Fluorine-expedited nitridation of layered perovskite Sr₂TiO₄ for visible-light-driven photocatalytic overall water splitting

Abstract

Photocatalytic overall water splitting is a promising approach for a sustainable hydrogen provision using solar energy. For sufficient solar energy utilization, this reaction ought to be operated based on visible-light-active semiconductors, which is very challenging. In this work, an F-expedited nitridation strategy is applied to modify the wide-bandgap semiconductor Sr₂TiO₄ for visible-light-driven photocatalytic overall water splitting. Compared to the conventional nitridation approach, F-expedited nitridation introduces the desirable integration of a high concentration of N dopant for strong visible light absorption and a low concentration of defects (i.e. Ti³⁺ and oxygen vacancies) for effective separation of photocarriers. After being coated with Ti-oxyhydroxide protection layer and deposited with RhCrO_y cocatalyst, the product from F-expedited nitridation can stably run photocatalytic overall water splitting with apparent quantum efficiency of 0.39% at 420 ± 20 nm and solar-to-hydrogen efficiency of 0.028%. These findings justify the effectiveness of F-expedited nitridation strategy and serve as a guidance to upgrade the photocatalytic activity of many other wide-bandgap semiconductors.

Photocatalytic overall water splitting is a promising approach for sustainable hydrogen production, but it remains challenging under visible light. Here, the authors report an F-expedited nitridation strategy for Sr₂TiO₄ that enables visible-light-driven photocatalytic overall water splitting.

Introduction

Green hydrogen production technologies involving only solar irradiation, water, and photocatalysts are considered to enable a zero-carbon economic model and the development of clean energy facilities^1–3. Such a promising scenario, nevertheless, awaits significant advancements in photocatalytic materials that can convert solar energy efficiently into hydrogen energy. Specifically, the photocatalytic materials should harvest appreciable amounts of solar photons to generate photocarriers (e⁻ and h⁺) which in turn, move to the surface to catalyze water-splitting reactions in competing with the unproductive photocarrier recombination⁴. Layered compounds have been recognized as promising candidates for photocatalytic overall water splitting (POWS) due to the charge separation driving force derived from their structural laminations^5–7. For instance, high POWS activities have been reported over A₄Nb₆O₁₇ (A = K, Rb)⁸, A_2-xLa₂Ti_3-xNb_xO₁₀·nH₂O (A = K, Rb, Cs; x = 0, 0.5, 1)⁹, A′₂ATa₂O₇ (A′ = H or K, A = La_2/3 or Sr)¹⁰, Sr₂M₂O₇ (M = Nb, Ta)¹¹ etc., whose layered architectures play a pivotal role in expediting photocarrier dissociation and migration. However, most layered compounds are wide-bandgap semiconductors thereby having intrinsic limitations on POWS efficiency in terms of solar energy conversions¹². Accordingly, it is highly desirable to exploit layered compounds with visible light sensitivity for POWS reactions.

As a typical Ruddlesden-Popper (RP) type layered perovskite, Sr₂TiO₄ has received extensive attention as it is a robust, easily producible compound comprising of cheap and low toxic elements. Aliovalent metal doping (e.g. Cr^13,14, Fe¹⁵, Mn¹⁶, Ag¹⁷, La/Cr¹⁴, La/Fe¹⁸, La/Rh¹⁹, etc.), non-metal doping (e.g. F²⁰, N^21,22, etc.), and mixed doping (e.g. Nb/N²¹, La/N²², Cr/F²³, etc.) tactics have been applied to extend the light absorption threshold of Sr₂TiO₄ into the visible light region. However, there is currently no viable approach to enable POWS activity over Sr₂TiO₄ under visible light illumination, which is the bottleneck for the development of this promising compound. Among various dopants studied, of note are N anions as they with high concentrations contribute to the hybridized 2p-orbital-based energy states of O and N, endorsing relatively higher mobility for the photo-generated holes. Doping N into metal oxides, i.e. nitridation, is normally conducted by high-temperature ammonolysis in which appropriate metal oxide precursors are calcined at elevated temperatures in the presence of flowing ammonia gas^24,25. This is a typical non-equilibrium solid-gas reaction that involves multiple diffusion steps and sluggish kinetic processes as the replacement of oxygen by nitrogen is very slow^26,27. A low N doping level (<1 at.%) at anion sites is usually observed^28,29 and there is a strong likelihood that the product conceives of both low visible light absorbance and plentiful unwanted defects in order to balance the charge disturbance raised by O/N substitution²⁶.

In this study, we adopted Sr₂TiO₃F₂ as a precursor for F-expedited nitridation of Sr₂TiO₄. The choice of a fluorinated precursor is based on the following considerations: (1) F is of a higher electronegativity than O. The presence of F can weaken the Ti–O bonds according to the inductive effect³⁰, facilitating O/N exchange. This is supported by the longer equatorial Ti–O bonds of Sr₂TiO₃F₂ (~1.9737 Å) than that of Sr₂TiO₄ (~1.9433 Å)^31,32. (2) F can serve as a co-dopant to N. Doping N and F together helps to minimize the charge variations at the anion sublattice, i.e. ${2O}_O^{\times }\to N_O^{\prime }+F_O^{\cdot }$. Our results show that the presence of F is critical for increasing the N uptake, facilitating crystal growth, and inhibiting detrimental defects like Ti³⁺ and oxygen vacancies in the product. Notably, F-expedited nitridation of Sr₂TiO₄ delivers superior photocatalytic performance in overall water splitting, liberating H₂ and O₂ in a 2:1 ratio, under visible light.

