Controlled transfer-reduction of functional groups with ethanol by their interaction strength with the oxygen vacancies

Abstract

Controlled reduction of functional groups remains a significant challenge in catalytic chemistry. In this study, we report a novel method for selectively reducing various functional groups using biomass-derived ethanol, mediated by oxygen vacancies in highly dispersed molybdenum oxide on nitrogen-doped carbon. Selectivity for the reduction of four representative substrates, sulfoxides, pyridine N-oxides, nitro compounds, and alkynes is successfully controlled via temperature, yielding the corresponding sulfides, pyridines, amines, and alkenes with high yields. Mechanistic studies indicate that the reactivity of these substrates is directly correlated with their interaction strength with O_v in molybdenum oxides, which effectively adsorb and activate both ethanol and the substrates via Columb interactions and facilitates hydrogen-atom transfer from ethanol to these functional groups. DFT calculations further confirm that O_v in Mo/CN-600 significantly lowers the energy barriers for the H transfer from α-C_sp3-H in ethanol to substrates (0.76 eV vs. 2.04 eV), thereby enhancing reactivity. This strategy provides a green and sustainable approach to the controlled reduction of functional groups with strong practical potential.

Selectively reducing functional groups without affecting others remains difficult in catalysis. Here, reduction of several groups using ethanol over oxygen-vacancy in molybdenum oxide on nitrogen-doped carbon, with temperature tuning selectivity via hydrogen transfer.

Introduction

Catalytic hydrogenation is a cornerstone of the chemical industry, driving essential transformations in bulk and fine chemical production^1–3. Molecular hydrogen (H₂) is a clean and sustainable reductant that can be produced by water splitting, with only water as a byproduct. Generally, supported metal nanoparticles (e.g., Pt/C, Pd/C) are employed to facilely activate H₂ molecules for reduction, while achieving high chemoselectivity remains challenging, especially for substrates containing multiple reducible functionalities or requiring partial hydrogenation^1,2. Typical examples include the selective reduction of nitro groups in 4-nitrostyrene without affecting C = C bonds, or the semihydrogenation of alkynes to alkenes. Therefore, great effort has been devoted to careful modification of supported metallic nanoparticles with the aim to get the high selectivity¹. For example, forming bimetallic nanoparticles (Ir₁Ni SAA, RuNi) is reported to be an effective way to modulate noble metal catalysts for the selective reduction of the -NO₂ group in 4-nitrostyrene to produce 4-vinylaniline^4,5, in which the non-noble metals selective adsorped -NO₂ group in a vertical mode, avoiding the hydrogenation of the C = C bond, and the noble metals functionalized in activating H₂ molecules.

Except for H₂, alcohols are among the most attractive environmentally benign hydrogen donors due to their renewable origin. Among them, ethanol stands out for its economic viability and carbon-neutral scalability from biomass³. Unlike secondary alcohols like iso-propanol, ethanol has been rarely explored in transfer hydrogenation reactions, because of its inherent inertness and susceptibility to catalyst poisoning (Fig. 1a). On the one hand, ethanol as a primary alcohol is much more inert than iso-propanol in transfer hydrogenation⁶. On the other hand, acetaldehyde, generated from ethanol dehydrogenation, can cause catalyst deactivation via metal-catalyzed decarbonylation, forming catalytically inactive metal-carbonyl species (Fig. 1b)^7,8. Existing methodologies for activating ethanol often rely on air- or moisture-sensitive noble metal complexes, encountering issues including dedicated preparation process, rigorous anaerobic/anhydrous conditions, and non-selective hydrogenation^8–10. For example, deep hydrogenation of C ≡ C was observed with Ir catalysts, failing to obtain C = C bonds with high selectivity (Fig. 1b)^9,10. These challenges highlight an urgent need for innovative catalytic systems that efficiently and selectively utilize ethanol as a sustainable hydrogen donor.

Fig. 1. Catalytic transfer hydrogenation reactions by ethanol.

a The challenge in ethanol-mediated transfer hydrogenation; b Previous works (Ir-catalysts for ethanol-mediated transfer hydrogenation); c This works (Oxygen vacancies-controlled transfer hydrogenation reaction).

Recently, reducible transition metal oxides such as CeO₂, ZrO₂, and MoO₃ have emerged as promising alternatives to supported metal nanoparticle catalysts for transfer hydrogenation^11–13. Among them, MoO₃ has been extensively investigated in hydrodeoxygenation reactions. A reverse Mars-van Krevelen mechanism was proposed involving the consumption of oxygen vacancies (O_v) by abstracting oxygen atoms from oxygenate substrates via strong interaction and their regeneration by thermal treatment in the presence of H₂ (typically above 400 °C)¹². This strong affinity of O_v sites for oxygen atoms arises from the coulombic interaction between electron-deficient O_v sites and electron-rich oxygen atoms bearing lone-pair electrons. Therefore, it was reasonable to imply that defective MoO₃ can efficiently adsorb and activate other electron-rich substrates or intermediates, including pyridine N-oxides, sulfoxides, nitro compounds, acetylide (from deprotonation of alkynes), and alkoxide (from deprotonation of alcohol). In this context, defective is promising for ethanol-mediated transfer hydrogenation of these functional groups, and the electron density on the substrate or intermediate may affect reaction activity, thereby offering a new avenue for modulating selectivity.

Our recent work showed that nitrogen-doped carbon support was promising for stabilizing transition metal sites, including single atoms, metallic nanoparticles, and metal oxides, exhibiting excellent catalytic performance in diverse reactions^13–15. Herein, we employed nitrogen-doped carbon to disperse oxygen-deficient molybdenum oxides by a simple pyrolysis method. In ethanol-mediated transfer hydrogenation, these defective molybdenum oxides demonstrated tunable catalytic performance, allowing for temperature-dependent switching of activity across a broad substrate scope (Fig. 1c), including various N-oxide, sulfur, nitro, and alkyne moieties with yields ranging from 91% to 99%. Experimental evidence indicated that O_v sites were responsible for the adsorption of these compounds, and the apparent activation energy was positively correlated with their adsorption strength at the O_v sites. DFT calculations further demonstrate that O_v in Mo/CN-600 plays a decisive role in reducing the energy barriers for ethanol-driven reduction of various functional groups. This work provides a novel insight into manipulating the hydrogenation of multifunctional group compounds and can inspire broad interest in designing defective metal oxides for transfer hydrogenation.

Results

Catalyst preparation and characterization

The nitrogen-doped carbon-supported molybdenum oxides were prepared by the pyrolysis of the dicyanamide and ammonium molybdate composite at four representative temperatures (500, 600, 700, and 800 °C) under inert atmosphere (Fig. 2a), and the as-prepared catalysts were labeled as Mo/CN-T, where T represents the pyrolysis temperatures. The color of the composite changed from white before the pyrolysis to brown after pyrolysis at 500 °C (Fig. S1). As discussed below, the brown color of Mo/CN-500 was due to the presence of the yellow g-C₃N₄. In contrast, other Mo/CN-T samples are black due to the formation of black amorphous nitrogen-doped carbon at or beyond 600 °C. The Mo content in Mo/CN-500 was determined to be 8.3 wt.% by inductively coupled plasma-atomic emission spectrometry (ICP-AES), and it significantly increased to 16.2, 18.4, and 19.2 wt.% for Mo/CN-600, Mo/CN-700, and Mo/CN-800, respectively (Table S1). The increase of Mo content in the catalysts at higher pyrolysis temperatures should be caused by the release of more versatile gases and the phase transition of molybdenum species (β-MoO₃ → MoO₂ → MoC) as evidenced by the following characterizations.

