Photoelectrochemical (PEC) water splitting has been considered as a promising technology for storing solar energy [1]. Since the start-up work of Fujishima and Honda on a TiO₂ photoanodes, many visible-light responsive semiconductor materials, such as α-Fe₂O₃, Ta₃N₅, and BiVO₄, have attracted intensive attention from researchers [2-4]. Among these materials, α-Fe₂O₃ is one of the most promising photoanode materials due to its high theoretical efficiency (about 16%), excellent photochemical stability, nontoxicity and low cost [5]. However, the PEC performance of α-Fe₂O₃ is hindered by some intrinsic drawbacks, including short hole diffusion length, poor electrical conductivity, severe surface states recombination and so on.

In order to address these limitations and achieve higher PEC performance, many methods have been carried out over decades [6-14]. Element doping with high valent cations such as Ti, Si, or Sn markedly enhances the plateau photocurrent of hematite photoanodes, due to the improvement of electron concentration and n-type conductivity [9, 15, 16]. High temperature annealing is another common strategy for targeting a larger photocurrent of hematite photoanodes by increasing crystallinity and activating dopants [17]. A combination of doping and high temperature annealing strategies can further improve the plateau photocurrent of hematite photoanodes, compared with applying a single strategy [16, 18]. However, an unfavorable onset potential shift of doped hematite photoanodes appeared after high temperature annealing. For this phenomenon of Ti doped hematite photoanodes, to the best of our knowledge, the underlying mechanism has not been explored to date and it deserves further investigation.

In this study, we have used different photoelectrochemical and structural characterization techniques to confirm that the change of onset potential comes from a surface effect. Through a combination of X-ray photoelectron spectroscopy (XPS) measurements and density functional theory plus Hubbard U correction (DFT + U) calculations, we further revealed that the surface Ti/Fe atomic ratio increased after high temperature annealing, which would decrease the amount of surface adsorbed hydroxide ions. As a result, the flatband potential (i.e., the theoretical onset potential) of Ti doped hematite photoanodes would positively shift after high temperature annealing, proven by Mott-Schottky measurements.

Herein, the titanium doped hematite photoanodes (denoted as Ti:Fe₂O₃) were prepared using a previously established method [19]. High temperature (denoted as HT in all figures of this study) annealing was performed on the as-prepared hematite film at 750 ℃ for 10 min. Fig. 1 presents the photocurrent-potential curves of Ti:Fe₂O₃ before and after high temperature annealing. After high temperature annealing, the plateau photocurrent of Ti:Fe₂O₃ doubled, indicating the enhanced photoactivity [17, 20]. Meanwhile, the photocurrent onset potential was significantly shifted anodically from 0.95 V_RHE to 1.05 V_RHE. This unfavorable shift suppressed the PEC response of Ti:Fe₂O₃ under low bias, compared to that of the samples without high temperature annealing treatment. Similar phenomena have been also observed in previous literatures [18, 20], which is consistent with this study.

X-Ray diffraction (XRD) data of Ti:Fe₂O₃ before and after high temperature annealing are shown in Figs. S1a and b (Supporting information), confirming the presence of hematite (JCPDS No. 330664). Peak positions and intensities are similar for both samples, indicating annealing treatment does not change the phase and crystallinity. Also, the grain size of both samples are almost the same, as determined by the full width at half-maximum (FWHM) of the (110) diffraction peaks. Figs. 2a and b show the scanning electron microscopy (SEM) surface images of Ti:Fe₂O₃ before and after high temperature annealing, respectively. No obvious changes of surface morphology and grain size were observed. Some nanoparticles are fused together and both of the samples exhibits crack structures, matching with the literature [19]. Additionally, the UV–vis absorption spectra were measured and the results are in Fig. 2c, suggesting no difference in the light absorption capability upon the high temperature annealing. Taken together, it is impossible that the unfavorable onset shift comes from the unchanged phase, crystallinity, grain size, morphology and light absorption capability. Moreover, control experiments showed that annealing at 750 ℃ for 10 min would not have a measurable change on the conductivity of fluorine-doped tin oxide (FTO) substrate, in line with previous reports [21].

Further, in order to separate any impact due to bulk change from the surface effects, photocurrent-potential measurements were carried out in the presence of an efficient hole scavenger (H₂O₂). The onset potential of both photoanodes in contact with H₂O₂ are identical shown in Fig. 2d, which indicates the difference of water oxidation onset potential of photoanodes before and after high temperature annealing is a consequence of surface effect [22, 23].

XPS measurements were used to characterize the composition change on the surface of the films before and after high temperature annealing, which would provide additional information about surface changes [24, 25]. Fig. 3 shows the binding energy of O 1s, Fe 2p, Ti 2p and Sn 3d for both samples. The peaks at 529.7 eV and 710.5 eV are assigned to the O^2- and Fe³⁺ ions in hematite, respectively [20]. Both photoanodes show the binding energy of Ti 2p at the peak of 457.9 eV, implying that Ti cations are in a different atomic environment from that in pure TiO₂ (binding energy of 459.0 eV) [20]. No signal of Sn⁴⁺ was detected on the surface of both samples, evincing that there is no Sn⁴⁺ on the hematite surface due to the diffusion from the FTO substrate. It should be noted that the Ti/Fe atomic ratio on the surface has an apparent increase after high temperature annealing obtained from the XPS data. Another notable change after high temperature annealing is the amount decrease of hydroxyl group (—OH) on the surface, as determined by the decrease of the peak intensities at about 532.5 eV [2, 26]. This was in accordance with Cao et al.'s work [27], when titanium was intentionally introduced into hematite lattices.

