Binary dopant segregation enables hematite-based heterostructures for highly efficient solar H2O2 synthesis

Dopant segregation, frequently observed in ionic oxides, is useful for engineering materials and devices. However, due to the poor driving force for ion migration and/or the presence of substantial grain boundaries, dopants are mostly confined within a nanoscale region. Herein, we demonstrate that core–shell heterostructures are formed by oriented self-segregation using one-step thermal annealing of metal-doped hematite mesocrystals at relatively low temperatures in air. The sintering of highly ordered interfaces between the nanocrystal subunits inside the mesocrystal eliminates grain boundaries, leaving numerous oxygen vacancies in the bulk. This results in the efficient segregation of dopants (~90%) on the external surface, which forms their oxide overlayers. The optimized photoanode based on hematite mesocrystals with oxide overlayers containing Sn and Ti dopants realises high activity (~0.8 μmol min−1 cm−2) and selectivity (~90%) for photoelectrochemical H2O2 production, which provides a wide range of application for the proposed concept.


Supplementary Note 1
The mechanism of the dopant segregation induced by electrostatic interactions to compensate ionic space charges at (sub)surface and/or grain boundaries (GBs) has been well established. [1][2][3][4] In this note, we summarize the main part of the space-charge theory related to our study.
Here, Fe2O3 is assumed to be composed of trivalent metal cations, Fe 3+ . It is also assumed that the Fe2O3 crystal has a free surface and GB (x = 0). The predominant lattice defects are considered as Schottky-type defects and showed based on the Kröger-Vink notation. The concentrations of vacancies for iron cation ( V ) and oxygen anion ( V •• ) and interstitial iron ( Fe ••• ) are described as The formation free energies , , and are defined relative to an interface. The denotes the electrostatic potential in the bulk far from the surface or GB and can be calculated by using the charge neutral condition, i.e., 3 V 2 V •• 3 Fe . Then, we can obtain (4) According to equations (4) and (5), the electrostatic potential in the bulk has a relatively low value for pure hematite. However, the potential in the space-charge layer will be remarkably modified by doping aliovalent cations to the lattice and thus provide a potential gradient to drive the dopant segregation in the space-charge layer.
For example, in the case of the Ti 4+ -doped Fe2O3 system, Ti 4+ is added as a donor in Fe2O3.
To reach the charge neutrality in the bulk, V will form. The concentration of V is readily determined by the condition Ti • 3 V . Therefore, the electrostatic potential between the bulk and surface can be deduced as equation (6), 3 Consequently, in the donor-doped case, the has a large value with positive sign, and thus the electrostatic potential in the space-charge layer ( ( )) rapidly decays. According to equation (1), the concentration of iron vacancies ( V ( )) in the space-charge layer is significantly reduced and thus leads to the strong accumulation of positively charged Ti • , which should be compensated by negative charges (e.g., O ) on the surface (Fig. 1a). It was recently found by the atom probe tomography analyses that doped titanium cations in hematite were segregated to the O-rich grains on the surface, 5 which is consistent with our finding that annealed Ti-Fe2O3 possesses more negative potentials on the edge (i.e., surface) as revealed by Kelvin prove force microscopy (KPFM) measurements ( Supplementary Fig. 2).
For the acceptor doping case, i.e., Sn 2+ -doped Fe2O3, the electrostatic potential can be also derived in the same manner. The effective negative charges by substituted tin anions will be compensated by oxygen vacancies via the relationship of Sn 2 V •• . The electrical potential for the acceptor doping case in the bulk is therefore given as ln (7) In contrast to the donor doping case, the has a large value with negative sign (Supplementary Fig. 1), which contributes to the acceptor segregation in the space-charge layer.
Based on the above consideration, the concentration of the dopants can be generally expressed as the exponentially varying simple Boltzmann distribution.
, exp (8) where and are the concentration and effective charge of the dopant, respectively. It is clear that dopant segregation is driven by the electrostatic potential in the space-charge layer.
Therefore, the increase of the electrical potential in the space-charge layer can provide a larger potential to promote the dopant segregation.
During thermal annealing of hematite MCs, oxygen vacancies are formed at the interfaces along with electrons as described by the following defect reaction where O is an oxygen ion at an oxygen site. The electrons are instantaneously captured by Fe 3+ to form Fe 2+ as 4 where Fe is Fe 2+ . Considering the electronic conduction of hematite is a result of polaron hopping, 678 the formation of Fe 2+ sites can improve the electrical conductivity, especially near the GB. Furthermore, the positively charged VO localized at the GB core provides an additional repulsive force against the dopant cations to accelerate the oriented migration of the dopant cations from the bulk to the surface without their accumulation to the GB regions.

