Ni redispersion from SiO₂ to molybdenum carbide creates dual interfaces to boost tandem CO₂ hydrogenation

Abstract

Interface engineering is crucial in the design of supported metal catalysts, as it significantly influences the catalytic process, particularly in terms of selectivity. Herein, we discover the redispersion of Ni nanoparticles from SiO₂ to molybdenum carbide (Mo₂C) being induced by the strong interaction between metal and Mo₂C. Parameters affect such migration are thoroughly investigated from carbon source, proximity between Ni and Mo₂C to activation atmosphere and temperature. The established dual interface on Mo₂C-Ni/SiO₂ catalyst exhibits excellent catalytic performance for CO₂ hydrogenation, readily shifting the selectivity from 91% CO on Ni/Mo₂C to near 100% CH₄ on Mo₂C-Ni/SiO₂. Density functional theory calculations further verify the interfacial synergy between Ni/Mo₂C and Ni/SiO₂ sites with low barrier for CO₂ activation and a subsequent successive hydrogenation of CO. These findings highlight the important role of strong metal-support interaction (SMSI) induced metal redispersion for the rational fabrication of dual interfaces, leading to a highly active catalyst for target products.

Engineering catalyst interfaces plays a key role in designing supported metal systems. Here, the authors report that Ni nanoparticles redisperse from SiO₂ to molybdenum carbide, forming dual interfaces that enhance tandem CO₂ hydrogenation.

Introduction

Supported metal catalysts are widely used in many industrial processes for the sustainable production of fuels and chemicals¹. Typically, the interaction between metal and support directly determines the structure and properties of the resulting interfaces, thereby exerting a substantial influence on catalytic processes, with selectivity particularly being affected^2–4. Therefore, much attention has been given to the precise tuning of the metal-support interface to attain desired products.

Distinct from the classical strong metal-support interaction (SMSI) effect, which typically characterized by reducible oxide encapsulation of metal nanoparticles, recent advances have expanded the scope of SMSI behaviors. Among these, metal redispersion represents a prominent SMSI-induced phenomenon under activation or reaction conditions^5,6. Note that the SMSI as the driving force, is able to fragment large metal cluster into high density assemblies by various approaches, such as oxidative fragmentation^7,8 and atomic trapping^9,10. As a result, the interfacial contact between metal and support undergoes change, influencing the activity and selectivity of synthesized catalysts¹¹. For example, a pioneering work performed by Jones et al.⁹ has revealed the restructuring of Pt on CeO₂ surface during oxidative activation, and emphasized the importance of Pt-CeO₂ interaction to incur the migration and redispersion of Pt via gas-solid route. Eventually, the movable Pt atoms could be effectively trapped by CeO₂ support to suppress sintering, resulting in an improved activity in CO oxidation as well as prolonged durability. In another study, Paredis et al.¹² observed the NO-induced redispersion of Pd over ZrO₂ at the low temperature range (≤120 °C)) of NO reduction process, leading to a shifted selectivity toward N₂O. These studies indicate that both the supporting material and redox environment play a pivotal role in the evolution of metal structure, thus influencing the performance of the reaction.

Transition-metal carbides (TMCs), molybdenum carbides (Mo_xC) in particular, have emerged as promising substrates for stabilizing metal nanoparticles (NPs), typically forming two characteristic configurations: epitaxial clusters with few atomic layer thickness^13–15 or isolated single atoms^16–21. Distinct from conventional metal oxides, the TMCs exhibit stronger interfacial chemical bonding with loaded metals under high-temperature reductive conditions^22–27, a unique attribute that drives structural reorganization through SMSI. Notably, this SMSI-driven redispersion phenomenon enables metal migration between different substrates. For example, our previous study demonstrated the redispersion of Pt NPs from carbon nanotubes (CNTs) to α-MoC_1-x substrate under reductive treatment²⁸. Subsequent work revealed a significant decrease in Ni particle size along with the altered electronic properties of relocated Ni species on the Mo₂C-Ni/Al₂O₃ catalyst, achieving 38% activity enhancement in dry reforming of methane^29,30. Similarly, Zhou et al. demonstrated Cu migration from SiO₂ to Mo₂CT_x, forming atomically dispersed Cu-Mo₂CT_x interfaces that boost methanol synthesis rates³¹. This analogous behavior also extends to nitride systems, in which the loaded Ni NPs undergo structural evolution into undercoordinated layer-like structure on γ-Mo₂N substrate³². Despite these advances, three critical knowledge gaps remain: (i) the atomic mechanisms governing the SMSI-driven metal redispersion between two distinct substrates are elusive; (ii) current systems face intrinsic limitations in reconciling high metal loading with atomic-level dispersion; (iii) the catalytic implications of dual interfacial systems are underexplored. To address these challenges, further investigation is required. Therefore, as motivated by this intriguing behavior, there is a call for studies on dual interfacial effect to gain precise control over the underlying mechanism of metal migration across supports as well as the influence of major factors, ultimately enabling the tailored catalytic systems to obtain target products in cascade reactions.

Herein, we report how to optimize the selectivity/activity in CO₂ hydrogenation by leveraging the SMSI effects to engineer the degree of metal dispersion as well as the amounts of metal-oxide/carbide interfaces. The Mo₂C-Ni/SiO₂ catalysts have been fabricated and taken as model systems for catalyzing methanation in CO₂ hydrogenation. Effects of various factors (carbon source, mixing method, activation temperature and atmosphere) that influenced the redispersion of Ni NPs from oxide to carbide surface are investigated in attempt to understand the driving forces behind the migration as well as discriminate the key factors for such redispersion. The process of Ni redispersion from SiO₂ to Mo₂C substrate was confirmed through direct visualization of structural evolution by high-resolution transmission electron microscopy (HRTEM) and electronic state characterization by X-ray photoelectron spectroscopy (XPS). This combined evidence demonstrates the redispersion of Ni NPs from oxide to carbide, leading to the generation of new Ni/Mo₂C interfaces. The synergy between the dual interface of Ni-Mo₂C and Ni-SiO₂ results in a doubled increase of CO₂ conversion and significantly shifts the product selectivity, from 46% CO over Ni/SiO₂ and 91% CO over Ni/Mo₂C to 97% CH₄ over the Mo₂C-10Ni/SiO₂ catalyst. The CH₄ formation rate on the optimized Mo₂C-10Ni/SiO₂ catalyst (127 μmol g_cat⁻¹ h⁻¹) is about 4.2 and 18.1 times higher than that of the 10Ni/SiO₂ (30 μmol g_cat⁻¹ h⁻¹) and Ni/Mo₂C (7 μmol g_cat⁻¹ h⁻¹) catalysts, respectively. After detailed characterizations and density functional theory (DFT) calculations, we discover that the observed changes in activity and selectivity originate from the tandem catalysis occurring at the established dual interfaces of Ni-Mo₂C and Ni-SiO₂, in which the activation of CO₂ initially occurs on the Ni/Mo₂C sites, resulting in the formation of CO^* and subsequently, these CO^* intermediates were successively hydrogenated on Ni/SiO₂ sites to produce CH₄. Findings from this study greatly broaden the applicability of the SMSI effects and provide insights into a promising approach to enhance catalysis by constructing dual interfaces.

