – 4560

www.chemsuschem.org

4556

⌫ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers

Figure 6. a) Equivalent circuit model used to fit EIS results obtained under 1 sun irradiation at higher potentials when indirect hole-transfer occurs. b) Chargetransfer (circles) and trapping (triangles) resistances of bare Fe2O3 (black) and CoFe-PB/Fe2O3 (red). c) Bulk and d) surface state capacitances of bare (black)
and CoFe-PB coated (red) Fe2O3 semiconductor photoanodes. All spectra were obtained at pH 7 (0.1 m KPi buffer) and 1 sun irradiation.

flat band potential and the doping density of hematite were
determined. As expected, the flat band potential, VFB = (0.5 ⌃
0.05) V versus RHE, is independent of the CoFe-PB catalyst,
which proves that CoFe-PB does not change the band positions. It differs by approximately 0.1 V from the photocurrent
onset potential in the presence of a hole scavenger (which
gives VFB = 0.6 V vs. RHE) (Figure 5) owing to inaccuracies in
the MS description for highly doped materials, as described by
Zandi et al.[50] A flattening or horizontal shift in the MS curves,
which was previously attributed to Fermi-level pinning,[17, 23] is
not observed here. The high annealing temperatures (800 8C)
employed in this study are partially thought to passivate surface states and unpin the Fermi level, an effect that is also
seen in the comparatively small photocurrent transients (Figure 3 a, b).
Figure 6 d shows the surface-state capacitance (CSS) values of
bare and CoFe-PB-modified hematite photoanodes. The magnitude of CSS is not affected by CoFe-PB; hence, the “catalyst”
does not passivate surface states. It can be seen clearly
though, that the deposition of CoFe-PB shifts the surface state
to lower potential. Hence, we conclude that the observed
cathodic onset potential shift originates from shifting the surface state, which initiates indirect charge transfer through the
surface states at a lower potential than for bare Fe2O3, and
ChemSusChem 2017, 10, 4552 – 4560

www.chemsuschem.org

this, in turn, is due to oxidation of the CoFe-PB catalyst. A similar effect was found for Ga2O3 overlayers on hematite.[28] Furthermore, the trends followed by the two capacitances included in the model validate the selection of this equivalent circuit
to fit our experimental results.
Although the energy-level alignment (Figure 2) does not
favor hole transfer to the catalyst in its ground state, an applied external potential can initiate charge transfer to the catalyst in the surface, and this creates the oxidized CoIIIFeIII states
in CoFe-PB.[46, 51] This transfer takes place at lower potentials
than hole transfer to the surface state, which could also explain the cathodic shift in the surface-state capacitance.
Another important issue is related to the interfacial adhesion
of Fe2O3 and the CoFe-PB catalyst. The larger the surface in
contact between the photoanode and the catalyst, the more
paths available for hole scavenging. As shown in a recent
study by Shao-Horn et al., water wetting strongly influences
the charge-transfer properties from/to the electrolyte/catalyst
interface and, therefore, crucially determines the dynamics of
the catalytic surface reaction.[52] A similar principle applies to
the “wetting” of the catalyst on the semiconductor surface,
which leads to hole transfer to the solution.
The micrographs in Figure 3 show that the nanoparticles are
essentially nonwetting the oxide surface, which thus limits the

4557

⌫ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
number of hole-transfer paths. The reasons behind the poor
wetting can be traced back to the crystal structure of the compounds. The most common surface termination of hematite is
hexagonal (Figure 1 b); however, the termination of PB is cubic,
and this leads to lattice mismatch. Having incommensurate
crystal facets minimizes the number of CoˇOˇFe bridges at
the interface, which are needed to enable hole transfer. Hence,
catalyst “wetting” is another aspect that may be studied in silico and is linked to interface engineering that is crucial to its
performance.

have filled electronic states between the valence band (VB) of
Fe2O3 and the HOMO of H2O, which are available for water oxidation (next to having accessible coordination sites). A better
strategy could involve shifting the VB of Fe2O3 to lower energies, which could be achieved by suitable dopants[36] or surface
modification with overlayers.[31]
In this line, analogous computational studies indicate that
the favorable alignment of a photoactive semiconductor and
catalyst (e.g., CoFe-PB/BiVO4) leads to a remarkable increase in
performance,[32] which corroborates the validity of our theoretical approach.