Fig. 2. Catalyst preparation and characterization.

a Procedure of the preparation of catalyst preparation [figure legends: black ball, grinding balls; white ball, dicyanamide; pink ball, (NH₄)₆Mo₇O₂₄·4H₂O.]; b XRD patterns of Mo/CN-T [Figure legends: club, g-C₃N₄ PDF#87-1526; heart, MoO₂ PDF#32-0671; diamond, MoO₃ PDF#47-1081; spade, MoC PDF#06-0546.]; c SEM image of Mo/CN-600, d TEM image of Mo/CN-600, e HR-TEM image of Mo/CN-600, f HAADF-STEM image of Mo/CN-600; g, h overlayed HAADF-STEM image and corresponding EDS mapping image of Mo/CN-600.

Powder X-ray diffraction (XRD) and Fourier transform infrared (FT-IR) spectroscopy analyses reveal that both the carbon nitride structure and Mo species strongly depend on the pyrolysis temperature (Figs. 1b, S2, and S3). In Mo/CN-500, the characteristic diffraction peaks of g-C₃N₄ at 2θ = 20.6° and 27.1°, corresponding to the (100) and (002) planes, are clearly observed (Fig. S2), which is consistent with the FT-IR signals (Fig. S3) of aromatic C = N/C-N (1200–1600 cm⁻¹) stretching and C-N (809 cm⁻¹) heterocycle vibrations¹⁶. However, these features weakened in Mo/CN-600 and disappeared in Mo/CN-700 and Mo/CN-800, indicating that the g-C₃N₄ framework was destroyed at temperatures above 600 °C. The pyrolysis temperature also governed the evolution of Mo species. β-MoO₃ is the dominant phase in Mo/CN-500, while partial transformation to MoO₂ occurred at 600 °C¹⁷. With further temperature increase, MoO₂ becomes the main species in Mo/CN-700, accompanied by the formation of MoC in Mo/CN-800 (Figs. 1b and S2, see Note 1 in the Supporting Information for further discussion). Raman spectra were also used to identify the Mo species in the Mo/CN-T catalysts. Three weak peaks with the Raman shifts around 656, 820, and 950 cm⁻¹ were assigned to the tri-coordinated oxygen (Mo₃-O), bi-coordinated oxygen (Mo₂-O, Mo-O-Mo), and MoO_x vibrations in Mo/CN-600 and Mo/CN-700, respectively (Fig. S4)¹⁸. While Mo/CN-800 only showed an additional weak peak at 992 cm⁻¹, attributed to the Mo-C vibration from MoC¹⁹, consistent with the XRD results. In conclusion, the Mo species changed from β-MoO₃ to MoO₂ as the pyrolysis temperature increased from 500 to 700 °C, and MoC was the dominant Mo species for Mo/CN-800. In addition, two prominent Raman bands at 1323 ~ 1341 cm⁻¹ and 1582 ~ 1595 cm⁻¹ were observed for the Mo/CN-600, Mo/CN-700, and Mo/CN-800 catalysts, corresponding to the D and G bands, which are associated with defective carbon and graphitic sp² carbon, respectively¹⁴. The calculated I_D/I_G ratios were 1.47, 1.36, and 1.28 for Mo/CN-600, Mo/CN-700, and Mo/CN-800 (Fig. S4). These results indicate that increasing the pyrolysis temperature promotes graphitization of the carbon framework in the Mo/CN-T catalysts. This trend is mainly attributed to the gradual decrease in nitrogen content at higher pyrolysis temperatures, which facilitates the formation of more ordered graphitic structures^14,15. N₂ adsorption-desorption analysis showed that the Mo/CN-T samples had interconnected micro/mesopores (Fig. S5). The smallest total surface area of 262.1 m²/g was observed for Mo/CN-500 and increased significantly to 838.2 m²/g when pyrolysis was performed at 600 °C (Mo/CN-600, Table S1). Further increasing the pyrolysis temperature resulted in a slight decrease in the total surface area (Table S1).

Transmission electron microscopy (TEM) image of Mo/CN-500 shows a thick-layered structure (Fig. S6a), while the other three Mo/CN-T (T = 600, 700, and 800) catalysts demonstrate a thin-layered structure (Figs. 2 and S6b, c). The representative scanning electron microscopy (SEM) image of Mo/CN-600 showed a loose structure with abundant pores (Figs. 1c and S7). Results from TEM and SEM images were consistent with the N₂ adsorption-desorption analysis. Mo/CN-500 with thick-layered structure had much smaller surface area than other three Mo/CN-T samples (Fig. S5 and Table S1). TEM image of Mo/CN-500 did not show visible molybdenum oxide nanoparticles (Fig. S6), while the small-sized β-MoO₃ nanoparticles were clearly observed in the TEM image of Mo/CN-600 with the average size in 2.7 ± 0.5 (Fig. 2d). The lattice fringe of 0.384 nm was measured in high-resolution TEM (HR-TEM) image of Mo/CN-600 (Fig. 2e), corresponding to the interplanar distance of (110) plane of β-MoO₃¹⁷, which was consistent with the XRD results (Fig. 2b). MoC nanoparticles with an average size of 2.0 ± 0.4 nm were observed in the TEM image of Mo/CN-800. In addition, high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images with Z contrast clearly showed the Mo species in bright color (Fig. 2f, g). Energy-dispersive spectroscopy (EDS) elemental mapping also revealed that the Mo, N, C, and O were uniformly distributed on the carbon support of the Mo/CN-600 catalyst (Fig. 2h).

The valence states of the surface elements in the catalysts were detected by X-ray photoelectron spectroscopy (XPS, see Note 2 in the Supporting Information for further discussion). Mo⁶⁺, Mo⁵⁺, and Mo⁴⁺ were identified in the Mo 3d XPS spectra for Mo/CN-500 and Mo/CN-600, whereas Mo²⁺ was the dominant species in Mo/CN-700 and Mo/CN-800 (Figs. 2a and S8 and Tables S2 and S3)¹⁴. The average Mo oxidation states were calculated from the XPS results and decreased from 5.1 and 5.2 to 3.6 and 3.1 with increasing pyrolysis temperature, indicating a gradual reduction in molybdenum oxides. The remaining high-valence species in Mo/CN-800 likely arise from surface oxidation of MoC during air storage²⁰. N 1 s XPS spectra of Mo/CN-500, Mo/CN-600, and CN-600 showed sp² aromatic N (C-N = C), tertiary N (-NC₃), amino groups (-C-NH_x), and g-C₃N₄ π excitation at 398.8, 399.6, 400.9, and 405.1 eV, respectively (Figs. S9 and S10 and Table S4)²¹. Nitrogen content decreased with higher pyrolysis temperatures due to N release, and g-C₃N₄-related features vanished above 600 °C, forming pyridinic and graphitic N in Mo/CN-700 and Mo/CN-800²². C 1 s spectra indicated g-C₃N₄ characteristic peak (288.4 eV, N-C = N) in Mo/CN-500 and CN-600, which diminished at higher temperatures, with C-C/C = C (284.7 eV) dominating in Mo/CN-600, Mo/CN-700, and Mo/CN-800 (Fig. S11 and Table S5)^20,23,24. A new peak at 283.4 eV could be fitted in Mo/CN-700 and Mo/CN-800, corresponding to Mo-C bonds, consistent with XRD phase transitions (Fig. 2b)^14,24. O 1 s spectra were deconvoluted into Mo-O (530.6 ~ 530.7 eV), oxygen vacancies (O_v, 531.4 eV), C = O/C-OH (532.2 eV), and C-O (533.3 ~ 533.4 eV) (Figs. 3b and S12 and Table S6)²⁵. O_v content decreased from ~20 at.% in Mo/CN-500 and Mo/CN-600 to 3.6 at.% and 1.5 at.% in Mo/CN-700, and Mo/CN-800, reflecting the transition from β-MoO₃ to MoO₂ and then MoC, leading to the gradual loss of oxygen vacancy sites at higher pyrolysis temperatures.