Results

Evidence and driving force for Ni migration from SiO₂ to Mo₂C

To evidence and elucidate the intrinsic driving force for Ni redispersion, a systematic study was performed by examining the influence of various factors (carbon source, mixing method, activation temperature, and atmosphere) involved in the preparation of Mo₂C-Ni/SiO₂ catalysts. Firstly, the activated carbon (C) and Mo₂C were individually chosen and mixed mechanically with the 10Ni/SiO₂ in a mass ratio of 1–3 before undergoing the reduction treatment with a gas mixture of 15%CH₄/H₂ at 590 °C. The transmission electron microscopy (TEM) analyses were employed to determine the cluster size of Ni. It was found that a small portion of Ni aggregated unevenly on SiO₂ over the Ni/SiO₂ reference, with an average particle size of 19.5 ± 0.4 nm (Figs. 1a and  S1). In contrast, the Ni-Mo₂C/SiO₂ catalyst exhibited well-distributed Ni-enriched dark dots on SiO₂, characterized by a decreased particle size of 13.8 ± 0.2 nm (Figs. 1b and  S2). While the estimated Ni particle size is about 18.1 ± 0.2 nm on the C-Ni/SiO₂ catalyst (Fig. S3), indicating a negligible impact on Ni distribution compared to the Ni/SiO₂ reference. Therefore, we can deduce that carbon (C) does not possess similar ability as Mo₂C does, an indication of weak interaction between Ni and C. Figure 1c shows the Rietveld-refined powder X-ray diffraction (PXRD) profiles of the prepared Ni-based catalysts. Notably, the Mo₂C-Ni/SiO₂ catalyst exhibited a broad diffraction peak with reduced intensity on Ni (111) reflection plane compared to the Ni/SiO₂ reference. The crystallite size of Ni decreased from 21.6 to 17.3 nm by calculating the reflection of Ni (111) plane (Fig. S4). Meanwhile, the estimated crystallite size of Ni obtained from CO pulse measurement showed a decrease upon mixing with Mo₂C (Table S1). The X-ray photoelectron spectroscopy (XPS) spectra of Mo 3d_5/2 (binding energy, B.E. = 227.7 eV) and Ni 2p_3/2 (B.E. = 855.1 eV) clearly evidenced the formation of Mo-C-Ni motifs (Fig. 1d–f)^29,33. The binding energy (B.E.) of Mo 3d_5/2 being characteristic of Mo₂C negatively shifted from 228.2 to 227.7 eV over the Mo₂C-Ni/SiO₂ catalyst, and correspondingly the B.E. of Ni 2p_3/2 being ascribed to metallic Ni⁰ showed a positive shift from 852.7 to 853.0 eV. The newly emerged peak at 855.1 eV in the Ni 2p_3/2 spectra, except for Ni⁰ and Ni²⁺ (856.3 eV, NiO) was assigned to Ni-C_x motif as confirmed in our previous studies^29,30. Notably, the Ni²⁺ content in the Mo₂C-Ni/SiO₂ catalyst (22%) exhibited a marked decrease compared to that in the pristine Ni/SiO₂ catalyst (47%). This reduction can be rationalized by the interfacial synergy between Mo₂C and Ni as induced by the migration of Ni from SiO₂ to Mo₂C, which diminishes the anchoring effect of SiO₂ support on Ni species. Consequently, the weakened metal-support interaction promotes the reduction of Ni²⁺ during catalyst activation process. This phenomenon aligns with our prior observation on the Mo₂C-Ni/Al₂O₃ system²⁹. Meanwhile, the low-valence Mo^δ+ (0 < δ < 2) species at B.E. of 227.7 eV were detected on the Mo₂C-Ni/SiO₂ catalyst, an indication of weakened Mo-C bonds, as a result of C being linked with Ni and the creation of Ni-C_x motifs. The assignment of 227.7 eV peak to Mo^δ+ aligns with our established methodology in an analogous system²⁹. Such valence change of Ni and Mo in the synthesized composite catalyst indicates the charge transfers from Ni to Mo₂C. The quantitative peak-deconvoluted data revealed that about 36% of Ni remains at electron-deficient states (Ni^σ+) after CH₄/H₂ activation on Mo₂C-Ni/SiO₂ catalyst (Table S2). The relative amount of Mo^δ+ (0 < δ < 2) increases to 44% after obtaining electrons. Thereafter, the relative amounts of Ni^σ+ and Mo^δ+ (0 < σ < 2, 0 < δ < 2) have been defined as descriptors to express the amounts of generated Ni-Mo₂C interfacial sites. Different from Mo₂C-Ni/SiO₂, the C-Ni/SiO₂ catalyst showed the formation of nickel carbides after CH₄/H₂ activation with the B.E.s of Ni 2p at 854.4 eV and C 1 s of 280.9 eV due to the strong Ni-C bonding (Fig. 1f).

Fig. 1. Structure and electronic states characterization of the as-prepared χ-Ni/SiO₂ catalysts.

a TEM image of Ni/SiO₂ catalyst. b TEM image of Mo₂C-Ni/SiO₂ catalyst. c PXRD patterns of Ni/SiO₂, Mo₂C-Ni/SiO₂, and C-Ni/SiO₂ catalysts with enlarged images marked in yellow. d–f detailed quasi in-situ XPS deconvolution of Mo 3 d, Ni 2p, and C 1 s spectra. g–i XPS semi-quantitative analysis of Mo₂C-Ni/SiO₂ prepared by different mixing methods, reduction temperatures, and gas atmospheres.