Conclusions
Light-harvesting semiconductors, which can transform sunlight
into an electric-field potential as the driving force to produce
fuels, are promising candidates for large-scale application of artificial photosynthesis technologies. However, they need to be
coupled to an appropriate catalyst for the reaction to be efficient and fast enough. Generally, electron–hole recombination
is faster than chemical transformations.
Beyond interfacial engineering requirements, there will be
an important contribution from the correct alignment between
the electronic levels from both the semiconductor and the catalyst. In this manuscript, we demonstrated how appropriate
level alignment could be used to shed some light on possible
charge-transfer pathways and, hence, to determine the applicability of a possible co-catalyst, as exemplified in the hematite/
cobalt hexacyanoferrate case.
Although the applied DFT analysis used the simplified model
of bulk structures only and did not include real electrochemical
interfaces, it clearly showed that there was an intrinsic mismatch, as the catalyst hole-acceptor level was below the valence band of hematite in the energy diagram. Consequently,
hole transfer to the catalyst was neutral or slightly uphill, and
thus, there was no thermodynamic pathway for the generated
holes in hematite to be transferred to the catalyst. This suggests that hole transfer to water, and thus water oxidation catalysis, is more likely to occur directly at the semiconductor surface. Nonetheless, different hole-transfer pathways, such as
tunneling or hopping between sub-band-edge states, may be
considered.[43, 44] In good agreement, our experiments indicated
that CoFe-PB decoration on top of hematite electrodes did not
lead to a relevant enhancement in the photoelectrocatalytic
performance. We assigned the small enhancement to longer
lifetimes of electron–hole surface recombination as a result of
the hole-scavenging character of the interface, which shifted
the surface-state capacitance to more cathodic potentials.
Water oxidation still preferentially occurred on the hematite
surface as observed in different hematite/catalyst systems.[23, 24]
Therefore, the theoretical model of simple energy diagrams
could predict the feasibility of this (and any other) junction. Interfaces may be engineered and improved, but if the process
is thermodynamically uphill, their improvement is not due to
an improvement in catalytic efficiency. For bare hematite in
particular, finding a suitable true catalyst is a difficult task, as
hematite has its valence band edge maximum very close to
the HOMO of water. An appropriate co-catalyst would need to
ChemSusChem 2017, 10, 4552 – 4560

www.chemsuschem.org

Experimental Section
Materials
Iron(II) chloride tetrahydrate (FeCl2·4 H2O, > 98 %), zirconyl chloride
octahydrate (ZrOCl2·8 H2O > 99 %), and potassium ferricyanide
(K3[Fe(CN)6], 99.0 %) were purchased from Sigma–Aldrich, and
cobalt chloride hexahydrate (CoCl2·6 H2O, 98.0 %) was purchased
from Fluka Analytical. Dimethyl sulfoxide (DMSO, 99.9 %) and hydrogen peroxide solution [H2O2, 30 % (w/w) in H2O] were obtained
from Sigma–Aldrich. The buffer solution was prepared from potassium phosphate monobasic and dibasic (KH2PO4, 99.0 %; K2HPO4,
98.0 %; Sigma-Aldrich). High-purity (Milli-Q) water was obtained
with a Millipore purification system (Synergy) and was used for all
solutions. Fluorine-doped tin oxide (FTO)-coated glass slides were
purchased from Hartford glass (15 W cmˇ2).

Synthesis of hematite electrodes
Thin-film hematite electrodes were prepared by a simple and costefficient electrodeposition method, based on a description by
Shaddad et al.,[53] but with varying calcination conditions as in
Refs. [50, 54]. Prior to deposition, FTO electrodes were ultrasonicated and then thoroughly cleaned with water and ethanol (isopropyl
alcohol). Zr-doped metallic Fe was deposited from a solution of
20 mm FeCl2·4 H2O and 0.9 mm ZrOCl2·8 H2O in DMSO by applying
a constant potential of ˇ20. V vs. Ag/AgCl (3 m KCl) for 6 min.
After carefully rinsing the films with Milli-Q water, the electrodes
were calcined in air by heating up to 800 8C for 9–10 min, which
was followed by rapid quenching at room temperature.

