Hydrogen induced interface engineering in Fe2O3–TiO2 heterostructures for efficient charge separation for solar-driven water oxidation in photoelectrochemical cells

Semiconductor heterostructure junctions are known to improve the water oxidation performance in photoelectrochemical (PEC) cells. Depending on the semiconductor materials involved, different kinds of junctions can appear, for instance, type II band alignment where the conduction and valence bands of the semiconductor materials are staggered with respect to each other. This band alignment allows for a charge separation of the photogenerated electron–hole pairs, where the holes will go from low-to-high valance band levels and vice versa for the electrons. For this reason, interface engineering has attracted intensive attention in recent years. In this work, a simplified model of the Fe2O3–TiO2 heterostructure was investigated via first-principles calculations. The results show that Fe2O3–TiO2 produces a type I band alignment in the heterojunction, which is detrimental to the water oxidation reaction. However, the results also show that interstitial hydrogens are energetically allowed in TiO2 and that they introduce states above the valance band, which can assist in the transfer of holes through the TiO2 layer. In response, well-defined planar Fe2O3–TiO2 heterostructures were manufactured, and measurements confirm the formation of a type I band alignment in the case of Fe2O3–TiO2, with very low photocurrent density as a result. However, once TiO2 was subjected to hydrogen treatment, there was a nine times higher photocurrent density at 1.50 V vs. the reversible hydrogen electrode under 1 sun illumination as compared to the original heterostructured photoanode. Via optical absorption, XPS analysis, and (photo)electrochemical measurements, it is clear that hydrogen treated TiO2 results in a type II band alignment in the Fe2O3–H:TiO2 heterostructure. This work is an example of how hydrogen doping in TiO2 can tailor the band alignment in TiO2–Fe2O3 heterostructures. As such, it provides valuable insights for the further development of similar material combinations.


Introduction
Solar energy-assisted splitting of water into its constituents, hydrogen and oxygen, in a photoelectrochemical (PEC) cell, represents a promising route to convert solar energy into more useful chemical fuels. 1,2 The water reduction process can produce hydrogen, where the required electrons can be generated via the water oxidation reaction at the surface of a photoanode through PEC reactions. However, the water oxidation reaction is a demanding electrochemical process, requiring an oxidatively robust and yet inexpensive semiconducting material as the photoanode. 3 Despite tremendous efforts, developing a highly active photoanode for water oxidation at low cost remains a signicant challenge. Aer the rst report on water splitting by Fujishima and Honda, 4 titanium dioxide (TiO 2 ) in its anatase phase has been extensively studied in solar-driven photocatalytic processes. [5][6][7] However, for practical applications, the large bandgap of TiO 2 ($3.2 eV) requires solar radiation with wavelength below 388 nm to create an electron-hole pair; thereby, limiting the overall efficiency of TiO 2 when illuminated under real Sun conditions. 8,9 Another material of high interest for PEC water splitting is hematite (a-Fe 2 O 3 ), due to its suitable band gap (2.1 eV), high stability over a wide range of pH and potentials, and low material cost. 10,11 In spite of these various advantages, the solar-to-hydrogen conversion efficiency of a-Fe 2 O 3 falls well below the theoretical maximum value (z12.9%) due to a number of factors such as (i) bulk charge recombination, (ii) interfacial carrier trapping and recombination, (iii) surface trapping and recombination, and (iv) improper band positions for unassisted water splitting 12 (see Fig. S1 † for further information).
