High-entropy photocatalyst enabling collaborative cascade C–N coupling for converting benzyl alcohol to imine

Abstract

The selective conversion of biomass-derived alcohols to nitrogen-containing organic chemicals, such as the highly valuable imine, is a significant challenge due to the complex tandem process of C–N coupling and multiple reaction pathways. The photocatalytic coupling of biomass-derived alcohols with NH₃ using renewable energy offers a sustainable way to produce imines, but requires the development of suitable photocatalysts. In this work, we present a promising high-entropy sulfide photocatalyst (NiSnCaZnInS) with integrated multifunctional sites. Using benzyl alcohol (BA) as an example, we achieved an imine yield of 1240 μmol ${\text{g}}_{\text{cat}}^{-1}$ for 5 hours, with a selectivity of 91.7%. The photocatalytic performance can be attributed to the special configuration of the high-entropy surface, which provides ideal catalytic sites for the cascade coupling reaction. This work highlights the great potential of high-entropy materials in photocatalytic reactions, providing a sustainable and effective approach to the photocatalytic conversion of biomass derivatives into valuable chemicals.

The design of multifunction reactive sites is pivotal for cascade reaction. Here, the authors prepared a high-entropy photocatalyst that enables direct C–N coupling of alcohols with NH₃ to form imines, achieving high benzyl alcohol-to-imine yield.

Introduction

Among nitrogen-containing chemicals, imines are vital and ubiquitous as key intermediates in the synthesis of pharmaceuticals, agrochemicals, amino alcohols and iminium catalysis^1–4. A large number of methods have been developed for imine synthesis, including the direct condensation of carbonyl substrates with amines, the self-condensation of primary amines upon oxidation, the oxidative dehydrogenation of secondary amines using various oxidants and hydrogenating condensation via nitrile reduction^5–7. However, these methods generally require acid/base additives or redox reagents and produce considerable liquid waste^8,9. In order to address the aforementioned limitations of this reaction, the direct synthesis of imines from more readily available chemicals via an atom-efficient and sustainable process has been recognized as a more environmentally benign and atom-economic approach¹⁰. Specifically, the direct coupling of alcohols with NH₃ was proposed as an alternative approach¹¹. Alcohols are highly desirable as substrates due to the promising developments in lignocellulose refining, which enable efficient access to diverse alcohols¹². However, the promising route of alcohol oxidation coupling with NH₃ suffers from severe conversion issues, as it requires complex cascade dehydrogenation and dehydration coupling processes^13,14. The construction of sites is a fundamental challenge for the aforementioned reaction, which involves multiple steps to accomplish efficient C–N coupling. Although cocatalyst loading has been extensively employed as a general strategy to tune active sites¹⁵, randomly dispersed cocatalysts do not ensure efficient cascade reactions for complicated C–N coupling. Thus, precise active site design is essential for promoting efficient C–N coupling.

Recently, photocatalytic dehydrogenation of alcohols to couple with amines for producing imines or second amine has realized comparable activity and selectivity^16–18. The reactions in these studies primarily rely on the oxidation of alcohols to generate the corresponding aldehydes, which then directly undergo subsequent condensation reactions with additional amines to achieve C−N coupling. The advantage of this approach lies in its relatively straightforward reaction process, but the amine source is limited to the rather unreactive anilines, causing the corresponding photocatalytic reactions typically proceed as asymmetric C–N coupling. The direct synthesis of imine from more accessible alcohols and NH₃ has promising advantages in terms of economy and sustainability. Nevertheless, this ideal reaction route involves multiple C–N coupling and redox reactions, which pose significant challenges to the supported active sites on conventional photocatalysts. Thus, facile photocatalysts that enable imine production through the partial ammoxidation strategy remain highly desirable^19–22. However, most of these advanced procedures often suffer from the poor yield and selectivity for the harsh requirements on multifunctional reactive sites^23–26. The high-entropy materials (HEMs) have emerged as a novel and promising functional material^27–29. In general, the homogeneous composition of various elements gives rise to multifunctional catalysts with a unique synergistic effect from the multi-metal components and the effect of lattice distortion^20–32. HEMs based on conventional photocatalysts can usually absorb light continuously, which is crucial for photocatalysis^33–36. Meanwhile, the various surface atom configurations of HEMs offer abundant active sites and diverse microstructures. The adjacent active sites at an atomic scale ensure highly efficient cascade reactions^37–39. Metal sulfides are the most widely utilized catalysts in the fields of photocatalysis and electrocatalysis due to their excellent semiconductor properties and unique electronic structures^40–43. Thus, high-entropy sulfides have remarkable potential to realize complex photocatalytic alcohol oxidation coupling with NH₃, although related research on high-entropy photocatalysts remains scarce.

To address the challenges of tandem reactions and selectivity control in direct partial ammoxidation from alcohol and NH₃ (Fig. 1a), we propose the use of a high-entropy sulfide (NiSnCaZnInS, denoted as NSCZIS) as a photocatalyst (see detailed rationale for element selection in Supplementary Note 1), using benzyl alcohol as a typical substrate. This design enables simultaneous reactant activation and intermediate conversion, ensuring an effective tandem catalytic process involving the dehydrogenation of alcohol, amination and dehydration coupling. An orthogonal experiment screens the optimized composition of the high-entropy samples from nine sulfides with different Ni/Sn/Ca/Zn/In/S atomic ratios (Supplementary Fig. 1 and Table 1). Optimized NSCZIS exhibits substantially higher activity in converting BA to N-benzylbenzaldimine (NBI), achieving a yield of 1240 μmol ${\text{g}}_{\text{cat}}^{-1}$ and a selectivity of 91.7% over a 5-hour period. Our study demonstrates the significant potential of a rationally designed high-entropy photocatalyst in sophisticated cascade reactions. This work offers a strategy for the sustainable production of a range of imines, advancing the conversion of alcohol ammoxidation to high-value nitrogen-containing chemicals efficiently.

Fig. 1. Schematic illustration for the BA conversion to NBI by catalysis processes.

a The proposed process from BA to NBI. b The comparison of the traditional route and our work.