Sequential CoFe-PB coating
The CoFe-PB catalyst was deposited by sequentially dipping the
hematite electrodes in reactant solutions of 0.02 m K3[Fe(CN)6] in
H2O and 0.04 m CoCl2 in H2O, as we recently reported for BiVO4
photoanodes.[32] First, the electrodes were dipped in [Fe(CN)6]3ˇ solution for 10–15 min with slow stirring, so that the negatively
charged iron cyanide complexes could bind to the Fe2O3 surface.
Afterwards, the electrodes were thoroughly rinsed with Milli-Q
water and were then dipped in the Co2 + solution, again for 10–
15 min with stirring to form CoFe-PB complex structures. The sequence was repeated at least two times to ensure significant
CoFe-PB deposition. In all shown measurements, 3–6 repetitions
were applied, which did not change the PEC behavior of the electrodes (see Figure S3 b).

4558

⌫ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Synthesis of CoFe-PB electrodes
For the sake of comparison, CoFe-PB/FTO electrodes were prepared by following the hydrothermal method described by Han
et al.[42] In this method, CoOx was first deposited by heating
Co(NO3)2 and carbamide in an autoclave at 1208 (10 h) and was afterwards derivatized to CoFe-PB in K3[Fe(CN)6] solution at 608 (1–
3 h).

ber was 8 î 10ˇ9 hPa. The data were evaluated by using Casa XPS
software. The energy corrections of the spectra were performed
considering a reference value of C 1s from the organic matter at
284.8 eV. UV/Vis spectra of the electrodes was recorded with a
Cary 300 Bio spectrometer (UV0911M213). Absorbance and band
gaps were calculated as described in the Supporting Information.

Computational details
Photoelectrochemical (PEC) measurements
PEC experiments were performed with an Eco Chemie Autolab potentiostat coupled with NOVA electrochemical software. A typical
three-electrode cell consisted of the hematite photoanode as the
working electrode, a Pt wire or mesh as the counterelectrode, and
a Ag/AgCl (3 m KCl) reference electrode. All potentials were converted into the pH-independent reversible hydrogen electrode
(RHE) by using the Nernst equation [Eq. (1)]:
V RHE à V Ag=AgCl á V 0
Ag=AgCl á 0:059 pH

Ö1Ü

with V 0
Ag=AgCl Ö3 m KClÜ à 0:21 V

To normalize the measured current (in A) to a current density j in
mA cmˇ2, the electrode geometrical areas were determined by the
graphical software ImageJ 1.50i. If not stated otherwise, the experiments were performed in a 0.1 m solution of potassium phosphate
(KH2PO4) buffer at pH (7 ⌃ 0.1). The pH was determined with a
CRISON Basic 28 pH meter. Hole-scavenger experiments were performed in 0.5 m H2O2 solution (pH 7), which was described to be
an effective hole scavenger by Warren et al.[15] A 450 W Xe arc
lamp with an AM 1.5 solar filter (Sciencetech Inc.) was used to simulate sunlight of 100 mW cmˇ2 (1 sun). If not otherwise mentioned,
cyclic voltammetry (CV) scans were typically performed at a scan
rate of 50 mV sˇ1 until a stable signal was reached. For the CoFe-PB
modified electrodes, approximately 2–3 CV scans were needed to
stabilize the signal, but not for bare hematite. All hematite electrodes were illuminated from the electrolyte from the top of the hematite surface. This ensured a small mean free path for photogenerated holes, as hematite is well known to have very small hole-diffusion lengths between 2–4[55] and 20 nm.[56] Impedance data were
collected between 10ˇ1 and 406 Hz by using a 20 mV amplitude
voltage perturbation and were analyzed with ZView software
(Scribner associates). Steady-state j–V curves were extracted by
monitoring the stabilized current at each applied voltage during
the impedance measurement.

