Gradient tantalum-doped hematite homojunction photoanode improves both photocurrents and turn-on voltage for solar water splitting

Hematite has a great potential as a photoanode for photoelectrochemical (PEC) water splitting by converting solar energy into hydrogen fuels, but the solar-to-hydrogen conversion efficiency of state-of-the-art hematite photoelectrodes are still far below the values required for practical hydrogen production. Here, we report a core-shell formation of gradient tantalum-doped hematite homojunction nanorods by combination of hydrothermal regrowth strategy and hybrid microwave annealing, which enhances the photocurrent density and reduces the turn-on voltage simultaneously. The unusual bi-functional effects originate from the passivation of the surface states and intrinsic built-in electric field by the homojunction formation. The additional driving force provided by the field can effectively suppress charge–carrier recombination both in the bulk and on the surface of hematite, especially at lower potentials. Moreover, the synthesized homojunction shows a remarkable synergy with NiFe(OH)x cocatalyst with significant additional improvements of photocurrent density and cathodic shift of turn-on voltage. The work has nicely demonstrated multiple collaborative strategies of gradient doping, homojunction formation, and cocatalyst modification, and the concept could shed light on designing and constructing the efficient nanostructures of semiconductor photoelectrodes in the field of solar energy conversion.

. Here, Yobs, Ycalc, Yobs-Ycalc, Bragg_position, Rp, Rwp, Rexp, S represent the experimental data, the calculated data, the difference of experimental and calculated data, Bragg's position, the profile factor, the weighted profile R factor, the expected R factor, goodness of fit, respectively. Note that lattice parameters, isotropic strain broadening, asymmetry, isotropic size broadening, and GauSiz are considered for the reliable profile matching. Note that lattice parameters, isotropic strain broadening, asymmetry, isotropic size broadening, and GauSiz are considered for the reliable profile matching. In order to have a consistent comparison, the first-order derivative method was developed by Grätzel group, 1 which is defined as the value, at which dJ/dV > 0.2 mA cm -2 V -1 from the obtained J-V curves. The method was adopted and confirmed reasonable by Wang 2 and Ye 3 groups. When photoanodes have a high J (over 3 mA cm -2 at 1.23 VRHE) and/or a strong signal noise, however, it is difficult to apply the criterion of dJ/dV > 0.2 mA cm -2 V -1 . In our case (a and b), it might be more reasonable to evaluate Von at which dJ/dV > 1 mA cm -2 V -1 (just above the strong noise of all samples), which gives Von of 0.81 (Ta:Fe2O3), 0.63 (Ta:Fe2O3@Fe2O3) and 0.54 VRHE (NiFe(OH)x/Ta:Fe2O3@Fe2O3), respectively. The results are almost the same as our extracted values (Fig. 5b)  Light absorption before and after optimal (60 min duration) cocatalyst modification (e). The nickel and iron chloride solution can be gradually hydrolyzed to form NiFe(OH)x in a dilute concentration. From SEM images (barely taken without any sputtering of noble metal), 20 min duration (a) gives an invisible deposition, but shows an improved Jph (2.84 mA cm -2 ). With 60 min duration, however, a thin and uniform layer of NiFe(OH)x was deposited (b) because some small nanoparticles can be seen. Besides, the poor contrast of image relative to the previous one also indicates the successful deposition of NiFe(OH)x, providing a maximum of Jph (3.22 mA cm -2 ). On the other hand, this thin and uniform NiFe(OH)x layer does not change any light absorption (e). When duration reaches 100 min, a great number of bigger nanoparticles can be observed clearly(c), which actually decreases the Jph (2.63 mA cm -2 ) due to too much an amount. XPS O 1s spectra (f) shows that the OH peak intenensity of homojunction modified with cocatalyst becomes clearly higher than that of bare one. Besides, SEM-EDS (g, h) detects the signal of Ni element (~0.41wt%). Our previous results demonstrated that this strategy deposited FeOOH cocatalyst on hematite successfully. Based on these results, it is reasonable that the cocatalyst is NiFe(OH)x. For both CTA (a) and HMA (b) samples, Fe2O3@Fe2O3 shows a limited increase of Jph and a negligible Von shift relative to bare hematite (Jph is slightly better for HMA), which demonstrates that the built-in electric field does not occur in this homojunction. Therefore, the sufficient amount of dopants (M n+ , n>3) in the core part is essential to construct an effective homojunction, in which built-in electric field would promote charge separation efficiently. Note that NiFe(OH)x cocatalyst overlayer on all the HMA photoanodes promotes higher photocurrent density than on the CTA photoanodes, which demonstrates the high quality of electrodes prepared by HMA is reflected not only in their own behavior but also in the performance with cocatalyst loading. XRD (c) shows the similar pattern and intensity of peaks except the stronger intensity of (110) peak for homojunction, indicating the a preferential growth for bare hematite compared to Ta:Fe2O3. The LHE (d) of homojunction shows a slight increase relative to bare hematite, suggesting its negligible contribution to light absorption.
where E(λ) is the solar irradiance at a specific wavelength (λ), IPCE(λ) is the photoresponse profile of the photoanode at a specific wavelength (λ) at 1.23 VRHE.

