Heteroepitaxial hematite photoanodes as a model system for solar water splitting

Heteroepitaxial multilayer Pt(111)/Fe2O3(0001) films were deposited on sapphire c-plane (0001) substrates by RF magnetron sputtering and pulsed laser deposition, respectively. The films were highly crystalline, displaying an in-plane mosaic of less than 1° and a homogenous surface morphology with roughness of ~3 Å. Ellipsometry and UV-Vis spectroscopy measurements were shown to be in excellent agreement with modelling, demonstrating that the optics of the system including absorption in the hematite layer are well described. For polycrystalline hematite photoanodes deposited on platinum, full characterization of the system is hampered by the inability to make measurements in alkaline electrolyte containing hydrogen peroxide (H2O2) due to spontaneous decomposition of H2O2 by the exposed platinum. The pin-hole free high quality of the heteroepitaxial films is demonstrated by the ability to make stable and reproducible measurements in H2O2 containing electrolyte allowing for accurate extraction of charge separation and injection efficiency. The combination of excellent crystalline quality in addition to the well characterized optics and electrochemical properties of the heteroepitaxial hematite photoanodes demonstrate that Al2O3(0001)/Pt(111)/Fe2O3(0001) is a powerful model system for systematic investigation into solar water splitting photoanodes.


Introduction
Hematite (α-Fe2O3) is a leading candidate for use as a photoanode in photoelectrochemical (PEC) cells due to its stability in aqueous solutions 1 , suitable bandgap energy 2 , and low cost 3 .However, the poor charge carrier mobility 4 and short life time 5 of photogenerated charge carriers combined with the low optical density results in significant bulk recombination and a charge collection length of 2-20 nm. 6One approach to mitigate this issue is by use of nanostructured photoelectrodes that decouple the light absorption pathway from the electrical transport pathway 7 .Another approach is to use thin films on metallic back reflector substrates in a resonant light trapping device structure. 8Despite these advances, state of the art photonoades 8,9,10 are limited to photogenerated current densities of approximately 4 mA/cm 2 .Significant improvements are needed to achieve higher efficiencies sufficient for commercialization.
Paving the way towards high efficiencies requires rational design by systematic investigation into hematite material properties and device performance.Use of single crystalline heteroepitaxial films with low defect concentration provide an ideal template for these studies.Since the substrate must be sufficiently conductive for use as a bottom contact, the most obvious choice for such a study is Pt(111) due to its close lattice match with α-Fe2O3 and thermal and chemical stability under oxidizing high temperature environments required for growth of high quality crystalline hematite.Indeed, heteroepitaxial deposition of hematite on single crystalline Pt(111) by molecular beam epitaxy (MBE) has been studied in detail 11,12,13 and one report has investigated the growth of a multilayer structure with successive heteroepitaxial growth of an α-Fe2O3 film on a Pt(111) buffer layer deposited on basal plane (0001) sapphire. 14Recently, Rioult et al. investigated the photoelectrochemical (PEC) properties of heteroepitaxial hematite films deposited on single crystal Pt(111) substrates and showed that the ability to deposit hematite films with the same crystalline structure upon dopant incorporation or thickness modification allows disentanglement of the intrinsic transport properties from overall mingled properties by exclusion of morphological and crystalline defect contributions. 15Epitaxial growth and electrochemical characterization of the metastable bixbyite polymorph of iron oxide has also been reported. 16 addition to crystallographic reproducibility and control, a model photoelectrode system should have the ability to be well characterized by a number of standard materials, optical, and electrochemical characterization techniques.Optical characterization is essential towards understanding how much light is absorbed in the hematite film and therefore, determining the maximum photocurrent that can be extracted.The use of a metallic reflective layer such as Pt as a back contact in place of a transparent electrode such as fluorinated tin oxide (FTO) slightly complicates optical characterization by preventing direct measurement of light absorption in the hematite film due to absorption in the platinum.However, it has been shown that the absorption and optics in such a system can be well described by optical models as long as the optical constants and layer thicknesses of the materials in the film stack are well defined. 8The use of homogenous heteroepitaxial films allows for precise extraction of the layer thickness and optical constants.
It is also important that a full collection of standard electrochemical characterization techniques should be performed on the model system.One challenge towards the use of platinum as a bottom contact for a photoelectrode is for measurements made with hydrogen peroxide (H2O2) containing electrolyte.When added to alkaline solution, H2O2 acts as a hole scavenger, eliminating the barrier for minority charge carrier injection at the semiconductor surface/liquid junction. 17As a result, this measurement is critical for calculation of both the charge separation and collection efficiency as well charge injection efficiency of hematite photoanodes.Under an applied bias, significant dark current is observed when Pt is exposed to electrolyte containing H2O2 due to spontaneous H2O2 decomposition.For polycrystalline hematite deposited on platinized silicon wafers (Si/SiO2/Pt), measurement with H2O2 was not possible due to unstable dark current behavior as a result of micro-pores and pinholes. 8Therefore, a high quality film devoid of pinholes is necessary for proper measurement of PEC properties in H2O2 containing electrolyte for a working electrode that uses Pt as a bottom contact.
Herein, we report a method for growth of heteroepitaxial hematite photoanodes with high crystalline quality by first using RF magnetron sputtering to deposit a heteroepitaxial platinum layer that serves as both a bottom contact and a back reflector on top of (0001) basal plane sapphire followed by growth of the hematite layer by pulsed laser deposition (PLD).The use of a heteroepitaxial Pt buffer layer eliminates the need for expensive single crystal platinum substrates.Additionally, PLD offers several advantages for the fabrication of high quality thin film photoanodes.The high rate of ablation results in stoichiometry transfer between the target and film, allowing for the study of different dopants or multilayer films simply by changing the target.The high energy ions resulting from the ablation allows growth of high quality epitaxial films at lower substrate temperatures than other deposition methods.In this work, a collection of standard optical and photoelectrochemical measurements, including stable measurement in H2O2 containing electrolyte, were performed and shown to be consistent with calculation, demonstrating that heteroepitaxial hematite deposited by PLD on platinized sapphire is a model system for investigation of hematite PEC properties.