Structural and optical characterization
Morphologies, particle sizes, and chemical compositions were determined by scanning electron microscopy (SEM) with a JSM-7000F
JEOL FEG-SEM system (Tokio, Japan) equipped with an INCA 400
Oxford EDX analyzer (Oxford, UK) and operating at 15 kV and a
JEM-2100 JEOL transmission electron microscope operating at
200 kV that also contained an INCA 400 Oxford EDX analyzer
(Oxford, UK). Prior to the SEM experiment, the samples were sputtered with a 2 nm thick layer of Pt. Surface analysis was performed
by X-ray photoelectron spectroscopy (XPS) by using a Specs
SAGE 150 instrument. The analyses were performed by using nonmonochromatic AlKa irradiation (1486.6 eV) at 20 mA and 13 kV, a
constant energy pass of 75 eV for overall analysis, 30 eV for analysis in the specific binding energy ranges of each element, and a
measurement area of 1 î 1 mm2. The pressure in the analysis chamChemSusChem 2017, 10, 4552 – 4560

www.chemsuschem.org

Density functional theory (DFT) calculations were performed by
using the Vienna Ab Initio Package (VASP)[57, 58] on the model structures of stoichiometric KCo[Fe(CN)6] and a-Fe2O3 (Figure 1). Although real CoFe-PB is a nonstoichiometric compound, its electronic structure is not expected to vary from the ideal KCo[Fe(CN)6]
structure.[59] However, a difference occurs owing to a change in the
magnetic configuration, which is discussed in the Supporting Information. As pure density functional theory has proven insufficient
to describe correctly the electronic structure of Prussian Blue type
materials[60] and as DFT + U cannot unambiguously predict an explicit U-term needed to compare materials with different transition-metal centers,[60, 61] a modified hybrid functional based on the
HSE03 functional,[62, 63] but including only 13 % of exact Hartree–
Fock (HF) exchange, was employed.[32] More information about the
functional is given in the Supporting information. Projector Augmented Wave (PAW) pseudopotentials with small cores, expanding
valence-subshell containing s and p electrons, ensured sufficient
flexibility and were used for all metal atoms in the lattice.[64, 65] For
structure optimizations, a G-centered k-point mesh was used, and
valence electrons were expanded in plane waves with kinetic energies up to 500 eV. Single-point calculations to obtain more accurate electronic structures were performed with a kinetic cutoff
energy of 600 eV and denser Monkhorst–Pack k-point grids with
3 î 3 î 3 (CoFe-PB) or 3 î 3 î 2 (Fe2O3) k-points.[66] Water was calculated with the same scheme. For this, a single H2O molecule in an
asymmetric box (14.5 ä î 15 ä î 15.5 ä) was solvated through a
continuum model. The implicit solvated water was represented
through the MGCM method.[67, 68]
All structures and calculations were uploaded to the ioChem-BD
database, from which they are openly accessible.[69–71]

Acknowledgements
We would like to acknowledge financial support from the European Union (project ERC StG grant CHEMCOMP no 279313); the
Spanish Ministerio de EconomÌa y Competitividad (MINECO)
through projects CTQ2015-71287-R, CTQ2015-68770-R and the
Severo Ochoa Excellence Accreditation 2014–2018 SEV-20130319; the Generalitat de Catalunya (2014-SGR-797, 2014SGR-199
and the CERCA Programme); University Jaume I through the
P11B2014-51 project; and the Generalitat Valenciana through the
Santiago Grisolia Program, grant 2015-031. Serveis Centrals at
UJI (SCIC) are also acknowledged. F.S.H. thanks the “LaCaixa”Severo Ochoa International Programme (Programa internacional
de Becas “LaCaixa”—Severo Ochoa) for a Ph.D. fellowship. We
thank BSC-RES for generous computational resources.

Conflict of interest
The authors declare no conflict of interest.