Results and discussion

Catalytic performance

Various high-entropy sulfide photocatalysts were prepared by a hydrothermal method based on different metal elements as illustrated in Fig. 2a. After thorough optimization, the ZnInS-derived high-entropy photocatalyst accomplished exceptional imine synthesis at mild conditions by direct photocatalytic BA ammoxidation that demonstrated various advantages over the traditional routes (Fig. 1b). The configurational entropy of our high-entropy sulfide is calculated to be 1.51 R based on the element contents (Supplementary Table 2), which is higher than the threshold (1.50 R) of high-entropy materials³⁷, and R is the universal gas constant (8.314 J K⁻¹ mol⁻¹). This manifests that our NSCZIS is high-entropy material rather than simple multi-component solid solutions or randomly doped systems. To evaluate the component–performance relationship, we prepared ZnInS, CaZnInS and NiSnZnInS sulfide photocatalysts, denoted as ZIS, CZIS and NSZIS, respectively. As the oxidation of BA to BAD is the initial step in ammoxidation of BA, we firstly evaluated the individual photocatalytic oxidation performance by only using BA as the initial substrate. The products were quantified by gas chromatography–mass spectrometry (GC–MS) (Supplementary Figs. 2, 3). The conversion rate of BA towards BAD over ZIS is only 336 μmol ${\text{g}}_{\text{cat}}^{-1}$ h⁻¹ (Supplementary Fig. 4). Owing to the introduction of hetero atoms, the product yield increases to 652 μmol ${\text{g}}_{\text{cat}}^{-1}$ h⁻¹ and 515 μmol ${\text{g}}_{\text{cat}}^{-1}$ h⁻¹ over CZIS and NSZIS. Specifically, the optimal activity of 730 μmol ${\text{g}}_{\text{cat}}^{-1}$ h⁻¹ is observed on NSCZIS. To validate the direct oxidation dehydrogenation of BA by photocatalytic process, we conducted control investigations. The O₂ could participate the oxidation reaction of alcohol by O₂ reduction to generate the •O₂^-, thus the degassing experiment was conducted to validate the role of O₂ on the oxidation of benzyl alcohol. The negligible variance in BA photocatalytic conversion to BAD between air conditions and degassed treatment could be attributed to the direct oxidation dehydrogenation by photogenerated holes and preclude the possibility of •O₂^- involvement (Supplementary Fig. 5a). The free participation of reactive oxygen species reasonably results in the insensitivity of BA oxidation to the existence of O₂. In addition, the production of BAD is severely inhibited as the additional triethanolamine (TEOA) is added, while the •OH trapping by tertiary butanol (TBA) shows merely mild inhibition on the photocatalytic activity (Supplementary Fig. 5b). The conversion results conclusively indicate that the dehydrogenation process is direct oxidation by photogenerated holes via a radical-free mechanism. Interestingly, the photocatalytic performance was significantly hindered when 5,5-dimethyl-1-pyrroline-N-oxide (DMPO) was used as a •CH(OH)Ph trapping agent, implying the initial cleavage of the α-C–H of BA in photocatalytic dehydrogenation reaction¹⁹.

Fig. 2. Photocatalytic benzyl alcohol conversion to imine.

a Element scheme of photocatalyst components. b Photocatalytic BA (benzyl alcohol) conversion to NBI (N-benzylidenebenzylamine) by NSCZIS, NSZIS, CZIS and ZIS under LED irradiation (400 nm, 5 W) for 5 h. BAD and BN represent benzaldehyde and benzylamine, respectively. Error bars mean ± standard deviations calculated from three independent measurements. c Time course of BA conversion to NBI on NSCZIS in the reaction period of 5 h. d The stability tests of NSCZIS for five successive cycles (each cycle is 5 h). e Selectivity of NSCZIS using various substituted benzyl alcohols under optimized conditions. Reaction condition: 10 mg catalyst, 7 mL H₂O, 3 mL NH₃·H₂O and 100 μl BA (0.97 mmol) were added in a 50 mL quartz reactor, the NH₃ concentration is ca. 4.5 mol L⁻¹. The reaction system was energetically stirred at 500 rpm and kept at 20 °C.

To elucidate the underlying mechanism for the variance on benzyl alcohol oxidation activity among these sulfides, we also investigated the behaviors of photogenerated carriers. The fluorescence decay dynamics revealed a significant prolongation of the lifetime from 6.84 ns (ZIS) to 39.39 ns (NSCZIS), indicating that the high-entropy effect induced by multicomponent configuration considerably facilitates the charge migration and separation efficiency (Supplementary Fig. 6). The electrochemical impedance spectra (Supplementary Fig. 7a) further validate the accelerated interfacial charge transfer and the smallest interfacial resistance over NSCZIS. Moreover, transient photocurrent response of NSCZIS (Supplementary Fig. 7b) also demonstrates a superior performance (50 μA cm⁻²) to the NSZIS (30 μA cm⁻²), CZIS (25 μA cm⁻²) and ZIS (17 μA cm⁻²). The lifetime and photocurrent of NSCZIS are near 6 and 3 times higher than those of ZIS, respectively, which are consistent with the conversion rates of BA for NSCZIS (730 μmol ${\text{g}}_{\text{cat}}^{-1}$ h⁻¹) and ZIS (336 μmol ${\text{g}}_{\text{cat}}^{-1}$ h⁻¹). These findings corroborate that the introduction of heteroatoms (Ni, Sn, Ca) collaboratively achieves the prominent enhancement in charge separation. Furthermore, the in-situ EPR signals using the DMPO as a probe molecule of photogenerated holes demonstrate that the optimal NSCZIS exhibits substantial intensity compared to pure ZIS, indicating that the optimal NSCZIS can generate more holes (Supplementary Fig. 8)⁴⁴.