To address the issues with solely a-Fe 2 O 3 and TiO 2 based oxide semiconductors and design a photoanode for PEC water oxidation with high solar-to-hydrogen conversion efficiency, the heterostructure of these two semiconductors has been recognized as an attractive candidate to enhance the photocurrents and lower the onset potential. 13,14 The combined properties of a-Fe 2 O 3 and TiO 2 allow the heterostructures to absorb a wider range of photons, thanks to the relatively narrow band gap of a-Fe 2 O 3 . 15 Also, the formation of heterostructures between a-Fe 2 O 3 and TiO 2 can allow band structure engineering to manipulate surface/interface properties for charge transfer/ separation, thereby enhancing the water oxidation performance. 16 In particular, the valence and conduction band alignment mechanisms at the interface are crucial for the separation of photogenerated charge carriers. 17 In both oxide semiconductors, the Fermi levels depend on the concentrations of the conduction electrons and hence on the oxygen vacancy concentrations. By appropriate adjustments of the Fermi levels via defect concentration, the valence band edge in the bulk of the semiconductor may be brought to a common equilibrium. Using the available data 18 of the electron affinities for Fe 2 O 3 (4.71 eV) and TiO 2 (4.33 eV), together with the generally accepted model of heterojunctions, the discontinuity at the conduction bands (CB) is estimated to be 0.38 eV while that for the valence bands (VB) is estimated to be 0.42 eV. Owing to the mutual positions of Fe 2 O 3 and TiO 2 conduction band edges, photogenerated electrons in TiO 2 can be easily transferred to Fe 2 O 3 and injected into the Indium Tin Oxide (ITO) substrate, and subsequently can migrate through the external electric circuit to reduce water at the cathode, thus suppressing detrimental recombination effects (Fig. 1a). On the other hand, the VB offset at the interface by 0.42 eV act as an energy barrier that blocks the hole transfer from the a-Fe 2 O 3 to the TiO 2 layer and prevents the water oxidation reaction on the surface of TiO 2 (Fig. 1a). This means the heterostructure between a-Fe 2 O 3 and TiO 2 forms a type I band alignment at the semiconductors' interface, and the hole transport from a-Fe 2 O 3 to TiO 2 is energetically impeded, as depicted in Fig. 1a and b.
However, in the case of a thin pours TiO 2 layer (<100 nm) over Fe 2 O 3 and at applied anodic potential, the depletion region and upward band bending in TiO 2 at Fe 2 O 3 -TiO 2 nanoheterostructure in contact with the electrolyte may extend into the a-Fe 2 O 3 . This high band bending in the thin TiO 2 overlayer, under the inuence of anodic potential, can provide a channel to transfer the photogenerated holes from Fe 2 O 3 to TiO 2 where they participate in the water oxidation reaction, thereby improving the water oxidation performance of the Fe 2 O 3 -TiO 2 photoanode. Barreca et al. 15 reported that the use of external potential could transfer of photogenerated holes from porous hematite to TiO 2 layers that can improve the PEC response under simulated light. However, in the case of thick and dense TiO 2 layer over a-Fe 2 O 3 , the mismatched band alignment can hinder the charge transfer process at Fe 2 O 3 -TiO 2 heterostructure, thereby constitute the limiting factor in PEC applications. Therefore, for swi transfer of photogenerated holes from VB of a-Fe 2 O 3 to the VB of TiO 2 , band edge and band gap tailoring in TiO 2 are required to form a type II band alignment at the Fe 2 O 3 -TiO 2 heterostructure interface. In our previous study, 8,9,17 we demonstrated that a high temperature hydrogen treatment in TiO 2 under partial pressure can modify the structural, optical, and electrical properties and signicantly improve the photocatalytic and photoelectrochemical performance. An upshi of the valence band of TiO 2 was achieved aer the high temperature hydrogen treatment at partial pressure that was useful for band edge engineering at TiO 2 -BiVO 4 heterostructure. However, a downshi of valance band maximum (VBM) was also obtained when TiO 2 thin lms were grown in situ in hydrogen plasma. 19 In this study, rst-principles calculations on a model Fe 2 O 3 -TiO 2 heterostructure show that the strain at the interface shis the valance and conduction band positions and affects the bandgap of the TiO 2 layer. Furthermore, the calculations show that neutral and positive charged H interstitials are energetically preferred under hydrogen treatment conditions. These interstitials introduce states in the bandgap of TiO 2 , which can assist in the transfer of holes through the TiO 2 layer. To challenge the theoretical predictions, well-dened planar Fe 2 O 3 -TiO 2 heterostructures were manufactured. The experimental results show that Fe 2 O 3 -TiO 2 initially forms a type I band alignment, prohibiting the transfer of holes through TiO 2 . However, aer hydrogen treatment, the Fe 2 O 3 -H:TiO 2 heterostructure seems to form a type II band alignment; thus, hole transfer becomes possible, leading to an enhanced PEC response.

