Integration of bio-inspired lanthanide-transition metal cluster and P-doped carbon nitride for efficient photocatalytic overall water splitting

Abstract Photosynthesis in nature uses the Mn4CaO5 cluster as the oxygen-evolving center to catalyze the water oxidation efficiently in photosystem II. Herein, we demonstrate bio-inspired heterometallic LnCo3 (Ln = Nd, Eu and Ce) clusters, which can be viewed as synthetic analogs of the CaMn4O5 cluster. Anchoring LnCo3 on phosphorus-doped graphitic carbon nitrides (PCN) shows efficient overall water splitting without any sacrificial reagents. The NdCo3/PCN-c photocatalyst exhibits excellent water splitting activity and a quantum efficiency of 2.0% at 350 nm. Ultrafast transient absorption spectroscopy revealed the transfer of a photoexcited electron and hole into the PCN and LnCo3 for hydrogen and oxygen evolution reactions, respectively. A density functional theory (DFT) calculation showed the cooperative water activation on lanthanide and O−O bond formation on transition metal for water oxidation. This work not only prepares a synthetic model of a bio-inspired oxygen-evolving center but also provides an effective strategy to realize light-driven overall water splitting.


INTRODUCTION
Green plants use a cubane-type {CaMn 4 O 5 } cluster for catalyzing the water oxidation reaction in the oxygen evolution center (OEC) of photosystem II (PSII) [1][2][3][4], which is a critical half reaction for converting sunlight energy into chemical energies stored in ATP and NADPH. Synergistic effect among the multi-metal centers of the OEC plays a key role for the high catalytic activity of PSII [5,6]. Ca 2+ serves to adsorb and activate the H 2 O molecule, while Mn with variable oxidation states in the cluster provides the oxidative equivalents. Nature chooses Ca and Mn as the elements to build the cluster, partly because of the availability of the two elements in the environment. To mimic nature, we can use any elements available to us. Lanthanide ions can be a better Lewis acid than Ca 2+ and Co is found to be a common element in water oxidation catalysts. As a result, a lanthanide-cobalt cluster may be a good biomimetic water oxidation catalyst [7]. Mimicking natural photosynthesis, light-driven overall water splitting to produce H 2 and O 2 including both the hydrogen evolution reaction (HER) and oxygen evolution reaction (OER) is a promising pathway for artificial conversion and storage of solar energy [8][9][10]. The OER side is usually the ratedetermining step. Inspired by the structure model of PSII, some interesting heterometallic cubane-like clusters have been designed and synthesized to act as bio-inspired water oxidation catalysts [11][12][13][14][15][16]. Sacrificial agents are used to test the performance of these catalysts. However, the natural OEC functions in an integrated system to optimize overall efficiency of a sequence of events including charge separation, charge transfer and catalytic reaction, which diminishes charge recombination. We envisioned that the synthetic biomimetic OECs should also be studied in an integrated system to reveal their true potentials and provide a better understanding of the synergistic effect in catalysis on an atomic level.
Various approaches have been put forward to build integrated systems for improving the separation and transportation of photo-generated electron-hole pairs [17][18][19][20][21]. For example, close connection between the catalytic center and the photosensitive center can effectively reduce charge recombination rates [22][23][24][25]. Assembling bioinspired OECs on the surface of two-dimensional (2D) layered semiconductor materials may enhance photocatalytic overall water splitting, which uses the connection between the OEC and the 2D materials as a junction to improve charge separation and fine tune surface electronic structure. Herein, we synthesized a heterometallic cluster LnCo II Co III 2 (LnCo 3 ), which is a structural analog to the CaMn 4 O 5 of PSII. By anchoring the bioinspired LnCo 3 as the OEC on phosphorus-doped (P-doped) graphitic carbon nitrides (PCN), we realized light-driven spontaneous overall water splitting to efficiently produce O 2 and H 2 . The NdCo 3 /PCN-c exhibited remarkable watersplitting activity with a high H 2 production rate of ∼297.7 μmol h −1 g −1 and O 2 evolution rate of 148.9 μmol h −1 g −1 under light irradiation. The photoexcited electron-hole pairs would be easily dissociated and transferred into the PCN and LnCo 3 to participate in HER and OER, respectively.