Fig. 3. Characterization of the as-prepared Mo samples and structural model of β-MoO₃.

a, b XPS spectra of the Mo 3d and O 1s in the Mo/CN-T; c The XANES curves of the Mo foil, MoO₃, Mo/CN-600 at Mo K-edge [figure legends: red, Mo/CN-600; blue, Mo foil; black, MoO₃.]; d The FT-EXAFS curves of the Mo foil, MoO₃, Mo/CN-600 at Mo K-edge; e The structural model of β-MoO₃.

The extended X-ray absorption near-edge structure (XANES) and X-ray absorption fine structure (EXAFS) were further conducted at Mo K-edge to probe the local environment of Mo in representative Mo/CN-600 catalyst. A well-defined pre-edge peak at 20006.9 eV (peak A) was found in the Mo K edge XANES spectra of Mo/CN-600 and MoO₃ reference (Fig. 3c), assigning to the direct 1s to 4d quadrupole-allowed transitions and the dipole-allowed 1s hybridized (5p, 4d) states²⁶. The absorption edge energy of Mo/CN-600 is close to the reference MoO₃. The two peaks at 20027.3 and 20039.1 eV (B and C) were discriminated in the absorption edge energy of the reference MoO₃²⁶, whereas the two peaks were not well resolved for Mo/CN-600. The broadening and overlap of peaks B and C in Mo/CN-600 indicated the presence of the disordered MoO₃ structure, which was caused by O_v²⁷.

Pseudo-radial distribution of the distances in Mo/CN-600 was further investigated with Fourier-transformed (FT) k³-weighted EXAFS. k³-weighted FT-EXAFS for MoO₃ reference shows two characteristic peaks at 1.2 and 1.5 Å (Fig. 3d), both of which are lower than the Mo-Mo bond at 2.4 Å in Mo foil. The peaks at 1.2 and 1.5 Å are assigned to Mo-O_I band (Fig. 3e, terminal oxygen atoms, -Mo=O) and Mo-O_II band (Fig. 3e, bridge oxygen atoms, Mo-O-Mo), respectively²⁷. A second major component was associated with the nearest Mo neighbors (Mo-Mo second shell distance). k³-weighted FT-EXAFS for Mo/CN-600 exhibited only one main peak at around 1.3 Å, which was close to the Mo-O_I band found in MoO₃ reference. While the peak for Mo-O_II is missing in k³-weighted FT-EXAFS for Mo/CN-600, suggesting that the bridge oxygen atoms in the Mo-O-Mo bond were lost. The absence of a peak associated with the second shell of Mo also indicated that the O in Mo-O-Mo had disappeared to form the Mo-O_v-Mo site. As well known to us, the Mo=O bond should have higher binding energy than the Mo-O-Mo bond, thus the bridge oxygen atom in Mo-O-Mo should be removed in priority by the in-situ formed reductive gases such as CO, NH_3, and H₂ from the decomposition of the precursors at high temperature²⁸. The EXAFS wavelet transform (WT) analysis (Fig. S13) further revealed that the O_v in Mo/CN-600 should be generated from bridging oxygen sites (Mo-O_II band). It was noted that the peak intensity at 656 and 820 cm⁻¹ in the Raman shift of the treated Mo/CN-600 catalyst by diphenyl sulfoxide became stronger (Fig. S4), while the peaks for the terminal oxygen coordinated Mo (-Mo=O) did not significantly changed. The increase of the peak intensity at 656 and 820 cm⁻¹ was due to the fill of the negative oxygen atom in diphenyl sulfoxide into the O_v sites, which suggested that the oxygen vacancies sites in the Mo/CN-600 catalyst were derived from the removal of the bridging oxygen atoms in Mo-O-Mo bonds, in agreement with the XAFS results.

Catalytic activity test

The catalytic performance of the Mo/CN-T catalysts was initially evaluated in the reduction of diphenyl sulfoxide by ethanol (Table 1). The catalytic activity significantly increased from Mo/CN-500 to Mo/CN-600, and then greatly decreased over Mo/CN-700 and Mo/CN-800 (Table 1, entries 1 ~ 4). Quantitative conversion of diphenyl sulfoxide into diphenyl sulfide was obtained over Mo/CN-600 after 4 h at 150 °C (Table 1, entry 2). H₂ was not detected in the gas phase after reaction, indicating that the transformation was a hydrogen atom transfer process. No reaction was observed over the Mo-free nitrogen-doped carbon support (Table 1, entry 5; abbreviated as CN-600, and the XRD, XPS, and ESR test shown in Figs. S2, S10, and S14), suggesting that Mo species were the active sites, whilst nitrogen-doped carbon materials or defective carbon/ nitrogen exhibit no catalytic activity in this system. Commercial MoO₃ and MoO₂ showed no activity in the reduction of diphenyl sulfoxide by ethanol (Table 1, entries 6 & 7), and the absence of 1′ was observed without the addition of diphenyl sulfoxide (Table 1, entry 8). Therefore, it can be concluded that the O_v sites in the defective MoO_x species are of great importance in the hydrogen atom transfer reaction.

Table 1.

Transfer hydrodeoxygenation of diphenyl sulfoxide over different catalysts^a

Entry	Catalyst	Conversion (%)	Selectivity (%)
			1	1′
1	Mo/CN-500	20.1	>99	>99
2	Mo/CN-600	100	>99	>99
3	Mo/CN-700	14.1	>99	>99
4	Mo/CN-800	6.2	>99	>99
5b	CN-600	0	–	–
6b	MoO3	0	–	–
7b	MoO2	0	–	–
8c	Mo/CN-600	0	–	–
9b	Mo/C-600	0	–	–

^aReaction conditions: diphenyl sulfoxide (0.5 mmol), catalyst loading (molar ratio of Mo to substrate = 7.0 mol%), ethanol (10 mL), 4 h, 150 °C, and N₂ (10 bar).

^b 50 mg of catalyst was used.

^c same as “a”, but not used diphenyl sulfoxide.

Then, the relationship between the content of the O_v sites and catalytic activity in Mo/CN-T catalysts was investigated. Electron paramagnetic resonance (EPR) technology was used to detect the O_v content in the catalysts. Two weak peaks with g values at g_⊥ = 1.961 and g_∥ = 1.988 are observed in the EPR spectra of the Mo/CN-T catalysts (Fig. 4), in association with typical five-coordinated Mo⁵⁺ species²⁸, while they are not present in the EPR spectra of commercial MoO₃ with Mo⁶⁺ and MoO₂ with Mo⁴⁺. Interestingly, a peak with g = 2.002 was only observed in the EPR spectrum of Mo/CN-500, assigning to the defective sites of g-C₃N₄ derived partial disorder (Fig. S14)²⁹. The peak with g = 2.005 was attributed to the signal of the O_v from molybdenum oxides in the Mo/CN-T catalysts (T = 500, 600, and 700)³⁰. Mo/CN-600 showed the strongest intensity of O_v signal (Fig. 4a), indicating it has the most abundant O_v sites. The intensity of O_v signal significantly weakened as the pyrolysis temperature increased and completely diminished at 800 °C due to the phase transition from MoO₂ to MoC, in good agreement with the O 1s XPS fitting results. It should be emphasized that XPS primarily probes surface chemical states and relative surface defect densities, whereas EPR reflects the presence and evolution of paramagnetic defect centers in the bulk. The apparent discrepancies between the two techniques are therefore rationalized by their intrinsically different sampling depths, sensitivities, and detection mechanisms^14,31.