Preparation of a-Fe 2 O 3 thin lms
The ultra-thin hematite lms were prepared onto indium-doped tin oxide (SnO 2 :In, ITO, PGO GmbH, sheet resistance < 20 U sq À1 ) substrate by using physical vapor deposition technique, followed by annealing of Fe coated ITO substrate in air atmosphere. 20 Briey, all the substrates were cleaned by using soap solution, acetone, and deionized water followed by ultrasonication in isopropanol for 5 minutes and nally with oxygen plasma for two minutes. Before deposition, 1/3 area (5 mm Â 10 mm) of the substrate was covered with a thermal tape for making the electrical contact later for PEC measurements. To fabricate the ultra-thin hematite lm (approximately 25 nm), a 10 nm thick Fe lms were deposited on ITO substrate by physical vapor deposition (PVD 225, Kurt J. Lesker, base pressure < 5 Â 10 À7 mbar) and annealed in air atmosphere at 350 C for 8 h with a heating rate of 4 C min À1 . The thickness of the Fe lm was measured in situ during deposition using a quartz-crystal microbalance monitor integrated in PVD system. Aer annealing, the samples were allowed to cool down to room temperature naturally.

Preparation of Fe 2 O 3 -TiO 2 and Fe 2 O 3 -H:TiO 2 heterostructures
Thin lms of TiO 2 were deposited onto ITO substrates and also on hematite coated ITO substrate by sol-gel technique. 6 In short, a transparent gel solution of titanium dioxide was prepared by mixing 3 ml titanium tetra-isopropoxide (TTIP, 97% pure) in 20 ml ethanol in the presence of diethanolamine. The solution was stirred for 4 h at room temperature to enhance the reaction rate between diethanolamine and TTIP, and nally, it was converted into a gel. This gel solution was applied on a-Fe 2 O 3 /ITO substrate and uniformly coated with the help of a spin coating unit at 3000 rpm for 1 minute. A thin layer of TiO 2 was achieved by the deposition of prepared gel over the a-Fe 2 O 3 / ITO substrates. Aer deposition, the prepared TiO 2 -Fe 2 O 3 heterostructure was allowed to dry for 10 min at 80 C and further annealed at 350 C for 4 h. The hydrogen treatment was carried out by annealing the Fe 2 O 3 -TiO 2 heterostructures at 300 C in 4% H 2 in Ar at atmospheric pressure for 6 h. We also prepared a hydrogen doped TiO 2 thin lm photoanode under the same annealing conditions to see the effect of hydrogen doping individually on optical, electrical, and PEC properties.

Material characterization
The chemical phase of the prepared samples was determined by using a confocal Raman microscope (alpha300 R; WITec) with a 488 nm laser pulse as an excitation source. The surface morphology of the bare and TiO 2 coated a-Fe 2 O 3 samples was examined by eld emission scanning electron microscope (FE-SEM) using a Zeiss Supra 60 VP microscope operated at an acceleration voltage of 10 kV. The optical absorption of all the samples was measured with the help of a Cary 5000 spectrophotometer (Varian). X-ray photoelectron spectroscopy (XPS) spectra were acquired in a PerkinElmer Phi 5500 setup (base pressure < 10 À10 mbar) using AlK a radiation of 1.4866 keV. The XPS spectra were shied using the Fe(2p 3/2 ) peak corresponding to 710.9 eV as a reference.