RESULTS AND DISCUSSION
LnCo 3 was synthesized by reacting Ln(NO 3 ) 3 , Co(Ac) 2 and bis-tris-propane (btp) in methanol solution. The LnCo 3 s are isostructural. Here we only describe the structure of NdCo 3 in detail. Single X-crystal structural analysis showed that NdCo 3 crystallizes in monoclinic P2(1)/n space group and contains one [NdCo II Co III 2 (btp-3H) 2 (Ac) 2 (NO 3 ) 2 ] + cation core, one nitrate ion and two guest water molecules. In the metal core, each Co 3+ is chelated by two N and three O atoms from one deprotonated btp-3H ligand, to form one stable [Co III (btp-3H)] unit (Supplementary Fig. 1 [26][27][28]. According to the theoretical calculation formula of valence bond, BVS = exp((R 0 −R)/B), the states of the metal ions and the pronation states of each oxygen and nitrogen atom of the organic ligand were calculated. As shown in Supplementary Tables 11 and 12, the calculated results show that the oxidation states of all the lanthanide ions are +3 and the oxidation states of Co1 and Co2 are +2 and +3; the O7, O9 and O10 in btp-3H ligands show the O 2− state, while the O6, O5 and O8 atoms exhibit OH − states.
Interestingly, the heterometallic cluster [NdCo II Co III 2 ] mimics the structure of CaMn 4 O 5 of PSII. Considering the monotonic change in radius and chemical properties of the lanthanides, it was an attractive choice for investigating the physical characteristics of the clusters [29]. In addition, due to the similarities in ionic radii and high coordination numbers of lanthanide ions and Ca 2+ , they can be exchanged in biological systems [30,31]. As shown in Fig. 1, topologically the NdCo 3 cluster can be viewed as the CaMn 4 O 5 missing one metal vertex from the cubane and adding one bridging-O atom between Nd 3+ and Co 3+ . In addition, the coordination mode of bridging-O in the NdCo 3 cluster is also very similar to that in CaMn 4 O 5 of PSII, except that the five bridging-O atoms are O 2− in biological CaMn 4 O 5 -cluster, while six bridging-O atoms come from the −OH groups of two btp-3H ligands in NdCo 3 . Notably, the mixed oxidation states of the cobalt ions (+2 and +3) in the NdCo 3 cluster are similar to the mixed oxidation states of manganese ions in CaMn 4 (+3 and +4), suggesting that the NdCo 3 cluster can be viewed as a synthetic model of the OEC [32][33][34][35]. Compared with the CaMn 4 O 5 , the NdCo 3 cluster shows high stability because of the presence of a chelating btp-3H ligand. High-resolution electrospray ionization mass spectrometry (HRESI-MS) of NdCo 3 in methanol shows main peaks in the range of 1242 to 1252, which corresponds to the Supplementary Fig. 3). This result indicates that the NdCo 3 cluster remains intact in methanol solution. The EuCo 3 and CeCo 3 clusters show the same crystal structure as NdCo 3 ( Supplementary Fig. 4).
Based on the stability of the cluster in methanol solution, anchoring NdCo 3 clusters on PCN was prepared as shown in Fig. 2a. Forty-five milligrams of prepared PCN was dispersed in methanol solution (3 mg/mL) with sonication and then transferred to the flask with stirring. One milliliter of NdCo 3 methanol solution (1, 2, 3, 5 and 7 mg mL −1 ) was dripped into the PCN suspension and refluxed for 12 h. The resultant precipitates were collected by filtration and dried at 70 o C overnight, resulting in NdCo 3 Table 6).
Transmission electron microscopy (TEM) shows that the obtained NdCo 3 /PCN exhibits the morphology of nanosheets (Fig. 2b). To determine the distribution of the NdCo 3 clusters, atomicresolution high-angle-annular-dark-field scanning transmission electron microscopy (HAADF-STEM) measurement was performed. The isolated bright dots in Fig. 2d can be assigned to NdCo 3 clusters. Elemental mapping analysis of the STEM images revealed that the Nd, Co and P atoms are uniformly distributed throughout the nanosheets (Fig. 2e), demonstrating good dispersion of NdCo 3 clusters on the PCN support.