The effects of mechanical strength of different mixing methods on the preparation of Mo₂C-Ni/SiO₂ catalysts were examined. Noticeably, the crystallite size of Ni obtained by ball milling (BM) and motor milling (MM) prepared Mo₂C-Ni/SiO₂ catalysts exhibits reduced particle size compared to their granular stacking (GS) counterparts across the (111) plane (BM: 17.3 nm; MM: 18.7 nm; GS: 20.6 nm), underscoring the critical role of mechanical strength on Ni dispersion (Fig. S5). Importantly, the TEM-derived particle size shows a reasonable agreement with the XRD-derived crystallite dimensions (Fig. S6, Table S3), which validates the reliability of our multi-scale characterization approach. The difference of Ni particle size between BM and MM-prepared samples may arise from the varied distance between Ni/SiO₂ and Mo₂C. In order to rule out the possible physical structure change caused by mixing process, a control experiment was conducted by ball-milling 10Ni/SiO₂ catalyst. It is worth noting that the crystallite size of Ni remained on the BM-treated 10Ni/SiO₂ catalyst compared to the as-synthesized one (Fig. S7). Also, the XPS results showed that the relative contents of formed Ni-C species and the ratio of Ni^σ+/Ni⁰ (0 < σ < 2) and Mo^δ+/Mo²⁺ (0 < δ < 2) of BM-prepared Mo₂C-Ni/SiO₂ are higher than that of the MM-prepared Mo₂C-Ni/SiO₂. Such findings imply that the enhanced mechanical strength promotes the proximity between Ni and Mo₂C sites, thus facilitating the migration of Ni and leading to an increased formation of Ni-Mo₂C interfacial sites. A linear correlation of Mo^δ+/Mo²⁺ or Ni^σ+/Ni⁰ with Ni crystallite size was established as shown in Fig. S8. Thus, we can infer that the proximity between Ni and Mo₂C is able to be adjusted from micro- to nano-scale based on the mechanical strength provided by different mixing methods, which in turn influences the dispersion of Ni NPs.

Next, the effects of gas atmosphere and reduction temperature as key factors were investigated. Both the TEM measurement and XRD profiles of Mo₂C-Ni/SiO₂ catalysts showed the decrease of Ni particle size under H₂ or CH₄/H₂ atmosphere, which demonstrates the redispersion of Ni under reducing gases (Figs. S9 and S10, Table S4). Meanwhile, the quasi in-situ XPS spectra showed a comparable distribution of valence states for the Mo₂C-Ni/SiO₂ catalyst treated with H₂ and CH₄/H₂, which is significantly distinct from the N₂-treated one. (Figs. 1f and  S11). Also, the TEM analyses revealed a reduction in Ni particle size from 19.5 ± 0.4 nm for the pristine Ni/SiO₂ catalyst to 14.7 ± 0.2 nm following mixing with Mo₂C and CH₄/H₂ treatment at 500 °C. Elevating the activation temperature to 590 °C resulted in a further decrease to approximately 13.8 ± 0.2 nm (Fig. S12). However, the crystallite size of Ni significantly increased to 19.9 ± 0.3 nm when reduced at 650 °C (Fig. S13, Table S5). This result suggests the effectiveness of Mo₂C on keeping Ni NPs from growing at an appropriate temperature range. However, the severe coalescence of Ni NPs on SiO₂ carrier was not able to be compensated by Ni redispersion above 590 °C. The XPS spectra (Fig. 1g, Table S2) revealed a temperature-dependent evolution of metal species. As the activation temperature increased from 500 to 590 °C, the Mo^δ+/Mo²⁺ ratio progressively increased from 1.59 to 1.67, reaching a plateau above 590 °C. Concurrently, the Ni^σ+/ Ni⁰ ratio showed a parallel enhancement from 0.81 to 0.85 within the same temperature window, indicative of Mo₂C-mediated interfacial interactions facilitating Ni redispersion. However, a decline in the Ni^σ+/Ni⁰ ratio was observed between 590 and 650 °C, because the thermodynamic driving force for Ni redispersion is overtaken by Ni⁰ agglomeration on SiO₂ support. This inversion suggests that while elevated temperatures initially promote the formation of Ni-Mo₂C interfacial sites through the enhanced metal migration kinetics, beyond 590 °C the dominant mechanism shifts to Ni agglomeration on SiO₂ surface, where the cohesive energy of metallic Ni clusters overcomes the dispersion forces generated by the Mo₂C matrix³⁴. Furthermore, the diffuse reflectance infrared Fourier-transform spectra (DRIFTS) were employed by using CO as a probe molecule to provide additional evidence on the electronic states of Ni species. The emerged Ni^σ+ adsorption features in 2120 ~ 2170 cm⁻¹ displayed a distinct blueshift under identical thermal treatment conditions (Figs. S14 and S15)³⁵. Overall, the above analysis suggests that the metallic Ni⁰ species become activated on the inert SiO₂ surface when the activation temperature is beyond its Hüttig temperature (T_Hüttig = 518 K). Also, part of the added Mo₂C was slightly reduced because of the electron transfer from the relocated Ni species that reside on the surface of Mo₂C^36–38.

As illustrated in Fig. 2a, the Mo₂C-Ni/SiO₂ catalyst was prepared using a two-step synthesis method. Three distinct areas surrounding a Mo₂C patch were randomly chosen and labeled as Regions I to III in Fig. 2b. Moreover, the high-magnification high-angle annular dark-field (HAADF-STEM) images with the simultaneous energy-dispersive X-ray spectroscopy (EDX) analyses, clearly revealed the spatial distribution of Ni and Mo species (Figs. 2c–e and  S16–S18). Noted that the elemental mappings showed a pronounced distribution of Ni within Si- or Mo-enriched regions, as exemplified in Region (I). Specifically, Ni and Mo signals were well correlated in the bright region above the boundary line in Fig. 2c, indicating that Ni species were highly dispersed at high density on Mo₂C surface. This observation is consistent with the EDX line-scanning results, in which the Ni signal increased in tandem with the Mo signal (Fig. 2f). In contrast, Ni species exhibited a more disordered distribution as small patches on the Si-enriched regions below the bright area. Fig. 2d, e further substantiated the distinct distribution of Ni on Mo₂C and SiO₂. Overall, this observation provides a compelling evidence of Ni redispersion from SiO₂ substrate onto the Mo₂C surface during the synthesis of Mo₂C-Ni/SiO₂ catalyst. To further verify the presence of Ni on Mo₂C, an acid treatment (3 vol% HF) was employed to selectively dissolve the Ni/SiO₂ component from the Mo₂C-10Ni/SiO₂ catalyst, thus confirming the formation of Ni-Mo₂C interfacial sites (Fig. S19). Additionally, EDX line-scanning analysis across the Mo₂C-Ni/SiO₂ interface boundary tacked the progressive evolution of Ni distribution with increasing treatment temperature. The resulting line profiles corroborated the XPS data, demonstrating a rise in the Ni/Mo ratio from 0.11 at 500 °C to 0.21 at 590 °C, followed by stabilizing at temperature exceeding 590 °C (Figs. S20–S23).

