Structural characterization of photoanodeThe NiO/CuxO/n-Si heterostructure was prepared by using EB evaporation to sequentially deposit a CuxO layer and a NiO layer on an n-Si substrate with native SiOx oxide. In this device configuration, the p-type NiO was employed to form a buried heterojunction with the n-Si, which also served as a protection layer due to its high chemical stability in alkaline media15,16,17. An ultra-thin CuxO layer was employed to modify the NiO/n-Si interface, which was expected to improve the band alignment between NiO and n-Si. To improve the OER kinetics, a self-healing NiCoFe-Bi co-catalyst was modified on the NiO/CuxO/n-Si photoanode by photo-assisted electrodeposition. To optimize the thicknesses of the CuxO and NiO layers, the PEC activity of the NiCoFe-Bi/NiO/CuxO/n-Si photoanodes with different layer thicknesses was tested using the cell configuration shown in Supplementary Fig. S1. The thicknesses of the CuxO and NiO layers were optimized to be 0.5 and 20 nm, respectively (Supplementary Fig. S2). The electrodeposition conditions of the NiCoFe-Bi co-catalyst were also optimized (Supplementary Fig. S3). The top-view scanning electron microscopy (SEM) images reveal uniform surface coverage by the EB-deposited NiO/CuxO layers and by the electrodeposited NiCoFe-Bi layer (Supplementary Fig. S4). The multilayer structure of NiCoFe-Bi/NiO/CuxO/n-Si was revealed by analyzing the cross-section of the electrode using high-resolution transmission electron microscopy (HRTEM) and the energy-dispersive X-ray spectroscopy (EDS) elemental mappings using TEM in scanning transmission electron microscopy (STEM) mode (Fig. 1). Notably, vertically aligned lattice columns are observed in the EB-deposited NiO layer (Fig. 1b, Supplementary Fig. S5). This indicates columnar growth, which could facilitate out-of-plane carrier transport. Due to its ultra-thin thickness, the CuxO layer cannot be directly distinguished in the TEM images or STEM–EDS mappings results (Supplementary Fig. S6). However, the STEM–EDS line scan detected the Cu signal at the NiO/n-Si interface (Supplementary Fig. S7). The optimized NiCoFe-Bi co-catalyst layer has a thickness of ~13 nm (Supplementary Fig. S5e), in which the Ni, Co, and Fe elements are uniformly distributed (Fig. 1c). It should be noted that the presence of the native SiOx layer is essential for achieving high PEC activity on n-Si photoanodes18,19. The removal of the SiOx layer led to a significant anodic shift of the photocurrent onset potential of the n-Si photoanode (Supplementary Fig. S8). Therefore, the native SiOx layer was kept in all the n-Si samples used in this study, unless specified otherwise.Fig. 1: Structural characterizations of NiCoFe-Bi/NiO/CuxO/n-Si heterostructure photoanode.a, b Bright-field TEM images of the cross-section of the sample. c Cross-sectional STEM–EDS elemental mappings of Ni, Co, Fe, Cu, Si, and O. d Dark-field STEM image of the film. e Overlapping the STEM image with the EDS mappings of the metallic elements.Photoelectrochemical performance of photoanodeThe PEC performance of the NiCoFe-Bi/NiO/CuxO/n-Si photoanode was tested in KOH electrolyte (pH 13.95 ± 0.05) containing potassium borate (KBi, 0.25 M) and iron sulfate (FeSO4, 50 μM) to facilitate the self-healing of the NiCoFe-Bi co-catalyst during the PEC test20,21. Figure 2a shows the photocurrent–potential (J–V) curves of the photoanodes after storing the NiO/CuxO/n-Si samples in the air for different times. Surprisingly, we found that the PEC activity of the NiCoFe-Bi/NiO/CuxO/n-Si photoanode gradually increased with the air exposure time. The freshly prepared NiCoFe-Bi/NiO/CuxO/n-Si photoanode (0 days) exhibited an onset potential (defined as the potential at which the current density reaches 1.0 mA cm−2) of ~0.94 V vs. RHE, a photocurrent density of ~29.2 mA cm−2 at 1.23 V vs. RHE, and a saturation photocurrent density of ~30.7 mA cm−2. With increasing exposure time of the NiO/CuxO/n-Si sample in the air, although the saturation photocurrent density did not increase, the onset potential shifted cathodically and reached a minimum of 0.88 V