Experimental 2.1 Fabrication of sapphire/Pt/ hematite film stack
The hematite films were deposited on sapphire c-plane (0001) oriented substrates and platinized sapphire c-plane substrates.Prior to deposition, the sapphire wafers were ultrasonically cleaned with soap, acetone, ethanol, and deionized water followed by dipping in Piranha solution (3 H2SO4:1 H2O2 by volume) and deionized water.The samples were then loaded into the vacuum chamber (PLD/MBE 2100, PVD Products), and pumped to a base pressure of 1 × 10 −7 Torr.The platinum film was deposited via RF magnetron sputtering from a 50 mm diameter target of pure (99.99%)Pt (Birmingham Metal).Two recipes were used for the deposition of platinum films.In both cases, the deposition was performed under 5 mTorr Ar pressure, 30 W forward power, and source to substrate distance of 75 mm.The depositions were performed at set-point temperatures of 500 °C and 700 °C.The films deposited at 500 °C were then subjected to a 2 h anneal at 900 °C under 5 mTorr Ar directly after deposition.The deposition rates were approximately 0.5 Å / s for the film grown at 500 °C.
Subsequent to platinum deposition and annealing, hematite films were deposited by pulsed laser deposition (PLD) from a 1 cation% Ti-doped Fe2O3 target.The hematite films were deposited using a PLD system equipped with a KrF (λ = 248 nm) excimer laser (COMPexPro 102, Coherent, GmbH).The films were deposited at a set-point temperature of 700 °C with a laser fluence of approximately 1.1 J / cm 2 , repetition rate of 3 Hz, source to substrate distance of 75 mm, and oxygen partial pressure of 10 mTorr.Hematite films deposited directly on sapphire were grown with identical conditions but at a set-point temperature of 800 °C.