Photoelectrochemical measurements
For electrochemical measurements, thin lms of a-Fe 2 O 3 , TiO 2 , H:TiO 2 , Fe 2 O 3 -TiO 2 and Fe 2 O 3 -H:TiO 2 heterojunctions were converted into the photoelectrodes with an active surface area of about 0.50 cm 2 . All the (photo)electrochemical measurements, current-voltage (I-V), capacitance-voltage (C-V), and electrochemical impedance spectroscopy (EIS), were conducted in a three-electrode conguration using a H-type PEC cell made of glass and tted with a at optical quartz window containing 0.1 M NaOH as electrolyte (pH ¼ 12.9). The PEC cell was controlled by using the Gamry Ref. 600 potentiostat and a solar simulator (SKU SS150, Sciencetech Inc.) with an output intensity of 100 mW cm À2 as an illumination source. Here, the prepared photoelectrodes were used as a working electrode, Pt wire as a counter electrode, and Ag/AgCl as a reference electrode. For better representation of our results, the Ag/AgCl reference potential was converted into the reversible hydrogen electrode (RHE) potential by using the following formula: Ag/AgCl is the standard potential of Ag/AgCl at 25 C (0.1976 V vs. the standard hydrogen electrode, SHE). The cyclic voltammetry sweep scans in the potential range between 0 and 2.0 V RHE at a scan rate of 10 mV s À1 were performed to obtain the I-V characteristics under dark and illumination. The photocurrent density (J ph ) has been calculated by subtracting the dark current from the current measured under illumination and devised by the geometrical area of the photoanode. Electrochemical impedance spectroscopic (EIS) measurements under illumination were carried out in the frequency range 10 5 and 0.1 Hz at applied potentials between 1.23 V RHE . Nyquist plots obtained under illumination were tted using the soware EIS spectrum analyzer. Mott-Schottky analysis was performed at an applied frequency of 1 kHz in the dark condition in a potential window between 0 and 1.5 V RHE . The obtained Mott-Schottky curves (1/C 2 versus V RHE ) were used to determine the donor density (N D ) and at band potential (V  ) by using the Mott-Schottky equation: where 3 o is the permittivity of the vacuum, 3 s is the dielectric constant of the hematite, q is the electronic charge, and kT/q is the thermal voltage (26 meV at room temperature). The donor density was calculated using the equation, , from the slope of the linear region, between 0.6 and 1.0 V RHE , of Mott-Schottky plots.

Computational methods
The rst-principles calculations were performed using density functional theory (DFT) as implemented in the VASP package. [21][22][23] The interaction between the valance electrons and the core follows the projector augmented wave (PAW) method. 24 PAW potentials with the valence states 1s for H, 2s and 2p for O, 3d and 4s for Fe, and 3d and 4s for Ti have been employed. A plane wave basis with a kinetic energy cut-off 700 eV was used. To improve convergence, a Gaussian smearing broadening of the Fermi surface of 0.1 eV was employed. The exchangecorrelation (XC) interaction was treated at the level of the generalized gradient approximation (GGA) using the XCfunctional of Perdew, Burke, and Ernzerhof (PBE). 25 In the DFT+U calculations, the rotationally-invariant scheme proposed by Dudarev et al. 26 and a U À J ¼ 4.3 eV on Fe atom and U À J ¼ 5.2 eV on Ti atom are employed in all our calculations. 27,28 The HSE06 method, as implemented in VASP, was employed to study H defects in anatase TiO 2 . 