Fig. 4. Analysing the role of oxygen vacancies by EPR and XPS O1s.

a EPR spectra of different catalysts; b The correlation of the TOF and O_v site abundancy in the Mo/CN-T catalysts; c, d Comparison of EPR signal before and after treated with substrates at reaction conditions, the Mo/CN-600-Benzyl alcohol, Mo/CN-600-iso-Propanol, Mo/CN-600-Ethanol, Mo/CN-600-Diphenyl sulfoxide, Mo/CN-600-Pyridine N-oxide, Mo/CN-600-Nitrobenzene and Mo/CN-600-Phenylacetylene represent the treated Mo/CN-600 catalyst by benzyl alcohol, iso-propanol, ethanol, diphenyl sulfoxide, pyridine N-oxide, nitrobenzene and phenylacetylene, respectively.

Furthermore, the catalytic activities (turnover frequency, TOF) of the Mo/CN-T catalysts were correlated with the surface concentrations of Mo⁵⁺ and O_v, as quantified by Mo 3d and O 1s XPS (Figs. 4b and S15). A nearly linear relationship between TOF and both Mo₅₊ and O_v abundance was observed. The apparent deviation of Mo/CN-500 from this trend can be attributed to its thick-layered structure and much smaller surface area (262.4 m² g⁻¹) compared with the other catalysts (718.6–838.2 m² g⁻¹), which limits the accessibility of active sites. In contrast, Mo/CN-600 combines a high surface area with abundant porosity, facilitating efficient mass transport of reactants and products, and thereby enhancing catalytic performance. These results indicated that both O_v/Mo⁵⁺ density and surface area jointly govern the catalytic activity of the Mo/CN-T catalysts. For comparison, carbon-supported MoO₂ without O_v sites (Mo/C-600), prepared by incipient wetness impregnation and calcination at 600 °C under N₂, exhibited no catalytic activity (Table 1, entry 9), further confirming the essential role of oxygen vacancies in the hydrogen atom transfer process.

The interaction between O_v in Mo/CN-600 and the substrates was further studied. Mo/CN-600 was first treated with the substrates under reaction conditions and then subjected to EPR characterization. The peak intensity of O_v in the treated Mo/CN-600 catalysts, whether by ethanol or diphenyl sulfoxide, decreased significantly in comparison with the untreated Mo/CN-600 (Fig. 4c), which could be due to the filling of the negative oxygen atoms in ethanol or diphenyl sulfoxide into the O_v sites. This phenomenon was also confirmed by Raman results. The intensity of Raman peaks at 656 and 820 cm⁻¹, corresponding to the bridge Mo-O-Mo bond, was noted to increase after treating Mo/CN-600 with diphenyl sulfoxide (Fig. S4)²⁸. However, the peaks associated with terminal oxygen coordinated Mo (-Mo=O) did not change significantly. These results suggested that the oxygen atoms in substrates should fill the bridged O_v sites (Mo-O_v-Mo) in accordance with the characterization results. Therefore, we can conclude that O_v in Mo/CN-600 plays a great role in the adsorption and activation of these substrates²⁸.

Optimization of the reaction conditions

Then, the effect of the reaction temperature on the reduction of diphenyl sulfoxide by ethanol was studied over Mo/CN-600 (Table S7). The reaction proceeded sluggishly at 90 °C, affording a low conversion in 5.4% after 4 h. Then the conversion gradually increased with the increase of the reaction temperature, and the quantitative conversion was achieved within 4 h at 150 °C (Table S7, entry 6). The reduction of diphenyl sulfoxide into diphenyl sulfide proceeded smoothly at 120 °C with the full conversion after 10 h (Fig. S16). Besides high activity, Mo/CN-600 also demonstrated excellent stability during the recycling experiments (Fig. S17). The XRD, XPS, SEM, and TEM images of the recovered Mo/CN-600 catalyst (Figs. S18 and S19 and Tables S1–S6) after fiveruns showed no significant change as compared with those from fresh sample (Fig. 2c–e). In particular, XPS results show that both the Mo species and the oxygen vacancy concentration remain largely unchanged after the catalytic cycles, further confirming the structural robustness and stability of the catalyst under the reaction conditions (Tables S1–S6). No Mo leaching into the reaction solution was detected by ICP-AES analysis (Table S1). This high stability should be due to the strong interaction between the nitrogen-doped carbon support with the molybdenum oxides.

Besides ethanol, other aliphatic alcohols were also used as hydrogen donors for the reduction of diphenyl sulfoxide (Table S8). Methanol, the simplest alcohol with the highest dehydrogenation enthalpy, also gave similar diphenyl sulfoxide conversion as ethanol (Table S8, entries 1 & 2). The conversion decreased in the order of ethanol, n-propanol, n-butanol, and iso-propanol. Our results differed from the commonly reported results, in which iso-propanol as the secondary alcohol was more active hydrogen donor than primary alcohols^6,8. This phenomenon was due to the different reaction mechanisms. For the supported metal nanoparticle catalysts, the reaction mechanism followed the dehydrogenation-hydrogenation process, in which hydrogen species from alcohols were first captured on the surface of metal nanoparticles and then transferred to the unsaturated groups in the substrates⁶. Therefore, iso-propanol as the secondary alcohol, demonstrated higher reactivity than the primary alcohols for its easier dehydrogenation. However, that was not the case in our method. The hydrogen atom transfer over Mo/CN-600 was promoted by the interaction between the oxygen atoms in alcohols and O_v in Mo/CN-600, in which the steric effect from alcohols may influence their reactivities. As discussed later, the key step in the deoxygenation of diphenyl sulfoxide by alcohols was the direct transfer of the α-C_sp3-H to diphenyl sulfoxide. Compared with primary alcohols in Table S8, iso-propanol with the largest steric hindrance should have the weakest interaction with O_v in Mo/CN-600, thus iso-propanol demonstrated the lowest reactivity in our study. To verify it, EPR signal of O_v over the treated Mo/CN-600 catalysts with iso-propanol or ethanol was compared. As expected, the peak intensity of O_v in the treated catalyst by ethanol was much weaker than that over the treated catalyst by iso-propanol (Fig. 4c), confirming that ethanol had a stronger interaction with O_v in Mo/CN-600 than iso-propanol.

Kinetic studies

Furthermore, kinetic experiments were conducted. As ethanol was in excess, its concentration on the catalyst surface can be considered constant. Log (rate) was plotted versus log[diphenyl sulfoxide], giving rise to a straight line (R² = 0.997), which indicated that the reduction of diphenyl sulfoxide should follow a pseudo-first-order reaction kinetics (Fig. S20a). According to the Arrhenius plot, the apparent activation energy of the reduction of diphenyl sulfoxide was calculated to be 88.8 kJ/mol over Mo/CN-600 (Fig. S20b). The dependence of alcohol concentration on the reduction of diphenyl sulfoxide was also studied. Because of the low boiling point of ethanol and high reaction temperature, its concentration in the reaction system could not be accurately controlled. Therefore, benzyl alcohol was used instead of ethanol to investigate the relationship between the reaction rate and the concentration of alcohol. The reactions were performed in toluene with an excessive amount of diphenyl sulfoxide. As shown in Fig. S20c, a straight line could also be fitted after plotting log(rate) versus log[benzyl alcohol], indicating that consumption of benzyl alcohol followed a pseudo-first-order reaction kinetics. This pseudo-first-order dependence found for diphenyl sulfoxide and benzyl alcohol was in accordance with the fact that the reaction was started by the direct interaction of the oxygen atoms in diphenyl sulfoxide and alcohols with the O_v in Mo/CN-600 as discussed above (Fig. 4)²⁸. Kinetic isotopic experiments of the reduction of diphenyl sulfoxide were performed to study the difference in the abstraction of the two H atoms (α-C_sp3-H and O-H) from ethanol. A low kinetic isotopic effect (KIE) value of 1.45 was obtained for CH₃CH₂OD, while that was 3.19 for CH₃CD₂OD (Fig. S20d). By deducting the contribution of -OD in CH₃CD₂OD, the approximate KIE value of CH₃CD₂OH to CH₃CH₂OH was calculated to be 2.74. These results indicated that the transfer of the α-C_sp3-H was significantly more difficult than the transfer of the active hydrogen (H⁺) in the hydroxyl group, suggesting that the former should be the rate-determining step.