Materials Characterization
The crystalline quality of the films were examined via x-ray diffraction (XRD; Rigaku Smartlab) using Cu Kα radiation, parallel beam optics, and a 2-bounce Ge(220) channel-cut monochromator.All scans were performed after precise alignment of the samples using the sapphire (0006) rocking curve scans.High resolution (HRXRD) θ-2θ scans were performed in standard geometry.Off axis phi scans were performed in skew-symmetric geometry to determine the sample in-plane alignment.The surface morphology of the films was characterized by atomic force microscopy (AFM, XE-100, PSIA Corporation) performed in tapping mode.The film thickness of a selected specimen was measured by cross-section TEM.The TEM lamella was prepared by focused ion beam (FIB) etching technique on a Strata 400S (FEI) with a gallium liquid metal ion source, a gas injection system and a micromanipulator (Omniprobe 200).After electron beam deposition of carbon and platinum, the film was protected by a platinum cap.The TEM lamella was cut free with trenches from both sides.After the lift-out was performed, the lamella was polished to ion transparency with currents down to 28 pA at 30 kV.Amorphization was diminished by low kV showering for several seconds at 2 kV.The thickness of the film was measured by cross-section TEM using a monochromated and aberration (image) corrected TEM (FEI Titan 80-300 kV S/TEM).

Optical Characterization
The film thickness and optical constants of the hematite films were measured by spectroscopic ellipsometry (VASE Ellipsometer, J. A. Woolam Co., USA).The amplitude and the phase were measured from 1200 to 300 nm at incidence angles of 65, 70, and 75°.VASE software was then used to fit the results and extract the optical constants.The transmittance (T) and reflectance (R) spectra of the hematite photoanodes were measured using a Perkin Elmer Lambda 950 UV/VIS spectrometer from 1200 to 300 nm.An integrating sphere was used to account for the scattered light.Measurements were performed for photoanodes deposited on both single side and double side polished sapphire.

Photoelectrochemical characterization
Photoelectrochemical measurements were carried out, for the most part, in "Cappuccino cells" electrochemical test cells equipped with a potentiostat (CompactStat, Ivium Technologies) as described elsewhere. 18All measurements were carried out in 1 M NaOH aqueous solution (pH = 13.6).The current was measured as a function of the electrode potential using a 3-electrode setup with an Ag/AgCl reference electrode and a platinum wire counter electrode with a scan rate of 10 mV / s.The applied potential was converted to the RHE scale using the Nernst equation.The current density was obtained by dividing the measured current by the electrode area immersed in the electrolyte (0.28 cm 2 ).This area is also commensurate with the illuminated area of the electrode upon exposure to light.Select measurements were also made in 1 M NaOH + 0.5 M H2O2 solution.A solar simulator (Sun 3000 class AAA solar simulator, ABET Technologies, AM 1.5G) was used for all measurements.
Capacitance-voltage measurements were performed with the same 3 electrode set up described above to obtain the Mott-Schottky plots.Measurements were taken at a number of frequencies from 50,000 Hz to 100 Hz and from 0.6 V vs RHE to 1.6 V Vs RHE.The flat band potential and carrier concentration were calculated as the average values extracted from multiple measurements taken between 260 and 1761 Hz.IPCE measurements were carried out in a three-electrode system with an Ag/AgCl counter electrode and a Pt counter electrode.A Zahner potentiostat was used with continuous light from a 1000 W Xenon lamp.The light from the lamp passed through a monochromator (Oriel) for IPCE measurements as a function of wavelength.The measurements were made in 1 M NaOH solution.