Extended X-ray absorption fine structures (EXAFS) of NdCo 3 and NdCo 3 /PCN-c were performed to probe the first coordination sphere of Co 3+ /Co 2+ metal centers. As displayed in Fig. 3a, the Co K-edge X-ray absorption near edge spectroscopy (XANES) of the NdCo 3 cluster gives a rising edge between that of CoO and Co 2 O 3 , indicating that Co centers in NdCo 3 have mixed oxidation states of +2 and +3, which is consistent with the crystal structure analysis. The XANES spectrum of Co centers in NdCo 3 /PCN is very similar to that of the isolated NdCo 3 cluster, suggesting that the Co oxidation states in NdCo 3 remain the same during the assembly on PCN. The pre-edge of the NdCo 3 and NdCo 3 /PCN-c showed that Co ions are maintaining an octahedral coordination [36]. As shown in Fig. 3b, the Fourier transform (FT) peak of the extended X-ray absorption fine structure (EXAFS) at 1.43Å contains both Co-O and Co-N coordination. An emerging peak at 1.59Å after anchoring the cluster on PCN could be ascribed to Co-P coordination with P from PCN. The EXAFS also confirms that no Co nanoparticles were formed in the reaction. A peak at a high R value (ca. 2.60Å) corresponds to the distance of Co...Co path, which is also present in the as-prepared cluster. As shown in Supplementary Fig. 6, the experimental and fitting FT-EXAFS curve of Nd 3+ (Nd L III Edge) in NdCo 3 /PCN-c can be perfectly matched, which indicates that Nd 3+ in NdCo 3 /PCN-c and the sample after reaction have the same coordination environment. The X-ray photoelectron spectroscopy (XPS) of NdCo 3 /PCN-c shows characteristic peaks of Nd 3d and Co 3d ( Supplementary Fig. 7). Co 2p XPS spectra show different oxidation states of Co ions in the NdCo 3 cluster on PCN nanosheets (Fig. 3c). The P 2p XPS spectra for the NdCo 3 /PCN-c sample displayed two peaks at 129.5 and 133.1 eV, which can be attributed to P with and without Co-P connections, respectively (Fig. 3d), and no peak related to Co-P showed up in the PCN spectra [36][37][38]. This XPS result suggests that P atoms coordinated with Co ion in the NdCo 3 cluster. According to the EXAFS and XPS results, the NdCo 3 cluster was anchored on the PCN through Co-P bonds and remained intact during the assembling process. The linker model of NdCo 3 /PCN-c was shown in Fig. 2a.
The photocatalytic overall water-splitting performances of NdCo 3 /PCN-c catalysts were evaluated in pure water without any sacrificial reagents under simulated solar illumination (see details in Supplementary Data). As shown in Fig. 4a, NdCo 3 /PCNc with different cluster loadings on PCN display different photocatalytic activities under light (λ > 300 nm) irradiation. With increased loading from 0.31 to 1.05 wt%, the photocatalytic activity of NdCo 3 /PCN-c improved because of the increased number of active sites for OER. The NdCo 3 /PCN-c with the loading of 1.05 wt% NdCo 3 shows the highest photocatalytic H 2 production rate of 297.7 μmol h −1 g −1 and O 2 production rate of 148.9 μmol h −1 g −1 , which is approximately 7.2 times that of PCN without loading clusters. Further increasing the loading of NdCo 3 to 1.55 and 2.03 wt% leads to slightly reduced H 2 and O 2 production rates, possibly due to competitive transfer of holes to adjacent clusters, which decreases the chance of transferring four electrons to the same OEC to complete the whole OER process. As shown in Fig. 4b, the time courses of simultaneous evolution of H 2 and O 2 gases of NdCo 3 /PCN-c display a constant H 2 /O 2 stoichiometric ratio of 2 : 1, suggesting the occurrence of overall water splitting. The NdCo 3 /PCN-c catalyst was recovered and reused four times without significant decrease in photocatalytic activity. A time course of H 2 and O 2 production of NdCo 3 /PCN-c under visible light irradiation (λ > 420 nm) was also studied. As displayed in Supplementary Figs 9 and 10, NdCo 3 /PCN-c shows about 210.4 μmol g −1 of H 2 production rate and 105.7 μmol g −1 of O 2 production rate in 12 hours under visible light irradiation. The quantum efficiency closely followed that of the ultraviolet-visible (UV-vis) absorbance trend, revealing that the reaction was driven by light absorption by the catalyst (Fig. 4c). Specifically, the NdCo 3 /PCN-c achieves a photocatalytic quantum efficiency of 2.0% at 350 nm and retains a quantum efficiency of 1.2% at the visible-light wavelength of 420 nm. To study the catalytic activity of NdCo 3 itself, the cyclic voltammetry (CV) and linear sweep voltammetry (LSV) measurements were performed in the cell equipped with three electrodes, working electrode, counter electrode (Pt plate) and reference electrode (Ag/AgCl) in 0.5 M NaAc/HAc buffer solution (pH = 6). As shown in Supplementary Fig. 11, the NdCo 3 cluster has obvious water oxidation catalytic activity. The overpotential for NdCo 3 is 325 mV to reach 1 mA cm −2 . The photocatalytic OER activity of the NdCo 3 cluster itself was also studied in 20 mL 0.5 M NaAc/HAc (pH = 8) buffer solution using 1 mM [Ru(bpy) 3 ]Cl 2 as photosensitizer and 5 mM Na 2 S 2 O 8 as sacrificial reagent. Under λ ≥ 420 nm light irradiation, the NdCo 3 cluster shows a photocatalytic O 2 production rate of 9.5 μmol h −1 g −1 , which is close to the 11.5 μmol h −1 g −1 O 2 evolutions of NdCo 3 /PCN-c ( Supplementary Fig. 12).