Fig. 2. Electron microscopic characterizations of the as-prepared Mo₂C-Ni/SiO₂ catalyst (BM, 590 °C, CH₄/H₂).

a Schematic illustration of Ni NPs redispersion during CH₄/H₂ treatment induced by Mo₂C. b–e STEM-HAADF images and the corresponding X-ray energy-dispersive spectroscopy (EDS) elemental maps of Ni, Mo, and Si (b: scale bar: 50 nm; c: scale bar: 10 nm; d, e: 20 nm). f EDX-line scaning of Mo₂C-Ni/SiO₂ catalyst.

Electronic interaction upon metal redispersion

The X-ray adsorption fine structure (XAFS) studies were performed to elucidate the electronic states of Ni species. All samples were properly treated in a glove box without exposure to air. Herein, the Mo₂C-1Ni/SiO₂ composite catalysts and 1Ni/SiO₂ reference with 1 wt% of Ni were evaluated, as it is more conductive to obtain precise structural and chemical information from XAFS analysis. The Ni/SiO₂ reference exhibits both characteristic metallic Ni⁰ and oxidized Ni²⁺, with electronic configurations aligning with the XPS results (Fig. S24). Also, the comparative analysis reveals that the low-temperature activated Mo₂C-Ni/SiO₂ catalyst (denoted as Mo₂C-1Ni/SiO₂-LT, LT = 400 °C) maintains electronic signatures nearly identical to that of the Ni/SiO₂ reference, which confirmed the absence of interfacial electronic coupling without adequate thermal activation. The Ni K-edge (8333 eV) X-ray absorption near-edge structure (XANES) spectra (Fig. 3a) of high-temperature activated Mo₂C-1Ni/SiO₂ catalyst (denoted as Mo₂C-1Ni/SiO₂-HT thereafter, HT = 590 °C) showed a positive shift of Ni K absorption edge compared to the low-temperature treated Mo₂C-1Ni/SiO₂ catalyst. Also, the intensity of “white-line” (8350.5 eV for Ni K edge) increased with increased activation temperature. Both phenomena indicate the enhanced charge transfer resulting from the formed Ni/Mo₂C interface, which is similar to the reported Co/γ-Mo₂N and Pt₁/FeO_x catalysts^39,40. Since most of the Mo ions still remain as carbide in Mo₂C-1Ni/SiO₂ catalysts, there is no obvious change on the Mo XANES (20000 eV) spectra (Fig. 3b). From the above analysis, we can deduce that the strong interaction between Ni and Mo₂C is able to induce the electronic perturbation of Ni species, leading to the formation of charge-redistributed Ni-Mo₂C interfaces, eventually the re-dispersed Ni could stay in cationic state. Similar phenomenon is also observed on the redispersion process of Pt or Rh NPs on CeO₂ substrate under oxidative atmosphere^9,41.

Fig. 3. Structure and electronic states characterization of Mo₂C-Ni/SiO₂ catalyst.

a Normalized Ni K-edge (8333 eV) XANES spectra of Mo₂C-1Ni/SiO₂ catalysts with different reduction temperatures. (Both the “white line” and “edge” are marked by a dashed line arrow; Ni foil and NiO were used as references) b Normalized Mo K-edge (20,000 eV) XANES spectra of Mo₂C-1Ni/SiO₂ catalysts. (Mo foil and Mo₂C were used as references). c Fourier transform Ni K-edge EXAFS spectra of Mo₂C-1Ni/SiO₂ catalysts. d Fourier transform Mo K-edge EXAFS spectra of Mo₂C-Ni/SiO₂ catalysts. e, f Ni K-edge wavelet transformation of Mo₂C-1Ni/SiO₂-LT and Mo₂C-1Ni/SiO₂-HT catalysts, respectively. (The insert of Fig. 2f is the contour plot of charge density difference upon the formation of Ni₁@Mo₂C sites; the charge accumulation and depletion regions are colored by yellow and blue, respectively).

Moreover, the Ni K-edge extended X-ray absorption fine structure (EXAFS) was employed to study the local chemical environment of Ni species. The fitting results (Figs. 3c and  S25, Table S6) revealed the presence of Ni-C, Ni-Mo, and Ni-Ni coordination on the Mo₂C-1Ni/SiO₂-HT catalyst. The mean bond length of Ni-C is about 2.07 Å, while the bond length of Ni-Mo is around 2.91 Å. According to a previous study²⁰, we can deduce that the Mo atoms are located beneath the subsurface layer of C. Notably, the average coordination number of Ni-Ni (CN_Ni-Ni) decreased from 4.3 to 1.9 as the reduction temperature increased over the prepared Mo₂C-1Ni/SiO₂ catalyst. This result again confirmed the redispersion of Ni on Mo₂C, which can occur at HT, and the anchored Ni atoms bond to Mo₂C surface via C as bridges. In addition, Mo K-edge (20,000 eV) spectra were analyzed (Figs. 3d and  S26, Table S7). Obviously, the CN_Mo-C-Mo decreased with increased reduction temperature from 9.5 to 8.7, perhaps due to the formation of Ni-C-Mo bonds, thus reducing the intensity of the second-shell scattering of Mo₂C. This result is in accordance with the Ni K-edge spectra. The EXAFS fittings were further validated by a wavelet transformation (WT) of the EXAFS spectra. The WT contour maps of the Mo₂C-1Ni/SiO₂-LT and Mo₂C-1Ni/SiO₂-HT catalysts are distinct, as shown in Fig. 3e, f, respectively. The component in the R range near 1.6 Å is attributed to the coordination between Ni and C/O, whereas the R at 2.1 Å is contributed by the Ni-Ni coordination based on the similar k-space intensity distribution. Interestingly, the R range appeared at ca. 2.6 Å belongs to the Ni-Mo bond, indicating the significant role of activation temperature on the extent of Ni NPs redispersion.

In addition, the ab-initio thermodynamic calculations proved that the Ni NPs prefer to be atomically dispersed over the Mo₂C(101) plane with a relatively large formation energy of −5.41 eV (Fig. S27a). The population analysis verified the charge transfer between Ni and Mo₂C (Ni: +0.18|e | ) and confirmed the energetically favorable connection between Ni and Mo₂C via Ni-C-Mo bonds. (Fig. S27b–d), Figs. S28–S30, Tables S8–S10).