Generally, there is a strong correlation between the variation of component configuration of photocatalysts and their performance. We thus further assessed the complicated ammoxidation reaction on alcohol conversion to imine by multicomponent photocatalysts (Fig. 2b, Supplementary Figs. 9–14, Supplementary Note 2). The pristine ZIS shows a low yield for BA conversion to imine with substantial byproducts due to the inadequate reactive sites (Supplementary Fig. 15). The introduction of NiSn or Ca leads to prominently enhanced imine synthesis yield of 710 or 678 μmol ${\text{g}}_{\text{cat}}^{-1}$ under the same conditions, respectively (Supplementary Figs. 16, 17). The results evidently indicate that the role of each element is pivotal for the catalytic reaction. Compared with those reference sulfides, NSCZIS delivers a substantially higher activity for the conversion of BA to NBI with a yield of 1240 μmol ${\text{g}}_{\text{cat}}^{-1}$ and selectivity of 91.7%, highlighting the special high-entropy effect for complicated photocatalytic reaction (Supplementary Figs. 18, 19). Physical mixture (NSS&ZIS) of NiSnS and ZnInS shows rapid decline in performance for NBI that could be attributed to the undesirable separation of reaction sites in atom scale (Supplementary Fig. 20). In addition, the NiSnS decorated CaZnInS (denoted as NSS@CZIS) was prepared and the photocatalytic conversion of BA to imine was also evaluated. The performance of NSS@CZIS is clearly reduced compared with the NSCZIS (Supplementary Fig. 21). The Ni-passivation experiment by KSCN demonstrates the evident inhibition for the passivation of Ni sites, thus verifying the role of Ni on the generation of benzylamine⁴⁵. The NiCaZnInS (NCZIS) and SnCaZnInS (SCZIS) both show decreased yields and poor selectivity (Supplementary Fig. 22). The above control experiments collaboratively validate the synergistic effect of Ni and Sn on facilitating the generation of benzylamine.

As shown in Supplementary Fig. 23 and Table 3, the NSCZIS exhibits a maximum quantum yield (AQY) of 5.2% at 400 nm for the photocatalytic conversion of BA to NBI. The AQY values subsequently decrease to 5.0% and 1.7% at 420 nm and 450 nm, respectively. A series of control experiments were carried out to evaluate the photocatalytic BA conversion to NBI (Supplementary Fig. 24). Only BA is detected when the light or catalyst is absent, suggesting the photocatalytic nature of our system. Interestingly, the formation of NBI is dramatically suppressed when water is replaced by anhydrous acetonitrile and BAD is almost the only product. This could be attributed to the fact that water acts as the proton donor to generate BN by photocatalytic hydrogenation of imine intermediate. Thanks to the efficient synergy of multiple reactive sites, NSCZIS exhibits the outstanding photocatalytic performance on imine synthesis, which is even much higher than that of most reported results by taking aniline and benzyl alcohol as substrates (Supplementary Table 4).

The conversion process of BA to imine on NSCZIS was further investigated by identifying the population of products in a period of 5 h (Fig. 2c). During the initial stage, BAD was dominantly produced on the surface of NSCZIS, along with small amount of benzylamine (BN) and imine. The generation of benzaldehyde is rapid in initial stage while the content decreased after 1 h, accompanied by an increase of BN. The phenomenon indicates that BAD may undergo further conversion to benzylamine through photocatalytic hydrogenation. The concentration of BAD and BN altered slightly, while the amount of imine continued to increase after 1 h, indicating that BAD and BN likely served as intermediates in imine synthesis from BA. The average conversion rates of BA are 538, 367, 445, 356 ${\mathrm{\mu }\mathrm{mol}\text{g}}_{\text{cat}}^{-1}$ h⁻¹ for NSCZIS, NSZIS, CZIS, ZIS, respectively (Supplementary Fig. 25). Moreover, the photocatalytic performance of NSCZIS was maintained after five consecutive photocatalytic cycles as depicted in Fig. 2d, fully indicating the stability of NSCZIS photocatalyst.

Furthermore, a series of extensive experiments were utilized to investigate universal availability of the as-prepared NSCZIS photocatalyst. As shown in Supplementary Fig. 12, the results of yield and selectivity remain satisfied in broad range of BA content from 100 μL to 1 mL in 10 mL reaction solution, which indicates that the NSCZIS could maintain satisfactory performance in the extensive range of substrate concentration. Furthermore, we evaluate the conversion reaction under Xe lamp illumination (λ > 420 nm) and the NSCZIS maintains a consistent performance for BA conversion to NBI, indicating that the experimental scheme can be available under general irradiation (Supplementary Fig. 26). For benzyl alcohols with different substituted groups (−Cl, −Br, and −CH₃) and furfuryl alcohol, the conversion selectivity could be retained at high levels for the corresponding imines (Fig. 2e). It is worth noting that the electronegativity of the substituent notably affects the yield of the corresponding imine. For the case of substituent BA with an electron-donating group (−CH₃), a considerable amount of the corresponding BAD is detected while the selectivity of target imine is decreased for the increased electrophilicity with substituent. In contrast, the yields of corresponding imines with an electron-withdrawing group (−Cl, −Br) are evidently reduced with decreased electrophilicity but the satisfactory selectivity is maintained.

Structural characterization of photocatalysts

Consequently, the introduction of multiple elements (Ni, Sn, Ca) into the ZIS ensures the formation of the uniform high-entropy sulfide (Fig. 3a). To disclose the chemical state and local atomic environment of the sulfide catalysts, X-ray absorption near-edge structure (XANES) and extended X-ray absorption fine-structure (EXAFS) experiments were performed. The Zn absorption edges in XANES spectra for sulfide samples are positioned between those of the Zn foil and ZnO due to the decreased electronegativity of sulfur. Impressively, Zn K edge XANES spectra show that the addition of multiple elements (Ni, Sn, Ca) can evoke a downshift of the absorption edge compared with the corresponding ZIS sample (Fig. 3b). The difference indicates an increase of the electron density around Zn atoms owing to the high-entropy effect. Comparable observations are also evident in Zn 2p XPS spectra. The binding energy of Zn 2p shows slight downshift with the incorporation of additional three elements (Supplementary Figs. 27–30). Notably, the major peak of EXAFS spectrum at ~1.8 Å of the identical samples (NSCZIS, ZIS) is attributed to the Zn−S scattering path at first coordination shell, which is obviously distinct from Zn−Zn in Zn foil at 2.31 Å and Zn−O in ZnO at 1.58 Å (Fig. 3c). The coordination number (CN) and the bond length of Zn−S bond in ZIS are 4.10 and 1.83 Å, respectively, which are similar to NSCZIS (3.94 and 1.76 Å), demonstrating the uniform distribution of multiple elements and the absence of oxides/metals in NSCZIS (Fig. 3d, Supplementary Table 5). Interestingly, the Zn K-edge EXAFS oscillation curve of NSCZIS shows a slight increase in amplitude compared with ZIS, clearly implying structural changes in the Zn atomic coordination environment. These XAFS results collectively validate the abounding of particular coordination environment via the high-entropy engineering, which is responsible for the photocatalytic performance towards BA oxidizing coupling to NBI.