29,30 A kinetic cutoff of 600 eV was applied in all calculations. The interface system is composed of a six-layer thick Fe 2 O 3 with a (0001) termination of p(2 Â 2), that is joined with a four-layer thick anatase TiO 2 with a (101) termination of p(1 Â 2). Owing to the lattice mismatch between Fe 2 O 3 and TiO 2 , the interface will introduce strain to the system. Here the lattice cell (a amounts to 10% compression strain on the a-axis and 0.6% compression strain to the b-axis, and an increase of the angle by 7.4%. During the relaxation of the interface system, the anatase TiO 2 reconstructs to amorphous TiO 2 (a-TiO 2 ). Here, the term amorphous is used to indicate that the reconstructed TiO 2 in the interface model can not be identied with any known phase (rutile, anatase, brookite, etc.) of TiO 2 . However, it should be noted that due to the periodic conditions, this phase should not be considered as genuinely amorphous TiO 2 . Defect calculations were carried out in a 2 ffiffiffi 2 p Â 2 ffiffiffi 2 p Â 1 (96 atoms) supercell of anatase TiO 2 , with a G-centered 1 Â 1 Â 1 k-point sampling. The effect of higher concentrations of defects is modeled by a ffiffiffi 2 p Â ffiffiffi 2 p Â 1 (24 atoms) supercell and a 1 Â 1 Â 1 (12 atoms) unit cell, using a 3 Â 3 Â 1 Monkhorst-Pack (MP) sampling. 31 All geometries were relaxed until the maximum force was less than 0.05 eV A À1 . The relative stability of the various defects in charged and/or neutral states is determined by the formation energy, where E t (D q ) is the total energy of the supercell containing a defect D in charge state q and E t (TiO 2 ) is the total energy of a perfect crystal in the same supercell. In the extreme O-rich limit, the m O is set to 1 2

Simulation results
The valence band offset (VBO) at the interface of Fe 2 O 3 and a-TiO 2 was calculated through the reference potential method originally introduced by Kleinman 32,33 where the reference potential use the macroscopically averaged electrostatic potential. 34 The calculated results are shown in Fig. 2. The VBO for the individual system, Fe 2 O 3 (0001) and a-TiO 2 thin lm, were also calculated and are shown in the ESI (Fig. S2). † The blue curve in Fig. 2   dz 00 w l 1 ðz À z 0 Þw l 2 ðz 0 À z 00 ÞV 0 ðz 00 Þ; where w l ðz À z 0 Þ ¼ , Q is the unit-step function, l 1 and l 2 are in the order of the (strained) thickness of hematite (0001) and amorphous TiO 2 thin lm along z, respectively.
where V 0 (z) is the xy-plane averaged electrostatic potential and d 1 and d 2 are the inter-planar distances along the z direction (normal to the interface). 35,36 The calculated V 0 (z) is the black line in Fig. 2. The VBO was calculated via, The calculated VBO for the individual system is À1.18 eV using DFT+U, which agrees well with the experimental observations ($0.5 eV). 38 However, the calculated VBO at the interface is around 1.36 eV, with the valence band edge of TiO 2 sitting below the Fe 2 O 3 , which is 0.18 eV higher than the separated systems indicating an interface dipole or double layer were created at the interface. The electrostatic potential across the interface is DV ¼ 1.65 eV.
We further studied H defects and O vacancy in anatase TiO 2 , including their formation energy and electronic structure. The formation energy is calculated using eqn (1), and the results are shown in Fig. 3a, which corresponds to the results from the modeled supercell of 96 atoms. It is clearly shown that the positively charged H interstitial is the most favorable (with negative formation energy) defect when the Fermi energy is located above the valence band edge and below 2.9 eV, while the H interstitial becomes stable when E f is larger than 2.9 eV (see Fig. 3). The formation energy of H interstitial is 0.04 eV. In contrast, O vacancies and substitutional H are less energetically favorable.