Catalytic performance in CO₂ hydrogenation

The catalytic performance of Mo₂C-10Ni/SiO₂ catalyst and several references was evaluated in CO₂ hydrogenation due to the structure-sensitive nature of this reaction^42–45. As shown in Fig. 4a, b, the 10Ni/SiO₂ catalyst exhibited a CO₂ conversion of 20.9% with CH₄ selectivity of 54% at 350 °C with a WHSV of 30,000 mL g⁻¹ h⁻¹. In addition, both the Mo₂C and Ni/Mo₂C references showed an enhanced CO₂ conversion due to their unique atomic arrangement and platinum-like behaviors, endowing an effectiveness for CO₂ adsorption and activation. Note that the Mo₂C-based catalysts presented an ultrahigh CO selectivity up to 91%, which is consistent with Zhang et al.¹⁷ observation. However, the limited surface areas of β-Mo₂C could impede the maximum Ni loading. How to resolve the intrinsic trade-off of metal loading and dispersion becomes an issue in catalyst design. To address this, we designed a Mo₂C-Ni/SiO₂ composite catalyst with Ni loading of 10% by taking advantage of the redispersion of Ni from SiO₂ to Mo₂C via the SMSI between Mo₂C and Ni. Impressively, the CO₂ conversion of the dual interfacial Mo₂C-10Ni/SiO₂ catalyst is the highest, more than two times that of the 10Ni/SiO₂ catalyst. Moreover, the selectivity shifts from CO over Mo₂C (~92%) and 10Ni/SiO₂ (~45%) towards CH₄ (~97%) over the Mo₂C-10Ni/SiO₂ catalyst at the same CO₂ conversion level (Fig. S31a). In contrast to Mo₂C, there is no improvement on the activity while adding C as additive (Fig. S32). Therefore, we propose that the enhanced CH₄ yield may result from the synergistic interplay between Ni-Mo₂C and Ni-SiO₂ interfaces that single interfacial system cannot be obtained. To further confirm this, we prepared another reference sample by physically mixing 10Ni/SiO₂ with 1Ni/Mo₂C without thermal pretreatment in CH₄/H₂ (denoted as 10Ni/SiO₂ + 1Ni/Mo₂C). It was found that the two interfaces that involved in the 10Ni/SiO₂ + 1Ni/Mo₂C reference (CO₂ conv. = 45.1%; CH₄ selec. = 95.5%) work synergistically in CO₂ hydrogenation, resulting in a comparable performance to that of the Mo₂C-10Ni/SiO₂ catalyst (CO₂ conv. = 48.6%; CH₄ selec. = 97.1%). Therefore, we can conclude that Ni redispersion induced by the SMSI between Mo₂C and Ni results in the formation of Ni-Mo₂C interfaces, establishing highly-dispersed catalysts at high metal loading with dual interfacial sites for CO₂ hydrogenation, which breaks the limitation of conventional preparation method.

Fig. 4. Catalytic performance of Mo₂C-Ni/SiO₂ catalysts in CO₂ hydrogenation.

a CO₂ conversion and CH₄ STY of Mo₂C-10Ni/SiO₂ catalyst with references at 350 °C, 0.1 MPa, and 30,000 mL g⁻¹ h⁻¹ with CO₂:H₂ ratio of 1:4. b Product (CO or CH₄) selectivities. c Effect of mixing methods in CO₂ hydrogenation. (DL-double layer stacking; GS-granular stacking, MM motor milling, BM ball milling. error bar represents standard deviation; the error bars in a, b represent the standard deviation from three or more independent experiments). d The corresponding product selectivities of various mixing methods. e Calculated TOFs of CH₄ formation over 10Ni/SiO₂ and Mo₂C-10Ni/SiO₂ catalysts (300 °C, 0.1 MPa, 80,000 mL g⁻¹ h⁻¹). f The calculated activation energy (E_a) for CH₄ formation of 10Ni/SiO₂ and Mo₂C-10Ni/SiO₂ catalysts. (275–350 °C, 0.1 MPa, and 80,000 mL g⁻¹ h⁻¹).

Followed by this, the effects of the integration manner were investigated (Figs. 4c, d and  S31b). It was found that both the CO₂ conversion and CH₄ selectivity increase in the order of double layer (DL-conv. 25.5%; CH₄:65.4%) <granule stacking (GS-conv. 27.7%; CH₄:72.2%) <motor milling (MM-conv. 38.2%; CH₄:89.4%) <ball milling (BM-conv. 48.6%; CH₄:97.1%), which is consistent with the increase in Ni-Mo₂C interfacial sites. As a result, a higher degree of redispersion resulting from distinct mixing methods was able to generate more Ni-Mo₂C interface, thus enhancing the formation of CH₄. Aside from the mixing method, other factors were investigated as well, including the mass ratio of Ni/SiO₂ to Mo₂C, activation temperature, and atmosphere (see details in Figs. S33 and 34). Therefore, we can reasonably deduce that the relative ratio between Ni-Mo₂C and Ni-SiO₂ and the activation process significantly impact the activity and selectivity of CO₂ hydrogenation by the SMSI-induced Ni redispersion. As depicted in Fig. S34, the obtained molar ratio of Ni/Mo was about 0.5, indicating that the redispersion degree of Ni was closely related to the amount of Mo₂C. Also, the CO₂ conversion exhibited a gradual increase as the activation temperature increased up to 590 °C and slightly dropped thereafter (Fig. S34). Only a good match between CO₂ activation and CO_x hydrogenation is able to enhance the CH₄ formation rate.