Substrate scope

The excellent catalytic performance of Mo/CN-600 in the reduction of diphenyl sulfoxide motivated us to explore its applicability toward other sulfoxides and pyridine N-oxides. As summarized in Table 2a, a wide range of aromatic and aliphatic sulfoxides were efficiently converted into the corresponding sulfides with excellent yields (Pro. 1 ~ 15). Electronic effects from substrates were apparent: 4, 4′-dimethyldiphenyl sulfoxide bearing an electron-donating group showed lower reactivity than 4, 4′-dichlorodiphenyl sulfoxide containing an electron-withdrawing group (Pro. 2 vs. 3). Phenyl alkyl sulfoxides exhibited comparable activity to diphenyl sulfoxides (Pro. 1 vs. 6), likely due to reduced steric hindrance which compensated for their lower electron density at the sulfoxide moiety. Cyclic and dialkyl sulfoxides were also quantitatively deoxygenated to the corresponding sulfides (Pro. 13 ~ 15), demonstrating the broad substrate scope. The transfer reduction of pyridine N-oxide and its derivatives with ethanol was then investigated. Pyridine N-oxide could be reduced at a significantly lower temperature (60 °C), affording a conversion of 5.3% after 4 h and a relatively low apparent activation energy of 74.2 kJ/mol (Table S9 and Fig. S21a). These results indicate that pyridine N-oxide was more easily reduced than diphenyl sulfoxide, which can be attributed to the stronger interaction between oxygen anions in pyridine N-oxide and O_v sites in Mo/CN-600. This conclusion is supported by the pronounced decrease in O_v EPR signal intensity after treatment with pyridine N-oxide (Fig. 4d). Various substituted pyridine N-oxides were smoothly converted into the corresponding pyridines at 100 °C with high to quantitative yields (Pro. 16 ~ 27).

Table 2.

Substrate scope for ethanol-mediated transfer reduction reactions

(a) Scope of substrates in the transfer deoxygenation of sulfoxides and pyridine N-oxides; Reaction conditions: substrates (0.5 mmol), Mo/CN-600 (20 mg), ethanol (10 mL), 120 °C (sulphoxides) or 100 °C (pyridine N-oxides) and N₂ (10 bar). (b) Scope of substrates in the transfer reduction of nitro compounds; Reaction conditions: nitro compounds (0.5 mmol), Mo/CN-600 (20 mg), iso-propanol (10 mL), 150 °C, and N₂ (1 MPa). (c) Scope of substrates in the transfer semi-hydrogenation of alkynes into alkenes; Reaction conditions: alkynes (0.5 mmol), the Mo/CN-600 catalyst (20 mg), ethanol (10 mL), 210 °C, and N₂ (1 MPa). Note: the yields in the bracket were the isolated yields.

In contrast, the transfer reduction of nitrobenzene proved to be considerably more challenging. The reaction initiated at 100 °C, giving only 4.3% conversion after 4 h (Table S9), and exhibited a much higher apparent activation energy of 100.4 kJ/mol compared with diphenyl sulfoxide (88.8 kJ/mol) and pyridine N-oxide (74.2 kJ/mol) (Fig. S21b). EPR analysis showed that O_v signal intensity remained relatively strong after treatment with nitrobenzene, indicating a much weaker interaction between O_v sites and nitrobenzene (Fig. 4d). Consequently, the reduction of nitrobenzene was significantly less favorable.

Moreover, nitrobenzene reduction with ethanol afforded aniline with low selectivity, accompanied by byproducts such as N-phenylethanimine and 2-methylquinoline (Table S10). These byproducts arise from condensation between in-situ-generated aniline and acetaldehyde, as well as subsequent cyclization reactions³². To avoid these side reactions, iso-propanol was employed as an alternative hydrogen donor. Owing to the steric hindrance imposed by the acetone formed during dehydrogenation, side reactions were effectively inhibited. As a result, nitrobenzene was converted to aniline in 95.4% yield at 170 °C after 11 h (Pro. 28). Under these optimized conditions, a broad range of nitro compounds were successfully reduced to the corresponding primary amines with good to excellent yields (Table 2b). Nitroarenes bearing electron-donating substituents were less reactive than those containing electron-withdrawing groups (Pro. 29 ~ 34 vs. Pro. 35 ~ 46), and steric effects were observed for o-, m-, and p-substituted substrates. The protocol was also effective for highly substituted nitroarenes, fused-ring systems, heteroatom-containing nitro compounds, and cyclic and aliphatic nitro substrates (Pro. 47 ~ 53). Notably, Mo/CN-600 exhibited excellent tolerance toward vulnerable functional groups such as vinyl, alkyne, halogen, nitrile, and especially iodine substituents. This chemoselectivity is attributed to the preferential adsorption of nitro groups and alcohols on O_v sites (Fig. 4d), where nitro groups can interact strongly through resonance structures containing negatively charged oxygen atoms, whereas nonpolar or weakly polar groups interact minimally^33,34.

Encouraged by the high tolerance toward vinyl groups, Mo/CN-600 was further evaluated for the semi-hydrogenation of alkynes to alkenes using ethanol as the hydrogen donor. Based on Lewis acid-base interactions between Mo⁵⁺ centers adjacent to O_v sites and anionic species, the deprotonated RC ≡ C⁻ intermediates were proposed to interact with O_v sites (Mo⁵⁺-O_v-Mo⁵⁺). Consistent with this hypothesis, treatment with phenylacetylene led to a detectable decrease in O_v signal intensity, albeit weaker than for oxygen-containing substrates, necessitating higher reaction temperature (160 °C, Table S9). Semi-hydrogenation of phenylacetylene proceeded at 210 °C to afford styrene with 96% selectivity at full conversion after 13 h (Pro. 54). Kinetic analysis revealed the highest activation energy (111.6 kJ/mol, Fig. S21c) among the examined transformations. Nevertheless, the method displayed broad applicability, enabling selective semi-hydrogenation of diverse alkynes to alkenes without over-reduction or interference from other functional groups (Table 2c). In short, the catalytic performance achieved with Mo/CN-600 is superior to that of other O_v-rich metal oxide catalysts (Table S11, entries 1 ~ 4 vs. 5 ~ 9) as well as homogeneous catalysts. Moreover, compared with a series of heterogeneous Mo-based catalysts, Mo/CN-600 in this work employs ethanol as the hydrogen donor (Table S11, entries 1 ~ 4 vs. 10 ~ 12), which is potentially a more environmentally benign alternative to hydrazine hydrate (N₂H₄·H₂O). In addition, defect sites and/or coordinatively unsaturated sites are likely the active centers responsible for this series of transfer hydrogenation reactions^13,35–38. This study systematically emphasizes the unique role of oxygen vacancies in promoting hydrogen atom transfer from alcohols, which differentiates our system from previously reported MoO₃- or Mo₄O₄N₃- catalysts^39–41.