vs. RHE after 21 days. Meanwhile, the filling factor of the J–V curve improved considerably with air exposure time, and the trend continued even when the onset potential reached the minimum. The combination of the reduced onset potential and the increased fill factor contributed to a significant improvement in the HC-STH (Fig. 2b, c). The HC-STH of the NiCoFe-Bi/NiO/CuxO/n-Si photoanode was improved from 2.23% for the freshly prepared sample (0 days) to 4.22% after 28 days. In contrast, for the NiCoFe-Bi/NiO/n-Si photoanode without a CuxO interlayer, no such improvement was observed after exposing the samples to the air under the same conditions. As shown in Fig. 2d, there was no obvious change in the shape of the J–V curves of the NiCoFe-Bi/NiO/n-Si photoanodes after air exposure. The HC-STH of the NiCoFe-Bi/NiO/n-Si photoanodes remained at a relatively low level of ~1.90% (Fig. 2e, f). These PEC results suggest that the enhanced PEC activity of the NiCoFe-Bi/NiO/CuxO/n-Si photoanode is likely caused by changes in the CuxO interlayer during air exposure.Fig. 2: Evolution of the PEC activity of NiCoFe-Bi/NiO/n-Si photoanodes with and without a CuxO interlayer.a J–V curves of NiCoFe-Bi/NiO/CuxO/n-Si photoanodes. b HC-STH curves calculated from the J–V curves in (a). c Change of the maximum HC-STH values of the NiCoFe-Bi/NiO/CuxO/n-Si photoanodes with time. d J–V curves of NiCoFe-Bi/NiO/n-Si photoanodes. e HC-STH curves calculated from the J–V curves in (d). f Change of the maximum HC-STH values of the NiCoFe-Bi/NiO/n-Si photoanodes with time. All the photoanodes were tested in KOH electrolyte (pH 13.95 ± 0.05) containing 0.25 M KBi and 50 μM FeSO4 under AM 1.5G simulated sunlight illumination. The error bars in (c) and (f) show the statistics for five samples. All J–V curves were not iR-corrected.Following the in situ transformation of the CuxO interlayerTo gain insight into the change of the CuxO interlayer in the NiO/CuxO/n-Si sample during air exposure, a high-energy HAXPES was employed to characterize the buried interfaces. Compared to the conventional XPS with a soft X-ray source (Al Kα, 1486.6 eV), the analytical information depth of the HAXPES with a high-energy X-ray source (Cr Kα, 5414.9 eV) is increased by a factor of ~3, making it possible to directly characterize the transformation of the CuxO interlayer buried under the NiO layer (20 nm) (Fig. 3a). As shown in Fig. 3b and Supplementary Fig. S9, Cu and Si signals from the buried interfaces were indeed directly detected by HAXPES in addition to the Ni signal from the top NiO layer. More importantly, the in situ transformation of the CuxO interlayer buried below the NiO layer was revealed by the changes of the Cu 2p core-level HAXPES spectra of the NiO/CuxO/n-Si samples with different air exposure times (Fig. 3b). For the freshly prepared sample (0 days) the Cu 2p3/2 peak is centered at 932.4 eV (Fig. 3c). As the air exposure time increases, the peak center of Cu 2p3/2 gradually shifts to 934.4 eV (Fig. 3b). The Cu 2p3/2 peak of the NiO/CuxO/n-Si samples after air exposure can be deconvolved into two peaks centered at ~932.4 and ~934.6 eV, which corresponds to Cu2O and CuO species, respectively22,23. With the increase of air exposure time, the relative intensity of the Cu2O peak gradually decreases while that of the CuO peak increases (Fig. 3d–g), and CuO becomes the dominant species after 28 days (Supplementary Table S1). In addition to that, the intensity of the satellite peak at around 940–945 eV, which is characteristic of CuO species (Supplementary Fig. S10), progressively increases as the air exposure time increases. All these results suggest the in situ transformation of the CuxO interlayer from Cu2O to CuO after long-term air exposure. Meanwhile, the core-level Ni 2p3/2 spectra of the NiO/CuxO/n-Si samples show no obvious change after air exposure (Supplementary Fig. S9c). It is thus evident that the increase in the PEC activity of the NiCoFe-Bi/NiO/CuxO/n-Si photoanode after air exposure is caused by the in situ transformation of the CuxO interlayer from Cu2O to CuO.Fig. 3: Changes in the photoelectron spectroscopic