Results and Discussion
Figure 1a shows the θ-2θ scan for a 32 nm thick hematite film deposited on the annealed Pt layer.The Laue oscillations (shown in  The in-plane alignment of the film stack is investigated by azimuthal φ-scans of the off axis peaks as shown in Figure 1c and verifies the epitaxial growth of the layers.The observation of 6-fold symmetry in the Pt(200) φ-scan indicates twinning of the Pt layer, consistent with other reports of vapor deposited Pt thin films deposited on c-plane sapphire. 19,20The Pt(200) reflections are rotated 30° with respect to the Al2O3 (104) reflections resulting in the orientational relationship Pt(111) || Al2O3(0001) and Pt[110] || Al2O3 [10-10] on the sapphire substrate.We observe twin domains in the hematite (-Fe2O3) film from the hematite (104) phi scan, possibly as a result of the Pt(111) layer twinning.However, twinned hematite has also been observed for films deposited on untwinned Pt single crystal substrates by MBE. 21For comparison, hematite films grown directly on sapphire produce a single isostructural orientational relationship.The full width half maximum (FWHM) of the peaks in the φ-scan are < 1°, demonstrating the excellent in-plane mosaic spread of the hematite films.Similar mosaic structure is observed for films deposited in the hematite thickness range of 10 -40 nm, indicating that the film quality remains unaffected in this thickness range.A cross sectional TEM image of a 21 nm thick heteroepitaxial hematite film grown on the annealed Pt layer is shown in Figure 1d.The thickness (61 nm) of the platinum layer is in agreement with the film thickness calculated from the Laue oscillations.
The microstructure of the hematite films was investigated using AFM.Initial attempts at growing the Pt layer at high temperature (700 -800 °C) and without an annealing process resulted in film with a high quantity of pores presumably as a result of incomplete coverage due to the formation of two different epitaxial island populations which do not fully coalesce.Further annealing of the Pt electrode or subsequent deposition of hematite resulted in enlarged pores, possibly related to solid state dewetting observed for Pt films deposited on sapphire 22 as shown Figure 2a.The depth of the pores measured by AFM were approximately 3 -4 nm, but very likely penetrated to the Pt layer as will be shown by the H2O2 measurements described below.By growing the films at lower temperature of 500 °C, and subsequently annealing at 900 °C, pore formation was completely suppressed and the platinum films possessed RMS roughness of ~3 Å.The Pt layer provided a high quality template for the subsequent growth of the hematite films which also showed RMS roughness of ~3 Å and homogenous morphology (Figure 2b).The high temperature anneal and subsequent deposition of the hematite films at 700 °C in 10 mTorr of O2 did not result in dewetting of the Pt film or in any kind of pore formation or roughening.Figures 3a and 3b show the measured reflectance data of the heteroepitaxial film stack deposited on single-side-polished and double-side-polished sapphire, respectively.The total absorption is given by A = 1 -R -T.The total absorption comprises contributions for the platinum and hematite films.Due to interference effects, the net absorption in the hematite film cannot be extracted directly from these measurements and must instead be calculated using an optical model described by Dotan et al 8 which requires knowledge of the layer thicknesses and optical constants.For a 21 nm thick heteroepitaxial hematite film deposited on the annealed Pt layer, the thickness value was extracted by ellipsometry and independently verified by TEM cross section, as shown in Figure 1d, suggesting proper extraction of hematite and Pt optical constants.
Employing a hematite thickness of 32 nm and a platinum thickness of 61 nm in the model calculations, a good fit to the reflection data is obtained on the single-side-polished sapphire as shown in Figure 3a.Due to slight transmission (0.5%) through the Pt layer, scattering from the rough back sapphire surface slightly distorts the measurement and there is deviation from the model at low wavelengths and around the band edge.By depositing a 25 nm thick hematite film on double-side-polished sapphire, an excellent fit to the data is obtained (Figure 3b), confirming that the optics of the system can be well modeled.Based on the net absorption in the film, we calculate a maximum photogenerated current density of 6.55 mA/cm 2 for the 32 nm thick film on platinum.The optical losses plotted in the figure are defined as absorption anywhere outside the hematite layer and consist primarily of the absorption in the platinum film as well as the light that is transmitted through the full stack.The PEC characteristics of the 32 nm thick hematite photoanode was measured in 1M NaOH solution by current density vs. potential (J-U) measurements shown in Figure 4.In Figure 4a, Jlight is the measured current density under illumination, Jdark is the dark current density, and the photocurrent is given by Jphoto = Jlight -Jdark.In Figure 4b, Jphoto is plotted as a function of the photovoltage (Vphoto) where Vphoto is calculated as described by Dotan et al. 23 , giving the photovoltaic performance of the device.Figure 4c yields the intrinsic photovoltaic power as a function of the photovoltage.The intrinsic photovoltaic power