4559

⌫ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Keywords: computational chemistry · electrochemistry
hematite · oxygen evolution catalysis · water splitting
[1]
[2]
[3]
[4]

[5]
[6]
[7]

[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]

[26]
[27]
[28]
[29]
[30]
[31]
[32]

[33]
[34]
[35]
[36]
[37]
[38]

·

K. Sivula, R. van de Krol, Nat. Rev. Mater. 2016, 1, 15010.
A. Fujishima, K. Honda, Nature 1972, 238, 37 – 38.
K. L. Hardee, A. J. Bard, J. Electrochem. Soc. 1976, 123, 1024 – 1026.
C. F‡brega, S. Murcia-Lopez, D. Monllor-Satoca, J. D. Prades, M. D. Hernandez-Alonso, G. Panelas, J. R. Morante, T. Andreu, Appl. Catal. B 2016,
189, 133 – 140.
T. W. Kim, K. S. Choi, Science 2014, 343, 990 – 994.
J. H. Kennedy, K. W. Frese, J. Electrochem. Soc. 1978, 125, 709 – 714.
R. M. Cornell, U. Schwertmann in The Iron Oxides: Structure, Properties,
Reactions, Occurrences and Uses, 2nd ed., Wiley-VCH, Weinheim, 2003,
pp. 9 – 38.
K. Sivula, F. Le Formal, M. Gr‰tzel, ChemSusChem 2011, 4, 432 – 449.
A. G. Tamirat, J. Rick, A. A. Dubale, W.-N. Su, B.-J. Hwang, Nanoscale
Horiz. 2016, 1, 243 – 267.
S. M. Ahmed, J. Leduc, S. F. Haller, J. Phys. Chem. 1988, 92, 6655 – 6660.
A. J. Bosman, H. J. Vandaal, Adv. Phys. 1970, 19, 1 – 117.
L. M. Carneiro, S. C. Cushing, C. Liu, Y. Sude, P. Yang, A. P. Alivisatos, S. R.
Leone, Nat. Mater. 2017, 16, 819 – 824.
M. Barroso, S. R. Pendelbury, A. J. Cowan, J. R. Durrant, Chem. Sci. 2013,
4, 2724 – 2734.
F. Le Formal, S. R. Pendlebury, M. Cornuz, S. D. Tilley, M. Gr‰tzel, J. R.
Durrant, J. Am. Chem. Soc. 2014, 136, 2564 – 2574.
H. Dotan, K. Sivula, M. Gr‰tzel, A. Rothschild, S. C. Warren, Energy Environ. Sci. 2011, 4, 958 – 964.
B. Klahr, S. Gimenez, F. Fabregat-Santiago, T. Hamann, J. Bisquert, J. Am.
Chem. Soc. 2012, 134, 4294 – 4302.
B. Klahr, S. Gimenez, F. Fabregat-Santiago, J. Bisquert, T. Hamann, Energy
Environ. Sci. 2012, 5, 7626 – 7636.
B. Iandolo, A. Hellman, Angew. Chem. Int. Ed. 2014, 53, 13404 – 13408;