Results

Different nitridation strategies

Considering the fact that direct nitridation of Sr₂TiO₄ by the conventional method (N-doped Sr₂TiO₄, denoted as Sr₂TiO₄-N, Fig. 1a) is subject to the slow kinetics due to the strong Ti–O bonds, Sr₂TiO₄ is firstly fluorinated into Sr₂TiO₃F₂ to weaken the Ti–O bonds so as to facilitate O/N exchange during nitridation (N, F co-doped Sr₂TiO₄, denoted as Sr₂TiO₄-NF, Fig. 1b). The Sr₂TiO₃F₂ has both interlayer and intralayer F anions in the structure which not only elongate the equatorial Ti–O bonds but also serve as the co-dopants to N. The morphology of product particles was inspected by field emission scanning electron microscope (FE-SEM, Supplementary Fig. 1). Although Sr₂TiO₄ and Sr₂TiO₃F₂ both contain irregularly shaped particles, their nitridation products reveal dissimilar particle morphologies. Specifically, Sr₂TiO₄-N is composed of perforated particles with rough surfaces. This has been attributed to the slow O/N exchanging rate that results in poor crystal growth during high-temperature ammonolysis²⁴. In contrast, Sr₂TiO₄-NF comprises plate-like particles with clear and smooth crystal facets, properly reflecting the crystal habit of a layered structure. Such distinct particle morphologies indicate that F helps to facilitate crystal growth during high-temperature ammonolysis.

Fig. 1. Different nitridation strategies.

a Direct nitridation of Sr₂TiO₄. b F-expedited nitridation of Sr₂TiO₄ using Sr₂TiO₃F₂. c UV–visible absorption spectra of Sr₂TiO₄ and Sr₂TiO₄-N. d UV–visible absorption spectra of Sr₂TiO₃F₂ and Sr₂TiO₄-NF, the insets show the digital photographs of sample powders. Source data for UV–visible absorption spectra are provided as a Source Data file.

Apart from the distinct particle morphologies, the products also reveal different light absorption capacities as illustrated in Fig. 1c, d. It is observed that the absorption edge of Sr₂TiO₄ occurs at 353 nm, indicative of a bandgap ~3.51 eV. The absorption edge position of Sr₂TiO₄-N closely resembles that of Sr₂TiO₄, with only a small absorption tail extending to 475 nm. In contrast, the light absorption range of Sr₂TiO₃F₂, as shown in Fig. 1d, is blue-shifted compared to Sr₂TiO₄. Interestingly, the absorption edge of Sr₂TiO₄-NF experiences a large band-to-band redshift to 482 nm. This is also reflected by their distinct powder colors, i.e. greyish green for Sr₂TiO₄-N (Fig. 1c inset) and bright yellow for Sr₂TiO₄-NF (Fig. 1d inset). The much stronger visible light absorption can be ascribed to higher N content in Sr₂TiO₄-NF than in Sr₂TiO₄-N (see the following section) as N contributes to additional energy states above the top of the O-based valence band. The bandgap values of these compounds are then deduced according to the threshold of their light absorption (Supplementary Fig. 2). N doping notably reduces the bandgap of Sr₂TiO₄, being beneficial for an enhanced solar energy harvest. The band-edge positions of Sr₂TiO₄-N and Sr₂TiO₄-NF were explored based on the ultraviolet photoelectron spectra (UPS) technique using synchrotron radiation (Supplementary Fig. 3). The onset of UPS valence band spectra corresponds to the top of the valence band (VB) relative to the work function whose value can be deduced according to the UPS secondary electron cutoff spectra³³. The band-edge positions can then be determined based on this information as well as the bandgap values. The sketched band-edge positions of Sr₂TiO₄-N and Sr₂TiO₄-NF are shown in Supplementary Fig. 4. Although both samples own a comparable work function (also indicated by Mott-Schottky analysis of Supplementary Fig. 5), Sr₂TiO₄-NF has higher levels of conduction band (CB) and VB positions than Sr₂TiO₄-N. This can be rationalized by the fact that a high N content in Sr₂TiO₄-NF uplifts the top of VB and F with a larger electronegativity contributing to the higher antibonding orbitals (i.e. CB). Nevertheless, both samples have straddled band-edge positions relative to water redox potentials, being thermodynamically feasible for overall water splitting.

Structure, composition, and coordination analysis

X-ray diffraction (XRD) analysis (Supplementary Fig. 6) suggests that Sr₂TiO₄-N and Sr₂TiO₄-NF exhibit similar patterns to the standard ones of Sr₂TiO₄ (JCPDS: 00-039-1471), indicating that the RP-type layered structure is maintained for both samples. This is also confirmed by Raman spectra that show identical vibration modes for Sr₂TiO₄, Sr₂TiO₄-N, and Sr₂TiO₄-NF (Supplementary Fig. 7). Rietveld refinements have been applied to investigate the crystal structure of Sr₂TiO₄-N and Sr₂TiO₄-NF. The refined XRD patterns are illustrated in Fig. 2a, b with the agreement factors included as the insets. The dopants, i.e. N and F, probably occupy randomly in the anion sublattice, as the agreement factors show no improvement in the case of ordered occupancies. The refined unit cell of Sr₂TiO₄-NF is slightly expanded when compared to Sr₂TiO₄ and Sr₂TiO₄-N, implying a higher N uptake in the presence of F (Supplementary Table 1). Transmission electron microscopy (TEM) analysis further indicates that each Sr₂TiO₄-NF particle is a single crystal with high crystallinity as revealed by the sharp selected area electron diffraction patterns as well as the well-defined lattice fringes in the high-resolution TEM images (Fig. 2c–e). All elements, including the N/F dopants, are uniformly distributed within the particles of Sr₂TiO₄-NF, as revealed by TEM-EDX element mapping analysis (Fig. 2f).

