Surface indium vacancies promote photocatalytic H₂O₂ production over In₂S₃

Abstract

Photocatalytic oxygen reduction reaction (ORR) represents a green and cost-effective means to produce the high-value industrial compound H₂O₂. However, the efficient H₂O₂ synthesis awaits in-depth knowledge of the ORR mechanisms for the design of highly active photocatalysts. In this work, surface indium vacancies (denote hereafter as V_In) have been introduced into In₂S₃, which significantly boosts the photocatalytic activity for H₂O₂ production, with an optimal H₂O₂ generation rate of 4.77 ± 0.05 mmol·h⁻¹·g_cat.⁻¹ under visible light illumination (λ ≥ 420 nm) and an apparent quantum efficiency (AQE) of 7.49 ± 0.01% at 420 ± 20 nm. Mechanistic analysis reveals the multifunctionality of V_In, such as enlarging the chemisorption capacity of O₂, enriching photo-generated electrons for ORR, endorsing high reducing power to photo-generated electrons, and promoting H₂O₂ desorption. These findings justify that surface cation vacancies open up new possibilities for H₂O₂ photosynthesis by leveraging their reaction dimensions in ORR.

Photocatalytic oxygen reduction reaction to produce H₂O₂ awaits in-depth knowledge for the design of active photocatalysts. Here, the authors identify that surface indium vacancies are the key to opening the activity of In₂S₃ for H₂O₂ photosynthesis.

Introduction

Hydrogen peroxide (H₂O₂) is an important platform chemical whose applications extend broadly across the field of the pharmaceutical industry, environmental remediation, and green energy economy, etc.^1–4 Conventionally, H₂O₂ is generated by the anthraquinone oxidation (AO) process, which has many drawbacks, including the usage of precious metal catalysts, large energy consumption, costly liquid-liquid extraction measures, and release of toxic waste, etc.⁵ Seeking alternative techniques for H₂O₂ production, therefore, is highly desired from the standpoint of energy-saving, cost-effectiveness, and environmental sustainability^5–8.

In this regard, H₂O₂ photosynthesis via oxygen reduction reaction (ORR) is highly attractive as it provides a green, sustainable, energy-conserving means to manufacture H₂O₂. Ideally, with only O₂ and water, H₂O₂ is continuously produced at the surface of a photocatalyst with no energy input other than solar insolation. Notwithstanding these promising scenarios, H₂O₂ photosynthesis is generally subject to low efficiency in terms of solar photon utilization even though a wealth of semiconductors are photocatalytically capable of H₂O₂ production, including metal chalcogenides, metal nitrides, metal organic frameworks (MOFs), and covalent organic frameworks (COFs), etc.^7,9–12 This is largely attributable to the discrepancy between the fast decay processes of photocarriers (i.e., photo-generated e^- and h⁺) and the slow reaction kinetics of ORR. For instance, the photocarrier lifetime of most semiconductors is in the range of pico- to nanoseconds (10⁻¹²s to 10⁻⁹s), which is several orders of magnitude shorter than the charge transfer rate of ORR (10⁻³s to 10¹s)^13–15. The majority of photocarriers, if not being separated instantaneously, would undergo unproductive recombination and therefore are inaccessible to ORR. How to bridge the huge gap of time scales between photocarrier lifetime and ORR kinetics is essential for efficient H₂O₂ photosynthesis, and it remains a great challenge. Apart from the elusive behavior of photocarriers, conventional semiconductors often lack a proper surface for H₂O₂ formation, being another factor for inefficient H₂O₂ photosynthesis. This is either because their surface contains limited adsorption sites for O₂ molecules or because the surface is in favor of H₂O₂ disproportionation¹⁶. Current strategy of loading cocatalysts to expedite ORR can lead to unwanted side effects. For instance, the presence of cocatalysts would not only increase the cost of H₂O₂ photosynthesis, block light penetration, but also diversify the ORR products (e.g. H₂O)¹⁷. New tactics to ameliorate the semiconductor surface for ORR without the aforementioned side effects are highly desired.

As common species in semiconductors, defects are of paramount importance as they are inherently associated with various physicochemical functions, including electrical, magnetic, and optical properties, etc^18–21. In the field of photocatalysis, defects are considered to be a critical factor that dramatically modulates the photocatalytic performance^22,23. This is because defects are involved intrinsically in the separation, migration, and transfer steps of photocarriers^13,21,24,25. However, there has been much debate on the role of defects in photocatalysis as contradictory impacts, i.e. both positive and negative ones, have been reported for systems that intentionally bring or eliminate defects^26–28. Recent studies have shown that the defects’ location matters and the surface defects are normally in favor of high activity levels, as opposed to the bulk ones^29–31. Nevertheless, how surface defects promote photocatalytic reactions needs further elucidation and in-depth mechanistic understandings of the role of surface defects are very helpful for the design and development of efficient photocatalysts.

Vacancy, as an important type of defects, is emerging as a hot research topic in photocatalysis as it brings many new possibilities in the photocarrier maneuver. Previous studies focused mostly on anion vacancies, such as S vacancies in ZnIn₂S₄^32–34, S vacancies in CdS^35,36, and N vacancies in g-C₃N₄^37,38, etc. However, cation vacancies, with distinct properties from anion vacancies in terms of charge, chemical environment, and energy level, etc., are much less explored. In this work, we have explored the role of surface indium vacancies (i.e. V_In) for H₂O₂ photosynthesis using In₂S₃ as a model compound. Such cation vacancies have played important roles in several critical steps of H₂O₂ photosynthesis, including O₂ adsorption, photocarrier separation and transferring, H₂O₂ desorption and disproportionation. These multifunctional roles endorse V_In as a key in opening the activity of In₂S₃ for H₂O₂ photosynthesis.

Results

Formation of indium vacancies

The V_In was introduced into In₂S₃ during hydrothermal synthesis with the aid of ethylene glycol (EG). EG is a bidental chelate compound that has a high coordination capacity to In cations. The presence of EG therefore can retain In cations in the solvent, resulting in a deficiency of In in the product In₂S₃ during hydrothermal reactions, as illustrated in Supplementary Fig. 1. Figure 1a illustrates the XRD patterns of In₂S₃ samples that were prepared using various EG contents. Although all patterns match with the standard ones of In₂S₃, the reflection intensity has a clear degradation by the increment of EG content. At high EG content (>90 vol%), some reflections, e.g. (116), (020) and (233) at 23.3°, 28.7°and 43.6°, respectively, are missing, indicative of much-weakened crystallinity of these In₂S₃ samples. The successful formation of V_In is affirmed by a signal at g = 2.005 in the EPR spectra (Fig. 1b). The intensity of the signal increases monotonically with the EG content, implying that the V_In concentration can be effectively tuned by controlling the amounts of EG used. This deduction is verified by the inductively coupled plasma mass spectrometry (ICP-MS) analysis which quantitatively determines the V_In content according to the stoichiometry between In and S, as shown in Fig. 1c and Supplementary Table 1. Notably, a V_In content as high as 46.7% can be attained by using 98 vol% EG. Nevertheless, further increasing the EG content, e.g. 100 vol%, fails to produce In₂S₃. Presumably, the crystal architecture of In₂S₃ collapses in the case of a severe shortage of In. Because V_In normally contains positive charges, the presence of V_In positively shifts the surface potential of In₂S₃, as is verified by the zeta potential measurements (Fig. 1d). Unless otherwise specified, the In₂S₃ prepared with 94 vol% EG is used to exemplify V_In-containing In₂S₃ and is denoted as In₂S₃-V as a discrimination to pristine In₂S₃. The formation of V_In defects is further supported by the UV-vis DRS spectra in Fig. 1e. The broad absorption tail above the absorption edge corresponds to the sub-bandgap absorption of defects, which is well observed by comparing In₂S₃-V with In₂S₃. The microstructures of as-prepared In₂S₃ samples are all characterized by flower-like morphologies and are porous according to BET surface analysis which are typically seen in metal sulfides prepared by hydrothermal method (see Fig. 1f, Supplementary Fig. 2, 3, Supplementary Table 2). Despite the similar morphologies, pristine In₂S₃ and In₂S₃-V are distinct under high-resolution TEM (HR-TEM) inspections. Crystal flaws, e.g. lattice dislocations, can be identified in the HR-TEM image of In₂S₃-V. These crystal imperfections are attributable to V_In which substantially impairs the periodicity of the crystal lattice, as indicated by the contrast intensity profile (Fig. 1g, h). Nevertheless, In₂S₃ and In₂S₃-V are essentially identical in terms of microstructures with a homogeneous distribution of In and S elements throughout the sample particles (Supplementary Figs. 4, 5).

Fig. 1. Structures of the In₂S₃ and In₂S₃-V.

a XRD patterns of In₂S₃ synthesized with different EG volume percentage (EG vol%), standard patterns of In₂S₃ ( JCPDS 00-005-0729) are included as vertical bars. b EPR spectra of In₂S₃ synthesized with different EG volume percentage. c V_In content in In₂S₃ as a function of EG volume percentage. d Zeta potential of In₂S₃ synthesized with different EG volume percentage. e UV-vis DRS spectra of In₂S₃ with and without V_In. f FE-SEM image of In₂S₃-V. g HR-TEM image of pristine In₂S₃. h Contrast intensity profile along the dashed red line of g. i HR-TEM image of In₂S₃-V, the label “T” marks the dislocation of crystal lattice. j Contrast intensity profile along the dashed red line of i. Source data for XRD, EPR, V_In content, zeta potential, and UV-vis DRS spectra are provided as a Source Data file.

The location of the indium vacancies

The chemical environment of In in In₂S₃ and In₂S₃-V was studied by X-ray absorption spectra (XAS). The presence of V_In leads to only a marginal shift of the absorption edge in the X-ray absorption near-edge structure (XANES) spectra (Fig. 2a), indicating that V_In has negligible impact on the valence of In. Similar observations are also noticed in the extended X-ray absorption fine structure (EXAFS) spectra. Both In₂S₃ and In₂S₃-V reveal a single In-S bond signal around 2.53 Å with almost identical coordination numbers, as revealed by the Fourier-transformed k³-weighted EXAFS spectra and wavelet-transformed EXAFS contour map (Fig. 2b, c, Supplementary Table 3). Thereby, the chemical environment of In is essentially the same in In₂S₃ and In₂S₃-V. Nevertheless, the intensity of the In-S bond signal is slightly enhanced when comparing In₂S₃-V with In₂S₃, suggesting stronger In-S interactions in the presence of V_In^39,40. Although XAS shows a collective similarity between In₂S₃-V with In₂S₃, XPS reveals some differences in the chemical environment of surface In. Compared with pristine In₂S₃, In₂S₃-V shows a clear red-shift in the binding energy of In 3 d states which is accompanied by a blue-shift of the S 2p states (Fig. 2d and Supplementary Fig. 6), corresponding to a lower positive charge on In. The peak shape of In 3d states also reveal some asymmetry in the presence of V_In, likely due to the fact that V_In brings inhomogeneity to the chemical environment of surface In cations. These In cations will experience different chemical environments when they are neighboring to or distant from V_In. However, this phenomenon is restricted to the top surface because removing the top surface layer by Ar⁺ etching eliminates the differences in the binding energy and mitigates the peak asymmetry (Fig. 2e, f). Accordingly, one can infer that the V_In defects are probably accommodated at the surface of In₂S₃-V. To verify this deduction, In₂S₃-V was probed by the synchrotron radiation photoelectron spectroscopy (SRPES) technique and was compared with In₂S₃. The SRPES has tunable photon energies and can provide atomic fraction information at variable depths^41–43. Fig. 2g, h illustrate the SRPES spectra of S 2p and In 3d states of In₂S₃-V with different photon energies, respectively. By comparing with pristine In₂S₃ (Supplementary Fig. 7) and considering the photoionization cross-section^41,44, one can determine the atomic fraction of both samples as a function of photon energies which correspond to the detection depth. It can be seen from the Fig. 2i that In₂S₃-V has a much lower In fraction and a much higher S fraction at the top surface than those of pristine In₂S₃ while it bears almost equivalent In and S content in the bulk with those of pristine In₂S₃. In other words, the In₂S₃-V has its V_In defects mainly at the upmost surface, being consistent with the previous deduction. As these In₂S₃ samples are characterized by relatively large surface area, it is favorable to accommodate V_In at the surface (Supplementary Table 2).

Fig. 2. Spectroscopic analysis of V_In.

a Normalized XANES spectra at In K-edge of In₂S₃, In₂S₃-V and In₂O₃, inset shows the enlarged region marked by the dashed rectangle. b Fourier-transformed EXAFS spectra at In K-edge of In₂S₃, In₂S₃-V and In₂O₃. Signals assignable to In-S and In-O bond are labeled. c Wavelet-transformed EXAFS contour map of In₂S₃, In₂S₃-V and In₂O₃. Signals assignable to In-In, In-S, and In-O bond are labeled. d XPS spectra at In 3 d state (3 d_3/2 and 3 d_5/2) of In₂S₃ and In₂S₃-V. e XPS spectra at In 3 d state (3d_3/2 and 3d_5/2) of In₂S₃ before and after Ar⁺ ion etching. f XPS spectra at In 3d state (3d_3/2 and 3d_5/2) of In₂S₃-V before and after Ar⁺ ion etching. g SRPES spectra at S 2p state (2p_1/2 and 2p_3/2) of In₂S₃-V with different photon energies. h SRPES spectra at In 3d state (3d_3/2 and 3d_5/2) of In₂S₃-V with different photon energies. i In and S atom fraction for In₂S₃ and In₂S₃-V as a function of photon energy. Source data for XANES, EXAFS, XPS, and SRPES spectra are provided as a Source Data file.

H₂O₂ photosynthesis

The photocatalytic properties in the application of H₂O₂ synthesis were then explored to elucidate the role of V_In. Figure 3a summarizes the photocatalytic activities of In₂S₃ with variable V_In content. It can be seen that V_In plays a positive role in H₂O₂ photosynthesis as introducing V_In significantly enhances the production rate of H₂O₂. At the optimal V_In content (~32.08%), the H₂O₂ production rate approaches 4.77 ± 0.05 mmol·h⁻¹·g_cat.⁻¹ which is almost three times as high as pristine In₂S₃ with an absence of V_In. The surface area probably has only marginal contribution here as only a moderate enlargement of surface area (~1.13 fold) is observed at this optimal V_In content (Supplementary Table 2). In general, the photocatalytic activity reveals a volcano-shape relation with V_In content. Such a nonlinear trend in the activity enhancement is probably due to the negative roles of defects that may offset the positive ones at high defect content⁴⁵. Apart from the higher H₂O₂ yield, the presence of V_In also modifies the H₂O₂ production profile, as shown in Fig. 3b. Compared with In₂S₃, In₂S₃-V demonstrates a much stable temporal increment in H₂O₂ yield, implying that H₂O₂ disproportionation is inhibited in the presence of V_In. This is quantitatively assessed by calculating the formation rate constant K_f and the decomposition rate constant K_d of H₂O₂ by fitting the profile using the following Eq. (1)^9,46:

$$\left[{\mathrm{H}}_2{\mathrm{O}}_2\right]=\frac{{\mathrm{K}}_{\mathrm{f}}}{{\mathrm{K}}_{\mathrm{d}}}\left(1-e^{{\mathrm{K}}_{\mathrm{d}}t}\right)$$ 1

Fig. 3. H₂O₂ photosynthesis.

a Photocatalytic H₂O₂ production rate as a function of V_In under visible light (λ ≥ 420 nm). b Temporal photocatalytic H₂O₂ production profile of In₂S₃ and In₂S₃-V. c K_f and K_d constant of In₂S₃ and In₂S₃-V. d Decomposition profile of 1 mM H₂O₂ in the presence of In₂S₃ and In₂S₃-V. e Action spectra of In₂S₃-V for H₂O₂ photosynthesis, UV-vis DRS spectra were included for comparisons. f Repeated usage of In₂S₃-V for H₂O₂ photosynthesis, sample powders were collected at the end of each testing cycle (~3 h) by centrifugation and were reused without any treatments. The error bars represent the standard deviations of three independent measurements. The error bars of wavelength indicate the spectral width of the monochromatic light. Reaction conditions: 5 mg photocatalyst powders, 60 mL deionized water as the solvent, 10 mL TEA as the sacrificial agent, a 300 W Xe lamp filtered with a cutoff filter (λ > 420 nm) as the light source, 293 K, continuous O₂ bubbling. Source data for photocatalytic performance are provided as a Source Data file.

Compared with In₂S₃, the K_f of In₂S₃-V increases by a factor of 1.8 and K_d decreases by a factor of 3, confirming that V_In helps to stabilize H₂O₂ (Fig. 3c). To further verify this point, H₂O₂ decomposition experiments were carried out in the presence of In₂S₃ and In₂S₃-V, respectively. As shown in Fig. 3d, In₂S₃-V delivers much lower catalytic activity in decomposing H₂O₂ than In₂S₃, affirming the role of V_In in stabilizing H₂O₂. For better understanding of the H₂O₂ synthesis procedures, the action spectra of In₂S₃-V, i.e. AQE vs wavelength, was collected, as plotted in Fig. 3e and tabulated in Supplementary Table 4. The AQE varies reasonably well with the UV-vis DRS spectra and has an onset around 700 nm, indicating that H₂O₂ generation by In₂S₃-V is photon-driven and proceeds via bandgap excitations. The slight deviation in the action spectra above 500 nm is attributable to the fact that defect absorption dominates the absorption profile at this spectral range and does not contribute to the generation of photocarriers. Notably, the AQE approaches as high as 7.49 ± 0.01 % at 420 ± 20 nm, being competitive among photocatalysts reported for H₂O₂ photosynthesis (Supplementary Table 5)^9,47–49. Meanwhile, the repeated testing cycles suggest that In₂S₃-V can maintain its activity for at least 5 cycles up to 15 h, and the XRD patterns and SEM image of the used In₂S₃-V remain almost unchanged, highlighting its good stability and durability for H₂O₂ synthesis (Fig. 3f and Supplementary Fig. 8). XPS and EPR analysis further suggest that the surface state and the V_In of In₂S₃-V are maintained before and after H₂O₂ photosynthesis (Supplementary Figs. 9, 10), affirming the robustness of the V_In. In addition, In₂S₃-V maintains its performance in various types of water (Supplementary Fig. 11) and is active even in pure water without a sacrificial reagent (Supplementary Fig. 12 and Table 5), suggesting good potentials for practical applications.

Reaction mechanisms

Given the much-enhanced activity in the presence of V_In, we sought to explore the reaction mechanisms for H₂O₂ photosynthesis. Firstly, the reaction pathways were investigated by isotope-labeling experiment and by varying the reaction conditions. Figure 4a shows the results of In₂S₃-V using H₂¹⁸O as the reaction solvent. The H₂O₂ generated in the system was decomposed into O₂ by MnO₂ which was then analyzed by mass spectra^50–52. Only ¹⁶O₂ was detected under this condition, indicating that H₂O₂ was produced by ORR pathways (O₂ + 2H⁺ + 2e^- → H₂O₂) rather than water oxidation reaction pathways (WOR, 2H₂O + 2 h⁺ → H₂O₂ + 2H⁺). This is further confirmed by the rotating ring-disk electrode (RRDE) measurements which suggest that WOR produces only O₂ rather than H₂O₂ at the surface of In₂S₃ and In₂S₃-V (Supplementary Fig. 13). In addition, H₂O₂ photosynthesis over In₂S₃-V was tested in the presence of different reaction mediums, including air, coumarin, Ar and NaIO₃, respectively. Compared with pure O₂, H₂O₂ yield undergoes appreciable decrement in air, complete annihilation in Ar and electron trapping agent NaIO₃, confirming that ORR is the only reaction pathway for H₂O₂ generation (Fig. 4b). In the presence of coumarin, a known reagent to capture •OH, the H₂O₂ yield is reduced by nearly 50%, suggesting that •OH is an important intermediate for H₂O₂ production^53,54. To elucidate the potential role of •O₂^-, the H₂O₂ yield was also evaluated in the presence of p-benzoquinone (PBQ, •O₂^- capture agent) using isopropyl alcohol (IPA) as a sacrificial agent. Under this condition, H₂O₂ production was completely interrupted (Supplementary Fig. 14), suggesting that •O₂^- is a critical intermediate for H₂O₂ synthesis. From the above control experiments, one can realize that H₂O₂ is generated through ORR with both •OH and •O₂^- as important intermediates and there are probably multiple reaction pathways during the generation of H₂O₂^53,54. This is further supported by EPR spin-trapping experiment. Using 5,5-dimethyl-1-pyrroline N-oxide (DMPO) as a spin-trapping agent, the EPR signals for the DMPO-•O₂^- and DMPO-•OH adduct can be identified for both In₂S₃ and In₂S₃-V (Fig. 4c, d). Considering the fact that H₂O₂ is generated via ORR, the presence of •OH is attributable to the O₂ reduction mediated processes⁵⁵. It is noteworthy that In₂S₃-V shows much stronger EPR signals for both radicals than In₂S₃, being accordant to their photocatalytic activity.

Fig. 4. Origin of H₂O₂.

a Mass spectra of the reaction products (¹⁶O₂, ¹⁸O₂) using isotope-labeled water (i.e. H₂¹⁸O) over In₂S₃-V. The inset shows the reaction pathways. b Photocatalytic H₂O₂ yield of In₂S₃-V under different reaction conditions. The error bars represent the standard deviations of three independent measurements. Reaction conditions: 5 mg catalysts, 60 mL deionized water, 10 mL TEA, under visible light illumination (λ ≥ 420 nm) at room temperature. c EPR spin-trapping experiments for In₂S₃ and In₂S₃-V using DMPO to detect •O₂^- (DMPO-•O₂^-). d EPR spin-trapping experiments for In₂S₃ and In₂S₃-V using DMPO to detect •OH (DMPO-•OH). Source data for mass spectra, photocatalytic yield of In₂S₃-V under different reaction conditions, and EPR spectra are provided as a Source Data file.

The catalytic role of V_In during ORR was then studied to understand the high activity of In₂S₃-V. Figure 5a compares the temperature-programmed desorption (TPD) profile of O₂ for both In₂S₃ and In₂S₃-V. In₂S₃-V not only delivers stronger TPD signals than In₂S₃ but also desorbs O₂ at higher temperatures. Besides, In₂S₃-V reveals a new desorption peak around 410 °C, attributable to O₂ adsorbed induced by V_In. These results suggest that V_In considerably enhances the chemisorption capacity of O₂ at the surface of In₂S₃. The positive role of V_In in ORR can also be realized by cyclic voltammetry (CV) measurements in Fig. 5b. The In₂S₃-V shows a cathodic ORR peak at a much lower potential than In₂S₃, indicative of much easier ORR reactions in the presence of V_In. The rotating disk electrode (RDE) test further reveals that In₂S₃-V owns a much higher ORR current than In₂S₃ at various rotational speeds (Supplementary Fig. 15), implying that V_In is in favor of fast ORR reaction kinetics. According to the Koutecky-Levich plots, the corresponding electron transfer numbers are determined to be 1.65 and 1.89 for In₂S₃ and In₂S₃-V, respectively, suggesting that V_In promotes H₂O₂ generation in a more favorable two electron pathway (Fig. 5c). To probe the reaction intermediates during H₂O₂ formation, in situ diffuse reflectance infrared Fourier transformed spectra (DRIFTS) were collected at various reaction time with humidified O₂ as the reactant. Figure 5d, e illustrates the DRIFTS results of In₂S₃ and In₂S₃-V. The DRIFTS of both samples are characterized by several common absorption peaks shortly after light illumination, assignable to various reaction intermediates. The absorption peaks around 820 cm⁻¹ and 1435 cm⁻¹ are attributable to the vibration and stretching mode of O-O bond of surface-adsorbed H₂O₂. The peak around 950 cm⁻¹ belongs to the O-O stretching mode of surface-adsorbed O₂^56,57. The peaks around 1110 cm⁻¹ and 1260 cm⁻¹ can be assigned to the absorption of •O₂^- and 1,4-endoperoxide intermediates, respectively^58,59. Compared with In₂S₃, In₂S₃-V demonstrates much stronger absorption signals at these positions, indicating that V_In contributes to higher concentration of reaction intermediates. Notably, a very strong absorption peak emerges at 888 cm⁻¹ for In₂S₃-V. This peak is not observable for pristine In₂S₃ therefore it most likely originated from the O-O stretching mode of V_In-adsorbed O₂ as it is only slightly red-shifted compared to the absorption peak of surface-adsorbed O₂ (950 cm⁻¹). This assignment is consistent with previous TPD results that V_In enhances the chemisorption capacity of O₂. From the above results, one can realize that V_In is deeply involved in the H₂O₂ generation processes, particularly in the O₂ adsorption stage. Such useful properties are attributable to the inherent character of V_In as the dynamic migration of V_In at the surface of In₂S₃ is unlikely to happen at the conditions for H₂O₂ photosynthesis, according to aberration-corrected transmission electron microscopic analysis and theoretical studies (Supplementary Figs. 16, 17). Meanwhile, in situ Raman spectroscopic analysis suggests that there is currently no evidence for the surface reconstruction during H₂O₂ photosynthesis (Supplementary Fig. 18).

Fig. 5. Catalytic role of V_In.

a O₂ TPD experiment over In₂S₃ and In₂S₃-V. b CV curves of In₂S₃ and In₂S₃-V at negative bias in the O₂-saturated solution. c Koutecky−Levich plots of In₂S₃ and In₂S₃-V based on RDE analysis. The electron transfer numbers (n) are deduced by fitting the plots. In situ DRIFTS spectra for H₂O₂ production over d In₂S₃-V and e In₂S₃. Absorption signals assignable to O₂ (O-O bond), H₂O₂ (HO-OH bond), •O₂^-, endoperoxide, and surface adsorbed H₂O₂ (HOOH*) are labeled. O₂ absorption signals associated with V_In is marked by the red arrow. Source data for TPD experiment, CV curves, Koutecky−Levich plots, and in situ DRIFTS spectra are provided as a Source Data file.

The behavior of photocarriers

The photocarrier (e^- and h⁺) behavior in the presence of V_In was then exploited to understand the photocatalytic performance of In₂S₃-V. Figure 6a compares the photoluminescence (PL) spectra of In₂S₃ and In₂S₃-V. Both samples reveal a broad PL emission band which culminates around 610 nm. Considering the bandgap of In₂S₃ (E_g ~ 2.2 eV) and stokes shift, the PL signal around 610 nm is assignable to the band edge emission of photocarriers. This PL signal is severely attenuated when comparing In₂S₃-V with In₂S₃, suggesting that radiative photocarrier recombination is inhibited in the presence of V_In. Time-resolved PL (TRPL) decay spectra indicate that In₂S₃-V owns a much slower PL decay process than In₂S₃ (Fig. 6b), corresponding to a much longer photocarrier lifetime. Using a double first-order exponential decay model (Supplementary Table 6), the amplitude-weighted average PL decay lifetime is determined to be 4.5 ns for In₂S₃-V which is almost five-fold longer than that of In₂S₃ (0.93 ns). The better photocarrier separation in the presence of V_In is also supported by 2,2,6,6-tetramethylpiperidinyloxy (TEMPO) electron-scavenger experiments⁶⁰. As shown in Fig. 6c, In₂S₃-V is more effective in quenching the triplet TEMPO signals than In₂S₃, corresponding to more photo-generated electrons available for surface reduction reactions. Electrochemical impedance spectra (EIS) show that V_In facilitates interfacial charge transfer as In₂S₃-V owns a smaller charge transfer resistance than In₂S₃, both in the dark and under visible light illumination (Fig. 6d). To further explore the photocarrier dynamics in the presence of V_In, femtosecond transient absorption (fs-TA) spectra were collected. The results of In₂S₃ and In₂S₃-V are summarized in the two-dimensional (2D) contour plots⁶¹, as illustrated in Fig. 6e, f. For In₂S₃, two fs-TA signals can be identified in the spectral range measured. The first one is a negative response in the spectral range from 500 nm to 600 nm. The second one is a positive response in the spectral range from 660 nm to 750 nm. As the PL emission is centered above 600 nm, the negative signal in the fs-TA spectra is attributable to the ground-state bleaching (GSB) rather than the stimulated emission (SE). The positive signal, on the other hand, belongs to the excited-state absorption (ESA) whose origin is most likely from photo-generated holes in the valence band (VB) according to photoelectrochemical (PEC) analysis in the following section. For In₂S₃-V, however, the GSB signal is not observable and the fs-TA spectra contain a broad ESA signal ranging from 540 nm to 750 nm. The absence of GSB is probably due to accumulated photocarriers in the presence of V_In that deliver strong positive ESA signals to counteract with negative ones. The broad ESA signal of In₂S₃-V, however, decays much faster than that of In₂S₃, as indicated by the time evolution of their TA spectra (Fig. 6g, h). Fitting the ESA decay profile at 710 nm by a double exponential function gives an average lifetime of 139.93 ps and 21.82 ps for In₂S₃ and In₂S₃-V, respectively (Fig. 6i and Supplementary Table 7). Because In₂S₃-V has a much longer TRPL decay lifetime than In₂S₃, its accelerated ESA decay implies the occurrence of sub-bandgap transition events for the photo-generated holes, for instance, transitions to the mid-gap states.

Fig. 6. Photocarrier separation and transfer.

a Photoluminescence (PL) spectra of In₂S₃ and In₂S₃-V. b Time-resolved PL (TRPL) decay spectra of In₂S₃ and In₂S₃-V at the emission of 610 nm, inset lists the fitted average PL decay lifetime (τ_av). c EPR spectra of TEMPO electron-scavenger experiments (TEMPO-e^-) over In₂S₃ and In₂S₃-V. d EIS spectra of In₂S₃ and In₂S₃-V under dark and visible light illumination. e 2D contour plots of fs-TA spectra for In₂S₃. f 2D contour plots of fs-TA spectra for In₂S₃-V. g Time evolution of the TA spectra for In₂S₃. h Time evolution of the TA spectra for In₂S₃-V. Signals belong to the ground-state bleaching (GSB) and the excited-state absorption (ESA) are marked by dashed lines. i TA decay kinetic traces for In₂S₃ and In₂S₃-V at 710 nm, insets lists the fitted average decay lifetime (τ_av). Source data for PL, TRPL, EPR, EIS, and fs-TA spectra are provided as a Source Data file.

The band structure

To rationalize the photocarrier behavior in the presence of V_In, the band structures of In₂S₃ and In₂S₃-V were determined both theoretically and experimentally. Firstly, the density of states (DOS) near fermi level (E_f) is calculated based on density functional theory (DFT). Figure 7a compares the DOS of In₂S₃ with and without V_In. The presence of V_In introduces additional mid-gap states above the valence band that are formed dominantly by S 3p orbitals (Supplementary Figs. 19, 20). This is probably due to the fact that removal of lattice In leads to some non-binding S 3p orbitals whose energy levels are higher than the bonding ones. These mid-gap states can trap/accept photo-generated holes which are helpful to prevent photocarrier recombination, as supported by previous PL and TRPL results. The trap/acceptance of photo-generated holes is verified by PEC analysis. As shown in Fig. 7b and Supplementary Fig. 21, the anodic photocurrent of In₂S₃-V is considerably smaller than that of In₂S₃ because of hole trapping/accepting. This can be understood by the fact that holes accommodated in the mid-gap states not only have weaker oxidizing power but also become more difficult to be transferred. As a further support to this point, TEMPO hole-scavenger experiments were performed and the results are summarized in Supplementary Fig. 22. In contrast to TEMPO electron-scavenger experiments (Fig. 6c), In₂S₃-V provides less photo-generated holes than In₂S₃ for surface oxidation reactions, presumably due to the trap/acceptance of holes by mid-gap states. These results help to identify the broad ESA signals of fs-TA spectra to holes in the VB due to their accelerated degradation in the presence of V_In. It also explains the increment of photo-generated electrons for In₂S₃-V during the TEMPO electron-scavenger experiments as photocarrier recombination is inhibited. To accurately determine the band edge positions, ultraviolet photoelectron spectra (UPS) were collected for In₂S₃ and In₂S₃-V using a synchrotron radiation source. Their UPS secondary electron cutoff (UPS-SEC) spectra and UPS valence band (UPS-VB) spectra are illustrated in Fig. 7c and d, respectively. The work function (WF) can be determined directly from UPS-SEC spectra by subtracting the photon energy (40 eV) with SEC energy. Band edge positions can then be deduced by combining bandgap value (Supplementary Fig. 23) with UPS-VB spectra which correspond to the energy gap between valence band top and WF. As schematically depicted in Fig. 7e, the conduction band edge (CB) of In₂S₃-V is negatively shifted by 0.6 V compared to In₂S₃. Presumably, the photo-generated electrons in In₂S₃-V will have a higher reducing power than those of In₂S₃, being beneficial for ORR reactions to produce H₂O₂. Notably, the valence band position of In₂S₃-V cannot fulfill the thermodynamic requirements of WOR pathways to produce H₂O₂, further suggesting the ORR is the only pathway for H₂O₂ generation. From the above analysis, it can be recognized that V_In has multifunction to the behavior of photocarriers: (1) it enriches photo-generated electrons by introducing S-based mid-gap states to trap/accept photo-generated holes; (2) it endorses high reducing power to photo-generated electrons by negatively shifting band edge positions.

Fig. 7. The band structure.

a DFT-calculated DOS for In₂S₃ with and without V_In. Mid-gap states associated with V_In are labeled by a dashed red cycle. b Photocurrent density of In₂S₃ and In₂S₃-V. c UPS secondary electron cutoff (UPS-SEC) spectra of In₂S₃ and In₂S₃-V. The incident photon energy (hν) is 40 eV. The work function (WF) is deduced by subtracting the photon energy with the cutoff energy. d UPS valence band (UPS-VB) spectra of In₂S₃ and In₂S₃-V. e Schematic illustration of UPS-deduced band edge positions (CB and VB) of In₂S₃ and In₂S₃-V. Fermi level (E_f) is denoted by dashed lines and the potential scale is defined by referring to the vacuum level at 0.0 V and to the standard hydrogen electrode at pH = 0. The processes of photocarrier generation, migration, recombination, and trapping by mid-gap states (S 3p) are indicated by the dashed arrows. Standard redox potentials of •O₂^-/O₂, H₂O₂/H₂O, and •OH/H₂O are also marked. Source data for DFT-calculated DOS, photocurrent, and UPS spectra are provided as a Source Data file. The optimized geometries of atoms for DFT calculation are provided in Supplementary data 1.

Theoretical calculations

To gain a deeper insight into the reaction mechanisms, the ORR to H₂O₂ was theoretically investigated by density functional theory (DFT) focusing on the role of V_In. The O₂ adsorption sites and geometries of surface atoms were optimized using In₂S₃ (010) surface according to the TEM analysis (Supplementary Figs. 4, 5). The adsorption sites with the highest absorption energy were used for subsequent calculations (Supplementary Fig. 24). Figure 8a and b compare the optimized geometries of ORR to H₂O₂ intermediates over In₂S₃ (010) surface without and with V_In, respectively. Different configurations of the intermediates can be noticed when the surface is in the presence and absence of V_In. Specifically, the oxygen of each intermediate adheres to the surface by “end-on” type (Pauling-type) in the absence of V_In and by “side-on” type (Yeager-type) in the presence of V_In. Apart from the adsorption configurations, the O₂ adsorption site also changes from a surface In site to two unsaturated In sites beneath V_In in the absence and presence of V_In, respectively (Fig. 8a and b). These differences presumably stem from the fact that V_In strongly promotes the adsorption of O₂, as revealed by previous O₂ TPD and DRIFTS results. As indicated by the plot of charge density differences in Fig. 8c, the “side-on” type configuration is more favorable for electron transfer from In₂S₃ to adsorbed O₂ than “end-on” type because there are now two oxygens bonded to the surface. In addition, the calculated free energy diagrams for ORR to H₂O₂ over In₂S₃ and In₂S₃-V are illustrated in Fig. 8d. In both cases, the ORR undergoes a sequence of energetically downhill steps to produce H₂O₂, indicating that both samples are in favor of H₂O₂ formation. Notably, the In₂S₃-V shows a larger energy decrease than In₂S₃ for the O₂ adsorption step, being consistent with previous observations that V_In promotes O₂ adsorption. However, both In₂S₃ and In₂S₃-V reveal an energetically uphill step for desorption of H₂O₂, suggesting that the detachment of H₂O₂ from the surface is the rate-determining step. Because of V_In, such an uphill step is considerably decreased when comparing In₂S₃-V with In₂S₃. These results well explain the higher photocatalytic activity of In₂S₃-V than In₂S₃ as difficulties in desorption of H₂O₂ not only slow down the reaction rate but also increase the likelihood of H₂O₂ disproportionation. Indeed, the in situ DRIFTS spectroscopic analysis reveals that In₂S₃-V shows much faster H₂O₂ desorption than In₂S₃ (Supplementary Fig. 25), confirming the critical role of V_In at the desorption stage of H₂O₂ production. Although DFT has limitations for modeling the exact material under catalytic conditions, the conclusions drawn from DFT are helpful to understand the role of V_In.

Fig. 8. Theoretical study of H₂O₂ generation.

The calculated geometries of ORR to H₂O₂ intermediates (*O₂, *OOH, and *H₂O₂) at the (010) surface of a In₂S₃ and b In₂S₃-V. c The calculated charge density differences and the corresponding charge accumulated during the formation of *O₂ intermediate over In₂S₃ and In₂S₃-V, electron accumulation and depletion area are represented by yellow and blue color. d The calculated free energy diagram for ORR to H₂O₂ over In₂S₃ and In₂S₃-V. The rate-determining step with maximal Gibbs free energy demand (∆G_max) is labeled. Source data for the calculated free energy are provided as a Source Data file. The optimized geometries of atoms for DFT calculation are provided in Supplementary data 1.

Discussion

Indium vacancies (V_In) have been introduced to In₂S₃ by an ethylene glycol-modulated hydrothermal method. The V_In is exclusively distributed at the surface of In₂S₃ and can significantly boost the photocatalytic activity of In₂S₃ for H₂O₂ synthesis. At the optimal V_In content, the defective In₂S₃ can deliver a H₂O₂ generation rate of 4.77 ± 0.05 mmol·h⁻¹·g_cat.⁻¹ under visible light illumination and an apparent quantum efficiency as high as 7.49 ± 0.01 % at 420 ± 20 nm. Further analysis reveals that V_In plays many important roles during H₂O₂ photosynthesis, including enhancement of O₂ adsorption, enrichment of photo-generated electrons, increment of reducing power of photo-generated electrons, and promotion of H₂O₂ desorption.

Methods

In₂S₃ synthesis

In₂S₃ was prepared by a hydrothermal method. Briefly, 0.4424 g InCl₃ (Aladdin, 99.9%) and 0.4508 g thioacetamide (TAA, Aladdin, 99%) were dissolved in 50 mL deionized water under constant magnetic stirring for 30 min. The resulting solution was sealed in a stainless autoclave for hydrothermal reactions⁶². The autoclave contains a Teflon lining with a volume of 100 mL. The typical reaction temperature and time are 180 °C and 12 h, respectively. At the end of reactions, the autoclave was cooled down to room temperature in air. The reacted suspensions were centrifuged to collect the product powders. To remove any unreacted substances, the product powders were rinsed repeatedly with deionized water and ethanol, and eventually dried in an oven at 80 °C for 12 h.

Preparation of In₂S₃-V

The In₂S₃ containing V_In (i.e. In₂S₃-V) was prepared by replacing a portion of the water with ethylene glycol (EG, Aladdin, 98%) during the hydrothermal synthesis. The In₂S₃-V was synthesized by using 47 mL EG and 3 mL H₂O. All other reaction conditions remained the same as the preparation of In₂S₃.

Materials characterization

The as-synthesized sample powders were characterized using various analytical techniques. X-ray diffraction (XRD) analysis was performed with a Bruker D8 Focus diffractometer, employing Cu K_α1 (λ = 1.5406 Å) and Cu K_α2 (λ = 1.5444 Å) as the radiation. Data were collected with a step size of 0.01° and a collection time of 10 s per step. A field emission scanning electron microscope (FE-SEM, Hitachi S4800, Japan) and a transmission electron microscope (TEM, JEOL JEM-2100, Japan) were used to examine the microstructures of In₂S₃ and In₂S₃-V. The dynamic migration of indium vacancies was inspected by an aberration-corrected scanning transmission electron microscope (AC-STEM, JEM-ARM300F, JEOL, Japan), operating at an acceleration voltage of 300 kV and 293 K. In situ Raman spectra were recorded using a Renishaw in Basis microscopic Raman spectrometer (Renishaw, Wotton-under-Edge, UK) with an excitation wavelength of 513 nm. Surface chemical compositions were analyzed by X-ray photoelectron spectroscopy (XPS, Thermo Escalab 250, Al K_α source), with all binding energies calibrated to the C 1 s peak at 284.7 eV. The UV-vis diffuse reflectance spectra (UV-vis DRS) were recorded using a JASCO V750 spectrophotometer (Japan), with BaSO₄ as the non-absorbing reference. The In content of the samples was determined by inductively coupled plasma optical emission spectrometry (ICP-OES, Agilent 5110, USA). Stead-state photoluminescence (PL) was measured with an FLS1000 spectrophotometer (Edinburgh Instruments, UK). Time-resolved photoluminescence (TRPL) spectroscopy was recorded on a spectrophotometer (PicoQuant FluoTime 300, Germany) under an excitation wavelength of 450 nm. Femtosecond transient absorption (fs-TA) spectra were performed on a pump-probe system (TIME-TECH SPECTRA, Femto-TA100-1, China) using an amplified femtosecond laser system (Coherent, 35 fs, 1 kHz, 800 nm). The oxygen chemisorption capacity of the catalysts was evaluated using a temperature-programmed chemisorption analyzer (Microtrac BELCat II, Japan). Specific surface areas were determined via the Brunauer-Emmett-Teller (BET) method using a TriStar 3000 instrument. The X-ray absorption near-edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) spectra were collected at BL11B beamline of the National Synchrotron Radiation Laboratory (NSRL) (Shanghai, China). Ultraviolet photoelectron spectra (UPS) and synchrotron radiation photoemission spectroscopy (SRPES) were conducted at the BL11U beamline of NSRL (Hefei, China). Superoxide radical (•O₂⁻), hydroxyl radical (•OH) and photocarriers were detected using a benchtop electron paramagnetic resonance (EPR) spectrometer (CIQTEK EPR 200 M, China) with an aid of a spin-trapping reagent. 10 mg catalyst was dispersed into 1 mL methanol and the 5-dimethyl-1-pyrrolidine N-oxide (DMPO, DOJINDO, AR) was used as the spin-trapping reagent to detect the •O₂⁻and •OH radicals^9,46,63. For the detection of photo-generated electrons, 10 mg of catalyst was dispersed into 1 mL of 20 μM TEMPO (Aladdin, 98%) aqueous solution which was then detected by EPR after visible light illumination for 3 s. For the detection of photo-generated holes, the reaction steps are the same except that 1 mL of 20 μM TEMPO acetonitrile solution was used.

Photocatalytic H₂O₂ production

The photocatalytic H₂O₂ production experiments were performed using a top-irradiation-type quartz reactor. In brief, 5 mg sample powders were distributed into 70 mL deionized water containing 10 mL triethylamine (TEA, Aladdin, 99%) as the sacrificial reagent. The suspensions were sealed into the reactor with continuous O₂ bubbling to ensure a sufficient supply of oxygen for the ORR. A water jacket was employed to maintain the reactor temperature at 20 °C. A 300 W xenon lamp (PLS-SXE300, Beijing Perfectlight, China) served as the light source, with visible light generated by applying a UV cutoff filter (λ ≥ 420 nm). The concentration of H₂O₂ was determined by the N,N-Diethyl-p-Phenylenediamine (DPD) method. The DPD method enables quantitative detection of the product H₂O₂ without interference from other byproducts during the process. Specifically, 1 mL solution was taken from the reactor at a given time point, filtered through a 0.22 μm membrane, and mixed with 0.4 mL of sodium phosphate buffer (PBS), 1.12 mL of distilled water, 50 μL of DPD (Aladdin, 99%), and 50 μL of horseradish peroxidase (POD, Aladdin, EnzymoPure^TM, ≥ 250 U/mg) solution. The DPD solution was prepared by dissolving 0.1 g DPD into 10 mL H₂SO₄ (Aladdin, 98%) aqueous solution (0.1 M). The POD solution was prepared by dissolving 5 mg POD into 5 mL distilled water. The sampled solution was vigorously shaken for 90 seconds before analysis by a UV-1800 spectrometer (SHIMADZU, Japan). The H₂O₂ concentration was quantified by measuring the absorbance at the characteristic peak of 552 nm. To assess the applicability of the H₂O₂ production, different water sources were used, including tape water and seawater. Seawater was simulated by the aqueous solution containing NaCl (24.5 g/L, Macklin‌, 99.5%), MgCl₂ (5.2 g/L, Aladdin, 99.99%), MgSO₄ (6.6 g/L, SCR, 98%), CaCl₂ (1.2 g/L, SCR, 96%) and KCl (0.7 g/L, Aladdin, 99%).

The apparent quantum efficiency (AQE) for H₂O₂ production was determined using the same setup under monochromatic light illumination. Monochromatic light was generated by filtering the output of a 300 W xenon lamp (PLS-SXE300, Perfectlight, China) with the bandpass filter at wavelength of 420 nm, 450 nm, 500 nm, 550 nm, 600 nm, and 700 nm, respectively. A quantum meter (Apogee MP-300) was used to calibrate the photon flux at each wavelength. The AQE was calculated using the following Eq. (2):

$$\mathrm{AQE}=\frac{n\times \mathrm{mole}s\mathrm{of}{\mathrm{H}}_2{\mathrm{O}}_2\mathrm{produced}\mathrm{per}\mathrm{hour}}{\mathrm{photon}\mathrm{flux}\mathrm{per}\mathrm{hour}}\times 100\%$$ 2

where n = 2 for H₂O₂ generation. The amounts of H₂O₂ produced and the photon flux of monochromatic light illumination are tabulated in Supplementary Table 4.

Isotope experiment

The isotopic experiment was performed to identify the origin of oxygen during H₂O₂ formation. The photocatalytic H₂O₂ production experiments were conducted in isotope-labeled water (H₂¹⁸O, Aladdin, 99.5%) whilst normal ¹⁶O₂ gas was bubbled into the reactor. The product hydrogen peroxide was decomposed into oxygen gas by MnO₂, whose mass number was then identified by the gas chromatograph-mass spectrometer (GCMS, ThermoFisher-2015, USA).