Fig. 3. Characterizations of high-entropy photocatalyst.

a Schematic illustration of the high-entropy photocatalyst. b Normalized XANES spectra of ZIS and NSCZIS at the Zn K-edge. c Fourier transform (FT) k²-weighted χ(k) function of the EXAFS spectra of ZIS and NSCZIS at the Zn K-edge. d Zn K-edge EXAFS curves fitting for NSCZIS in R space (inset: Zn K-edge EXAFS curves fitting for ZIS in R space). e HAADF-STEM image by aberration-corrected TEM. The lattice angles indicated by the white lines preserve the hexagonal characteristics. f Corresponding geometric phase analysis GPA, of NSCZIS. g Corresponding EDS elemental mapping images of Ni, Sn, Ca, Zn, In and S.

XAFS spectra of Ni K-edge, In K-edge, and Sn K-edge were also collected to investigate the structure of NSCZIS. As illustrated in Supplementary Fig. 31, the XANES spectrum of Ni shows an evidently positive shift as comparing to that of Ni foil, indicating that the valence state of Ni atoms in NSCZIS is close to NiO. The XANES spectrum of In is close to that of In₂O₃, indicating that the valence state of In atoms in NSCZIS is slightly smaller than +3 (Supplementary Fig. 32). A similar Sn K-edge XANES spectrum of NSCZIS with SnO₂ is also observed (Supplementary Fig. 33). These results confirm the electronic regulation caused by high entropy effect, thereby influencing the XANES spectral features at the In and Sn K-edge. The corresponding EXAFS spectra provide further insights into the local coordination environment of Ni, In, and Sn, the prominent peaks at 2.21 Å, 2.51 Å, and 2.67 Å are ascribed to the single Ni−S, In−S, and Sn−S bonds within the first coordination shell, respectively. The coordination numbers of Ni, In, and Sn with S atoms are calculated to be 2.98, 3.98, and 1.99, respectively (Supplementary Table 5).

The structure of the as-prepared photocatalysts with multiple components was further examined using X-ray diffraction (XRD), scanning electron microscopy (SEM), high-resolution transmission electron microscopy (HRTEM) and energy-dispersive spectroscopy (EDS). The detailed structure analysis is presented in Supplementary Note 3. Compared with the ZIS, the sulfides exhibit similar feature in the X-ray diffraction peaks with the introduction of hetero atoms (Supplementary Fig. 34), which is also confirmed by HRTEM (Supplementary Figs. 35–37). The XRD patterns of NSCZIS remain highly similar to that of pristine ZIS, with only subtle peak shifts/broadening that indicate substitution of heteroatoms into the cation sites of ZIS framework. Therefore, the overall crystal structure of ZIS remains well preserved due to the substituted incorporation does not induce a phase transformation or noticeably alter the long-range lattice structure. EDS analysis of four sulfides demonstrates the even distribution of each element, confirming the uniform surface among every sulfide sample (Supplementary Figs. 38–41). The atomic-resolution high-angle annular dark-field scanning TEM (HAADF-STEM) was employed to study the atomic configuration of the NSCZIS (Fig. 3e). The atom arrangement reveals a hexagonal structure. Specifically, entropy-addition leads to random fluctuations in crystal lattice. The clear lattice distortion is probably due to the incorporation of multiple transition elements in the lattice. Furthermore, the atomic strain distribution pattern clearly indicates obvious lattice distortion across whole region due to the variation in ionic radius (Fig. 3f). The corresponding elemental mapping indicates clear arrangement of every element on atomic scale (Ni, Sn, Ca, Zn, In and S) in the sample, substantiating the existence of these elements within metal atom columns and the formation of high entropy sulfide (Fig. 3g).

To evaluate the stability of NSCZIS during photochemical testing, we conducted ICP-OES analysis to detect the metal content in the solution after photocatalytic test. The elemental analysis of the post-reaction solution demonstrates that the leaching of metals (Ni, Sn, Ca, Zn, In) is negligible after photocatalytic reaction (Supplementary Table 6). The SEM images show the same morphology for NSCZIS after the photocatalytic reaction (Supplementary Fig. 42). The results of XRD and STEM images show negligible variance for NSCZIS before and after the photocatalytic reaction (Supplementary Figs. 43, 44). Moreover, Ca 2p, Zn 2p, In 3d, and S 2p in post-reaction XPS spectra are nearly consistent with the pristine NSCZIS (Supplementary Fig. 45), implying the main framework remain stable for post-reaction NSCZIS. Specifically, the peaks of Ni 2p, and Sn 3d binding energy of NSCZIS demonstrate both negative shift after photocatalytic reaction as compared with the pristine sample, suggesting that the sites of Ni and Sn are reduced by photogenerated electron.

Exploration of the reaction mechanism

The surface properties were further evaluated to understand the differences of BA ammoxidation among various sulfide photocatalysts. As CO₂-TPD is primarily used to probe adsorption sites on catalyst surface, it can provide indirect insights into the surface properties of the catalyst that can significantly influence the adsorption of reactant in dehydrogenation reaction. The adsorption of CO₂-TPD can reflect the capability of BA dehydrogenation on catalysts to some extent due to the fact BA and CO₂ both possess intrinsic acidity. CO₂-TPD experiments show a dominant peak centered at 318 °C for the NSCZIS, while the main peak shift to a low temperature centered at 305 °C with an evident decrease for NSZIS (Fig. 4a). Moreover, CZIS exhibits a weak CO₂ desorption peak at 294 °C, and ZIS shows a smaller desorption peak located at 283 °C. The gradual increase of desorption temperature suggests that the introduction of Ca element could enhance the adsorption compared with the ZIS, which is conducive to the dehydrogenation of BA. The significantly increased adsorption intensity of NSCZIS indicates its more abundant active sites (Supplementary Table 7). Meanwhile, the adsorption experiment also indicates the promoted capability of NSCZIS for BA adsorption (Supplementary Table 8). These results suggest a strong synergistic effect between the NiSn and Ca in NSCZIS for CO₂ adsorption. For the ZnInS-based photocatalysts, the Zn atoms on the surface are usually considered as active sites for the adsorption and activation of reactant⁴⁶. The more intensive adsorption peak of NSCZIS implies that the introduction of multiple elements (Ni, Sn, Ca) facilitates the adsorption of alcohols⁴². This is further supported experimentally by the fact that the NSCZIS exhibits much higher photocatalytic activity than other three samples.