MoBT-Schottky analysis was performed to gain further insight into changes after the high temperature annealing. Fig. 4a exhibits Mott-Schottky plots of both samples and the V_fb (denoted as flatband potential) of Ti:Fe₂O₃ positively shifted about 100 mV after high temperature annealing. V_fb was denoted as the theoretical onset potential, as one would expect a photocurrent at any potentials positive of V_fb theoretically [28]. So it reveals that it is the positive shift of V_fb that induce the unfavorable onset potential after high temperature annealing.

According to previous reports, a shift of flatband potential on hematite photoanodes is generally attributed to changes of surface states and/or surface ions adsorption [19, 29]. Firstly, in order to investigate the effects of high temperature annealing on the surface states properties, we carried out photoelectrochemical impedance spectra (PEIS) measurements of the photoanodes before and after high temperature annealing. Representative Nyquist plots shown in Fig. S2 (Supporting information) exhibits apparently two semicircles and the low frequency semicircle is attributed to the surface states. The values for C_ss (surface states capacitance) extracted from fitting the PEIS data under different biases are displayed in Fig. 4b, showing Gaussian behavior for both photoanodes before and after high temperature annealing [30]. While the peak intensity of C_ss did not have an apparent decrease/ increase after high temperature annealing, the peak shifted 100 mV to a more positive potential, in good agreement with the shift of V_fb. Therefore, it suggests that the amount and relative location of surface states in band gap does not change after annealing treatment and surface states of Ti:Fe₂O₃ are not responsible for the change of V_fb and onset potential.

Secondly, we will investigate the effect of surface ions' adsorption on the flatband potential shift. When a metal oxide photoelectrode is immersed in an aqueous solution, H⁺ and OH^- ions in the solution will successively adsorb/desorb from the surface and form a region called Helmholtz layer. The amount of adsorbed H⁺ and OH^- ions on the electrode surface (i.e., inner Helmholtz layer) would change the potential drop across the Helmholtz layer and then shift the flatband potential of the electrode [31]. With a testing electrolyte pH of 13.6, the net total charge adsorbed at a hematite photoanode should be negative. Reducing the amount of surface adsorbed OH^- ions would then decrease the Helmholtz voltage and positively shift the flatband potential of hematite photoanodes.

Recalling the XPS data, the amount decrease of surface hydroxyl group (i.e., adsorbed OH^- ions) was observed after high temperature annealing, which could induce the anodic shift of flatband potential and onset potential. At the same time, we note that the decrease of surface hydroxyl group is correlated to the increase of surface Ti doping. Similar phenomenon was obtained in previous work [32]. To furtherly investigate whether it would decrease the hydroxide ions' adsorption capacity when inducing more Ti doping on the surface, we applied DFT computations to calculate the adsorption energy change before and after introducing Ti atom on the hematite surface.

Two surface models, (110) and (110) + Ti_Fe, were adopted to perform the OH^- adsorption energy calculations. Since PEC water oxidation of hematite electrodes is under high pH condition, here we used oxygen-terminated (110) surfaces, at which the surfaces were deprotonated [33]. For Ti doped hematite surface model, we denoted the surface substitutional Ti atom as a center and calculated OH^- adsorption energies bonding with the central Ti atom and nearest-neighbor Fe atoms on the surface, including 3 terminal sites and 3 bridge sites. Similar calculation were performed on the pristine (110) surface model. Figs. 5a and b show the most stable structures of OH^- bonding with the central Ti/Fe atom on (110) and (110) + Ti_Fe surfaces, respectively. Table 1 lists adsorption energies of one OH^- bonding with different reaction sites on these two surface models. The OH^- adsorption energies on (110) surface are always smaller than that on the (110) + Ti_Fe surface, suggesting that Ti doping would decrease the OH^- adsorption capacity on the hematite (110) surface, in line with our XPS data. Our study reveals that, due to the surface Ti/Fe atomic ratio increase caused by high temperature annealing, the surface OH^- adsorption capacity would decrease and it would induce the anodic shift of flatband potential and the unfavorable onset shift. We hope that the future investigations on Ti doped hematite photocatalyst may be enlightened from this work and the unfavorable onset shift caused by high temperature annealing can be overcome.

In conclusion, we have investigated the effect of high temperature annealing on the water splitting onset potential of Ti:Fe₂O₃ photoanodes. XPS measurements have shown that surface Ti/Fe atomic ratio increases and adsorbed hydroxide ions amount decreases, as a result of high temperature annealing. Further DFT calculation suggests that more surface Ti doping would decrease the surface OH^- adsorption energy, which is in line with the XPS data. The following flatband potential shift caused by hydroxide ions adsorption capacity change was also observed under Mott-Ti:Fe₂O₃ photoanodes.

Acknowledgment

This work was supported by the National Natural Science Foundation of China (Nos. 21473090, U1663228). Supercomputing facilities of the High Performance Computing Center (HPCC) of Nanjing University are also acknowledged.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.cclet.2018.01.022.