characteristics of the CuxO interlayer after air exposure.a Differences in analytical depth between HAXPES with a Cr Kα hard X-ray source and conventional XPS with an Al Kα soft X-ray source. b Core-level Cu 2p HAXPES spectra of the NiO/CuxO/n-Si samples after different air exposure times. c–g Fitting of Cu 2p spectra at 0, 7, 14, 21, and 28 days. UPS spectra of the Cu2O (h) and CuO (i) deposited on Si substrate.The in situ transformation of the CuxO interlayer likely follows a grain boundary facilitated oxygen ion diffusion process, as shown in Supplementary Fig. S11. When oxygen molecules are adsorbed on the surface of the NiO/CuxO/n-Si sample, they are polarized by the vacant anion lattice sites on the NiO surface, resulting in the formation of \({{{{\rm{O}}}}}_{2{{{\rm{ads}}}}}^{-}\) ions24,25. The formed \({{{{\rm{O}}}}}_{2{{{\rm{ads}}}}}^{-}\) ions on the NiO surface could gradually diffuse into the bulk of the film through the grain boundaries. Owing to the unique structure of the EB-deposited NiO layer consisting of vertically aligned columns (Fig. 1b, Supplementary Fig. S5), there are a large number of grain boundaries penetrating the NiO layer. It is generally accepted that ion diffusion in crystalline solids proceeds more rapidly along grain boundaries than through the lattice26,27. Therefore, it is likely that the \({{{{\rm{O}}}}}_{2{{{\rm{ads}}}}}^{-}\) ions diffuse along the grain boundaries vertically and then gradually oxidize the CuxO interlayer along the horizontal direction (Supplementary Fig. S11).The proposed mechanism for the diffusion of oxygen ions is supported by the slight shift of the Ni 2p3/2 and O 1s HAXPES spectra toward higher binding energies. As shown in Supplementary Fig. S12, the core-level Ni 2p3/2 HAXPES spectra of the NiO/CuxO/n-Si samples were fitted with four peaks, two for the main feature and two for the satellite peak. Among the main features, peak 1 at 854.0 eV is attributed to the Ni2+ ions in the Ni–O bonding, whereas peak 2 at 856.1 eV is ascribed to the Ni3+ ions induced by the surface adsorbed \({{{{\rm{O}}}}}_{2{{{\rm{ads}}}}}^{-}\) ions28. Correspondingly, the O 1s HAXPES spectra of the NiO/CuxO/n-Si samples were fitted with two peaks originated from the lattice oxygen in the Ni–O bonding at 529.4 eV and the surface adsorbed oxygen species (i.e., \({{{{\rm{O}}}}}_{2{{{\rm{ads}}}}}^{-}\)) at 531.0 eV (Supplementary Fig. S13)29,30. From the deconvolved spectra, it is observed that the relative intensity of the peaks associated with the surface adsorbed oxygen ions (peak 2) for both the Ni 2p3/2 and O 1s spectra gradually increase with air exposure time (Supplementary Figs. S12 and S13). These results are consistent with our proposed mechanism for the polarization of oxygen molecules on the surface and the gradual diffusion of \({{\mbox{O}}}_{2{\mbox{ads}}}^{-}\) ions along the grain boundaries of the NiO layer.The effect of the in situ transformation of the CuxO interlayer on the interface energetics between n-Si and NiO was investigated using ultraviolet photoelectron spectroscopy (UPS). CuxO film was deposited on a Si substrate and the XPS analysis revealed that the as-deposited film was Cu2O, which was completely oxidized to CuO after long-term air exposure (Supplementary Fig. S14). The corresponding UPS spectra of the Cu2O and CuO layers were plotted in Fig. 3h, i. By subtracting the cut-off energy (Ecutoff) of the secondary electrons from the excitation energy of the He I line (21.22 eV), the Fermi levels (EF) of Cu2O and CuO are obtained at 4.71 and 5.10 eV below the vacuum level (Evac), respectively. The low binding energy edges of the UPS spectra show that the valence bands (EV) of Cu2O and CuO are below their EF by 0.46 and 0.68 eV, respectively. This suggests air exposure significantly changes the band positions of the CuxO interlayer. In contrast, the band positions of the NiO film are unchanged before and after air exposure (Supplementary Fig. S15), with the EF remaining at 4.48 eV below the Evac and the EV at ~0.92 eV below the EF. Combining these data with the known bandgaps (Eg) of the Cu2O (2.1 eV), CuO (1.5 eV), and NiO (3.6 eV) from the literature31,32 and the band positions of n-Si (electron affinity energy: 4.05 eV, work function: 4.27 eV, and Eg: 1.12 eV), the band diagrams of NiO/n-Si, NiO/Cu2O/n-Si, and NiO/CuO/n-Si heterojunctions are obtained (Supplementary Fig. S16). When NiO film is directly deposited on n-Si, the more positive EF of n-Si compared to NiO results in a built-in field and upward band bending at the NiO/n-Si interface, which provides the driving force for the transfer of photogenerated holes from n-Si to the electrode surface. When the Cu2O interlayer is introduced between NiO and n-Si, the more negative EF of Cu2O than NiO leads to an increase in the built-in field and band bending at the Cu2O/n-Si interface. The in situ transformation of Cu2O to CuO further increases the EF difference and hence the band bending at the CuO/n-Si interface. Therefore, it is expected that the introduction and in situ transformation of the CuxO interlayer leads to the formation of a higher built-in electric field and greater band bending within the heterojunction photoanode.Effect of the in situ transformation of the CuxO interlayerThe effect of the CuxO interlayer on the band bending at the NiO/n-Si interface was studied by preparing solid-state NiO/CuxO/n-Si heterojunctions using the as-prepared sample (with a Cu2O interlayer) and sample after long-term (~2 months) air exposure (with a CuO interlayer). The change in flat band potential (Efb) of the solid-state heterojunctions was studied by Mott–Schottky (M–S) analysis under dark conditions (Fig. 4a). The effective Efb of the NiO/n-Si heterojunction is 0.41 V, which increases to 0.46 V with a Cu2O interlayer and further increases to 0.52 V with a CuO interlayer. The slope of the fitted line in the M–S diagram allows the calculation of the n-Si doping density (ND) and the further obtaining of the effective potential barrier height (ΦB) of the heterojunctions (calculation details in Supplementary Note 1). The calculated ND of the n-Si is about 4.3 × 1015 cm−3, consistent with the resistivity (~1 Ω cm) of the n-Si wafer. The effective ΦB of the NiO/CuO/n-Si heterojunction reaches 0.74 eV, exhibiting a higher value compared to those of NiO/Cu2O/n-Si (0.68 eV) and NiO/n-Si (0.63 eV) heterojunctions (Supplementary Table S2).Fig. 4: Effect of the CuxO interlayer on the electrochemical properties of n-Si photoanode.a The M–S plots of n-Si-based solid-state heterojunctions with different layered structures measured at 100 kHz in the dark. b The photovoltage of the n-Si photoanodes with different layered structures. c Carrier lifetimes derived from OCP-decay curves at the light on–off transient for n-Si photoanodes with different layered structures. d PEIS of n-Si photoanodes with different layered structures measured at 1.0 V vs. RHE under AM 1.5G simulated sunlight illumination. Gray lines show the fitting of the PEIS data using the three-RC-unit equivalent circuit model shown in the inset. The charge separation efficiency (e) and charge injection efficiency (f) of n-Si photoanodes with different layered structures. All the photoanodes are modified with NiCoFe-Bi co-catalyst and tested in KOH electrolyte (pH 13.95 ± 0.05) containing 0.25 M KBi and 50 μM FeSO4. All J–V curves were not iR-corrected.The ΦB determines the theoretical maximum photovoltage (Vph, max) that can be generated by the heterojunction33,34. The actual photovoltage (Vph) of the n-Si photoanode was obtained by comparison with the onset potential of a non-photoactive p++-Si electrode (Fig. 4b). The onset potential (at 1 mA cm−2) difference between the NiCoFe-Bi/NiO/p++-Si electrode and the NiCoFe-Bi/NiO/n-Si photoanode is 540 mV, revealing that the photovoltage produced by the NiCoFe-Bi/NiO/n-Si photoanode is ~540 mV. Similarly, the photovoltages of the NiCoFe-Bi/NiO/Cu2O/n-Si and NiCoFe-Bi/NiO/CuO/n-Si photoanodes are estimated to be about 568 and 624 mV, respectively. Impressively, the photovoltage produced by the NiCoFe-Bi/NiO/CuO/n-Si photoanode is comparable to those achieved with the state-of-the-art p+n-Si-based photoanodes18,35,36.Similar Vph changes were obtained from the open-circuit potential (OCP) decay curves of the n-Si photoanodes in the dark and under illumination (Supplementary Fig. S17). The Vph of the photoanode without the CuxO interlayer is 530 mV, which increases to 560 and 620 mV with Cu2O and CuO interlayers, respectively. The carrier lifetime is derived from the OCP-decay curve (calculation details in Supplementary Note 2)37,38,39. As shown in Fig. 4c, the NiCoFe-Bi/NiO/CuO/n-Si photoanode exhibits the smallest carrier lifetime compared to the other photoanodes upon turning off the light, indicating a reduction in charge trapping at the interface. Moreover, photoelectrochemical impedance spectroscopy (PEIS) was employed to compare the charge transfer kinetics in the n-Si photoanodes with different layered structures40,41. As shown by the Nyquist plots in Fig. 4d, the radii of the double-semicircle curve for the NiCoFe-Bi/NiO/CuO/n-Si photoanode are significantly smaller than those for the other photoanodes, indicating reduced hindrance of charge transfer to the surface reaction sites. The Nyquist plots are fitted using a typical two-RC-unit equivalent circuit with the addition of a heterojunction RC-unit (inset of Fig. 4d). The detailed fitting parameters are listed in Supplementary Table S3. The introduction of the CuxO interlayer into the NiCoFe-Bi/NiO/n-Si photoanode leads to a dramatic reduction in the trapping resistance (Rtrap) and charge transfer resistance (Rct). Especially for the photoanode with a CuO interlayer, the Rtrap and Rct are approximately one order of magnitude lower than those in the NiCoFe-Bi/NiO/n-Si photoanode. These results suggest a reduction in charge recombination and an increase in the charge transfer rate across the interfaces.The bulk charge separation efficiency (ηsep) and surface injection efficiency (ηinj) of the n-Si photoanodes with different layered structures were further characterized (calculation details in Supplementary Note 3)34,42. The J–V curves of three different photoanodes were measured in the same electrolyte used for the PEC test with the addition of 0.5 M H2O2 as a hole scavenger (Supplementary Fig. S18). The absorption photocurrent density (Jabs) of each photoanode was calculated from the number of photons reaching the n-Si absorber layer (Supplementary Fig. S19). In Fig. 4e, the potential dependent ηsep curves show that the NiCoFe-Bi/NiO/CuO/n-Si photoanode achieves more efficient separation of photogenerated carriers at lower potentials (0.3–1.0 V vs. RHE). The potential dependent ηinj curves in Fig. 4f further reveal that the holes efficiently injected into the electrolyte from the NiCoFe-Bi/NiO/CuO/n-Si photoanode at lower potentials (0.85–1.3 V vs. RHE).Taking together, the mechanism for the enhanced PEC activity of the NiO/n-Si heterojunction photoanode by the CuxO interlayer is proposed. In the dark, the buried junction formed at the interface between n-Si and p-type metal oxides generates a built-in electric field and energy band bending with a barrier height of ΦB. Under illumination, the EF of n-Si splits, producing quasi-Fermi level for electrons (EF,n) and quasi-Fermi level for the holes (EF,p), the potential difference of which determines the photovoltage. The higher the ΦB generated by heterojunction in the dark, the higher the photovoltage obtained under illumination. Consequently, lower external bias is required for the injection of holes to participate the water oxidation reaction7,33. As illustrated in Fig. 5, the introduction and in situ transformation of the CuxO interlayer allow the NiO/n-Si heterojunction to acquire a higher ΦB, which provides a greater driving force for the separation and transport of photogenerated carriers. Consequently, a higher photovoltage is generated under illumination and the efficiency of PEC water oxidation is greatly improved. The PEC performance of n-Si photoanodes with and without a CuxO interlayer further confirms the critical role of the CuO interlayer (Supplementary Fig. S20). The pristine n-Si photoanode exhibits no PEC activity, with almost no photocurrent at applied potentials below 1.60 V vs. RHE. The