Fig. 2. Crystal structure analysis.

a Refined XRD patterns of Sr₂TiO₄-N, agreement factors (R_p, R_wp, and χ²) are inserted. b Refined XRD patterns of Sr₂TiO₄-NF, agreement factors (R_p, R_wp, and χ²) are inserted. c TEM image of Sr₂TiO₄-NF particles, inset shows the selected area electron diffraction patterns of a single particle in the upper part. d High-resolution TEM image of Sr₂TiO₄-NF particle along [001] direction, inset shows the projected crystal structure along [001] direction. e High-resolution TEM image of Sr₂TiO₄-NF particle along [010] direction, inset shows the projected crystal structure along [010] direction. f TEM-EDX element mapping images of Sr₂TiO₄-NF particles, all elements (Sr, Ti, O, F, N) are uniformly distributed in two typical particles. Source data for refined XRD are provided as a Source Data file.

Element content analysis suggests that N and N/F are detected respectively in Sr₂TiO₄-N (Fig. 3a) and Sr₂TiO₄-NF (Fig. 3b), indicative of successful N doping and N/F co-doping. Accordant with previous expectations, the N content in the product is increased by a factor of 2.4 in the presence of F even though the nitridation conditions are kept the same for both samples. The oxygen vacancies, however, are decreased by a factor of 9 in the presence of F. These observations are analog to N/F co-doped TiO₂³⁴ and N/F co-doped SrTiO₃³⁵. The chemical formula of Sr₂TiO₄-N and Sr₂TiO₄-NF can be written as Sr_1.99TiO_3.86N_0.05□_0.09 and Sr_1.99TiO_3.78N_0.12F_0.09□_0.01, respectively, where oxygen vacancies are represented as empty squares ‘□‘. Accordingly, Ti has an average valence of ~+3.9 in Sr₂TiO₄-N and ~+4.0 in Sr₂TiO₄-NF. The surface of sample powders is then studied by X-ray photoelectron spectroscopic (XPS, Supplementary Fig. 8) techniques. The XPS Ti 2p spectra involve a single spin-orbit doublet for all samples, being assignable to 2p_3/2 and 2p_1/2 states of Ti⁴⁺ species^17,36. It is noteworthy that Ti³⁺ species, observed in N/F co-doped SrTiO₃³⁵, are not detected here. This might correlate with the high Sr content in Sr₂TiO₄ that helps to stabilize Ti⁴⁺ species according to the inductive effect^37,38. The successful N doping for Sr₂TiO₄-N and Sr₂TiO₄-NF is confirmed by a weak peak around 396 eV in XPS N 1s spectra which is assigned to N³⁻ species in the lattice³⁹. Likewise, the existence of F⁻ species in Sr₂TiO₄-NF is also suggested by a strong peak around 685 eV in XPS F 1s spectra which is the same as its precursor Sr₂TiO₃F₂⁴⁰. The XPS O 1s spectra of all samples comprise two overlapping peaks centered around 529 eV and 531 eV, assignable to lattice oxygen (O²⁻) and surface hydroxyl groups (OH⁻), respectively⁴¹. The strong signal for surface hydroxyl groups is consistent with the high hydrophilicity of Sr₂TiO₄-based compounds^14,18,19.

Fig. 3. Coordination states of elements.

Element content of a Sr₂TiO₄-NF and b Sr₂TiO₄-N by ICP, ONH, and ion chromatograph analysis, deduced chemical formula are shown at the bottom (oxygen vacancies are represented by empty squares ‘□‘); c EPR spectra of Sr₂TiO₄-N and Sr₂TiO₄-NF measured at 100 K. d XAS spectra at Ti K-edges of Sr₂TiO₄-N and Sr₂TiO₄-NF, the shading areas correspond to the pre-edge and edge region; e XAS spectra at Ti L-edges of Sr₂TiO₄-N and Sr₂TiO₄-NF, L₃ and L₂ denote signals due to spin-orbit interactions whilst t_2g and e_g denote signals due to octahedral crystal field splitting; f XAS spectra at O K-edges of Sr₂TiO₄-N and Sr₂TiO₄-NF, O_2p-t_2g, O_2p-e_g, and O_2p-Ti_4sp denote the hybridized states between O 2p and Ti 3d and 4sp orbitals. Source data for element content, ESR spectra, and XAS spectra are provided as a Source Data file.