Fig. 4. Mechanism analysis of photocatalytic reaction.

a CO₂-TPD (temperature programmed desorption) profiles over NSCZIS, NSZIS, CZIS and ZIS. b In-situ DRIFTS (diffuse reflectance infrared fourier transform spectroscopy) on the optimal NSCZIS under irradiation condition using BA and NH₃ as the reactants. c Free-energy diagrams for BA dehydrogenation to BAD on ZIS and NSCZIS. d Free-energy diagrams for benzylimine hydrogenation to BN on ZIS and NSCZIS. e In-situ Raman spectra of NSCZIS under photocatalytic BA conversion.

Meanwhile, in-situ infrared (IR) spectroscopy indicates the adsorption behavior of BA species on various sulfides in dark condition. The obvious signal located at 3600 cm⁻¹ is assigned to the characteristic signal of free BA molecule. The centered position of peak gradually shifts to low wavenumber due to the adsorption on sulfides (Supplementary Fig. 46). Owing to the better adsorption capability of the NSCZIS for BA, a maximum shift of adsorbed BA signal can be observed for the multiple components derived high-entropy effect. The in-situ IR adsorption result aligns well with the CO₂-TPD experiment, validating that the high-entropy strategy is available to promote the activation of BA. Furthermore, in-situ attenuated total reflection surface enhanced infrared absorption spectroscopy (ATR−SEIRAS) was performed to thoroughly identify the reaction species in photocatalytic reactions (Fig. 4b). Under light irradiation, typical IR absorption signals appear immediately at 1700 and 2892 cm^–1, which are attributed to the stretching vibrations of C=O and C–H in BAD, respectively¹⁹. Notably, the characteristic peak at 1640 cm^–1 is observed that is attributed to the C=N stretching vibration of the intermediate *Ph–CH = NH (benzylimine)⁴⁷, signifying that the benzylimine forms through dehydrated condensation of BAD and NH₃, and the gradually increased signal indicates its accumulation before hydrogenation conversion. Specifically, the shoulders located at 1598 and 3305 cm⁻¹ gradually evolve from 6 min under illumination, which can be assigned to the bending vibration and stretching vibrations of N–H bond of BN⁴⁸. As the reaction progressed, the evolved peak at 1665 cm^–1 corresponds to the C=N stretching vibration of the NBI⁴⁷, confirming that the conversion of BA to NBI is accompanied by the generation of BAD, benzylimine and BN as intermediates.

To gain insights into the reaction process of BA to NBI, we systematically calculated the reaction pathways over the ZIS and NSCZIS model surfaces (Supplementary Fig. 47). The detailed description of the configuration within the DFT model is provided in Supplementary Note 4 and Data 1. The reaction generally involves four steps: (i) BA oxidation dehydrogenation to BAD, (ii) the dehydration condensation of BAD and NH₃ to benzylimine, (iii) hydrogenation of benzylimine to BN, and (iv) subsequent formation of NBI from dehydration condensation of BAD and BN. As the dehydration condensation is favorable in thermodynamics and kinetics for (i) and (iv), the reaction could proceed spontaneously and rapidly. Thus, we considered (i) and (iii) to be crucial and performed theoretical evaluations for the two processes.

We firstly conducted Gibbs free energy calculations to gain insights into the reaction process of BA oxidation dehydrogenation over ZIS and NSCZIS (Supplementary Figs. 48, 49). As shown in Fig. 4c, the BA adsorption step is more thermodynamically favorable on NSCZIS as evidenced by the more negative Gibbs free energy (−1.81 eV vs. −1.45 eV), which indicates the advantage of high-entropy configuration in facilitating BA adsorption. The conversion of *PhCH₂OH to *PhCH₂O is thermodynamically unfavorable due to the positive energy barrier, and the less positive value indicates that the NSCZIS is more capable of dehydrogenation than ZIS (4.15 eV vs. 4.73 eV), which is consistent with the higher performance of BA oxidation on NSCZIS. Meanwhile, the *PhCHO formed from *PhCH₂O intermediate on NSCZIS has a smaller energy barrier of 2.8 eV than that of 3.0 eV on ZIS, suggesting the preferable ability in BA oxidation.

For the process of benzylimine hydrogenation to BN, the calculated results show that the adsorption of benzylimine on NSCZIS is much stronger than that on ZIS (−0.50 eV vs. 0.37 eV), which suggests that the activation of benzylimine is more available by NSCZIS (Fig. 4d, Supplementary Figs. 50, 51). The calculations also showed that the conversion of *PhCHNH to *PhCHNH₂ is both thermodynamically and kinetically unfavorable, while the process for the initial hydrogenation needs to overcome large barriers of 0.82/0.88 eV. Meanwhile, the *PhCH₂NH₂ formed from *PhCHNH₂ intermediate on NSCZIS has a smaller energy barrier of 0.76 eV than that of 0.90 eV on ZIS. The reaction barriers for the intermediate hydrogenation process on NSCZIS are always lower than those of ZIS, implying the remarkable hydrogenation capability of NSCZIS, which plays a competitive role in the generation of BN. While the computational approach provides consistent conclusion with experimental results, it is limited by the simplification in the modeling and structural interpretation for high-entropy system.