The close values of the photocatalytic O 2 production rates of NdCo 3 and NdCo 3 /PCN-c suggest that the OER rate is still the rate-determining step for the overall water splitting of NdCo 3 /PCN-c.
The TEM image and HAADF-STEM image of NdCo 3 /PCN-c after photocatalysis show that the morphology of NdCo 3 /PCN remained unchanged after the photocatalytic reaction (Supplementary   13). ICP-MS studies revealed that less than 0.3% of clusters leached into the solution after a reaction of 12 h, indicating the stability of NdCo 3 /PCN-c. The stability of clusters is not only due to the chelating effect of the bis-tris-propane ligand but also due to the Ln 3+ ions that stabilize the 3d-4f cubane structure [39,40]. To verify the role of the lanthanide on the photocatalytic activities, the isostructural EuCo 3 /PCN and CeCo 3 /PCN clusters were also studied. As shown in Fig. 4d, under light (λ > 300 nm) irradiation, EuCo 3 /PCN and CeCo 3 /PCN show the photocatalytic H 2 production rate of 279.1 and 274.5 μmol h −1 g −1 respectively, which are close to that of NdCo 3 /PCN-c. The time courses of H 2 and O 2 evolutions of EuCo 3 /PCN and CeCo 3 /PCN in 12 hours under light (λ > 300 nm) irradiation are displayed in Supplementary Figs 14 and 15.
To exclude the contribution of other species for the catalytic activity in this system, the control experiments-by combining CoO, Co 3 O 4 and Co(Ac) 2 , and Nd(NO 3 ) 3 with PCN as the photocatalysts respectively for overall water splitting based on the same method as that of NdCo 3 -were performed. As shown in Supplementary Fig. 16, CoO/PCN, Co 3 O 4 /PCN and Nd(NO 3 ) 3 /PCN showed very low catalytic activity. Although the Co(Ac) 2 /PCN can give rise to a significant capability of water splitting (H 2 production rate of 126.4 μmol h −1 g −1 ), compared with the activity of [Co(Ac) 2 + Nd(NO 3 ) 3 ]/PCN under the same conditions, the NdCo 3 /PCN-c shows much higher performance with 297.7 μmol h −1 g −1 . These control experiments suggested that the NdCo 3 itself boosts the activity in the system.
Electrochemical impedance spectroscopy (EIS) Nyquist plots and the transient photocurrent were measured to characterize the electron-hole transfer efficiency. NdCo 3 /PCN-c has a much smaller semicircle diameter and lower interfacial chargetransfer resistance than that of PCN, demonstrating the enhanced interfacial charge transfer of NdCo 3 /PCN-c ( Supplementary Fig. 17). Consistently, NdCo 3 /PCN-c has better photocurrent responses under irradiation than that of PCN (Supplementary Fig. 18). Photoluminescence (PL) and the time-resolved fluorescence spectra of PCN and NdCo 3 /PCN-c were performed ( Supplementary  Fig. 19a). They were monitored at 430 nm under irradiation by a 368 nm laser at room temperature. Time-resolved fluorescence spectra revealed average lifetimes of approximately 2.17 and 1.91 ns for NdCo 3 /PCN-c and PCN, respectively (Supplementary Fig. 19b).
The photocatalytic H 2 or O 2 production reaction in the presence of a hole acceptor or electron acceptor could be performed to reveal more details about the two processes. The photocatalytic H 2 evolution of NdCo 3 /PCN-c was enhanced in the presence of CH 3 OH as a hole acceptor, as compared to that without sacrificial agent ( Supplementary Fig. 20), indicating that the intrinsic catalytic activity of the HER side is higher than that exhibited in overall water splitting. The rate-determining step is thus likely on the OER side. However, the photocatalytic O 2 evolution of NdCo 3 /PCN-c in the presence of AgNO 3 as an electron acceptor was slower than that without sacrificial agent ( Supplementary Fig. 21), which suggests that the hole injection into NdCo 3 is not the rate-determining step in the OER [41]. We thus conclude that the OER rate is still limited by catalysis.