Diffused reflectance infrared Fourier transform spectroscopy (DRIFTS)

Diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) technique was employed to identify intermediates formed during H₂O₂ photosynthesis. 5 mg of the sample powder was placed onto a Harrick IR cell, which was purged with humidified O₂ as reactants for H₂O₂ production. A temperature control unit was used to hold the cell temperature around 40 °C. A 300 W xenon lamp (AuLight CEL-HXF300/CEL-HXUV300, China), equipped with a UV cutoff filter (λ ≥ 420 nm) was used as the light source. The light was directed onto the sample powder through an optical fiber. Before illumination, the system was kept in the dark for 30 minutes to achieve desorption-adsorption equilibrium. DRIFTS spectra were recorded at various time intervals using a Nicolet iS10 spectrometer (Thermo Scientific, USA). For the probe of product desorption, the light source was removed after 30 min illumination and DRIFTS spectra at the region of H₂O₂ absorption were collected at various time points.

Photoelectrochemical (PEC) and electrochemical measurements

Photoelectrochemical (PEC) measurements were performed using photoelectrodes prepared via a drop-coating method. Specifically, 10 mg of sample powders were dispersed in 1 mL of ethanol containing 20 μL of Nafion solution (Aladdin, ~5% in a mixture of lower aliphatic alcohols and water). The mixtures were sonicated for 30 minutes to achieve uniform dispersion. The resulting slurry (200 μL) was drop-coated onto a fluorine-doped tin oxide (FTO) glass substrate. The coated substrate was then dried under ambient conditions. PEC measurements were conducted using a three-electrode configuration on an electrochemical workstation (CHI660E, CH Instruments Ins, China). The prepared photoelectrode served as the working electrode. Pt foil (10 × 10 mm, CH Instruments Ins, China) and an Ag/AgCl electrode (CH Instruments Ins, China) were used as the counter and reference electrodes, respectively. A 0.5 M Na₂SO₄ (SCR, 99%) aqueous solution (pH = 6 ± 0.2) was employed as the electrolyte. The electrolyte was prepared by dissolving 71.02 g Na₂SO₄ into 1 L deionized water. 10 mL trimethylamine (TEA) was added in the case to study the TEA oxidation reactions over sample surface. The light source was the same as that used in the photocatalytic H₂O₂ production experiments, with incident light regulated by an electronic timer and shutter (GCI-73, DAHENG, China). The electrochemical impedance spectra (EIS) were recorded at a bias of −0.2 V versus Ag/AgCl over the frequency range of 1 MHz to 0.01 Hz in Na₂SO₄ aqueous solution (0.5 M). The electrolyte resistance is 0.16 ± 0.03 Ω. The potentials were referenced to the reversible hydrogen electrode (RHE) using Eq. (3):

$$V_{\mathrm{RHE}}=V_{\mathrm{Ag}/\mathrm{AgCl}}+0.0591\times \mathrm{pH}+0.1976$$ 3

The average electron transfer number during the ORR was determined through rotating disk electrode (RDE) measurements performed on a Pine AFMSRXE 1523 advanced electrochemical system in a three-electrode configuration. To prepare the working electrode, 5 mg of the sample powder was dispersed in 1 mL of anhydrous ethanol containing 25 μL of Nafion aqueous solution to fabricate the catalyst ink. 10 μL ink was drop-cast onto the RDE surface (0.20 cm²) and allowed to dry naturally. An Ag/AgCl electrode and a platinum wire electrode were employed as the reference and counter electrodes, respectively. The rotating ring disk electrode (RRDE) measurement was performed in Ar atmosphere (Jiaya Chemicals, 99.999%) with a rotating speed of 1000 rpm under a ring potential of 0.6 V and - 0.23 V vs Ag/AgCl for the detection of H₂O₂ and O₂, respectively. Linear sweep voltammetry (LSV) was performed in an oxygen-saturated phosphate-buffered saline solution (PBS, 1 M, pH = 6.8 ± 0.2) at a scan rate of 10 mV/s to evaluate the electrochemical performance. The slopes of the Koutecky–Levich (K–L) plots determined the average number of electrons (n) in the ORR via the corresponding Eqs. (4–5)^64–66:

$$j^{-1}=j_k^{-1}+B^{-1}{\omega }^{-1/2}$$ 4

$$B=0.2nF{\nu }^{-1/6}C{D}^{2/3}$$ 5

where j is the current density, j_k is the kinetic current density, ω is the rotating speed (rpm), F is the Faraday constant (96485 C·mol⁻¹), ν is the kinetic viscosity of water (0.01 cm²·s⁻¹), C is the bulk concentration of O₂ in water (1.26 × 10⁻³ mol·cm⁻³), and D is the diffusion coefficient of O₂ (2.7 × 10⁻⁵ cm²·s⁻¹).

Theoretical calculations

The spin-polarized DFT calculations are performed within a Generalized Gradient Approximation (GGA) based on the Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional by using the Vienna ab initio simulation package (VASP)^67,68. The applied energy cutoff was set as 400 eV. The effect of the van der Waals interactions was considered by the DFT-D3 method proposed by Grimme et al⁶⁹. All atoms were allowed to relax until the forces were less than 0.02 eV/ Å^–1 and total energies were converged to 10⁻⁵ eV. Compared with the pure surface model of In₂S₃ (010), the In vacancy model removes one In and one S atom from the surface layer. The bottom layers of In₂S₃(010) surface model are fixed at the corresponding bulk structure during structure optimization. The vacuum regions are set at least 15 Å to eliminate the effect of interlayer interaction. The structure optimization and electronic property calculations are performed with 1 × 3 × 1 and 2 × 6 × 1 k-point grids according to the MonkhorstPack scheme⁷⁰.

The Gibbs free adsorption energies ∆G in each step are expressed by following Eq. (6):

$$\mathrm{\Delta }G=\mathrm{\Delta }E+\mathrm{\Delta }{E}_{\mathrm{ZPE}}-T\mathrm{\Delta }S$$ 6

where ∆E, ∆E_ZEP, ∆S, and T stand for the difference of adsorption energy, zero point energy, entropy, and temperature, respectively.

The climbing image nudged elastic band (CI-NEB) method was utilized to calculate the migration barriers of In ions⁷¹.

Author contributions

X.X. led the project. Q.Z. designed and performed experiments. J.S. and J.L. collected the fs-TA data. G.Lin collected the TEM data. G.Li provided help with DRIFTS measurements. Z.Z. and W.W. did the DFT calculations. Q.Z. analyzed the data and wrote the manuscript. X.X. revised the manuscript. All authors were involved in the data discussion.

Peer review

Peer review information

Nature Communications thanks Changlin Yu, and the other anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data supporting the findings of this study are reported in the main text or the Supplementary Information. Raw data are provided as a Source Data file. Source data are provided with this paper.

Competing interests

The authors declare no competing interest.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-65538-w.