Although XPS spectra suggest no significant difference in terms of the surface state, Sr₂TiO₄-N and Sr₂TiO₄-NF are essentially different in defect concentrations, as revealed by EPR and X-ray absorption spectra (XAS). Both Sr₂TiO₄-N and Sr₂TiO₄-NF show detectable EPR signals at g values of ~2.004 and ~1.977 (Fig. 3c), corresponding to oxygen vacancies and Ti³⁺ species^42,43, respectively. Notably, Sr₂TiO₄-NF has much weaker EPR signals than Sr₂TiO₄-N, corresponding to a much lower concentration of both defects. This is further supported by XAS. XAS spectra at Ti K-edges show that the transition energy of Sr₂TiO₄-NF is positively shifted compared to Sr₂TiO₄-N at both pre-edge and edge region (Fig. 3d), indicating a higher valence state of Ti in Sr₂TiO₄-NF than in Sr₂TiO₄-N. Similar observations are also noticed in XAS spectra at Ti L-edges which reveal transitions from occupied Ti 2p orbitals to unoccupied Ti 3d orbitals (Fig. 3e). The two doublets, arising from spin-orbit interactions (2p_3/2(L₃), 2p_1/2(L₂)) and octahedral crystal field splitting (t_2g, e_g), shift slightly to higher energy when comparing Sr₂TiO₄-NF with Sr₂TiO₄-N, affirming again the higher Ti valence in the former. Moreover, the e_g states at Ti L-edges are known to be more sensitive to octahedral distortion than t_2g states due to the fact that d_z² and d_x²_-y² orbitals are pointing towards ligand anions⁴⁴. It can be seen from Fig. 3e that e_g peaks of Sr₂TiO₄-N have a higher level of asymmetry than those of Sr₂TiO₄-NF. In other words, the d_z² and d_x²_-y² orbitals of Ti 3d states are less degenerated in Sr₂TiO₄-N, being consistent with its higher concentration of defects (e.g. oxygen vacancies) that break the octahedral symmetry. In addition, the XAS spectra at O K-edges also provide useful information for the Ti states due to the strong hybridization between O 2p and Ti 3d and 4sp orbitals. Figure 3f shows the XAS spectra at O K-edges for Sr₂TiO₄-NF and Sr₂TiO₄-N. The pre-edge absorption at the low-energy side (<540 eV) is assignable to the transitions from O 1s orbitals to the hybridized O 2p and Ti 3d orbitals which split into O_2p-t_2g and O_2p-e_g hybridized states due to octahedral crystal field splitting⁴⁵. Compared to Sr₂TiO₄-NF, Sr₂TiO₄-N has a poorly resolved O_2p-t_2g state, probably because of its high Ti³⁺ concentration that quenches the transition by filling electrons into the empty t_2g orbitals. These results jointly confirm that Sr₂TiO₄-NF has a much lower defect concentration than Sr₂TiO₄-N, being consistent with the composition analysis. The decrement of defect concentration is fully consistent with earlier expectations: the presence of F not only facilitates O/N exchange by weakening Ti–O bonds but also balances the charge variations by serving as a co-dopant to N, both of which reduce the risks of defect formation.

Photocarrier separation

The photocarrier behavior in Sr₂TiO₄-N and Sr₂TiO₄-NF were initially examined through photoluminescence (PL) spectroscopy by cooling the sample to 10 K to suppress the phonon-assisted nonradiative recombination⁴⁶. For both samples, PL spectra showed a broad emission band in the visible light region, corresponding to a variety of radiative photocarrier recombination events. The broad PL emission band can be deconvolved into a series of emission peaks that have different origins, as shown in Fig. 4a. The peaks centered below 500 nm are probably due to interband emissions as their photon energy exceeds the light absorption threshold of Sr₂TiO₄-N and Sr₂TiO₄-NF. The strong peak around 500 nm can be assigned to the band-edge emissions as it is close to the light absorption edges. The rest peaks above 500 nm are attributable to the defect-mediated emissions within the bandgap because of their low photon energy. Similar defect-mediated emissions have also been noticed in N/F doped and N/F co-doped TiO₂ and SrTiO₃, and originated from Ti³⁺ species and oxygen vacancies that are singly and doubly ionized^47,48. Notably, Sr₂TiO₄-NF is characterized by much weaker emission intensity at this region than Sr₂TiO₄-N, being accordant with its lower concentration of Ti³⁺ and oxygen vacancies. Density Functional Theory (DFT) calculation further suggests that Ti³⁺ and oxygen vacancies tend to form additional energy states below the conduction band and N dopants help to uplift the top of the valence band (Fig. 4b and Supplementary Figs. 9–12). The energy levels of these defect states can then be deduced based on PL data. A schematic illustration for Sr₂TiO₄-N is depicted in Fig. 4c. Although Sr₂TiO₄-N is feasible for overall water splitting, the separation of photocarriers is severely impaired by defects, i.e. Ti³⁺ and oxygen vacancies. These defects align relatively deep below the conduction band and thereby are efficient trapping/recombination centers for photocarriers, being detrimental to photocatalytic activity. It is generally considered that defects with shallow energy states (i.e. shallow defects) may play a positive role in photocarrier separation, particularly when they are located at the surface⁴⁹. However, defects with deep energy states (i.e. deep defects) generally undermine photocarrier separation as photocarriers can be tightly trapped and are unlikely to escape from these defects for surface reactions⁴⁹. This is particularly true for oxygen vacancies here as they form deep defects whose energy states lie close to or lower than the water reduction potential. Presumably, the trapped electrons by oxygen vacancies can hardly escape and would have an inadequate driving force for water reduction reactions. From these considerations, co-doping N/F into Sr₂TiO₄ is more preferable than doping N alone due to a lower concentration of both Ti³⁺ and oxygen vacancies that is beneficial for photocarrier separation.