Mechanism study

The starting reaction temperatures and the activation energies for the reduction of the four representative substrates both increased in the following order: pyridine N-oxide, diphenyl sulfoxide, nitrobenzene, and phenylacetylene (Table S9). As discussed above, the interaction strength of the substrates with O_v also increases in the same order. These results indicated there was a direct correlation between the reactivities of substrates and their interaction strength with the O_v (Figs. 4 and S15). XRD, HR-TEM, and other studies indicate that the β-MoO₃ (110) crystal plane exhibits structural stability^42,43. Therefore, DFT calculations were performed to study these reduction reactions by ethanol on the (110) crystal face of a β-MoO₃ model with O_v [hereinafter referred to as β-MoO₃-O_v (110)]. The reduction of pyridine N-oxide, with the lowest energy barrier, and the reduction of phenylacetylene, with the highest energy barrier, were used as the two representative reactions for DFT calculations (Fig. S21). After removal of the bridge oxygen atom in Mo-O-Mo, the distance between the two Mo atoms of β-MoO₃ (110) increased from 3.934 to 5.866 Å (Fig. S22). Ab initio molecular dynamics simulation was further conducted to verify the proposed models, particularly under finite temperature conditions relevant to the reaction environment. As shown in Fig. S23 (see Note 3 in the Supporting Information for further discussion), both perfect and oxygen-deficient β-MoO₃ surfaces only indicated minor local distortions without bond breaking or lattice collapse, remaining structurally intact under thermal fluctuations, thereby confirming the reliability of the as-proposed models.

First, particular attention was paid to the significant role of O_v in these transformations over the perfect and defective β-MoO₃ (110) crystal faces. As shown in Figs. 5 and S24, employing temperature and saturated vapour pressure for free energy correction has a limited effect on DFT calculations. It was surprising to note the adsorption energies of ethanol and pyridine N-oxide on the perfect β-MoO₃ (110) surface were even higher than those on the β-MoO₃-O_v (110), due to the strong affinity of them to Mo atom (Fig. 5). In this case, the bridge oxygen atom was destroyed after the adsorption of pyridine N-oxide by the connection of the oxygen atom in pyridine N-oxide with the Mo atom (Fig. 5a, Structure I-2). In both cases, the dissociation of “H” from the hydroxyl group (-OH) in ethanol to the basic site (O atom) in β-MoO₃ was an easy step with low energy barriers (E_a = 0.01 eV for β-MoO₃ (110) and E_a = 0.13 eV for β-MoO₃-O_v (110), Fig. 5, TS1 vs. TS1*), in agreement with the isotopic kinetic experiments (Fig. S20d). In both cases, the N-O bond in pyridine N-oxide was cleaved, and the oxygen atom bonded to the Mo atom. The cleavage of the N-O bond in pyridine N-oxide was also easy on a perfect β-MoO₃ (110) surface with an E_a of 0.13 eV and on a β-MoO₃-O_v (110) surface with an E_a of 0.16 eV (Fig. 5, TS2 vs. TS2*). The released pyridine molecule directly desorbed from the perfect β-MoO₃ (110) surface, but it preferred to transfer to the neighboring adsorbed ethanol on the β-MoO₃-O_v (110) surface (Fig. 5, I-5 vs. I-5*). The H atom from α-C_sp3-H in ethanol then transferred to the “N” atom in the neighbor pyridine molecule on the β-MoO₃-O_v (110) surface to generate the pyridine-H species with the energy barrier of 0.76 eV (Fig. 5b, TS3*). In the meantime, acetaldehyde desorbed from the β-MoO₃-O_v (110) surface. After the release of acetaldehyde, the pyridine-H species bonded to the Mo atom on β-MoO₃-O_v (110) surface (Fig. 5b, I-8*).

Fig. 5. DFT calculation.

a, b Simulated pathways for the reaction between C₅H₅NO and CH₃CH₂OH on β-MoO₃ (110) and β-MoO₃-O_v (110); c The DFT calculated free energy diagram for the reaction between C₅H₅NO and CH₃CH₂OH on β-MoO₃ (110) and β-MoO₃-O_v (110) [figure legends: red, β-MoO₃-O_v (110); black, β-MoO₃ (110).]. Energy barriers (red values) and reaction energies (black values) are also given for the associated molecular transformations (Legend: red, O atoms; blue, Mo atoms; white, H atoms; and brown, C atoms, respectively).

After detaching from pyridine N-oxide to the Mo atom, the oxygen could abstract an H atom in pyridine-H species via TS4* (E_a = 0.38 eV), and the generated pyridine molecule would desorb from the β-MoO₃-O_v (110) surface. Water molecule finally formed via the transportation of the initial adsorbed H atom on the basic “O” site from -OH in ethanol to the surface -OH group on the Mo site (TS5*, E_a = 1.07 eV/103.3 kJ mol⁻¹), and desorbed from the β-MoO₃-O_v (110) surface with the desorption energy of 0.03 eV to finish the catalytic cycle. In contrast to the reaction on the β-MoO₃-O_v (110) surface, the H atom on the “O” atom in the Mo-O-Mo moiety first transferred to the desorbed oxygen atom on the Mo site (from pyridine N-oxide) (TS3, E_a = 1.36 eV; Fig. 5a, I-6 ~ I-7), following by the H atom from α-C_sp3-H in ethanol (TS4, 2.04 eV; Fig. 5a, I-7 ~ I-8). Worse still, the desorption energy of water on the perfect β-MoO₃ (110) surface was also very high, up to 3.22 eV. It is notable that for the α-C_sp3-H transfer process, the energy barrier on the β-MoO₃-O_v (110) surface is significantly lower (Fig. 5, TS3* vs. TS4; 0.76 eV vs. 2.04 eV), which can be attributed to enhanced charge redistribution, more delocalized electronic reorganization, and vacancy-induced electronic states near the Fermi level that effectively buffer electron transfer (Figs. S25–S27).

According to the DFT calculation, it can be concluded that the O_v greatly lowered the energy barriers of abstracting two H atoms in ethanol molecule, especially for the H atom from α-C_sp3-H. The presence of O_v between two Mo atoms provides the free room and electronic flexibility for the pyridine-mediated transfer of the H atom from α-C_sp3-H to the desorbed “O” atom on the Mo sites at the other side. On the meanwhile, the flexible geometry of the β-MoO₃-O_v also benefited the transfer of the H atom on the basic “O” atom site in the catalyst to the desorbed “O” atom on the Mo sites.

Furthermore, the semi-hydrogenation of phenylacetylene by ethanol on the β-MoO₃-O_v (110) surface was also studied by DFT calculation. As shown in Fig. S28, the adsorption energy of phenylacetylene on β-MoO₃-O_v (110) surface was lower than that of pyridine N-oxide (−0.77 vs. −1.61 eV; Fig. S28, Structure I-2** vs. Fig. 5b, Structure I-2*), in consistence with the EPR results (Fig. 4d). On the meanwhile, phenylacetylene and ethanol adsorbed on the two Mo atom sites with O_v between them. DFT calculation revealed that the highest energy barrier of 1.44 eV (140.0 kJ mol⁻¹) was the transfer of the H atom from α-C_sp3-H in ethanol (Fig. S28, TS2**), which was much higher than that of the process in pyridine N-oxide (Fig. 5b, TS5*). Therefore, DFT calculation also confirmed the transfer hydrogenation of phenylacetylene was much more difficult than the transfer hydrodeoxygenation of pyridine N-oxide, which is in good accordance with our experimental data (Table S12).