reaches a maximum at a value of 0.19 mW / cm 2 at an applied potential of 1.5 V vs RHE, corresponding to a power conversion efficiency of 0.19% from solar simulated radiation power of 100 mW / cm 2 to electric power of 0.19 mW/cm 2 .The intrinsic solar to chemical (ISTC) conversion efficiency of the device is plotted in Figure 4d.The maximum ISTC reaches 0.13% at a photocurrent density of 0.56 mA / cm 2 and potential of 1.54 V vs RHE.The effect of the microstructural quality on the measurements in 1M NaOH + 0.5 M H2O2 electrolyte is demonstrated in Figure 5a which shows the PEC characteristics of a 21 nm thick hematite film deposited on the high temperature as deposited (without anneal) porous Pt layer.Significant and non-stable dark currents do not allow for reliable extraction of Jphoto as shown by the mismatch of the current under chopped light (Jchopped) measurements with the Jdark and Jlight measurements.This suggests that the electrolyte is in contact with the Pt as a result of the pore formation in the unannealed Pt layers shown in Figure 2a.However, when the hematite is deposited on the annealed epitaxial Pt electrode, the dark current is significantly reduced and stable as shown by the excellent agreement between Jchopped, Jlight and Jdark, allowing for photocurrent extraction in H2O2 containing electrolyte solutions, giving further evidence to the suppression of pores and pinholes.The dark current further reduces with the increasing thickness of the hematite layer, suggesting better surface coverage.For a 32 nm thick hematite film, the dark current is negligible across a potential range of 0.6 to 1.5 V vs RHE.Steady state chronoamperometry measurements were made on the 32 nm thick hematite film from 0.6 to 1.7 V vs RHE in the H2O2 containing electrolyte to reduce any further instability resulting in an excellent match between the Jphoto plateau for both NaOH and NaOH + H2O2 electrolyte solutions as shown in Figure 5b.Reliable extraction of the photocurrent in H2O2 containing electrolyte solution allows calculation of the charge injection and separation efficiencies of the hematite photoanodes which are plotted in Figure 5c.The charge separation and collection efficiency of the hematite photoanodes is calculated by  ℎ  2  2 /  where  ℎ  2  2 is the photocurrent in H2O2 solution.  was taken to be 6.55 mA/cm 2 as calculated from the optical data shown in Figure 3.The charge separation and collection efficiency spans from approximately 0% to 12% over a potential range of 0.6 to 1.7 V vs RHE.The charge injection efficiency of the hematite photoanode is given by  ℎ  2  /  ℎ  2  2 where  ℎ  2  is the photocurrent in 1M NaOH solution.
Below 1.2 V vs RHE, charge injection is limited by surface recombination.Above 1.2 V vs RHE, the charge injection efficiency increases until 1.7 V vs RHE, where all charge that reaches the interface is injected into the solution.For comparison, charge separation and injection efficiencies are shown for a 21 nm thick film in Figure 4d.Although this film absorbs less light with a maximum possible photogenerated current density of 5.85 mA/cm 2 , the total photocurrent and charge separation and injection efficiencies are improved.Nevertheless, since the stability range for the H2O2 measurement with this film thickness is only between 0.8 V to 1.5 V vs RHE, the 32 nm film is better suited as a model electrode for fundamental investigation, allowing H2O2 measurement across a larger potential range.Figure 6 shows the IPCE measurements of the 32 nm thick hematite photoelectrode taken at potential of 1.6 V vs RHE.The measured IPCE reaches a maximum of approximately 10% at the shortest wavelength of 350 nm.The solar spectrum of the AAA solar simulator used in the measurement is multiplied by the IPCE to yield the number of photons which contribute to water oxidation.The IPCE was then integrated over the spectrum to give an integrated current density value of 0.64 mA / cm 2 consistent with the value of 0.67 mA / cm 2 measured at an applied potential of 1.6 V vs RHE under solar simulated light (Figure 4).Capacitance-potential (C-U) measurements were performed on the 32 nm thick heteroepitaxial hematite photoanode in the dark and the extracted Mott-Schottky plot taken at frequency of 1761 Hz is shown in Figure 7.The results were fitted to the Mott-Schottky equation using a dielectric constant 24 of 30 with a correction term added to account for the capacitance of the Helmholtz layer. 25Due to the low surface roughness (3 Å) of the hematite films, accurate measurement of the device area is possible and an area correction factor is not needed.The flat band potential and carrier concentration were extracted from multiple measurements taken between 260 and 1761 Hz, yielding 0.20 ± 0.01 V vs RHE and (5.5 ± 0.3) x 10 20 cm -3 , respectively.While this flat band potential is lower than most other reports for polycrystalline hematite photoanodes 26 , it is similar to the flat band potential reported for single crystal Zr-doped hematite. 27Despite this favorable low flat band potential, the onset potential of the hematite photoanodes is rather high, approximately 1.2 V vs RHE, for reasons that are not quite clear yet.Future work will require increasing both the charge separation and collection efficiency towards higher photocurrents and increasing the charge injection efficiency to reduce the overpotential for water oxidation.