Fig. 4. Separation of photocarriers.

a PL spectra of Sr₂TiO₄-N and Sr₂TiO₄-NF by the 420 nm excitation at 10K. b DFT-calculated DOS of Sr₂TiO₄ with different defects, i.e. Ti³⁺, oxygen vacancies (V_O), nitrogen at the oxygen site (N_O), and fluorine at the oxygen site (F_O), the fermi level is adjusted properly for better comparisons. c Schematic illustration of photocarrier behaviors in Sr₂TiO₄-N, i.e. generation, transportation, being captured by defects, defect-mediated radiative recombination, Ti³⁺ species, and oxygen vacancies that are singly and doubly ionized are labeled as Ti³⁺, V_O^•, and V_O, respectively. d Time-resolved PL spectra of Sr₂TiO₄-N and Sr₂TiO₄-NF, sample powders are excited by 420 nm photons at 10K, and amplitude-weighted average PL decay lifetime (τ_a) is shown as inset. e Integrated PL emission intensity as a function of temperature and the fitted exciton binding energy (E_b) curves for Sr₂TiO₄-N and Sr₂TiO₄-NF. f SPV spectra of Sr₂TiO₄-N and Sr₂TiO₄-NF. Source data for PL, DFT-calculated DOS, TRPL, integrated PL, and SPV spectra are provided as a Source Data file.

As a support to the above statements, time-resolved photoluminescence (TRPL) decay curves and PL emission intensity at different temperatures are analyzed and fitted based on a double first-order exponential decay model (equation (1))⁵⁰ and Arrhenius relation (equation (2))⁵¹, respectively:

$$A\left(t\right)=A_1\exp \left(\frac{-t}{{\tau }_1}\right)+A_2\exp \left(\frac{-t}{{\tau }_2}\right)+A_0$$ 1

$$I\left(T\right)=\frac{I_0}{1+A{e}^{-E_b/k_{B}T}}$$ 2

where A(t) is PL intensity at time t, I(T) is integrated PL intensity at temperature T, E_b is the exciton binding energy, k_B is the Boltzmann constant. The amplitude-weighted average PL decay lifetime (τ_a) is then determined by equation (3):

$${\tau }_a=\frac{A_1{\tau }_1^2+A_2{\tau }_2^2}{A_1{\tau }_1+A_2{\tau }_2}$$ 3

It can be seen from Fig. 4d and e that Sr₂TiO₄-NF reveals a longer TPRL decay lifetime (τ_a = 9.93 ns) and a smaller exciton binding energy (E_b = 28 meV) than Sr₂TiO₄-N (τ_a = 6.47 ns, E_b = 42 meV). The E_b of Sr₂TiO₄-NF is comparable to the thermal energy (25 meV at 293 K), indicating that excitons can be readily dissociated into free photocarriers within Sr₂TiO₄-NF at ambient conditions. Correspondingly, Sr₂TiO₄-NF exhibits a notably higher surface photovoltage (SPV) signal compared to Sr₂TiO₄-N, as illustrated in the SPV spectra (Fig. 4f). The larger photocurrent (Supplementary Fig. 13) and slower open-circuit voltage decay process (Supplementary Fig. 14) also revealed that Sr₂TiO₄-NF has more efficient photocarrier separation efficiency than Sr₂TiO₄-N, resulting in more effective quenching of the triplet EPR signals of TEMPO under light illumination (Supplementary Fig. 15).

Photocatalytic overall water splitting

After loading Rh-Cr bimetallic oxide (i.e. RhCrO_y) as a cocatalyst, Sr₂TiO₄-NF, characterized by wide spectral response and low defect concentration, demonstrates the capability to achieve photocatalytic overall water splitting with a stoichiometric ratio of hydrogen to oxygen of 2:1 under visible light illumination (λ ≥ 420 nm, Supplementary Fig. 16). However, apart from H₂ and O₂, there was a simultaneous production of N₂ during the experiment, suggesting that oxidative self-corrosion reactions occurred over Sr₂TiO₄-NF. To address this issue, Sr₂TiO₄-NF was coated with a thin protection layer of amorphous Ti-oxyhydroxide (denoted as TiOXH) against self-corrosion⁵². This treatment not only inhibited N₂ evolution but also facilitated H₂/O₂ evolution by a factor of 2.4 as illustrated in Supplementary Fig. 16. The TiOXH is reported to have dual functions: (1) it serves as a permeation barrier for O₂, therefore suppressing the occurrence of water-splitting reverse reactions; (2) it facilitates hole injection from photocatalyst to the TiOXH layer, therefore avoiding excessive buildup of holes at the photocatalyst’s surface and reducing the likelihood of photocorrosion^52,53. Therefore, TiOXH was coated onto sample powders after loading RhCrO_y unless otherwise specified. In contrast to Sr₂TiO₄-NF, Sr₂TiO₄-N was completely inactive for POWS reactions under the same conditions (Fig. 5a). Applying additional annealing treatment has no improvement on the photocatalytic activity of Sr₂TiO₄-N (Supplementary Fig. 17). This is in good accordance with many other N-doped metal oxides that remain inert to POWS or are only moderately active for water-splitting half-reactions^39,54. In addition, swapping the fluorination and nitridation orders for sample synthesis, i.e. fluorination after nitridation, substantially annihilates the POWS activity both under visible light and simulated sunlight (Supplementary Fig. 18). Structure and composition analysis reveals that this modification generates an F-rich and N-poor sample with a high similarity to Sr₂TiO₄-N (Supplementary Fig. 18). From these control experiments, one can realize that F serves to enhance the N uptake during nitridation, and N rather than F plays the vital role for POWS activity. The successful POWS reactions over Sr₂TiO₄-NF justify the effectiveness of the F-expedited nitridation strategy in opening up the photocatalytic potential of otherwise inactive semiconductors. The POWS activity of Sr₂TiO₄-NF was further investigated by tuning some important parameters including RhCrO_y content, TiOXH content, and catalyst dosage (Supplementary Fig. 19). Under optimized conditions, Sr₂TiO₄-NF delivers apparent quantum efficiency (AQE) as high as 0.39% at 420 ± 20 nm and solar-to-hydrogen (STH) efficiency as high as 0.028%. These values are quite competitive when compared to the reported photocatalysts that are active to the visible-light-driven POWS (Supplementary Table 2 and Supplementary Fig. 20). Moreover, the activity and AQE at specific wavelengths show clear wavelength-dependence, closely matching the UV-vis DRS spectra (Fig. 5b and Supplementary Fig. 21). The stability of photocatalytic overall water splitting was evaluated by extended usage under both visible light and simulated AM1.5 G illumination. As shown in Fig. 5c, d, Sr₂TiO₄-NF does not show activity attenuation after four consecutive cycles, indicating superior stability of the material, which is further verified by the XRD, XPS, and SEM before and after the reaction (Supplementary Figs. 22–24).