Next, we have studied the formation energies of different concentrations of defects in anatase TiO 2 using different supercells, and the results are shown in Fig. 3b. Here, the formation energy plot is shown for O-rich conditions and the Fermi energy is set to 3.0 eV. As seen from Fig. 3b, the two most stable defect congurations are neutral and positively charged H interstitials, H i , where the positively charged H interstitial is the most favorable defect at high concentrations. The neutral H interstitial is the second state conguration when H concentration larger than 4 at%. The other defects are less stable as their formation energies are very high (larger than 3 eV). At high concentration (8 at%), the substitutional H s becomes more stable than O vacancies. Interestingly, under O-poor conditions, the H s is as stable as the H i (the formation energy plot is shown in Fig. S3 †).
When constructing the interface, a strain is induced automatically due to the lattice mismatch. Therefore, we also studied the stain effects on defect congurations for defect concentration of 4 at%. The calculated formation energy of different defects under biaxial strain is shown in Fig. 3c. It is interesting to notice that under biaxial tensile strain, the stable defect conguration changes to H i . Under compressive strain, positively charged H i + remains as the most stable defect conguration. As shown in Fig. 3c Fig. S5 †).
Under biaxial strain, due to lattice mismatch, the valence band maximum and conduction band minimum (CBM) of pure anatase (32 atoms supercell) under different strain are calculated and plotted in Fig. 4a. It is clear that both VBM and CBM decrease in energy as strain increases from compressive to tensile strain. When strain increases to 2%, the value of VBM decreases around 0.3 eV. Similarly, the CBM decreases around 0.4 eV. Therefore, the band gap is shrinking as strain increases. When doped with H i + , the VBM position is located below the VBM of pure anatase under 2% tensile strain, while the VBM remains unchanged with doped with H i (shown in Fig. S4 †). Hydrogen doping also introduces distortions to anatase TiO 2 , which results in the formation of the amorphous TiO 2 at the interface. To shed further light on this, H doped amorphous TiO 2 was studied with the PBE+U method, keeping the same parameters used for the interface. Doping with H i is energetically favorable (shown in Fig. S4 †) with a formation energy of À0.4 eV for single H i . The calculated DOS of H doped systems is shown in Fig. 4b. Compared with pure a-TiO 2 , single H i introduces one occupied localized state on top of the valence band edge about 0.5 eV high in energy. The doping with H does not seem to change the averaged electrostatic potential, as can be seen from Fig. S5. † Adding two H i introduce two localized states in the gap, but with slightly different energies. As more interstitial hydrogens are incorporated, we speculate that these states will eventually form energy levels in the range of up to $0.7 eV above the valence band edge. These energy levels work as the hole acceptance levels and help the transport of photoexcited holes generated from the Fe 2 O 3 side to the H:TiO 2 electrolyte side, where they can take part in the water oxidation reaction.

Experimental results
To test our hypothesis and to scrutinize the rst-principles results experimentally, we have fabricated Fe 2 O 3 -TiO 2 heterojunction electrodes in two steps. First, we deposited an ultrathin a-Fe 2 O 3 (25 nm) lm on ITO substrate by thermal evaporation of Fe followed by air annealing at 350 C. Further, the asprepared a-Fe 2 O 3 thin lms were coated with TiO 2 by sol-gel spin coating method followed by thermal annealing at 400 C. In as-prepared Fe 2 O 3 -TiO 2 heterostructure, the band edge engineering was achieved by hydrogen doping (Fe 2 O 3 -H:TiO 2 ) by further annealing the Fe 2 O 3 -TiO 2 heterostructure in 4% H 2 in Ar at 300 C at atmospheric pressure for 6 h.