Femtosecond time-resolved transient absorption (fs-TA) spectroscopy was used to detect the ultrafast excited state dynamics of the system (Supplementary Fig. 22) [42,43]. The dynamics in the femtosecond-picosecond (fs-ps) time scale can be fitted to a five-component exponential model as shown by the time trace at 520 nm (excited at 360 nm) ( Fig. 5a and b). In the initial 100 ps, negative signals due to ground state bleach (GSB) are prominent, which reflect the behavior of holes on the valence band (VB). Evolution of the initial GSB signal can be described by three time constants: τ 1 = 1.02 ps, τ 2 = 4.46 ps, τ 3 = 69.8 ps for PCN, and τ 1 = 0.14 ps, τ 2 = 1.76 ps, τ 3 = 54.5 ps for NdCo 3 /PCN-c. The τ 1 and τ 2 correspond to initial vibrational cooling of energetic holes, and τ 3 may be attributed to the hole transfer process to the surface trap site. Compared to the PCN sample in the early 100 ps, NdCo 3 /PCN-c has shorter relaxation times of all the three components. The accelerated hole transfer rate may be related to fast hole transfer to the NdCo 3 cluster in NdCo 3 /PCN-c. After a few hundred picoseconds, electrons and holes complete the transfer to trap states. The TA signals of both PCN and NdCo 3 /PCN-c samples showed a significant signal growth on the timescale of hundreds of ps to ns (τ 4 ). This growth may be due to stimulated emission (SE) from trap states, which are supported by fluorescence lifetimes on an ns timescale. The NdCo 3 /PCN-c sample shows a longer τ 4 time than PCN (τ 4 = 1.84 ns for NdCo 3 /PCN-c vs. τ 4 = 1.24 ns for PCN) because hole transfer to the cluster competes against populating the surface trap sites and thus delays the emissive electron-hole recombination. The fifth time constant (τ 5 = 2.39 ns for NdCo 3 /PCN-c vs. τ 5 = 1.81 ns for PCN) represents the fluorescence process, which is consistent with the result from time-resolved fluorescence ( Supplementary Fig. 19b). The τ 5 's determined by TA spectra are not very accurate due to the limited number of points on the long waiting time. Based on these analyses, we can understand the reason why NdCo 3 /PCN-c improves higher photocatalytic performance (Fig. 5c). The cluster not only acts as a reaction center for water oxidation but also suppresses electron-hole recombination due to fast hole injection into the clusters. This efficient hole transfer leads to an increase in hole utilization and finally improves the overall efficiency.
To further investigate the catalytic OER, spin polarized DFT + U calculations were carried out using the VASP software [44]. NdCo 3 clusters can easily lose two Ac − ligands from the Nd 3+ and Co 3+ ions in aqueous solution, resulting in two coordination unsaturated sites (CUS). A series of geometrical optimizations reveal that the CUS of one Co (III) ion prefers to coordinate with PCN by the anchoring site of P atom (d P-Co = 2.436Å) with an adsorption energy of −1.01 eV, while another Co 3+ (CUS) can serve as the catalytic center. One water molecule adsorbs on the CUS of Co 3+ ions with a d Co-O of 2.191Å and adsorption energy (E ads ) of −0.74 eV. Both the charge density difference and electron localization function (ELF) value (around 0.5) suggest that the 3d orbital of Co 3+ (CUS) ions effectively overlaps with the 2p orbital of O w , resulting in the formation of one weak coordination bond. Partial density of states curves of spin up and spin down indicate that spin carriers (e.g. Nd and Co ions, P and N of C 3 N 4 ) present apparent spin polarization at the vicinity of Fermi level ( Supplementary Fig. 25 Oxidation states of metal ions are poorly described by Bader charge calculation [45]. As a result, magnetic moments were evaluated to assist oxidation state identification, considering their similar coordination field from oxygen atoms [46]. The spin states of each intermediate were investigated by using symmetry-broken calculations (Supplementary Table 5). Initially, both Co (P) and Co (CUS) ions feature low spin state Co III ions, while the middle divalent Co(M) ion features high spin state Co II ion. After two hole injections to the cluster, the water molecule attached to the Co III (CUS) ion is deprotonated and converted to * OH. The injected two holes bring the Co III (P), Co II (M) and Co III (CUS) ions to the oxidation states of +4, +3 and +3, respectively. Then, another hole injection induced the deprotonation of * OH, yielding bridging * O in Co IV (CUS)-oxo-Nd III (spin magnetization of 2.875 μ B on Co). The Co IV (CUS)-oxo is electron deficient, which is an ideal target for a nucleophilic attack by a second water in a concerted process of forming one O−O bond and one O-Nd III coordination while losing one proton upon injection of another hole. A bridging * OOH species and high oxidation state Co V (CUS)-hydroperoxyl-Nd III (spin magnetization of −2.648 μ B on Co) is formed. Finally, the liberation of O 2 from the cluster with concomitant deprotonation and reduction of the cluster to regenerate its initial oxidation state happens at low activation energy. Generally speaking, the Ln 3+ ion stabilizes negatively charged intermediates, and the other two Co ions that are not directly attached to water molecules serve as hole reservoirs to store oxidation equivalents and thus avoid the building up of too high an oxidation potential on one Co ion. All four metal ions in the cluster synergistically catalyze the water oxidation ( Supplementary  Fig. 28).

CONCLUSION
In summary, we demonstrated a bio-inspired lanthanide-transition metal cluster as an oxygenevolving center anchored on PCN for efficient photocatalytic overall water splitting. The obtained LnCo 3 clusters not only display high stability but also show excellent oxygen-evolving activity. The combination of LnCo 3 clusters and PCN achieves efficient separation of electrons and holes and enables rapid production of H 2 and O 2 . Mechanistic investigation shows synergistic effects of lanthanide ion and variable-valence Co ions in the oxygenevolving reaction. This work not only prepares a synthetic model of a bio-inspired oxygen-evolving center but also develops an avenue to designing efficient catalysts for overall water splitting by coupling bio-inspired clusters and photoactive supports.

Synthesis of P-doped C 3 N 4 photocatalysts
A mixture of 0.5 g of the prepared C 3 N 4 and 0.25 g NaH 2 PO 2 was ground with motar. Then, the mixture was heated to 350 o C in 2 o C/min in a muffle furnace and then heated for 2 h in a N 2 atmosphere. The resultant precipitate was ultrasonicated and washed with water and ethanol twice, collected by filtration and dried at 70 o C overnight.

Synthesis of NdCo 3 /PCN photocatalysts
Forty-five milligrams of PCN was dispersed in methanol solution (3 mg/mL) with sonication, and then transferred to a flask with stirring, then 1 mg, 2 mg, 3 mg, 5 mg and 7 mg NdCo 3 clusters in 1 mL methanol were dropped into the suspension and refluxed for 12 h, respectively. The resultant precipitate was collected by filtration and dried at 70 o C overnight.

Photocatalytic reactions
The photocatalytic experiments were performed via a photocatalytic evaluation system (CEL-SPH2N, CEAULight, China) in a 300 mL Pyrex flask. A 300 W Xenon arc lamp with a wavelength range of 300-800 nm was used as the light source. The focused intensity on the flask was ∼200 mW · cm −2 . In a typical photocatalytic experiment, 40 mg of photocatalyst was suspended in aqueous solution. Before irradiation, the system was vacuumed for 10 min via the vacuum pump to completely remove the dissolved oxygen. The evolved gases contents were analyzed by gas chromatography (GC7920, CEAULight, China). The apparent quantum efficiency was measured under identical photocatalytic reactions. Single wavelength 365 nm, 420 nm, 450 nm, 500 nm and 600 nm filters were employed as the light sources to trigger the photocatalytic reactions, respectively.

Photochemical studies
Cyclic voltammograms (CV), EIS data, photocurrent and the Mott-Schottky spots were recorded using electrochemical workstation (CHI 760E, Shanghai Chenhua). The Indium tin oxide glasses with samples were served as the working electrodes. EIS measurements were recorded over a frequency range of 100 kHz-200 kHz with ac amplitude of 20 mV at 0 V. Water was used as the supporting electrolyte. The Mott-Schottky plots were also measured over an alternating current frequency of 1000 Hz, 1200 Hz and 1500 Hz. These three electrodes were immersed in the 0.2 M Na 2 SO 4 aqueous solution (pH = 6.6).
All other experimental details, as well as TA spectroscopy characterizations and the DFT calculations, are provided in the Supplemental Experimental Procedures.