To ensure the measurements were maintained in a kinetic-relevant regime, the CO₂ conversion was limited to 15% with a relatively high space velocity of 80,000 mL g⁻¹ h⁻¹ at various temperatures. The calculated approach to equilibrium (η) is about 0.16, which aligns with the recommended threshold (η < 0.2) for kinetic control and suggests that the experimental condition applied is far from equilibrium⁴⁶. All rate measurements were thermodynamically corrected, which confirmed the kinetic control throughout the experimental range (Fig. S35, Tables S11–S14). As displayed in Fig. 4e and Table S15, the calculated turnover frequency (TOF; normalized by the number of CO chemisorption sites) of CH₄ formation of the Mo₂C-10Ni/SiO₂ catalyst (7.1E-04 s⁻¹) is about 7.8x’s higher than that of the 10Ni/SiO₂ catalyst (9.0E-05 s⁻¹), indicating the improved intrinsic activity of Ni sites due to their altered electronic properties. Taking the observed Ni redispersion from SiO₂ to Mo₂C and the established electron transfer into account, we propose that the increased TOF originates from synergistic effects combining better Ni exposure and optimized electronic structure. Moreover, the apparent activation energy (E_a) of Mo₂C-10Ni/SiO₂ and 10Ni/SiO₂ catalyst for CO₂ hydrogenation was determined to be 67.7 ± 1.1 and 85.2 ± 0.7 kJ mol⁻¹, respectively (Table S12). The obtained E_a over 10Ni/SiO₂ catalyst is in good agreement with the literature reported value^47–49. Importantly, the activation energy (E_a) of CH₄ formation exhibited a substantial reduction from 93.8 ± 0.6 kJ mol⁻¹ over Ni/SiO₂ to 67.6 ± 1.1 kJ mol⁻¹ for the Mo₂C-Ni/SiO₂ catalyst, demonstrating that Mo₂C incorporation effectively lowers the energy barrier for methane generation (Fig. 4f, Table S16). In contrast, the E_a decreased for CO formation was less pronounced, shifting from 79.8 ± 0.7 to 72.4 ± 3.2 kJ mol⁻¹ upon Mo₂C addition (Fig. S36). This differential effect highlights the preferential enhancement of CH₄ pathway kinetics by Mo₂C-containing sites. The variation in the activation barrier of 10Ni/SiO₂ and Mo₂C-Ni/SiO₂ catalysts implies the facile activation of CO₂ on Mo₂C-Ni/SiO₂, which may result from the Ni redispersion-induced electronic and geometric structure change on active Ni sites.

Reaction mechanism exploration

As shown in Fig. 5a–h, we performed a two-step temperature-programmed surface reaction (TPSR) to explore the reaction mechanism of CO₂ hydrogenation over the Ni/SiO₂, Mo₂C-Ni/SiO₂, and Mo₂C catalysts. Initially, a 2%CO₂/He mixture (50 mL min⁻¹) was introduced at 300 °C after activation. Noted that a small amount of CO(g) was immediately produced on the Ni/SiO₂ catalyst (Fig. 5a). While no gaseous CO was detected over the Mo₂C-Ni/SiO₂ catalyst (Fig. 5d), either because of the strong interaction of CO^* to the active sites or an alternative CO₂ activation route. To verify the potential intermediates, we performed a temperature-programmed desorption (TPD) under He flow after CO₂ dissociation (Fig. 5c, f). Notably, the attached intermediates gradually desorbed as gaseous CO with increased temperature on both catalysts. For Ni/SiO₂ catalyst, a broad CO desorption peak was detected, which starts at around 400 °C; while only one CO desorption peak appeared with the onset temperature of 500 °C on the Mo₂C-Ni/SiO₂ catalyst, which implies that the interaction between CO^* and Mo₂C-Ni/SiO₂ surface is stronger than that of the Ni/SiO₂ catalyst. Also, the amount of desorbed CO on Mo₂C-Ni/SiO₂ is about two times that of the Ni/SiO₂ catalyst, indicating that more CO^* species are preserved on the former catalyst. In sharp contrast, the Mo₂C alone exhibited a spontaneous CO₂ dissociation, forming CO(g) as a major product, indicating the weak interaction between CO and the O* occupied Mo carbide surface (Fig. 5g). This result is similar to that of the Ni/Mo₂C reference (Fig. S37). Followed by this, the system was switched to 5%H₂/He (50 mL min⁻¹) at the same temperature after purged with N₂ (50 mL min⁻¹, 30 min) (Fig. 5b, e). Note that the amount of generated CH₄ in the 2nd step of hydrogenation over Ni/SiO₂ was less than that of Mo₂C-Ni/SiO₂, which is consistent with the TPD results as well as our experimental data. It suggests that the produced CH₄ mainly originates from the hydrogenation of adsorbed CO. Similar to the Mo₂C-Ni/SiO₂ catalyst, the physically mixed 10Ni/SiO₂ + 1Ni/Mo₂C reference also exhibits a considerable formation of CH₄ as a result of the two-step tandem process on the involved dual interface (Fig. S38). Meanwhile, the in-situ DRIFT experiments were carried out to probe surface intermediates by co-feeding CO₂ and H₂ mixture (1:4) over the Mo₂C-Ni/SiO₂ catalyst (Fig. S39). It was found that more CO^* intermediates are detected at low temperature, as indicated by the bands located at 2141, 2075, 2030, and 2010 cm⁻¹, which are attributed to the adsorbed CO^* on the Ni^σ+, Ni⁰, and Mo₂C sites over the Mo₂C-Ni/SiO₂ catalyst, respectively. The emergence of CO^* indicates the strong CO₂ dissociation ability of the Mo₂C-Ni/SiO₂ catalyst, even at a relatively low temperature. This result is in agreement with the two-step MS experiment. Also, the intensity of CH₄ signal at 3016 cm⁻¹ increases with increased reaction temperature, most likely due to the successive hydrogenation of CO^*.

Fig. 5. Insights into the reaction mechanism over Mo₂C-Ni/SiO₂ catalyst.

a–h Two-step MS investigation of Ni/SiO₂ (a–c), Mo₂C-Ni/SiO₂ (d–f), and Mo₂C (g, h) catalysts. CO₂ dissociation at 300 °C (a, d, g), followed by H₂ treatment (b, e, h) (Products were detected by online mass spectrometer) or a temperature-programed desorption under He flow (TPD: c, f). i Energy barrier (Ea) for CO₂ and H₂ dissociation on Ni₄/SiO₂, Ni₁/Mo₂C (101), and Mo₂C models. j Adsorption energy of CO. k Reaction paths for CO₂ hydrogenation over Mo₂C-Ni/SiO₂ catalyst.

To get more insights on the catalytic function of involved active sites, the Ni₄/SiO₂, Ni₁/Mo₂C(101), and Mo₂C(101) models were built and used as references, representing Ni/SiO₂, Ni/Mo₂C, and Mo₂C catalysts, respectively. The energy barrier of these three catalyst models for CO₂ and H₂ dissociation (E_diss,CO2 & E_diss,H2) was calculated and compared, as displayed in Fig. 5i. The E_diss,CO2 of Ni₁/Mo₂C(101) (0.46 eV) is lower than those of Ni₄/SiO₂ (1.83 eV) and Mo₂C(101) (0.63 eV), indicating that the Ni/Mo₂C catalyst is favorable for CO₂ activation. The value of E_diss,H2 follows the trend of Mo₂C (0.31 eV) > Ni₁/Mo₂C (0.17 eV) > Ni₄/SiO₂ (0.15 eV), indicating the outstanding performance of Ni₄/SiO₂ in H₂ activation. We then investigated the important successive step by calculating CO adsorption energy (E_ads,CO) over those three models (Fig. 5j). The results show that the Ni/SiO₂ model binds CO much stronger (−2.30 eV) than Ni/Mo₂C (−1.95 eV) and Mo₂C(101) (−1.58 eV), which is consistent with the CO-TPD measurement. Apart from this, Fig. S40 displays the CO desorption behavior on Ni/SiO₂, Mo₂C, and Mo₂C-Ni/SiO₂ surface. Clearly, a new CO-desorption peak (ca. 242 °C) appeared on the Mo₂C-Ni/SiO₂ catalyst, indicating the formation of new active sites, which might originate from the Ni-Mo₂C domains. Similar result was also observed by Ma et al.⁵⁰ on a tandem Pt-Mo₂C/C catalyst. Therefore, we can deduce that the electronic interfacial sites provided by Ni-Mo₂C interface showed the potential of stabilizing undercoordinated Ni^δ+ species for the selective cleavage of C = O bond, while the structural interfacial sites of Ni-SiO₂ maintained the high-density metallic ensembles for efficient H₂ dissociation. The presence of Mo₂C modulated the interaction between metal and support by preventing the over stabilization of intermediates on Ni/SiO₂ sites. This bifunctional synergy, in which the Mo₂C dictates electronic states and SiO₂ controls nanostructure, resolves the persistent activity-selectivity-lifetime trade-off.

Extensive DFT-based calculations were performed to highlight the synergy between Ni/SiO₂ and Ni/Mo₂C dual interfaces in CO₂ hydrogenation (Fig. 5k). Ni atoms are stabilized by two-fold coordinated C atoms on Mo₂C (101) plane as shown on the insert of Fig. 3f. These partially charged and undercoordinated Ni atoms are more efficient than surface Mo as reaction sites for CO₂ and H₂ activation. The initial barrier for CO₂ dissociation (CO₂^* + 4H^* → CO^* + O^* + 4H^*) on Ni/Mo₂C interface is about 0.63 eV, which is comparable to that over bare Mo₂C (101) surface (Fig. S41). Also, this step is 0.69 eV exothermic, forming stable Ni-carbonyl in-situ. Moreover, we notice that the hydrogenation of adsorbed O^* can be regarded as the most sluggish step on the Ni/Mo₂C (101) and Mo₂C (101) surface with energy barrier of 1.18 and 1.49 eV (TS3 and TS5, Fig. 5k), respectively, but is vital to keep the reaction proceed. Originating from the Ni/Mo₂C interface, the previously formed Ni-carbonyl sites could convert the adsorbed CO₂ and H₂ at an energy barrier of 0.46 and 0.17 eV (TS6 and TS4, Fig. 5k), respectively, which are significantly lower than those over Mo₂C (101) plane and Ni/SiO₂ interface. Furthermore, CO generated by CO₂ dissociation will be shifted to a surface Mo atom that is not available for CO adsorption/desorption on Mo₂C(101), where CO desorbs easily at a barrier of 0.85 eV (step 16-17) at reaction conditions. Compared to Mo₂C(101), this value decreased about 1.0 eV. In contrast, we found that the rate determine step of CO₂ hydrogenation over Ni₄/SiO₂ model is CO₂ dissociation, which needs to overcome an energy barrier of 1.83 eV though the reaction is exthermic by 0.58 eV. These results indicate that CO₂ can be activated easily on Ni/Mo₂C interface to facilitate the breakage of C = O bonds as well as the CO desorption. Driven by the aforementioned strong CO adsorption ability, the released CO would migrate to Ni₄/SiO₂ interface for further hydrogenation, eventually forming CH₄ as final product. Given the experimental and computational results, we can propose that the synthesized Mo₂C-Ni/SiO₂ catalyst constitutes a tandem catalytic system for CO₂ hydrogenation as induced by the SMSI between Mo₂C and Ni, in which the CO₂ reduction could initially occur on the Ni/Mo₂C interface while the subsequent hydrogenation of CO takes place on the Ni/SiO₂ interface.

Discussion

In summary, we report the redispersion of Ni NPs from SiO₂ substrate to Mo₂C under reduction conditions. The established Ni-C-Mo bonds stabilize Ni on Mo₂C surface in electron-deficient states as well as decrease the particle size of Ni nanoparticles. This results in the creation of Ni/Mo₂C interface, which works synergistically with the original Ni/SiO₂ interface to yield a dual interfacial system for CO₂ hydrogenation. A detailed tandem catalytic mechanism, involving interfacial synergistic catalysis over the Mo₂C-Ni/SiO₂ catalyst, has been elucidated. Both the experimental studies and theoretical calculations substantiate that the Ni-Mo₂C interfacial sites are catalytically active for CO₂ activation, resulting in an increased formation of CO species, which are readily hydrogenated on Ni-SiO₂ interface to produce CH₄. This work opens up a new avenue for designing dual interface systems to enable a precise control on the reactivity and selectivity of catalytic process.

Methods

Catalyst synthesis

Preparation of Mo₂C-Ni/SiO₂ catalysts: The Ni-based catalysts (denoted as Ni/SiO₂) were synthesized by incipient wetness impregnation (IWP) with Ni nominal loading in the range of 1–10 wt% on silica support. Molybdenum carbide (denoted as Mo₂C) as the second component was synthesized by a temperature programmed reduction as follows: Firstly, the ammonium heptamolybdate ((NH₄)₆Mo₇O₂₄·4H₂O) was heated to 500 °C in a muffle furnace and held for 4 h to prepare MoO₃ precursor. After that, the obtained MoO₃ was heated from room temperature (RT) to 300 °C in 20 vol% CH₄/H₂ mixture with a ramping rate of 5 °C/min, then increased from 300 to 700 °C at a ramping rate of 1 °C/min and held for 2 h at 700 °C. The resulting material was quickly quenched to room temperature and passivated by 1%O₂/Ar mixture (20 mL/min) for 4 h. The investigated materials were further prepared by mixing Ni/SiO₂ with Mo₂C with a mass ratio of 5:1–1:3. Finally, the prepared mixture is subjected to a 15%CH₄/H₂ treatment at designated temperature (400–650 °C) for 2 h. The resulting samples were denoted as Mo₂C-Ni/SiO₂.

Preparation of Ni/Mo₂C reference catalysts: Ni/Mo₂C catalysts were prepared by co-precipitation and carburization for comparison. Prior to air exposure, the Ni/Mo₂C catalyst was typically passivated in 1%O₂/N₂ flow for 2 h.