Fig. 5. Photocatalytic activity and stability.

a Temporal gas evolution (H₂, O₂ and N₂) over Sr₂TiO₄-NF and Sr₂TiO₄-N under visible light illumination (λ ≥ 420 nm). The samples are coated with TiOXH (1 wt%) after loading RhCrO_y (0.5 wt%). b Action spectra (AQE vs. λ) of Sr₂TiO₄-NF for POWS reactions, the error bars correspond to the bandwidth of the band-pass filter used, UV-vis DRS spectra are included for comparisons. c Extended cycles of POWS reactions over Sr₂TiO₄-NF under visible light illumination (λ ≥ 420 nm). d Extended cycles of POWS reactions over Sr₂TiO₄-NF under AM1.5 G (100 mW·cm⁻²). Reaction conditions: 0.4 g catalysts, 100 mL deionized water. Source data for gas evolution and action spectra are provided as a Source Data file.

Discussion

In summary, the RP-type layered perovskite with a wide spectral response and high-efficiency POWS activity have been obtained through F-expedited nitridation of Sr₂TiO₄. The presence of F not only increases the N uptake but also facilitates crystal growth during high-temperature ammonolysis. Notably, the involvement of F for nitridation reduces the concentration of deep-level defects in the product, such as Ti³⁺ and oxygen vacancies, which can seriously undermine the activity. Correspondingly, Sr₂TiO₄-NF demonstrates a much stronger visible light absorption and much-improved photocarrier separation capabilities than Sr₂TiO₄-N. As a result, Sr₂TiO₄-NF displays a stable POWS activity under both visible light and simulated sunlight illumination, delivering an AQE as high as 0.39% at 420 ± 20 nm and STH efficiency as high as 0.028%. These findings provide a useful guideline for the modification of the wide-bandgap semiconductors that are potentially active for POWS. For a more general application of the F-expedited nitridation strategy, an F-involving amorphous precursor can be easily prepared by the Chimie douce method, which shall work in a similar way as the crystallized precursor Sr₂TiO₃F₂ used in this work and will be our future study.

Methods

Preparation of Sr₂TiO₄ powders

Strontium nitrate (1.9142 g, SCR, 99.5%), citric acid (8.6888 g, Aladdin, 99.5%), and tetrabutyl titanate (1.5496 g, Aladdin, 99%) were dispersed in a mixture of ethylene glycol (17 mL, Aladdin, GC grade) and ultrapure water, and stirred continuously until fully dissolved. The transparent solution was evaporated (573 K, 5 h) to a brown resin, followed by pyrolysis (823 K, 15 h) to remove organic components. The resulting white powders were thoroughly ground and pressed into pellets at 5 MPa, and further calcined at 1373 K for 10 h. The final product was ground again for characterization and further processing.

Conventional nitridation of Sr₂TiO₄

The Sr₂TiO₄ powders (0.2 g) were loaded into a tube furnace using an alumina boat and subjected to heating 1273 K for 48 h under flowing ammonia gas (200 mL·min⁻¹, Jiaya Chemicals, 99.999%). Subsequently, the product powders were allowed to cool to room temperature naturally, followed by thorough rinsing with deionized water and drying in an oven overnight, and denoted as Sr₂TiO₄-N. A double annealing sample has been prepared, i.e. annealing in air at 573 K for 14 h and 623 K for 14 h with intermediate grinding before nitridation. This sample was denoted as Sr₂TiO₄-N2 in order to study whether or not the annealing history improves the photocatalytic activity.

F-expedited nitridation of Sr₂TiO₄

Sr₂TiO₄ powders (1.0 g) and PVDF (0.2 g, Arkema, M.W. ~1,100,000) were mixed and pressed into pellets, followed by stepwise calcination at 573 K for 14 h and 623 K for 14 h with intermediate grinding. The product powders were identified to be Sr₂TiO₃F₂. The subsequent high-temperature ammonolysis process was consistent with the conventional nitridation method, except for the use of Sr₂TiO₃F₂ powders as a precursor. The final product is denoted as Sr₂TiO₄-NF. In addition, the Sr₂TiO₄-N sample was also fluorinated in a similar way using PVDF, i.e. fluorination after nitridation, to produce a control sample denoted as Sr₂TiO₄-N@F.