To estimate the redox driving force, the energy band information of these photocatalysts was also determined. The driving force of band edge is adequate to the hydrogenation by means of H₂O as hydrogen source (Supplementary Figs. 52−54). The hydrogenation reaction (HGR) performance for benzylimine intermediate using H₂O as hydrogen source was assessed by electrochemical experiments. The linear sweep voltammetry (LSV) results demonstrate that a high overpotential is required to realize the HGR for ZIS due to the sluggish catalytic ability for H₂O activation and following hydrogenation of imine to BN (Supplementary Fig. 55). However, an evident increase in the current density is observed with the introduction of NiSn for NSCZIS or NSZIS, indicating that the NiSn is much more efficient for hydrogenation reaction. Thus, the hydrogenation of benzylimine intermediate to produce benzylamine is favorable in the presence of NiSn for NSCZIS or NSZIS. Moreover, in-situ Raman spectroscopy shows that the character peaks of NSCZIS located at 220, and 318 cm⁻¹ can be assigned to the characteristics of the Ni(Ⅱ)−S bonds (Fig. 4e)⁴⁹. The spectrum fades slightly when the visible light is irradiated on the NSCZIS, suggesting the partial reduction of Ni(Ⅱ) in the initial process, further confirming the role of reduction reaction site. Meanwhile, the partial reduction of Ni species by photogenerated electrons can be further confirmed by in-situ EPR experimental results (Supplementary Fig. 56), in which the reduction of Ni²⁺ into Ni⁺ during photocatalytic process can be verified by the clear signal peak (g value = 2.005) attributed to paramagnetic Ni (I) species⁵⁰. The post-reaction XPS (Supplementary Fig. 45) and in-situ EPR experiments both indicate that Ni-Sn likely serves as a reduction site as undergoing partial reduction activation at the early stage of the reaction. Consequently, the generation rate of BN is relatively slow compared to that of BAD during the initial stage, resulting in the accumulation of BAD. After activation on NSCZIS, the accelerated production of BN and further condensation of BAD and BN to produce the final NBI promote the rapid consumption of BAD in the later stages of the reaction.

To verify the reaction mechanism for photocatalytic BA to NBI, isotope experiments were further performed to verify the conversion mechanism, in which H₂O or D₂O acted as solvent and hydrogen source simultaneously. The obtained products were analyzed by GC–MS. The strong D-substituted NBI (m/z) peak (196) and benzyl (m/z) peak (92) were observed in the case of D₂O, while only normal NBI (m/z) peak (195) and benzyl (m/z) peak (91) exist in H₂O case (Fig. 5a, Supplementary Fig. 57). These results conclusively indicate that the generation of BN is accomplished by the reductive amination of BAD and H₂O acts as hydrogen source to participate in the hydrogenation reaction. The generated BN would couple with BAD by the dehydration condensation to produce imine. The isotope experiment adding additional ¹³C-substituted BAD was also conducted to confirm the condensation mechanism for imine generation. The emergence of the NBI (m/z) peak (197) and benzyl (m/z) peak (92) aligns well with the ¹³C-labeled imine, indicating that BN is derived from BAD and acts as a key intermediate in the formation of NBI (Fig. 5b, Supplementary Fig. 58). In addition, ¹H NMR result of D₂O experiment shows significantly decreased signal compared with normal counterpart for the D substitution in C–N (Fig. 5c, d). Furthermore, ¹³C NMR spectra of the above products show a significantly enhanced signal (69 ppm) that is assigned to the ¹³C-labeled α-C of benzyl in ¹³C-labeled NBI in the case of BA + α−¹³C BAD, while the signal is very weak in the normal experiment using bare BA as the feed source (Fig. 5e, f). This demonstrates that the ¹³C atoms in C–N of imine come entirely from BAD, implying the BN-involved condensation mechanism for the origin of C–N in NBI.

Fig. 5. Isotope-labeled measurements.

a Mass spectra of NBI using D₂O as the solvent on the optimal NSCZIS. b Mass spectra of NBI using equivalent BA and α−¹³C BAD as reactants on the optimal NSCZIS. c, d ¹H NMR spectra of the products using H₂O or D₂O as the solvent on the optimal NSCZIS. BA, BAD and NBI represent benzyl alcohol, benzaldehyde and N-benzylidenebenzylamine, respectively. e, f ¹³C NMR spectra of the products using only BA or equivalent BA+α−¹³C BAD as reactants on the optimal NSCZIS. The peaks marked by dashed box are assigned to the α-C of benzyl in NBI. g NBI yield using H₂O or D₂O as solvent on the optimal NSCZIS. h Schematic illustration of the mechanism for the BA conversion to NBI by NSCZIS photocatalyst.

To further verify the significance of BN formation, a mixture of H₂O and D₂O at equivalent volume was used in photocatalytic experiment to identify whether the generation of BN was the rate-limiting step. The abundance ratio of normal NBI yield to that of deuterated NBI on NSCZIS was 2.5 (Fig. 5g). Furthermore, we performed photocatalytic hydrogenation experiment to determine the rate of BN formation by adding additional BAD to ensure the adequate content of benzylimine intermediate. The hydrogenation reactions were conducted by utilizing D₂O and H₂O as hydrogen source, respectively. As revealed by the evolution results of BN (Supplementary Fig. 59), the hydrogenation rate is significantly slowed by replacing H₂O with D₂O. The value of kinetic isotope effect calculated by kinetic constants (k_H/k_D) is 2.43. Thus, hydrogenation of benzylimine intermediate to BN is inferred to be the rate limiting step^51,52. The above results conclusively indicated that H₂O participated in the reaction process of BN generation followed by the condensation with BAD, thus the BN-involved mechanism for imine synthesis could be illustrated in Fig. 5h.

In summary, we developed a high-entropy sulfide photocatalyst (NSCZIS) for the highly efficient photocatalytic conversion of BA to NBI, which exhibits an exceptional NBI yield of 1240 $\mathrm{\mu }\mathrm{mol}{\mathrm{g}}_{\mathrm{cat}}^{-1}$ for 5 h, as well as a selectivity of 91.7%. This represents the first instance of photocatalytic oxidation coupling of benzyl alcohol and NH₃ to produce NBI. The photocatalytic mechanism on NSCZIS involves the oxidation of BA to BAD; then, benzylamine is generated by the hydrogenation of the benzaldehyde-NH₃ coupling intermediate. Ultimately, the second coupling, involving the condensation of benzylamine and benzaldehyde, results in NBI production. Extensive in-situ characterizations demonstrated that introducing Ca modulates the oxidation of BA, while NiSn species promote the hydrogenation of the benzylimine intermediate. This work provides a strategy for the photocatalytic conversion of benzyl alcohol into valuable imines using a high-entropy photocatalyst to achieve exceptional activity and selectivity.