deposition of a NiO layer forms a p–n heterojunction with the n-Si, which provides a driving force for the separation of photogenerated carriers (Fig. 5a). Nevertheless, the onset potential of the photoanode is still high because there is a substantial barrier (860 mV) for the injection of holes to participate in the OER (Fig. 5d). With the introduction of the Cu2O interlayer, the degree of band bending in the heterojunction is increased (Fig. 5b) and higher photovoltage is generated under illumination, which reduces the OER potential barrier to 830 mV (Fig. 5e). The transformation of Cu2O to CuO further increases the degree of band bending (Fig. 5c), which ultimately leads to the reduction of OER potential barrier to 780 mV (Fig. 5f). With the lowering of OER potential barrier, the onset potential undergoes a cathodic shift from 0.98 V vs. RHE for the NiCoFe-Bi/NiO/n-Si photoanode to 0.88 V vs. RHE for the NiCoFe-Bi/NiO/CuO/n-Si photoanode (Supplementary Fig. S20a). Consequently, the NiCoFe-Bi/NiO/CuO/n-Si photoanode achieves an HC-STH of 4.51%, which is 2.3 times higher than that of the NiCoFe-Bi/NiO/n-Si photoanode (Supplementary Fig. S20b). Consistently, the n-Si photoanodes without co-catalyst modification exhibited similar PEC performance variations, with the HC-STH of NiO/CuO/n-Si photoanodes (3.76%) being 2.4 times higher than that of NiO/n-Si photoanodes (1.54%) (Supplementary Fig. S21).Fig. 5: Band diagrams of n-Si photoanodes with different layered structures in the dark and under illumination.a–c Band diagrams of the n-Si photoanodes with different layered structures in the dark. d–f Quasistatic energy profile of the n-Si photoanodes with different layered structures in contact with the electrolyte under continuous illumination. Evac is the vacuum level, EC is the conduction band minimum, EV is the valence band maximum, EF is the Fermi level, EOER is the OER potential, ΦB is the barrier height caused by band bending, Vph is the open-circuit photovoltage, EF,n and EF,p are the quasi-Fermi levels of electrons and holes under illumination.Reactive e-beam deposition of CuO interlayerThe above results show that the CuO interlayer plays a central role in improving the PEC water oxidation activity of the NiO/n-Si heterojunction photoanode. However, the in situ transformation process is too long for practical device fabrication. To achieve the direct deposition of the CuO interlayer, we have developed a reactive EB deposition process by introducing O2 gas into the deposition chamber. The PEC performance of the NiCoFe-Bi/NiO/CuxO/n-Si photoanodes prepared under different O2 gas flow rates for the reactive deposition of CuxO interlayer are shown in Fig. 6a. The HC-STH of the photoanodes gradually increases with increasing O2 gas flow rates, achieving a maximum value of 4.56% at a flow rate of 17 sccm (Fig. 6b), similar to that achieved after long-term air exposure (Supplementary Fig. S20b). Meanwhile, the XPS characterization of CuxO film deposited at 17 sccm O2 flow reveals essentially single-phase composition of CuO (Supplementary Fig. S22). These results reveal the effectiveness of the reactive deposition in transforming Cu2O into CuO and further demonstrate the key role of the CuO interlayer in improving the efficiency of the NiO/n-Si heterojunction photoanode. To the best of our knowledge, the HC-STH of the optimized NiCoFe-Bi/NiO/CuO/n-Si photoanode is the highest value reported to date for n-Si photoanodes (Supplementary Fig. S23, Supplementary Table S4).Fig. 6: PEC performance of NiCoFe-Bi/NiO/CuO/n-Si photoanode.a J–V curves of NiCoFe-Bi/NiO/CuxO/n-Si photoanodes with CuxO interlayer deposited at different oxygen partial pressures. b HC-STH curves calculated from the J–V curves in (a). c Statistics of the HC-STH of n-Si photoanodes with different layered structures. The average HC-STH and the standard deviations are shown in the box chart. d Stability of the NiCoFe-Bi/NiO/CuO/n-Si photoanode measured at an applied potential of 1.2 V vs. RHE. e J–V curves of NiCoFe-Bi/NiO/CuO/n-Si photoanode before and after the stability test. All the photoanodes are modified with