Angew. Chem. 2014, 126, 13622 – 13626.
N. Yatom, O. Neufeld, M. C. Toroker, J. Phys. Chem. C 2015, 119, 24789 –
24795.
N. Yatom, Y. Elbaz, S. Navon, M. C. Toroker, Phys. Chem. Chem. Phys.
2017, 19, 17278 – 17286.
F. Ning, M. Shao, S. Xu, Y. Fu, R. Zhang, M. Wei, D. G. Evans, X. Duan,
Energy Environ. Sci. 2016, 9, 2633 – 2643.
W. Liu, H. Liu, L. Dang, H. Zhang, X. Wu, B. Yang, Z. Li, X. Zhang, L. Lei,
S. Jin, Adv. Funct. Mater. 2017, 27, 1603904.
B. Klahr, S. Gimenez, F. Fabregat-Santiago, J. Bisquert, T. Hamann, J. Am.
Chem. Soc. 2012, 134, 16693 – 16700.
M. Barroso, A. J. Cowan, S. R. Pendelbury, M. Gr‰tzel, D. R. Klug, J. R.
Durrant, J. Am. Chem. Soc. 2011, 133, 14868 – 14871.
L. Badia-Bou, E. Mas-Marza, P. Rodenas, E. M. Barea, F. Fabregat-Santiago, S. Gimenez, E. Peris, J. Bisquert, J. Phys. Chem. C 2013, 117, 3826 –
3833.
D. K. Zhong, D. R. Gamelin, J. Am. Chem. Soc. 2010, 132, 4202 – 4207.
G. Wang, Y. Ling, X. Lu, T. Zhai, F. Qian, Y. Tong, Y. Li, Nanoscale 2013, 5,
4129 – 4133.
L. Steier, I. Herraiz-Cardona, S. Gimenez, F. Fabregat-Santiago, J.- Bisquert, S. D. Tilley, M. Gr‰tzel, Adv. Funct. Mater. 2014, 24, 7681 – 7688.
E. Aharon, M. C. Toroker, Catal. Lett. 2017, 147, 2077 – 2082.
K. Ulman, M. T. Nguyen, N. Seriani, R. Gebauer, J. Phys. Chem. 2016, 144,
094701.
O. Neufeld, N. Yatom, M. C. Toroker, ACS Catal. 2015, 5, 7237 – 7243.
F. S. Hegner, I. Herraiz-Cardona, D. Cardenas-Morcoso, N. LÛpez, J. R.
Galµn-MascarÛs, S. Gimenez, ACS Appl. Mater. Interfaces 2017, DOI:
10.1021/acsami.7b09449.
J. Conradie, J. Phys. Conf. Ser. 2015, 633, 012045.
A. V. Marenich, J. Ho, M. L. Coote, C. J. Cramer, D. G. Truhlar, Phys. Chem.
Chem. Phys. 2014, 16, 15068 – 15106.
C. P. Kelly, C. J. Cramer, D. G. Truhlar, J. Phys. Chem. 2006, 110, 16066 –
16081.
Z. D. Pozun, G. Henkelman, J. Chem. Phys. 2011, 134, 224706.
A. Fujimori, M. Saeki, N. Kimizuka, M. Taniguchi, S. Suga, Phys. Rev. B
1986, 34, 7318 – 7328.
M. Catti, D. Valerio, R. Dovesi, Phys. Rev. B 1995, 51, 7441 – 7450.