Based on the experimental results, DFT results, and previous reports^13,35, the plausible mechanism was proposed for the reduction of sulfides, pyridine N-oxides, nitro compounds as well as the semi-hydrogenation of alkynes by ethanol over Mo/CN-600. Taking the reduction of pyridine N-oxide as an example, the possible mechanism is illustrated in Fig. 6. Initially, pyridine N-oxide and ethanol adsorb on the O_v site between two Mo atoms in Mo-O_v-Mo by filling the negative oxygen atoms in the substrates into the O_v sites. Meanwhile, the H atom from the -OH group in ethanol transfers to the basic “O” atom sites in β-MoO₃-O_v, such as Mo-O_II, to generate the adsorbed alkoxide ions (CH₃CH₂O⁻). Then, the oxygen atom in pyridine N-oxide is released and bonded to the Mo atom according to the above DFT calculation. Then two H atoms from α-C_sp3-H in CH₃CH₂O⁻ and the bonded H atom in the basic “O” sites transfer to the desorbed “O” atom from pyridine N-oxide on the Mo site to generate one water molecule. Finally, the desorption of the formed water molecule from the catalyst surface finishes the catalytic cycle.

Fig. 6.

A possible mechanism for the transfer reductions over the Mo/CN-600 catalyst with ethanol (Taking the transfer reduction of pyridine N-oxide as an example).

The application of our developed method

Then, the excellent catalytic performance of Mo/CN-600 was compared with commercial noble Pd/C and Pt/C catalysts. It was noted that commercial Pd/C and Pt/C catalysts delivered no reaction or negligible conversion over the four representative reactions with ethanol as the hydrogen donor at the same reaction temperatures (Table S13), suggesting O_v in Mo/CN-600 played a key role in the activation of alcohols as well as these functional groups in these substrates to promote these transfer reductions. However, that was not the case for the Pd/C and Pt/C catalysts, which requires to overcome high energy barrier to cut the hydrogen atoms directly from alcohols to the surface of Pd and Pt nanoparticles via the traditional dehydrogenation-hydrogenation process. As well known to us, Pt/C is robust in the activation of H₂ molecules at room temperature⁴⁴. The reduction of the four representative substrates by H₂ was also performed over Pt/C. Except for diphenyl sulfoxide, quantitative conversions of other three substrates were achieved over Pt/C catalyst (Table S14). The failure in the hydrogenation of diphenyl sulfoxide over Pt/C was most likely due to the poison of Pt/C, because of the strong affinity of surfur atom in diphenyl sulfoxide to the Pt sites⁴⁵. As shown in Fig. 7a, Pt/C catalyst showed no selectivity towards these transformations. The hydrogenation of pyridine N-oxides gave 59.6% yield of pyridine, 40.4% yield of tetrahydropyridine (Table S14). Both the vinyl group and the nitro group in 4-nitrostyrene were reduced over Pt/C with H₂ (Table S14). Semi-hydrogenation of phenylacetylene also failed over Pt/C, giving the quantitative yield of ethylbenzene with the full hydrogenation of the alkyne group (Table S14). The failure in controlled reduction of these substrates over Pt/C was due to the random adsorption of these substrates on the surface of Pt nanoparticles, and efforts to modulate the noble metal catalysts to selective adsorption and controlled reduction of the functional groups were devoted (Table S15)^4,5,46–51. Obviously, our developed method demonstrated distinct advantages, including the simple preparation of the catalyst with low cost, the use of biomass-derived ethanol as the hydrogen donor, and the controlled reduction of different functional groups at an appropriate temperature. For example, although alkynes and nitroarenes can both be reduced as disrcibed above, the chemo-selective reduction of nitro group rather than alkynyl group in 3-nitrophenylacetylene 3-ethynylaniline could be easily realized by choosing an appropriate temperature at 150 °C, only overcoming the energy barrier for the tranfer-reduction of -NO₂ group (Table 2, Pro. 36).

Fig. 7. Gram-scale synthesis of bioactive molecules.

a Compare with Pt/C catalyst and application; b Application of our work. ^a Nitro compounds (5.0 mmol), Mo/CN-600 (100 mg), iso-propanol (40 mL), and 150 °C; ^b Step 1: same as “a”, Step 2: 2-(4-methoxyphenyl) benzimidazole (5.0 mmol), pyridine (10 mL), and 100 °C.

Finally, the application of our method in the gram-scale synthesis of bio-active molecules was studied (Fig. 7b). 4′-Aminochalcone, an anti-cancer drug⁵², was successfully obtained by the chemo-selective hydrogenation of the nitro group in 4′-nitrochalcone without affecting the vinyl and carbonyl groups (Pro. 70). N-(2-methyl-5-nitrophenyl)-4-(3-pyridyl)-2-pyrimidine amine (Pro. 71) is a key intermediate for Imatinib with an annual consumption amount of $561 million in 2023⁵³, and it was obtained in a yield of 95% via reduction of the corresponding nitro compound. Entacapone with the application of Parkinson’s disease was successfully functionalized to the ammino product (Pro. 72; 1.2 × g) in a yield of 91% without the reduction of ketone, vinyl, and nitrile groups. Pomalidomide and Lenalidomide with the use in the treatment of Multiple Myeloma, were attained in yields of 94% from the reduction of corresponding nitro compounds (Pro. 73 & 74; 1.2 ~ 1.3 × g)⁵³. A two-step method was applied to synthesize Erlotinib in a yield of 91% (1.2 × g, Pro. 75)⁵⁴, the epidermal growth factor receptor-tyrosine kinase inhibitor. The first key step is the chemo-selective reduction of nitro group rather than alkynyl group in 3-nitrophenylacetylene into 3-ethynylaniline in a yield of 97% (Fig. S29), followed by the coupling reaction with 4-chloro-6,7-bis(2-methoxyethoxy) quinazoline via nucleophilic substitution to deliver Erlotinib⁵⁴.

Discussion

In summary, we have developed an effective and simple method for temperature-regulated reduction of functional groups using ethanol as the hydrogen donor over Mo/CN-600 with abundant O_v. Comprehensive analyses revealed that O_v in molybdenum oxide played a decisive role in the adsorption and activation of substrates, greatly promoting the hydrogen atom transfer from alcohols to the functional groups. Only some substrates, including sulfoxides, pyridine N-oxides, nitro compounds, and alkynes, that can generate the structures with anion atoms either by conjugated model or by the release of free H⁺, can be effectively reduced, because there exists an interaction between these substrates and the O_v. Moreover, those functional groups can be controlled in the reduction at different temperatures according to the strength of the interaction between the functional groups with O_v. DFT calculations further confirm that O_v in Mo/CN-600 significantly lowers the energy barriers to the reduction of functional groups by ethanol. The controlled reduction of certain functional groups makes Mo/CN-600 a promising platform for the synthesis of bioactive molecules. Our developed method showed some distinct advantages, including the use of heterogeneous non-noble metal oxide catalysts with facile preparation and low cost, the use of biomass-derived alcohols as the hydrogen donor, the controllable reduction of the functional groups, and the resistance to the heteroatoms.

Methods

Materials

Ammonium molybdate tetrahydrate [(NH₄)₆Mo₇O₂₄·4H₂O] and dicyanamide were purchased from Aladdin Chemicals Co., Ltd. (Shanghai, China). All sulfoxides, pyridine N-oxides, nitro compounds, and alkynes were commercially supplied and used directly without further purification. All reagents were used as received without further purification.

