Nanostructured Fe2O3 electrodes for solar-induced water splitting
Hen Dotan, advised by Prof. Avner Rothschild, Department of Materials Science and Engineering,
Technion
Society needs clean and renewable fuels to replace fossil fuels. One of the most promising alternative
fuels is hydrogen. The main barriers in the transition to hydrogen economy are the lack of direct hydrogen
sources and the lack of effective means for hydrogen storage and transportation. Solar-induced water photoelectrolysis is one of the most promising technologies for hydrogen production from renewable and abundant
resources, water and sunlight. My research aims at understanding the fundamental limitations and improving
the performance of semiconductor photoelectrodes for water photo-electrolysis. Using photovoltaic (PV) cells
to drive water electrolysis by an electrolyzer is technically possible but too expensive due to the high cost of
commercial electrolyzers. A more feasible approach is combining the electrolyzer and PV cell together into a
monolithic cell and at the same time replacing the expensive materials used in the electrolyzer (e.g., Pt- and
Ir-loaded electrodes and nafion membrane in PEM electrolyzers) with inexpensive ones. This can be done
using semiconductor photoelectrodes based on water-stable metal oxides such as iron oxide (Fe2O3, obtained
from rust) coupled in tandem (i.e., in series) with conventional PV cells, wherein the counter electrode can be
made of graphite. Photoelectrodes for water photo-electrolysis should meet quite a few requirement: they
should be stable in aqueous solutions under conditions of water oxidation (photoanodes) or reduction
(photocathodes); they should have a favorable band gap energy of about 2 eV to absorb a large enough
fraction of the solar spectrum while at the same time produce high enough photovoltage for driving the water
splitting reaction; and they should be made of non-toxic and inexpensive materials. These complicated
requirements make material selection and design the enabling key for solar-induced photo-electrolysis as a
feasible route for hydrogen productions.
Hematite (α-Fe2O3) is one of the few materials which meet thus complicated requirements; however
its transport properties are poor and the technological benchmark (10%) was not meet yet. Hematite band gap
energy is 2.1 eV, this implies that hematite photoanodes can theoretically get as high as 15.5% if they are to
absorb every photon below the absorption edge (590 nm) while have a quantum efficiency of 100%. This
means that every absorbed photon below the absorption edge should contribute to the water decomposition
reaction without letting the photogenerated carriers get lost by recombination processes. Recombination
processes occur in the bulk and at the surface of the photoelectrode. Nanostructuring plays a critical role in
mitigating bulk recombination losses by reducing the length that the photogenerated (minority) carriers must
cross to reach the electrode/electrolyte interface. Given the small diffusion length of minority carriers (holes)
in hematite photoanodes, estimated to be in the range of several nm to 20 nm, the critical dimensions (grain
size, nanowire diameter, film thickness) must be smaller the 50 nm at most. Surface recombination accrues at
the electrolyte/hematite interface and the hematite/substrate interface. The electrolyte/hematite recombination
can be reduced by two means: catalysis which increase the injection rate to the electrolyte, or surface
passivation which reduce surface traps concentration. The hematite/substrate recombination can be reduced
by adding an underlayer (under the hematite above the substrate) which allows only electrons to pass from the
hematite to the substrate.
In the last decade much progress has been achieved by producing nanostructured hematite
photoanodes and improving their water photo-oxidation rate using nano-catalysts. Most of the progress has
been achieved largely by trial and error, and we are still lacking detailed understanding of the fundamental
limitations that would ultimately set the limit in solar to hydrogen conversion efficiency using hematite
photoanodes. Likewise, quantitative understanding of the complicated relationship between material
properties such as doping level and transport properties, microstructure and nanostructure, and
photoelectrochemical performance is still lacking. This research aims to rectify these deficiencies and feel up
these gaps. This knowledge will then be used to design and fabricate highly efficient hematite photoanodes,
aiming at meeting the 10% efficiency benchmark. To achieve these ambitious goals we are working on both
the theoretical and experimental fronts in close collaboration with Prof. Michael Grätzel of EPFL and other
leading experts in photo-electrochemistry from TUD, UW, EMPA, UPorto, UiO and Eni.
The low quantum efficiency of Fe2O3 electrodes is typically attributed to the low rate constant for
water oxidation1 and the short (≤ 20 nm) diffusion length of the photogenerated minority carriers.2 One of our
first goals was to identify and understand the rate limiting step. During an exchange visit at EPFL in 2009 (2
months) we developed a new electrochemical characterization method that distinguishes between bulk and
surface recombination processes using H2O2 as a hole scavenger.3 Comparison of the photocurrent obtained
for the same electrode under the same conditions using electrolytes with or without H2O2 allows the extent of
electrolyte/hematite recombination to be quantified. The power of such an analysis is that it able to quantify
the electrolyte/hematite recombination as a function of potential for hematite photoanodes with different
morphologies as demonstrated in Figure 1.
C
D
Figure 1. SEM micrographs of nanostructured Fe2O3 photoanodes produced (at EPFL) by APCVD (A) and USP (B). Yield of charge separation (C) and
injection (D) of the APCVD and USP Fe2O3 electrodes as function of applied potential in 1 M NaOH solution.
By using this method we found that at low anodic potentials the rate constant for water oxidation is
very low and limits the photocurrent, but at high potentials it increases substantially and the injection yield
reached more than 90%. This method showed that the bottleneck for water photoelectrolysis with state of the
art nanostructured hematite photoanodes is bulk recombination.
On my second exchange visit at EPFL in 2011 (ten months), we significantly improved the method.
This enabled us to dig deeper into the complexity of the solar-induced water photoelectrolysis by comparing
the photoelectrochemical performance of the hematite photoanode in contact with redox couple (ferricyanide)
to the decomposition of H2O2 and water4. By comparing the electrode behavior in the presence of these redox
couples we are able to identify the illusive processes underlying the photoelectrochemical reaction at the
electrode/electrolyte interface, such as electron injection to the electrolyte, and quantify different potential
losses arising from Fermi level pinning and other effects. This method provides new insights important for
understanding the device performance. We applied this method to identify the so-called dead layer effect at
the hematite/substrate (FTO) interface. We found that electron injection from the substrate to the hematite and
the electrolyte are responsible for at least 20% loss in efficiency for thick hematite photoanodes and up to
100% for very thin films (less than 15 nm)5. We were able to significantly reduce thus losses by introducing a
very thin niobium oxide layer between the hematite and the substrate and achieving the highest APCE
(Absorbed Photon to Current Efficiency) ever achieved by hematite. In addition to the practical uses of this
new analysis, it also provides an experimental method to verify our complete theoretical model for
photoelectrochemical water photo-electrolysis using stable semiconductors and sun light. The formulation and
verification of this model is now in progress. We believe this model will provide the needed pathway for
understanding the relation between the hematite properties and its photoelectrochemical performance.
In order to reduce the bulk recombination loss we devised a new strategy that employs ultrathin films
with a dense (non-porous) microstructure designed as light trapping structures. This innovative approach
enabled us to break the record for hematite photoanodes and got us an article in Nature Materials.6 At film
thickness of few tens of nm the separation yield is quite high (as high as 60%, see Figure 2a) and the surface
recombination loss is small compared to nanostructured porous electrodes with high surface area.
Furthermore, the use of dense films produced by pulsed laser deposition (PLD) enables us to precisely control
important parameters such as doping level and interface structure that are not controllable using APCVD,
spray pyrolysis, and other chemical deposition methods used for fabricating nanostructured porous electrodes.
There are two main problems with ultrathin film photoelectrodes: the electron injection from the substrate to
the hematite and electrolyte and the low optical density of the films. The first was significantly improved by
introducing the niobium oxide intermediate layer, but the last required a unique design. This design is based
on an optical design for light trapping in ultrathin films using interference effects.6 This design was modelled
and tested with thin films deposited (by PLD) on platinized silica wafers and on conducting glass substrates
for comparison. Photographs of these electrodes are shown in Figure 2(b), and the measured photocurrent
density is shown in Figure 2(c). As predicted by our model calculations (depicted in lines in figure 2c) the
photocurrent density is significantly improved by this optical design, but the high absorption of the platinum
substrate had limit the photocurrent to only 1.5 mA/cm2. In order to achieve a new efficiency record the
platinum substrate was replaced with highly reflective silver-gold alloy substrate. In addition, a light
retrapping V-shape design was used and new efficiency record of 4 mA/cm2 was achieved. Furthermore, our
model calculations suggest that much higher efficiency is possible with addition of over and under layers.
This remarkable improvement is enabled by inventing a novel means for light trapping in ultrathin films, an
invention which we are now patenting.
(a)
(b)
(c)
Figure 2. (a) The separation yield as a function of thickness of thin film Fe2O3 photoelectrodes produced by PLD. (b) Pictures of the photoelectrodes (on platinized silica and TEC15, substrates). (c) Photocurrent density as a function of film thickness for Fe2O3 thin films deposited by PLD
on conducting glass (TEC15) substrates (blue dots) or platinized silica wafers (black squares). Calculated photocurrent densities for Fe2O3 electrodes
deposited platinized silica are shown in curves.
We invented new experimental methods to analyze the photoelectrochemical water splitting process.
Using these experimental methods we identified and quantified the different loss mechanisms in this
complicated process. Once the loss mechanisms were identified we figured out pathways to overcome them.
An extremely thin hematite photoanodes (less than 30 nm thick) provides a way to reduce bulk recombination
while using reflective substrates overcomes the absorption problem. The interface and surface recombination
losses are mitigated by adding underlayers and catalysts. We believe that reducing these recombination losses
using an ultrathin hematite photoanode with an optimal optical design, improved interface design and
catalysts will enable us to outperform the incremental improvements that have been achieved over the past
decade, reaching the elusive 10% soalr to hydrogen conversion efficiency benchmark.
1
Dare-Edwards, M. P.; Goodenough, J. B.; Hamnett, A.; Trevellick, P. R. Journal of the Chemical Society,
Faraday Transactions 1: Physical Chemistry in Condensed Phases 1983, 79, 2027.
2
Kennedy, J. H.; Karl W. Frese, J. Journal of The Electrochemical Society 1978, 125, 709.
3
Hen Dotan, Kevin Sivula, Michael Grätzel, Avner Rothschild and Scott C. Warren, Energy Environ. Sci.,
2011, 4, 958-964.
4
Hen Dotan, Nripan Mathews, Takashi Hisatomi, Avner Rothschild, Michael Grätzel (in preparation)
5
Takashi Hisatomi, Hen Dotan, Morgan Mertens Stefik, Kevin Sivula, Avner Rothschild, Subodh Mhaisalkar,
Michaël Grätzela and Nripan Mathews Advanced Materials 24, 2699-2702 (2012).
6
Dotan, H., O. Kfir, et al. (2012). "Resonant light trapping in ultrathin films for water splitting." Nat Mater
advance online publication.