To determine the crystal structures of a-Fe 2 O 3 , TiO 2 , and Fe 2 O 3 -TiO 2 heterostructure thin lms and possible phase changes aer hydrogen doping in TiO 2 , Raman spectroscopy was performed, and its results are shown in Fig. S7a. † In pristine a-Fe 2 O 3 , normally seven phonon modes are expected in the Raman spectrum: namely two A 1g modes (225 and 498 cm À1 ) and ve E g modes (247, 293, 299, 412, and 613 cm À1 ), but the 293 and 299 cm À1 bands can only be resolved at temperatures #100 K. In the present case, six of the observed bands (A 1g , 223 and 498 cm À1 ; E g , 244, 294, 410, and 607 cm À1 ) are detected, which agrees with the formation of the hematite phase (a-Fe 2 O 3 ). On the other hand, the Raman spectra of both pristine and hydrogen doped TiO 2 samples exhibit well-resolved TiO 2 Raman peaks at 144 cm À1 (E g ), 398 cm À1 (B 1g ), 515 cm À1 (E g ), and 640 cm À1 (E g ), indicating that anatase is the predominant species, except for 147 cm À1 (B 1g ), which is suppressed by a much stronger E g peak at 144 cm À1 . No phase change aer hydrogen doping in TiO 2 was observed (ESI Fig. S7b †). The Raman spectra of TiO 2 -Fe 2 O 3 heterostructure shows the signicant up-shi of E g mode of TiO 2 from 141 to 155 cm À1 and also other bands of TiO 2 . The shi of Raman bands by 14 cm À1 (inset of ESI Fig. S7b †) is probably due to the strain and defect states at the junction with a-Fe 2 O 3 and lattice mismatch. Morphological characteristics were investigated by scanning electron microscopy (SEM) (ESI Fig. S8 †) and transmission electron microscopy (TEM) (ESI Fig. S9 †). SEM images of pristine a-Fe 2 O 3 electrodes showed uniform deposition of a thin layer of a thickness of about 25 nm without any inter-particle pores, as shown in Fig. S8. † Optimized thickness of TiO 2 layer (thickness < 10 nm) over the a-Fe 2 O 3 layer on ITO clearly showed a smooth and dense coverage, as shown in Fig. S8c. † First, we optimized the temperature for optimized hydrogen doping with respect to the reduction in band gap and upshi of band edge position. Fig. 5a  700 nm measured under the same experimental conditions are shown in the ESI (Fig. S10). † In pristine TiO 2 samples, a steep increase in absorption at wavelengths shorter than $373 nm can be attributed to the intrinsic band gap of crystalline TiO 2 . The hydrogen doping of the TiO 2 sample show a signicant shi of absorption edge from higher wavelengths down to 490 nm in the visible light absorption. The calculated values of the band gap energy show that the band gap of the pristine TiO 2 thin lms is approximately 3.30 eV, slightly higher than that of bulk anatase TiO 2 . However, the onset of light absorption in hydrogen doped TiO 2 thin lm is lowered to about 2.5 eV.
To validate the reduction of band gap energy, we have carried out the XPS measurement on both pristine and hydrogen doped TiO 2 thin lms to understand the surface chemical bonds, chemical composition, peak position, and hence the electronic properties. Therefore, the Ti 2p, O 1s, and valance band spectra were measured and analyzed for pristine and hydrogen doped TiO 2 thin lms. The Ti 2p core level XPS spectrum, Fig. 5b, shows that the Ti 2p 3/2 peak is at 458.8 AE 0.1 eV, which was attributed to Ti 4+ states for both the samples. However, aer hydrogen doping, a slight asymmetric narrowing in the Ti 2p 3/2 peak can be seen. This narrowing in the peak upon hydrogen doping is likely due to a decrease in Ti 3+ states, concurrent with the reduction in oxygen content, and to a disorder due to the thermal annealing of TiO 2 in hydrogen environment. 39 However, no reduction in the Ti 4+ was observed in TiO 2 aer hydrogen doping. Fig. 5c shows the valence band spectra obtained for the pristine and hydrogen treated TiO 2 thin lms. The VBM were calculated by linearly extrapolating the peaks to the baselines. The VBM for pristine TiO 2 thin lm was observed at 1.93 eV below the zero potential energy point. However, in the case of hydrogen doped TiO 2 lms, VBM was obtained at 1.27 eV. Therefore, a valence band edge shi by $0.66 eV was observed towards the Fermi level. Moreover, small bands occurred above the VBM, and small bands existed close to the binding energy at zero. In light of the rst-principles calculations, the up-shi of the VBM is suggested to be the result of defects states above the valance band induced by interstitial hydrogen in the TiO 2 thin lms. Fig. 5d represents the O 1s spectra of pristine and hydrogen doped TiO 2 samples. The peak at a binding energy of 529.6 AE 0.1 eV corresponds to O-Ti bonds in TiO 2 in both the samples. However, in hydrogen doped samples, an additional peak can be seen at 531.6 AE 0.1 eV, which is probably due to oxygen vacancies, formation of hydroxyl groups, and Ti-OH bonds. 39,40 From Fig. S13b and c, † where the XPS spectra of all four samples are shown, it is further visible that Fe 2 O 3 -H:TiO 2 displays the same narrowing of the Ti 4+ peak as H:TiO 2 , and a more accentuate OH À peak in the oxygen 1s spectrum, correlating the formation of the iron oxide heterostructure with the generation of oxygen vacancies and/or hydroxyl groups.
The PEC measurement in the form of photocurrent densities vs. applied potential (J ph vs. V RHE ) curves are shown in Fig. 5e for pristine and hydrogen doped TiO 2 samples under 1 sun illumination. The calculated value of J ph for pristine TiO 2 photoanode was obtained $14 mA cm À2 at 1.23 V RHE . However, hydrogen doping in TiO 2 enhanced the PEC response signicantly and reached a J ph to $27.4 mA cm À2 at 1.23 V RHE . It is EIS plots from the Fe 2 O 3 -H:TiO 2 heterostructure is much smaller than that from Fe 2 O 3 -TiO 2 , which indicates that the hydrogen treatment of the TiO 2 layer changes the charge distribution in Fe 2 O 3 -H:TiO 2 . The new charge distribution results in a lower magnitude of the equivalent series resistance in Fe 2 O 3 -H:TiO 2 heterojunction photoanodes indicating strongly improved charge transport properties as compared to Fe 2 O 3 -H:TiO 2 .
Finally, in order to assess the durability of the photoanodes, chronoamperometry was performed on the Fe 2 O 3 -TiO 2 and the Fe 2 O 3 -H:TiO 2 electrodes (as presented in Fig. S12 in the ESI †): the resulting photocurrent density, under a bias voltage of 0.5 V RHE and an illumination of 1 sun, was found to be stable around a value of 0.4 mA cm À2 for up to 500 minutes for Fe 2 O 3 -H:TiO 2 , with no sign of decreasing, while the non-treated sample displayed a lower photocurrent (0.23 mA cm À2 ) at the start, which halved aer less than 300 minutes of continuous use. This sustained stability of Fe 2 O 3 -H:TiO 2 electrodes conrms the applicability of the device.

Conclusion
First-principles calculations predict that hydrogen treatment induce states above the valance band, which can transfer holes through the heterostructure junction. The predictions are validated with experimental observation obtained on sol-gel grown TiO 2 thin lms. Optical absorption, XPS analysis, and (photo) electrochemical measurements show that the tailoring in optical band gap, shi in valence band position, and change in electrical properties in hydrogen doped TiO 2 results in a type II band alignment in the Fe 2 O 3 -H:TiO 2 heterostructure. The Fe 2 O 3 -H:TiO 2 heterostructure reduces the electron-hole recombination sharply at the junction and improves the water oxidation performance. This study shows that hydrogen treatment can enhance the photoelectrochemical response of the Fe 2 O 3 -TiO 2 heterostructure, thanks to the formation of type II band alignment at the interface junction. The theoretical and experimental strategies can be applied to other oxides based heterostructures and might become important tools for engineering efficient and stable photoelectrodes.

Conflicts of interest
There are no conicts to declare.