Sample characterizations

X-ray powder diffraction (XRD) patterns of sample powders were recorded using a Bruker D8 Focus diffractometer (Bruker, Germany). The General Structure Analysis System (GSAS) software package was employed for Rietveld refinement. A UV-Vis spectrophotometer (JASCO V750, Japan) was adopted to collect the UV–visible diffuse reflection spectra. The microstructures of sample powders were monitored using a field emission scanning electron microscope (FE-SEM, JSM-7900F, Japan) and a transmission electron microscope (TEM, JEOL JEM-2100, Japan). Inductively coupled plasma optical emission spectrometry (ICP-OES, PE 8300, USA) was used to determine the content of Sr and Ti within the samples. The O and N content of samples was determined using an oxygen-nitrogen-hydrogen analyzer (ONH2000, USA). The F content within the sample powders was determined by dissolving the sample powders in aqua regia which was then analyzed using an ion chromatograph (ICS-1100, USA). X-ray photoelectron spectra (XPS) were collected using an X-ray photoelectron spectrometer (Thermo Escalab 250, USA). Brunauer-Emmett-Teller (BET) surface area of sample powders was evaluated by a surface analyzer (TriStar II 3020, Micromeritics, USA). A Raman spectrometer (Renishaw InVia, UK) was employed to record the Raman spectra of different samples. A laser with a wavelength of 514.5 nm was used for sample excitation during spectra collection. For the photoluminescence (PL) spectroscopic and time-resolved PL (TRPL) decay spectroscopic analysis, the samples were cooled down to the targeted temperature by a cryostat system (ARS DE-202, USA). A picosecond pulsed laser at a wavelength of 420 nm was then introduced to excite the sample powders and the PL and TRPL signals generated were collected by a fluorescence spectrometer (PicoQuant FluoTime 300, Germany). For the electron paramagnetic resonance (EPR) spectroscopic analysis, the samples were frozen down to 100 K using a liquid nitrogen variable temperature unit. The EPR spectra were then collected by an X-band benchtop EPR spectrometer (CIQTEK EPR200M, China). The hard X-ray absorption spectra of sample powders were recorded in the Shanghai Synchrotron Radiation Facility (SSRF) at Beamline BL11B. Ultraviolet photoelectron spectra (UPS) and the soft X-ray absorption spectra were collected at the Catalysis and Surface Science Endstation (Beamline BL11U) of the National Synchrotron Radiation Laboratory (NSRL, Hefei, China). The photon energy for UPS is 40.0 eV. The samples were electrically biased by –5 V for the collection of the secondary electron cutoff (SEC) spectra. The work function was deduced based on the energy difference between the photon energy and SEC binding energy. For the surface photovoltage (SPV) analysis, the samples were enclosed into a photovoltaic cell which was then irradiated using a Xenon lamp (Perfect Light, PLS-LAX500, China). The output of the lamp was controlled by a monochromator (Zolix, SBP500, China) and a shutter (Stanford Research Systems, SR540, USA) whose chopping frequency was set at 23 Hz. The signal generated by the photovoltaic cell was strengthened by a lock-in amplifier (Stanford Research Systems, SR830, USA) before being collected for analysis. To probe the photo-generated electrons in the sample powders under light illumination, 2,2,6,6-tetramethylpiperidinyloxy (TEMPO) was used as an electron scavenger and was monitored by EPR. The photoelectrochemical (PEC) properties of sample powders were investigated by a standard three-electrode setup in a commercial single-compartment PEC cell comprising the photoelectrode, Pt foil (1 × 1 cm), and Ag/AgCl electrode as the working, counter, and reference electrode, respectively^55,56. The Ag/AgCl reference electrode was calibrated by measuring the reversible hydrogen electrode (RHE) potential of Pt foil in an H₂-saturated electrolyte. The photoelectrode used for PEC measurements was fabricated by the electrophoretic method⁵⁷: two pieces of clean fluorine-doped tin oxide (FTO) glass (1 × 3 cm) were immersed in parallel, with a separation distance of 1 cm and conductive sides facing inward, into 40 mL acetone solution containing 40 mg samples and 10 mg iodine. A potentiostatic control (Keithley 2450 Source Meter, USA) was applied to create a constant bias of 10 V between the two pieces of FTO glass for 3 min. This treatment deposited ~ 5 mg sample powders onto FTO glass with an area of 1 cm². The as-deposited FTO glass was heated at 473 K for 2 h in N₂ to eliminate adsorbed iodine. The as-deposited FTO glass was then dipped into TiCl₄ methanol solution (0.01 M) and was calcined at 623 K for 15 min. This measure helps to strengthen the interconnection between sample particles deposited on the FTO glass. The PEC measurements were conducted using a Zahner electrochemical workstation. KH₂PO₄/K₂HPO₄ buffer solution (pH = 7.3 ± 0.2) was used as the electrolyte. The buffer solution was prepared by dissolving 0.1361 g KH₂PO₄ (Aladdin, 99.5%) and 0.1742 g K₂HPO₄ (Aladdin, 99%) into 10 mL deionized water. The buffer solution was deaerated and stored in a sealed volumetric flask before use. A 300 W Xenon lamp (Perfect Light, PLX-SXE300, China) was employed as the light source whose output was filtered through an ultraviolet cut-off filter (λ ≥ 420 nm). An electronic timer and shutter (DAHNG, GCI-73) were introduced to generate chopped light illumination. The linear scan voltammetry (LSV) of the photoelectrode was recorded under chopped light illumination from negative potential to positive potential at a scan speed of ~20 mV/s. It is worth mentioning that the potential reported in this work was not iR corrected. For better comparisons, the potentials were adjusted based on the Nernst relation (equation 4) for the reference to the reversible hydrogen electrode (RHE) potential:

$$V_{\mathrm{RHE}}=V_{\mathrm{Ag}/\mathrm{AgCl}}+0.0591\mathrm{V}\times \mathrm{pH}+0.1976\mathrm{V}$$ 4

The flat band potential of sample powders was determined by the Mott-Schottky (MS) analysis and the capacitance was deduced from electrochemical impedance spectra using an a. c. signal of 1000 Hz and an amplitude of 10 mV. The electrochemical impedance spectra were recorded using the same setup for PEC measurement.