ChemSusChem 2017, 10, 4552 – 4560

www.chemsuschem.org

[39] G. Dr‰ger, W. Czolbe, J. A. Leiro, Phys. Rev. B 1992, 45, 8283 – 8287.
[40] R. J. Lad, V. E. Henrich, Phys. Rev. B 1989, 39, 13478 – 13485.
[41] Y. Ma, P. D. Johnson, N. Wassdahl, J. Guo, P. Skytt, J. Nordgren, S. D.
Kevan, J.-E. Rubensson, T. Bçske, W. Eberhardt, Phys. Rev. B 1993, 48,
2109 – 2111.
[42] L. J. Han, P. Y. Tang, A. Reyes-Carmona, B. Rodriguez-Garcia, M. Torrens,
J. R. Morante, J. Arbiol, J. R. Galan-Mascaros, J. Am. Chem. Soc. 2016,
138, 16037 – 16045.
[43] Y. W. Chen, J. D. Prange, S. D¸hnen, Y. Park, M. Gunji, C. E. D. Chidsey,
P. C. McIntyre, Nat. Mater. 2011, 10, 539 – 544.
[44] J. R. Avila, M. J. Katz, O. K. Farha, J. T. Hupp, J. Phys. Chem. C 2016, 120,
20922 – 20928.
[45] R. O. Lezna, R. Romagnoli, N. R. de Tacconi, K. Rajeshwar, J. Phys. Chem.
B 2002, 106, 3612 – 3621.
[46] N. R. de Tacconi, K. Rajeshwar, R. O. Lezna, Chem. Mater. 2003, 15,
3046 – 3062.
[47] J. Bisquert, S. Gimenÿz, L. Bertoluzzi, I. Herraiz-Cardona in Photoelectrochemical Solar Fuel Production: From Basic Principles to Advanced Devices (Eds.: S. Gimenÿz, J. Bisquert), Springer International, Cham, Switzerland, 2016, Chap. 6.
[48] J. E. B. Randles, Discuss. Faraday Soc. 1947, 1, 11 – 19.
[49] L. Bertoluzzi, P. Lopez-Varo, J. A. Jimenÿz Tejada, J. Bisquert, J. Mater.
Chem. A 2016, 4, 2873 – 2879.
[50] O. Zandi, A. R. Schon, H. Maihibabaei, T. W. Hamann, Chem. Mater. 2016,
28, 765 – 771.
[51] R. MartÌnez-GarcÌa, M. Knobel, J. Balmaseda, H. Yee-Madeira, E. Reguera,
J. Phys. Chem. Solids 2007, 68, 290 – 298.
[52] B. Han, K. A. Stoerzinger, V. Tileli, A. D. Gamalski, E. A. Stach, Y. ShaoHorn, Nat. Mater. 2017, 16, 121 – 126.
[53] M. N. Shaddad, M. A. Ghanem, A. M. Al-Mayouf, S. Gimenez, J. Bisquert,
I. Herraiz-Cardona, ChemSusChem 2016, 9, 2779 – 2783.
[54] A. Annamalai, A. Subramanian, U. Kang, H. Park, S. H. Choi, J. S. Jang, J.
Phys. C 2015, 119, 3810 – 3817.
[55] K. Sivula, R. Zboril, F. Le Formal, R. Robert, A. Weidenkaff, J. Tucek, J. Frydruch, M. Gr‰tzel, J. Am. Chem. Soc. 2010, 132, 7436 – 7444.
[56] M. P. Dare-Edwards, J. B. Goodenough, A. Hemnett, P. R. Trevellick, J.
Chem. Soc. Faraday Trans. 1 1983, 79, 2027 – 2041.
[57] J. Hafner, G. Kresse, The Vienna Ab-Initio Simulation Program VASP: An
Efficient and Versatile Tool for Studying the Structural, Dynamic, and Electronic Properties of Materials, Plenum Press, New York, 1997, pp. 69 – 82.
[58] G. Kresse, J. Furthm¸ller, Phys. Rev. B 1996, 54, 11169 – 11186.
[59] J. C. Wojdel, I. D. R. Moreira, S. T. Bromley, F. Illas, J. Chem. Phys. 2008,
128, 044713.
[60] F. S. Hegner, J. R. Galµn-MascarÛs, N. LÛpez, Inorg. Chem. 2016, 55,
12851 – 12862.
[61] M. Capdevila-Cortada, Z. Lodziana, N. Lopez, ACS Catal. 2016, 6, 8370 –
8379.
[62] J. Heyd, G. E. Scuseria, J. Chem. Phys. 2004, 120, 7274 – 7280.
[63] J. Heyd, G. E. Scuseria, M. Ernzerhof, J. Chem. Phys. 2003, 118, 8207 –
8215.
[64] P. E. Blçchl, Phys. Rev. B 1994, 50, 17953 – 17979.
[65] G. Kresse, D. Joubert, Phys. Rev. B 1999, 59, 1758 – 1775.
[66] H. J. Monkhorst, J. D. Pack, Phys. Rev. B 1976, 13, 5188 – 5192.
[67] M. Garcia-Ratÿs, N. LÛpez, J. Chem. Theory Comput. 2016, 12, 1331 –
1341.
[68] M. Garcia-Ratÿs, R. GarcÌa-Muelas, N. LÛpez, J. Phys. Chem. C 2017, 121,
13803 – 13809.
[69] ioChem-BD software, http://www.iochem-bd.org/, Institute of Chemical
Research of Catalonia (ICIQ), 2014.
[70] M. çlvarez-Moreno, C. de Graaf, N. LÛpez, F. Maseras, J. M. Poblet, C. Bo,
J. Chem. Inf. Model. 2015, 55, 95 – 103.
[71] F. S. Hegner, link to database: Photoanodes_Fe2O3_CoFe-PB, https://
doi.org/10.19061/iochem-bd-1-51.

Manuscript received: August 15, 2017
Revised manuscript received: September 29, 2017
Accepted manuscript online: October 2, 2017
Version of record online: November 7, 2017

4560

⌫ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim