Interfacial oxygen vacancies yielding long-lived holes in hematite mesocrystal-based photoanodes

Hematite (α-Fe2O3) is one of the most promising candidates as a photoanode materials for solar water splitting. Owing to the difficulty in suppressing the significant charge recombination, however, the photoelectrochemical (PEC) conversion efficiency of hematite is still far below the theoretical limit. Here we report thick hematite films (∼1500 nm) constructed by highly ordered and intimately attached hematite mesocrystals (MCs) for highly efficient PEC water oxidation. Due to the formation of abundant interfacial oxygen vacancies yielding a high carrier density of ∼1020 cm−3 and the resulting extremely large proportion of depletion regions with short depletion widths (<10 nm) in hierarchical structures, charge separation and collection efficiencies could be markedly improved. Moreover, it was found that long-lived charges are generated via excitation by shorter wavelength light (below ∼500 nm), thus enabling long-range hole transfer through the MC network to drive high efficiency of light-to-energy conversion under back illumination.

The synchrotron-based X-ray total scattering and pair distribution function (PDF) analyses of the samples. The peak at around 2 Å is composed of short Fe−O (1.94 Å) and long Fe−O (2.12 Å) distances. 4 The peak at around 3 Å is composed of the first neighbor edge-sharing (2.97 Å) and face-sharing (2.89 Å) Fe−Fe distances. The first neighbor corner-sharing Fe−Fe pairs show two peaks at 3.39 Å and 3.70 Å. 5 After the annealing of as-synthesized Fe2O3 MCs at 700 °C, there was no apparent change in the peaks corresponding to Fe−O and Fe−Fe distances, possibly due to the low concentration of oxygen vacancies in MCs (see black and green lines). In contrast, a significant difference in the peaks was observed for Ti-modified Fe2O3 MCs before and after the annealing (see blue and red lines). By the annealing of as-synthesized Ti-Fe2O3 MCs, all the peaks became narrow and similar to those of pure hematite, suggesting that most of doped Ti ions are segmented from the bulk. Notably, the difference in the peak intensities between annealed Fe2O3 (green line) and Ti-Fe2O3 (red line) MCs implies that a small portion of Ti ions are doped in hematite. For Fe2O3 MCs, no apparent difference in Fe-O and Fe-Fe bond distances was observed between the samples before and after the annealing. This is probably due to the fact that the amounts of VO in annealed MCs are still too small to be detected. Whereas, the Ti modification results in a significant intensity decrease and broadening of the peaks. After the annealing, The component 1 is considered to be mainly attributed to hematite according to the similarity of Fe-L2,3 and O-K spectra with the reference. 7 For component 2, the shift of Fe-L2,3 peaks to lower energy loss and the suppression of the pre-peak intensity at ca. 527 eV of O-K spectrum indicate the presence of Fe 2+ species. These changes can be explained by the formation of 4-coordinated Fe 2+ . 8,9 Here, we propose that their possible origin is not only the oxygen vacancies at the interfaces in MCs, but also Fe2-xTixO3 (e.g., ilmenite (FeTiO3)), judging from the Ti-L2,3 spectral profile. 10 The latter is considered to be present at the interface between hematite and TiO2 (see Supplementary Fig. 7) and play a key role in the formation of the TiO2 overlayer during the annealing at 700 °C. The component 3 is mainly TiO2. Considering the similarity of Ti-L2,3 and O-K spectra with the reference, 10,11 the typical transition temperature from anatase to rutile at ca. 600 °C, 12 and XRD data (see Supplementary Fig. 8), rutile phase would be a potential candidate. The formation of rutile TiO2 on hematite has been reported in the literature. 4 It should be noted  The multiexponential decay curves were fitted using a nonlinear least-squares method with a three-component decay law described by I(t) = a1exp(−t/ 1) + a2exp(−t/ 2) + a3exp(−t / 3). The average lifetime (<τ>) was than evaluated using the equation:

Supplementary Note
All the applied voltage has been converted into the potential vs. RHE via the Nernst equation: ERHE represents the converted potential vs. RHE, and EAg/AgCl is the applied potential vs.
Ag/AgCl, E 0 Ag/AgCl is 0.189 V at ambient temperature (25 o C), pH of the electrolyte is 13.6.
The carrier density (Nd) can be calculated according to the following equation: where e0 is the electron charge, ɛ is the dielectric constant of hematite and ɛ0 is the permittivity of vacuum, d(1/C 2 )/dV is the slop of the obtained Mott-Schottky curve.
The density of surface state (NSS) can be calculated according to the following equation:

Supplementary Discussion
The correlation between Nd and VO concentration on surface estimated from the XPS O 2p (Supplementary Fig. 19), as well as current density at 1.23 V vs. RHE (Fig. 3a), is summarized in Supplementary Fig. 32 Fig. 9c and 19). Therefore, the ligandto-metal charge transfer (LMCT) transition (O 2p → Fe 3d) centered at 3.2 eV 54 can effectively occur when being excited by short-wavelength light. A recent study employing time-resolved microwave conductivity suggested that the 355 nm laser excitation of pure and metal-doped hematite films produces more mobile charges than excitation with 532 nm light that is not well matched with the LMCT transition. 55 Recently, the excitation-wavelength-dependent lifetime of the photoexcited electrons was explained in terms of polaron-hopping theory. 56 The higher excitation energy provides more excess energy to the lattice, i.e., fewer polarons being formed by the nonthermal phonon bath, and thus the hopping rate of the polarons in equilibrium with mobile carriers will increase, resulting in the increased hopping radius and lifetime of carriers. However, this model cannot solely explain our finding that the nanosecond PL lifetimes were observed only for the MC samples.