Cocatalyst deposition

The cocatalyst was deposited onto the photocatalyst according to the following procedures⁵³: specifically, 0.4 g of the photocatalyst was dispersed in an aqueous solution containing rhodium(III) chloride trihydrate (Aladdin, 98%) and chromium(III) nitrate nonahydrate (Aladdin, 99%) and evaporated at 343 K. The resulting powder was subjected to a calcination process at 623 K for 1 h under flowing N₂ (~100 mL·min⁻¹, Jiaya Chemicals, 99.999%). To prevent the photocorrosion, an amorphous layer of Ti-oxyhydroxide (TiOXH) was deposited after samples were loaded with RhCrO_y⁴⁹: briefly, 18 µL titanium tetraisopropoxide (Aladdin, 95%) and 50 µL aqueous H₂O₂ solution (SCR, 30 wt% in H₂O) were dissolved in 2 mL distilled water. The solution was then added to the sample aqueous suspension (0.1 L). TiOXH was deposited onto the sample powders by illuminating the above suspensions using a 300 W Xenon lamp (Perfect Light, PLX-SXE300, λ ≥ 300 nm) for 10-15 h.

Photocatalyst evaluation

The performance of POWS reactions was evaluated in a commercial photocatalytic testing system (Perfect Light Labsolar-6A, China)⁵⁸. The system employed a top-illumination-type quartz reactor for measurements whose temperature was maintained by a water jacket at 281 K. Typically, 0.4 g sample powders, loaded with RhCrO_y and coated with TiOXH, were suspended into 0.1 L deionized water. After the air in the system was thoroughly evacuated, a 300 W Xenon lamp (Perfect Light, PLX-SXE300, China) was employed for light illumination. An ultraviolet cut-off filter (λ ≥ 420 nm) or an AM1.5G filter was applied to generate visible light or simulated sunlight. The gas content in the reactor was examined at targeted time points by an online gas chromatograph (SHIMADZU, GC2014C, Japan). Ultrapure Argon gas (Jiaya Chemicals, 99.999%) was used as the carrier gas.

Apparent quantum Efficiency (AQE) and solar-to-hydrogen conversion efficiency (STH) measurements

The monochromatic light (420 nm, 450 nm, 500 nm, 550 nm, and 600 nm) was generated by filtering the output of a 300 W Xenon lamp using a band-pass filter. The photon flux of each monochromatic light and the bandwidth of individual band-pass filters was measured by a quantum meter (Apogee MP-300, USA).

The AQE was determined by the following equation (5):

$$\mathrm{AQE}=\frac{n\times \mathrm{moles}\mathrm{of}\mathrm{evolved}\mathrm{gas}\mathrm{per}\mathrm{hour}}{\mathrm{moles}\mathrm{of}\mathrm{photon}\mathrm{flux}\mathrm{per}\mathrm{hour}}\times 100\%$$ 5

where n = 2 and 4 for H₂ and O₂, respectively.

The STH was calculated using the following equation (6):

$$\mathrm{STH}=\frac{\mathrm{Evolved}\mathrm{hydrogen}\mathrm{amount}\times \mathrm{\Delta }{G}_r}{P\times S}\times 100\%$$ 6

where ∆G_r is the reaction Gibbs energy for the photocatalytic reaction (~237 kJ·mol⁻¹), P is the power of simulated sunlight illumination (100 mW·cm⁻²), and S is the illumination area (~28.3 cm²).

Theoretical calculations

The electronic structures of Sr₂TiO₄ containing different dopants and defects were calculated based on density functional theory (DFT). The Vienna ab initio simulation package (VASP) was adopted for the calculation^59,60. The generalized gradient approximation (GGA) and Perdew–Burke–Ernzerhof (PBE) functional were chosen for the description of the exchange-correlation between electrons^61,62. The projector augmented wave (PAW) pseudopotentials were introduced for calculation⁶³. The plane-wave cutoff energy was 450 eV. A 2 × 2 × 1 tetragonal unit cell (a = b = 7.77 Å, c = 12.59 Å) was built as the structure model for calculation. The structures were relaxed to reduce the force on each atom at a value lower than 0.02 eV/Å. The atomic coordinates of the optimized crystal structures can be found in Supplementary Data 1. A Monkhorst-Pack k-points mesh of 8 × 8 × 3 was sampled.

Author contributions

G.L. and X.X. led the project. J.Y. designed and performed experiments. R.L. and Y.L. collected the PL and TRPL data. X.X. analyzed the data and wrote the manuscript. J.H. and G.L. revised the manuscript. All authors were involved in the data discussion.

Peer review

Peer review information

Nature Communications thanks Jinlin Long, Zaicheng Sun 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 interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-024-55748-z.