Methods

Chemicals and materials

Zinc chloride (ZnCl₂, 99%) and indium chloride tetrahydrate (InCl₃·4H₂O, 99%) were purchased from Shanghai Macklin Biochemical Co., Ltd. Thioacetamide (TAA, AR) was purchased from Sinopharm Chemical Reagent Co., Ltd. Calcium nitrate tetrahydrate (Ca(NO₃)₂·4H₂O, 99%) was purchased from Sinopharm Chemical Reagent Co., Tin(IV) chloride pentahydrate (SnCl₄·5H₂O, 99%) and nickel chloride hexahydrate (NiCl₂·6H₂O, 99%) were purchased from Shanghai Macklin Biochemical Co., Ltd. Benzyl alcohol (C₇H₈O, 99%) was purchased from Shanghai Macklin Biochemical Co., Ltd. Ammonium hydroxide (NH₃·H₂O, 25–28 wt%, AR) was purchased from Shanghai Titan Scientific Co., Ltd. All of chemical reagents were used as received without further purification.

Synthesis of NSCZIS

0.0475 g NiCl₂·6H₂O, 0.3506 g SnCl₄·5H₂O, 0.2361 g Ca (NO₃)₂·4H₂O, 0.1363 g ZnCl₂, 0.2932 g InCl₃·4H₂O and 0.601 g TAA were dissolved into 30 mL water. After vigorous stirring for 30 min, the precursor solution was sealed in a 100 mL Teflon autoclave and heated at 180 °C for 12 h, after which the resulting precipitates were repeatedly washed with water and absolute ethanol and subsequently dried under vacuum at 60 °C overnight.

Synthesis of NSZIS

0.0475 g NiCl₂·6H₂O, 0.3506 g SnCl₄·5H₂O, 0.1363 g ZnCl₂, 0.2932 g InCl₃·4H₂O and 0.601 g TAA were dissolved into 30 mL water. After vigorous stirring for 30 min, the precursor solution was sealed in a 100 mL Teflon autoclave and heated at 180 °C for 12 h, after which the resulting precipitates were repeatedly washed with water and absolute ethanol and subsequently dried under vacuum at 60 °C overnight.

Synthesis of CZIS

0.2361 g Ca (NO₃)₂·4H₂O, 0.1363 g ZnCl₂, 0.2932 g InCl₃·4H₂O and 0.601 g TAA were dissolved into 30 mL water. After vigorous stirring for 30 min, the precursor solution was sealed in a 100 mL Teflon autoclave and heated at 180 °C for 12 h, after which the resulting precipitates were repeatedly washed with water and absolute ethanol and subsequently dried under vacuum at 60 °C overnight.

Synthesis of ZIS

0.1363 g ZnCl₂, 0.2932 g InCl₃·4H₂O and 0.601 g TAA were dissolved into 30 mL water. After vigorous stirring for 30 min, the precursor solution was sealed in a 100 mL Teflon autoclave and heated at 180 °C for 12 h (Anhui Kemi Instrument Co., Ltd. BFC-1200-1L), after which the resulting precipitates were repeatedly washed with water and absolute ethanol and subsequently dried under vacuum at 60 °C overnight.

Catalysts characterization

Scanning electron microscopy (SEM, Hitachi S4800) and transmission electron microscopy (TEM, Tecnai G2 F20) were employed to characterize the morphology and structural feature for the synthesized samples. Powder X-ray diffraction (XRD) analysis was performed on the composite material using the Rigaku D/Max2500PC diffractometer. X-ray photoelectron spectroscopy (XPS) measurement was conducted using Thermo Science 250 Xi system. Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC-HAADF-STEM) was conducted on the Thermo Fisher Titan Themis G2 60-300 with an acceleration voltage of 200 kV. Electrochemical measurements were performed on an electrochemical workstation (CS310M). ¹H NMR and ¹³C NMR spectra were recorded on an Agilent 600 NMR spectrometer using CDCl₃ as solvent. Ultraviolet–visible absorption spectra were examined using a UV-vis spectrometer (SHIMADZU, UV-2600i), with barium sulfate serving as the standard reference. Electron paramagnetic resonance (EPR) spectra were collected on a Bruker A200 spectrometer with a center field at 3362 G and a sweep width of 120 G at room temperature.

In-situ Raman spectra were obtained by using the WITec alpha 300 R confocal Raman microscope equipped with a 600 grooves/mm diffraction grating⁵³. The excitation source is a 633 nm solid-state laser operated at 20 mW with an attenuation filter. The calibration was performed by a standard silicon wafer at 520 cm⁻¹. The quartz cuvette (3 × 1 × 1 cm) sealed reaction solution and NSCZIS sample coated indium tin oxide (ITO) plate (2 × 10 mm, Nippon Sheet Glass, 14 Ω·sq⁻¹) was utilized as reaction apparatus. The spectra were collected continuously under UV irradiation of 5 W UV LED during in-situ measurements.

In-situ DRIFTS studies were conducted using a Nicolet iS50 FTIR spectrometer (Thermo Scientific). First, a certain amount of sample was placed on Si crystal semi-cylinder with chemically deposited Au film in a DRIFTS cell. Subsequently, reaction solution was introduced into the cell and the background spectrum was collected as a reference. The spectra were collected every 3 min with a resolution of 8 cm⁻¹ under dark condition or after a certain irradiation time using a 300 W Xe lamp equipped with a cut-off filter of 420 nm.

The X-ray absorption fine structure (XAFS) spectra of Zn K-edge were obtained by a commercial laboratory XAFS setup (Table XAFS-500A, Specreation Instruments Co., Ltd.)⁵⁴. The experiment was performed in transmission mode, X-rays were generated from an X-ray tube at 25 kV and 20 mA. The monochromated beam was produced using Si (733) spherically bent crystal analyzers (500 mm radius) mounted on a Rowland circle with a 250 mm radius.

The In and Sn K-edge X-ray absorption spectra were collected at the ID26 beamline of the European Synchrotron Radiation Facility (ESRF, Grenoble, France). The measurements were performed in transmission mode with a monochromatic X-ray beam of ~1 mm × 1 mm in size. Ionization chambers were used to record both the incident and transmitted intensities. The energy calibration was performed using In and Sn metal foil reference simultaneously measured in transmission mode. X-ray absorption fine structure (XAFS) measurement of Ni K-edge was performed at beamline 46U1A of the High Energy Photon Source (HEPS). The Ni K-edge spectra were collected in fluorescence mode, using Si (311) double-crystal monochromator. All spectra were collected in ambient conditions using ionization chamber.