Catalyst preparation

Typically, dicyanamide (8.0 × g) and ammonium molybdate tetrahydrate (140.0 mg) was homogeneously mixed by ball milling method. Then, the obtained composite was placed on an alumina boat in a quartz tube furnace and heated from room temperature to the set temperature with a ramping rate of 2 °C·min⁻¹ under nitrogen atmosphere and then remained for 2 h. Finally, it was cooled down to 400 °C at 10 °C·min⁻¹, and then to room temperature naturally in the air. The as-prepared catalysts were denoted as Mo/CN-T, where T represents the pyrolysis temperature. For comparison, the CN-600 catalyst was also prepared by the same procedure as for the preparation of Mo/CN-600 without (NH₄)₆Mo₇O₂₄·4H₂O. The Mo/C-600 catalyst was synthesized via incipient wet impregnation method of ammonium molybdate tetrahydrate (110.0 mg) onto active carbon (0.3 × g), followed by calcination at 600 °C under N₂ atmosphere to achieve a comparable Mo content to that of the Mo/CN-600 catalyst.

Catalyst characterization

Transmission electron microscopy (TEM) was performed on a Talos F200X microscope operated at 200 kV, with samples dispersed in ethanol and deposited onto copper grids. Powder XRD patterns were recorded on a Bruker D8 Advance diffractometer using Cu Kα radiation over a 2θ range of 10° ~ 80° at a scanning rate of 0.016° s⁻¹. N₂ adsorption-desorption isotherms were measured at 77 K on a V-Sorb 2800 P analyzer after degassing the samples at 150 °C for 12 h. XPS was conducted on a Thermo VG ESCA MultiLab-2000 system with monochromatized Al Kα radiation (1486.6 eV), and binding energies were calibrated to the C 1s peak at 284.6 eV. Raman spectra were collected using a Thermo Fisher DXR spectrometer with 532 nm excitation. The Mo content was determined by ICP-AES (IRIS Intrepid II XSP). FT-IR spectra were obtained using the KBr pellet method, and EPR measurements were carried out on a Bruker EMXnano spectrometer.

The X-ray absorption spectra were collected on the beamline BL07A1 in NSRRC, and were provided technical support by Ceshigo Research Service “www.ceshigo.com”. The radiation was monochromatized by a Si (111) double-crystal monochromator. XANES and EXAFS data reduction and analysis were processed by Athena software. The obtained XAFS data was processed in Athena (version 0.9.25) for background, pre-edge line, and post-edge line calibrations. Then Fourier transformed fitting was carried out in Artemis (version 0.9.25)⁵⁵. The k³ weighting, k-range of 3 ~ 12 Å⁻¹ and R range of 1 ~ 3 Å were used for two shells fitting and 1 ~ 4 Å for 3 shells fitting of MoO₃.

General procedure of the transfer hydrogenation reactions

In a typical run, diphenyl sulfoxide (0.5 mmol), the Mo/CN-600 catalyst (20 mg), and ethanol (10 mL) were placed in a 40 mL stainless steel autoclave. The air in the autoclave was removed by flushing with N₂ for 10 min and finally pressured with 10 bar N₂. The system was heated from room temperature to the set reaction temperature within ca. 4 min. Then the reaction was started for a certain time with a magnetic stirring rate of 1000 RPM. After the reaction, the vessel was cooled in an ice-bath and the pressure was released slowly to atmospheric pressure. The products were diluted with acetonitrile and analyzed by GC. Furthermore, the residue was purified by column chromatography on silica gel (200–300 mesh), eluting with a mixed solvent of n-hexane and ethyl acetate to afford the desired products. The isolated products were further identified by NMR spectra and GC-MS. For other types of transfer hydrogenation reactions, the same procedure was applied with different kinds of the substrates.

Analytic methods

Product analysis was carried out using an Agilent 7890 A gas chromatograph equipped with an HP-5 cross-linked capillary column (30 m × 0.32 mm × 0.4 μm) and a flame ionization detector. N₂ was employed as the carrier gas at a flow rate of 40 mL·min⁻¹. The injector and detector temperatures were both set at 300 °C. The column temperature was programmed from 50 °C (held for 1.5 min) to 300 °C (held for 3 min) at a heating rate of 10 °C·min⁻¹. Product contents were quantified using an internal standard. Product molecular weights and structures were analyzed by GC-MS (Thermo Scientific ISQ 7000), and structural identification was further confirmed by ¹H and ¹³C NMR spectroscopy (Bruker TCI III 600 MHz).

Catalyst recycling experiments

After the reaction, the Mo/CN-600 catalyst was collected from the reaction mixture via centrifugation, and the spent Mo/CN-600 catalyst was washed with water and ethanol for three times, respectively. After this treatment, the recovered Mo/CN-600 catalyst was tested for the next run under identical reaction conditions.

Computational methods

All DFT calculations were carried out with generalized gradient approximation of the Perdew–Burke–Ernzerhof functional using the VASP (VASP 5.4.1) simulation package⁵⁶. The approach of project-augmented-wave⁵⁷ was exploited to describe the interaction between core-electron and valence electron. According to the XRD and HR-TEM results (Fig. 2), the supercell of (110) crystal face of β-MoO₃ with O_v was constructed as the surface model. The X-ray absorption fine structure spectroscopy (XAFS) and EPR technologies demonstrated that the substrates adsorbed on the O_v sites (Figs. 3 and 4). A plane-wave basis set was expanded with the kinetic cutoff energy of 400 eV on valence electronic states. K-point sampling parameters of two systems namely, β-MoO₃ and β-MoO₃-O_v were tested using pyridine N-oxide (C₅H₅NO) chemisorption energy with 1 × 1 × 1, 2 × 2 × 1, and 3 × 3 × 1, and a k-point sampling of 2 × 2 × 1 was adopted with the convergence value less than 0.1 eV. The Methfessel-Paxton-approach with a Fermi smearing width of 0.1 eV was used to determine partial occupancies on electronic states. The effect of van der Waals (vdW) interaction is significant for the reaction mechanism in our previous study^58,59. Therefore, the DFT-D3 method of Grimme et al.^60,61 was utilized to calculate all the energetics and structures of the intermediates and transition states. MoO₃ (110) surface was chosen as the DFT model base on our XRD analysis. All surfaces were built using periodic slabs of four layers. The bottom two layers of atoms were fixed, and the top two layers were relaxed with a 15 Å vacuum region. The thickness of the slab was verified to be sufficient. In order to be consistent with our experimental data, we constructed O_v through removing oxygen atoms on the perfect MoO₃(110) surface. The energetics and structures were calculated using the conjugate-gradient algorithm until the Hellman–Feynman forces on each atom less than 0.01 eV/Å. The transition states (TSs) were searched using the method called a constrained optimization scheme^62,63. The transition states were confirmed by two rules: (i) all forces on atoms have been optimized to be less than 0.01 eV/Å; (ii) the total energy is a maximum along the reaction coordinate but a minimum with respect to all other degrees of freedom. All the optimized initial states, transition states, and final states were verified by frequency analysis⁶⁴ using a finite displacement method to the mass-weighted Hessian matrix.

Author contributions

Z.L.Y. conducted the experiments and discussed the work. Q.J.T. and Y.R.R. performed some experiments; Z.L.Y. and Z.H.Z. conceived the work, designed the experiments, and analysed the data; Z.H.Y. performed the DFT calculations; J.G.W. supervised the DFT calculation. Z.L.Y., P.Z. and Z.H.Z. co-wrote and commented the manuscript. Z.H.Z. supervised the whole project.

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.

Data availability

The data supporting the findings of this study are available within the article and its Supplementary Information files. All other relevant raw data can be obtained from the corresponding author upon request. 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/s41467-026-71308-z.