Conclusions
Heteroepitaxial hematite photoanodes with excellent crystalline quality have been deposited on platinized sapphire by a combination of RF magnetron sputtering pulsed laser deposition.Both the platinum and hematite layers are rotationally twinned with an in-plane mosaic of less than 1°.Pore formation in the layers was prevented by depositing the Pt layer at lower temperature and subsequently annealing at high temperature.This resulted in pin-hole free films as evidenced by the ability to make measurements in H2O2 containing electrolyte.A number of photoelectrochemical and optical characterization measurements were performed on the hematite photoanodes.The electrochemical measurements were all consistent with one another and the optics of the system could be well described by calculation, showing that Al2O3(0001)/Pt(111)/-Fe2O3(0001) heteroepitaxial photoanodes are a model system for investigation of hematite properties.
Figure 1a shows the θ-2θ scan for a 32 nm thick hematite film deposited on the annealed Pt layer.The Laue oscillations (shown in Fig 1b) up to the 15 th order around the Pt(111) Bragg reflection indicate the excellent crystallinity of the Pt layer and flatness of the ordered interface.

Figure 2 .
Figure 2. AFM images of 32 nm thick hematite film deposited at 700 C on (a) the Pt layer grown at 700 C and (b) the Pt layer grown at 500 C and subsequently annealed at 900 C.The scan size is 10 x 10 µm.

Figure 4 .
Figure 4. (a) Current density vs. Potential (J-U) measurements in 1M NaOH of a 32 nm thick hematite film deposited on annealed Pt layer.(b) The photocurrent as a function of the photovoltage.(c) The intrinsic photovoltaic power as a function of the photovoltage.(d) The ISTC efficiency of the photoanode as a function of the photocurrent density.

Figure 5 .
Figure 5. (a) J-U measurements in 1M NaOH + 0.5 M H2O2 for hematite films deposited on asdeposited and annealed Pt films.(b) Comparison of Jphoto measured in 1M NaOH and 1M NaOH + 0.5 M H2O2 for 32 nm thick hematite film deposited on annealed Pt film.(c) Charge injection and separation efficiency of a 32 nm thick hematite film deposited on annealed Pt film.(d) Charge injection and separation efficiency of a 21 nm thick hematite film deposited on annealed Pt film.

Figure 6 .
Figure 6.IPCE of 32 nm thick hematite film deposited on annealed Pt layer at an applied potential of 1.6 V vs RHE and the AAA solar simulator spectrum plotted vs wavelength.The total integrated photocurrent as a function of wavelength is shown.The IPCE and photon flux were multiplied together to show the number of photons which contribute to water oxidation and is denoted by the shaded region.

Figure 7 .
Figure 7. Mott Schottky plot taken at 1761 Hz with linear fit to the data for 32 nm thick heteroepitaxial Fe2O3 film deposited on annealed Pt film.