Product identification

The organic products were qualitatively identified by Agilent 7890B GC/7000C MS and quantitatively analyzed by Agilent 7890A gas chromatography. The BA conversion and NBI selectivity were calculated according to Eqs. (1) and (2).

$$\mathrm{Conversion}\left(\%\right)=\left({\mathrm{C}}_0-{\mathrm{C}}_{\mathrm{BA}}\right)/{\mathrm{C}}_0\times 100\%$$ 1

$$\mathrm{Selectivity}\left(\%\right)=2{\mathrm{C}}_{\mathrm{NBI}}/\left({\mathrm{C}}_0-{\mathrm{C}}_{\mathrm{BA}}\right)\times 100\%$$ 2

Here C₀ is the initial number of moles of BA; C_BA and C_NBI are the number of mole concentrations of residual BA and corresponding NBI, respectively.

Photochemical measurements

Generally, 10 mg catalyst, 7 mL H₂O, 3 mL NH₃·H₂O and 100 μl BA (0.97 mmol) were added in a 50 mL quartz reactor, the NH₃ concentration is ca. 4.5 mol L⁻¹. Before the reaction, the reactor was sealed without removing air or replacing air by inert gas. Subsequently, the above mixture was irradiated by a CEL-LAB200E7 Parallel Light Chemical Reaction Instrument (Beijing China Education Au-light Co., Ltd.) for 5 h, the light intensity is 13.8 mW·cm^–2 measured by the irradiation meter (CEL-NP2000-2A). The reaction system was energetically stirred at 500 rpm and kept at 20 °C. After the reaction, the mixed liquid products were extracted by ethyl acetate and filtrated for the quantitative analysis using GC (Agilent 7890 A) equipped with an FID detector and an HP-5 capillary column. Statistical analyses of three repeated experiments were conducted using Origin 2021 and the calculated results of mean value and ±standard deviation (SD) are plotted in the figure as mean value and error bar.

For the blank experiments, the procedure is the same as above experiment except for the removal of photocatalyst or irradiation. The quenching experiments of active species were conducted to investigate the mechanism during photocatalytic BA dehydrogenation to BAD by adding extra 0.1 mmol of different scavengers and other parameters keep consistent with general experiment.

As the extraction by ethyl acetate is necessary for the detection of products, long-term stability testing is conducted through five parallel reactions by using samples from the same batch. The reaction tube was removed from the multichannel reactor every hour, followed by the extraction and quantitative analysis. The procedure for the cyclic experiments is identical to the normal tests except the recycle procedure. After the reaction, the photocatalyst was washed sequentially with deionized water and ethanol and collected by centrifugation for two times. The final sample was dried in a vacuum oven for further recycling experiments, maintaining a cycle time of 5 h.

Test of apparent quantum yield (AQY)

To determine the wavelength dependence of the apparent quantum yield (AQY), samples were irradiated by a commercial 300 W Xenon lamp (PLS-SXE300+/UV, Beijing Perfectlight Technology Co., Ltd.) equipped with an appropriate band pass filter. AQY was calculated according to Eq. (3):

$$\mathrm{AQY}=\frac{4\times \mathrm{NBI}\mathrm{Number}\mathrm{of}\mathrm{molecules}}{\mathrm{Number}\mathrm{of}\mathrm{incident}\mathrm{photons}}\times 100\%$$ 3

$$=\frac{4\times \mathrm{M}\times N_A\times \mathrm{c}\times h}{\mathrm{P}\times \mathrm{t}\times \lambda }\times 100\%$$

Where M is the molar amount of NBI molecules, N_A is Avogadro constant, c is the speed of light, h is the Planck constant, P is the intensity of the irradiation and was measured by CEL-NP2000-2A, t is the time of illumination, and λ is the wavelength of the band pass filter. The irradiation area is a circular shape with a diameter of 2 cm, and geometry area is 3.14 cm².

DFT calculations

Density functional theory (DFT) calculations were conducted using the generalized gradient approximation (GGA) and Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional, implemented in the Vienna Ab Initio Simulation Package (VASP). VASP utilizes plane wave basis sets to span reciprocal space and employs the projector augmented wave (PAW) method. The calculation parameters were set as follows: energy cutoff of 400 eV, energy convergence criterion of 1 × 10⁻⁴ eV, force convergence criterion of 0.03 eV/Å, and a Monkhorst–Pack k-point grid of 1 × 1 × 1. The calculations employed a flat plate model, with a vacuum of approximately 20 Å placed above the plate to minimize interactions between periodic images. The slab model of ZIS was constructed by a 3 × 3 × 1 supercell of ZnIn₂S₄ (001) with a single stoichiometric layer. The slab model of NSCZIS was generated by introducing a Ni, Ca atom to the Zn vacancies and Sn atom to the In vacancy in a supercell of ZnIn₂S₄ (001). In both models, the thickness along the c-axis was set to 35 Å to prevent interactions between periodic images.

Supplementary information

Supplementary Information ^{(4MB, pdf)}

41467_2026_75878_MOESM2_ESM.pdf ^{(61.5KB, pdf)}

Description of Additional Supplementary File

Supplementary Data 1 ^{(235.9KB, txt)}

Transparent Peer Review file ^{(5.1MB, pdf)}

Source data

Source Data ^{(16.7MB, xlsx)}

Author contributions

N.C.Y. designed the experiments and performed the synthesis. Y.S. performed the photocatalysis and photo-electrochemical experiments. Z.A.C. performed the in-situ DRIFTS, Raman and XAFS tests. Y.S. and Z.A.C. jointly completed the related graphic design and drawing. Z.W.C. conducted the DFT calculations. N.C.Y. and Y.S. co-wrote the original draft. S.Y., H.Q.C. and Y.J.X. reviewed the manuscript. S.Y. and Y.J.X. discussed the data and commented on the manuscript. All the authors discussed the results and contributed to the writing of the paper.

Peer review

Peer review information

Nature Communications thanks the anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.”

Funding

N.C.Y., S.Y. and Y.J.X. disclose support for the research of this work from the National Natural Science Foundation of China (22372001, 22502002, 52261135635), the National Science Fund for Outstanding Young Scholars of Anhui Province (2408085Y008), and Starting Fund for Scientific Research of High-Level Talents of Anhui Agricultural University (rc382108). Y.S., Z.A.C., Z.W.C., and H.Q.C. declare no relevant funding.

Data availability

The processed data generated in this study are provided in the main text, supplementary information, and Source Data files. 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 https://doi.org/10.1038/s41467-026-75878-w.