Supplementary Note 2
The width of space charge layer (W) can be calculated according the following equation, where ɛ is the dielectric constant of hematite (80), Vbi is the built-in potential which can be calculated by subtracting the flat-band potential from the applied potential, Nd is the carrier density. Based on equation (11) where d(1/C 2 )/dV is the slop of the obtained Mott-Schottky curve.
The carrier density (determined by the Mott-Schottky curves in Supplementary Fig. 13 This increase is in fact the same trend as has been discussed by Xu et al.; 10 namely, introducing the Hubbard correction generally weakens adsorption. Furthermore, the latter value is in good agreement with the previously reported value (~1.5 eV), 11 which was obtained by using PBE + U (U = 4.3 eV) and 1/12 monolayer (ML).
With ΔGO* = 3.45 eV computed with U, which also agrees well with the above reference, 11 the O2 evolution reaction is predicted to dominate over the H2O2 evolution reaction in Fe2O3. 5 As soon as the coverage increases from 1/12 ML, we found a large increase in ΔGOH* with U = 4.3 eV to 1.74-1.88 eV on average. The computed ΔGOH* for 2/12 ML was found to be 1.74 eV on average, indicating a free energy of 2.15 eV is required to further attach the second OH to the 1/12 ML surface. This is an important insight, because it is strongly indicated that this step has a large thermodynamic barrier and is unlikely to occur at an extra bias of 1.76 eV, ideal for the H2O2 evolution reaction. These results also support the experimental evidence that Fe2O3 is an O2 evolution catalyst ( Supplementary Fig. 26).

Supplementary Note 4
The production of H2O2 via water oxidation reaction (WOR) is a two-electron process with a standard redox potential (E°) of 1.76 V, which is 0.53 V higher than that for O2 generation from four-electron reaction (equations (13) and (14)).
The production of H2O2 is a formidable challenge because the two-electron pathway is thermodynamically less favorable than O2 evolution via four-electron WOR. It requires the two-hole transfer from the bulk to the surface of photoanode semiconductor to activate the PEC WOR for H2O2 production from the electrolyte.
In 2016, Sayama's group reported that the PEC H2O2 performance of a BiVO4/WO3 photoanode can be significantly enhanced in HCO3 − electrolyte. 12 Furthermore, they found that different overlayers of SiO2, ZrO2, TiO2, and Al2O3 can increase the PEC H2O2 production performance of the BiVO4/WO3 photoanode in the presence of HCO3 − electrolyte. 13 According to their explanation, the weak acidic sites on the overlayer surface more effectively adsorb the weakly basic HCO3 − . HCO3 − (CO3 2− ) acts as a hole acceptor and can be oxidized to unstable 14 Supplementary Fig. 31, the onset potentials measured in phosphate buffer solution and NaOH solution are much lower than that measured in NaHCO3 solution. In addition, gaseous oxygen was linearly generated in these electrolytes, suggesting the prioritized water oxidation to O2 via the four-hole process without the presence of HCO3 − .  Fig. 8a), which is due to the energetic splitting of t2g-and eg-orbitals in the TiO6-octahedral crystal field. 18 The crystal field splitting measured at the P3/2 level (~2.3 eV) is close to that reported for ilmenite-type SnTiO3 structure (~2.4 eV). 19 In the O-K edge ( Supplementary Fig. 8b), the ionized edges at around 510-520 eV are corresponding to the M4 and M5 delayed edges of Sn, respectively. 20 The O-K edge is split into two main signals at 538.2 and 529.0 eV for the surface region, and 538.3 and 528.2 eV for the bulk region, respectively. The first peak, which represents the existence of oxygen, is much lower than the second peak related to the hybridized transition between oxygen and neighbored metal ions, thus suggesting the presence of VOs in both the surface and bulk regions. 21 The energy difference between these two peaks for the bulk region (~10.1 eV) is very close to the reported value of pure hematite, 22 while that for the surface region (~9.2 eV) is much lower (i.e., the formation of heterostructure). This value is also larger than those reported for tin oxides (4.2-6.3 eV) and titanium oxides (2.5 eV), 18 thus excluding the 12 possibility of the formation of single-metal oxides as the final main products on the surface.
However, we noticed that the Sn 3d XPS depth profiles show that the Sn 3d signals slightly shifted to higher binding energies when the sample was etched by Ar for 60-120 s (i.e. 0.83-1.66 nm depth) ( Supplementary Fig. 8d), while Ti ions possess the same oxidation state from the surface to the depth of ~7 nm as indicated by the non-shifted Ti 3d signals ( Supplementary   Fig. 8c). These results suggest the possibility of existing oxidized Sn ions at the outer surface region (below 2 nm depth). The Sn 3d peaks located at relatively lower energies observed before Ar etching are probably due to the surface adsorbates. Based on the Sn K-edge FT-EXAFS spectra measured in CEY mode (Fig. 2i), such oxidized Sn species might come from a small amount of SnO2 at the surface. Sn-Fe2O3 is the same as that of the reference SnO2 sample (Fig. 4d), proving the formation of a SnO2 overlayer. For the annealed SnTi-Fe2O3, the Sn-Sn bonding has a shorter radial distance than that of SnO2, which is due to the formation of Sn-Ti coordination (i.e., SnTiO3-x). 23

Supplementary Tables
Supplementary Table 1 were adapted from Refs. [30] and [28]. [a] The index corresponds to that of Supplementary Fig. 30.