NiCoFe-Bi co-catalyst and tested in KOH electrolyte (pH 13.95 ± 0.05) containing 0.25 M KBi and 50 μM FeSO4 under AM 1.5G simulated sunlight illumination. All J–V curves were not iR-corrected.For the optimized NiCoFe-Bi/NiO/CuO/n-Si photoanode, the steady-state measurement confirms that the photocurrent density reaches 1.0 mA cm−2 at 0.88 V vs. RHE, and a stable photocurrent density of ~60 μA cm−2 is generated at a potential as low as 0.84 V vs. RHE (Supplementary Fig. S24). The saturation photocurrent density reaches ~30 mA cm−2, which is consistent with the integrated photocurrent density (30.2 mA cm−2) obtained from incident photon-to-electron conversion efficiency (IPCE) data (Supplementary Fig. S25). Statistical plots of the HC-STH values for n-Si photoanodes with a CuO interlayer (ten samples) demonstrate good reproducibility in PEC performance, with an average HC-STH of 4.45 ± 0.08%, compared with 2.27 ± 0.08% for the ones with a Cu2O interlayer and 1.85 ± 0.07% without an interlayer (Fig. 6c, Supplementary Fig. S26). The stability of the optimized NiCoFe-Bi/NiO/CuO/n-Si photoanode was tested at 1.20 V vs. RHE under AM 1.5G simulated sunlight (Fig. 6d). The photocurrent density of the photoanode reached 29.8 mA cm−2 and no significant decay was observed during the 100-h stability test. The long-term stability of the NiCoFe-Bi/NiO/CuO/n-Si photoanode is mainly attributed to the self-healing properties of the NiCoFe-Bi co-catalyst. During the water oxidation of Fe-based OER catalysts, the active intermediates of Fe in high valence states (FeVIO42−) are continuously leached out, resulting in the loss of active centers and the reduction of catalytic activity21. In the electrolyte containing Fe ions and BO33− ions, Co in the NiCoFe-Bi co-catalyst has a catalytic effect on the redeposition of Fe active centers, which effectively compensates for the loss of Fe active centers and achieves the dynamic equilibrium between leaching and redeposition20. A quantitative comparison between the photocurrent and the amount of oxygen evolved during the PEC test confirms nearly unity Faraday efficiency for OER (Supplementary Fig. S27). The J–V curves of the photoanode before and after the stability test (Fig. 6e) reveal that there is no change in the onset potential of the photoanode. In contrast, most previously reported n-Si photoanodes exhibited a degraded onset potential after the stability test (Supplementary Table S5), leading to a decreased HC-STH. The slightly decreased saturation photocurrent of our device after the stability test is probably due to the aging of the Fe-containing KOH electrolyte (i.e., iron oxidation), which lightly blocks sunlight in the visible spectrum. Nevertheless, the HC-STH of the photoanode remained nearly unchanged after the 100-h stability test, with an initial value of 4.34% and a final value of 4.31% (Supplementary Fig. S28). The efficiency and stability of the photoanode compare favorably with the previous literature on n-Si-based photoanodes (Supplementary Table S5).In summary, this work demonstrated the effectiveness of HAXPES in probing the interfacial properties of the buried interface and guiding the design of a more efficient PEC device. The HAXPES analyses revealed the in situ transformation of the CuxO interlayer in a NiO/n-Si heterojunction from Cu2O to CuO, which was responsible for the significantly improved PEC efficiency of the heterojunction photoanode after long-term air exposure. The introduction and in situ transformation of the CuxO interlayer optimized the interfacial energetics by creating a higher degree of band bending in the heterojunction, thus improving the photovoltage achieved under illumination. These results motivated us to develop a reactive EB deposition process for the direct deposition of a CuO interlayer, yielding a HC-STH of 4.56% for the fabricated NiCoFe-Bi/NiO/CuxO/n-Si photoanode. Our work not only demonstrated an effective interfacial modification strategy for NiO/n-Si heterojunction photoanode but also highlighted HAXPES as a powerful tool to characterize buried interfaces in solar energy conversion devices.