Supplementary Fig. 24| OCP transient decay profiles for Ta:Fe2O3,
Ta:Fe2O3@Fe2O3, NiFe(OH)x/Ta:Fe2O3@Fe2O3. The OCP transient decay evaluates the surface recombination between trapped electrons and reaction intermediates rather than the bulk recombination and this process is very fast (within ns-ms domain), while the scale of OCP decay is usually in several minutes. 10 Generally, OCP is very positive in dark due to a largely upward band bending, while OCP will be more cathodic under illumination due to flattening of the energy band by photoexcited carriers. ΔOCP (OCPdark -OCPlight), also known as photovoltage, represents the amount of band bending under illumination with respect to that in the dark condition. Homojunction formation boosts ΔOCP value compared to that of Ta:Fe2O3, indicating that homojunction can generate an additional built-in electric field. At the transient time from illumination (quasi-equilibrium of flattened energy band) to the dark (equilibrium of bent energy band), the charge recombination strongly depends on the spatial charges built in the photoanode/electrolyte junction. The strong band bending enables a large amount of spatial charges in the depletion region and thus significant charge recombination occurs at the transient when the illumination is removed. As such, a fast OCP-decay is expected.
The ratio of evolved O2 and H2 is close to stoichiometry, and F.E.'s of O2 and H2 evolution reaction is 92.3 % and 95.8 %, respectively.
Supplementary Fig. 28| Representative two-and three-RC-unit equivalent fitting models. A typical two-RC-unit equivalent circuit generally consists of three resistances and two capacitors: a series resistance (Rs, essentially small and constant) by the electrolyte, external contact and conductive substrate layer, a trapping resistance (Rtrap) at surface states by the trapping holes, a charge transfer resistance (Rct) at semiconductor-liquid junction, a bulk capacitor of space charge region (Cbulk), and a surface states capacitor (Css). Three-RC-unit equivalent circuit is used to fit the cocatalyst-modified homojunction beyond 1.1 VRHE where a third circle shows up clearly (see Supplementary Fig. 28c). Ta:Fe2O3, homojunction shows an almost same shape of Nyquist plot at earlier potential (at least 100 mV in advance, see highlighting arrows), which is consistent with the measured Von. At the same potential, homojunction shows the smaller total resistance than that of Ta:Fe2O3, confirming the beneficial effect of charge separation. By cocatalyst modification, the total resistance is further reduced (< 1.1 VRHE) by greatly improved surface charge transfer. including recombination process at μs-ms and water oxidation process at ms-s. 12 To compare the timescale of starting water oxidation, the turning point is extracted from biphasic decay curve. The shorter the turning point time is, the more photoholes the water oxidation employs effectively.

Supplementary
Supplementary Fig. 36| Bias-dependent TAS for Ta:Fe2O3, Ta:Fe2O3@Fe2O3, and NiFe(OH)x/Ta:Fe2O3@Fe2O3 photoanodes in 1M NaOH. Bias-dependent TAS results display that the long-lived holes appear at lower bias (0.7-0.8 VRHE) in homojunction photoanode than that of the Ta:Fe2O3, in accordance with the negative shift of onset potential. However, no long-lived holes were observed at lower bias (0.7-0.8 VRHE) after NiFe(OH)x modification. This result may be related to two reasons. One is the hole transfer from homojunction to NiFe(OH)x. The other is the hole absorption spectra in NiFeOOH overlapped with the photoholes in Fe2O3 (the reactive center in NiFeOOH electrocatalyst is Fe-site), which cannot be distinguished from the hole in hematite. 13 Accordingly, at applied bias of 1.3 VRHE (Fig. 6d)