Catalyst characterization and CO₂ hydrogenation evaluation

X-ray diffraction (XRD): Powder X-ray diffraction (XRD) data were collected using a Rigaku SmartLab 9 kW diffractometer with Cu Kα radiation (λ = 1.5406 Å) with a step size of 0.02° and counting time of 8 s per step in the range of 2θ = 10-80°. The Diffraction patterns were compared to powder diffraction files from the Joint Committee on Powder Diffraction Standards (JCPDS) database. The crystallite size was estimated from the broadening peak of (111) plan for Ni by Scherrer equation. All the catalysts were passivated before measured.

Quasi in-situ X-ray photoelectron spectroscopy (XPS): The information of surface structure of the as-prepared catalysts was collected by XPS of Axis Ultra Imaging Photoelectron Spectrometer (Kratos Analytical). The prepared Mo₂C-Ni/SiO₂ and their references were activated in a continuous flow fixed-bed quartz reactor and treated under the target feed gas. Treated catalysts were made into small tablets (6.0 mm diameter) and placed on the sample holder without exposure to air in a glovebox. The sample was then introduced into the ultra-high vacuum chamber for XPS measurements without air exposure. 284.6 eV was used as an internal binding energy standard of C 1 s to correct for surface charging shift. XPSPEAK41 was used for spectra deconvolution by a nonlinear least-squares method with a combination of Lorentzian (20%) and Gaussian (80%) peak shapes.

Electron microscopy characterization: The high-resolution transmission electron microscope (HR-TEM) images of the prepared catalysts were recorded on a (JEOL) JEM 2200FS electron microscope with an accelerating voltage of 200 kV. Samples were prepared by dispersing in ethanol and sonicated for 300 s. A drop of the suspension was placed on a 230-mesh copper grid coated with Formvar-carbon film and then dried in a vacuum chamber. The clusters/particles identified in the TEM images were measured and counted to yield a size distribution that was fitted to a Gaussian distribution (100–200 particles were counted for statistical analysis) to determine the average size of the cluster/particle. All image analysis was conducted with Digital Micrograph software.

CO-adsorbed diffuse reflectance infrared spectra (CO-DRIFTs) and in-situ DRIFT: DRIFT spectroscopy of CO adsorption was employed on a Thermo Nicolet iS50 spectrometer with a mercury cadmium telluride (MCT) detector cooled by liquid nitrogen to estimate the electronic state of Ni and Mo species. All catalysts were prepared by diluting with quartz sands at a mass ratio of 1:5 prior to DRIFTS measurements to minimize the intermolecular dipole interaction of adsorbed CO. All the tests were performed in Harrick Praying Mantis high temperature reaction chamber with ZnSe windows. A thermocouple mount that allowed direct measurement of sample surface temperature. Each recorded spectrum was averaged over 32 scans in the region of 4000-650 cm⁻¹ with a nominal 32 cm⁻¹ resolution. Prior to CO adsorption, the precursor was pre-treated by 15%CH₄/H₂ at 590 °C for 2 h, unless otherwise indicated. After that, the flow was switched to N₂ until the temperature decreased to 17 °C at a flow rate of 40 ml (STC)/min to clean the catalyst surface and then a background spectrum was recorded. CO adsorption was conducted at 17 °C by introducing CO into the system for above 20 min to ensure the complete saturation of CO gas molecules on catalyst surface. Finally, N₂ was introduced to flush the physiosorbed CO. Several spectra were collected by 30 s interval until the spectra were stabilized.

X-ray absorption fine structure (XAFS): Ni L₃ edge (8333 eV) and Mo K-edge (20000 eV) XAFS were measured at BL14W beamline of Shanghai Synchrotron Radiation Facility (SSRF) in transmission/ fluorescence mode using Lytle detector to collect the data. Ni foil and NiO were used as standards for Ni K-edge spectra; Mo foil and MoC were used as standards for Mo K-edge spectra. All the samples were pre-activated in 10%H₂/Ar at 350 °C for 1 hour and sealed in a chamber under argon protection in a glove box.

Two-step temperature programmed surface reaction (TPSR): The 1st step is CO₂ dissociation—The activated samples were exposed to 2%CO₂/He (50 mL min⁻¹) at a proper temperature for about 10 min. The effluent gases were measured online using a Mass Spectrometer (MKS-Cirrus3, USA). After N₂ puring (50 mL min⁻¹), the samples were reduced in 5%H₂/He (50 mL min⁻¹) in the 2nd step hydrogenation. In the control experiment, the 2^nd step hydrogenation was replaced by a temperature-programmed desorption (TPD) under He flow to observe the desorbed species.

CO temperature-programmed desorption (CO-TPD): Temperature programmed desorption of CO (CO-TPR) was carried out on Micromeritics AutoChem 2920. The sample (100 mg) was reduced at 400 °C by 5%H₂/Ar flow (30 mL min⁻¹) for 1 h to remove passivation layer. After cooling dow, the CO adsorption was performed at 40 °C for 1 h. The sample was then flushed with N₂ (30 mL min⁻¹) until the baseline is stable. The desorption was conducted from 40 to 650 °C with a ramping rate of 10 °C/min. Dosorbed species were monitored by mass spectrometer (MS).

CO Chemisorption: This technique was used to determine the number of active sites (#CO, μmol g⁻¹) on a Quantachrome ChemBETPulsar analyzer. Samples were purged with He at 100 °C for 30 min to remove moisture, and then reduced by 5%H₂/Ar flow at 350 °C for 1 h to remove the passivation layer. Afterwards, the CO uptake was measured using pulsed chemisorption at 35 °C by taking thermal conductivity detector (TCD) as detector. The turnover frequency (TOF), defined as the number of CO₂ converted per surface site per second, was calculated based on the following equation:

$$\mathrm{TOF}=\frac{F_{CO2,inlet}\cdot X_{CO2}}{\mathrm{\#}CO\cdot M}$$ 1

where the F_{CO2, inlet} represented the inlet flow rate of CO₂ (mol s⁻¹), M represents the molecular weight of CO₂ (g mol⁻¹).

To determine Ni particle size (assuming cubic geometry), CO chemisorption measurements were performed on a Micrometritics AutoChem III 2930. Approximately 100 mg of the sample was loaded into a U-shaped quartz tube. The sample was prepared in situ under different conditions. After cooling to 35 °C and purging with high-purity He, the TCD signal was allowed to stabilize for 30 min. The sample was then exposed to ten pulses of 10% CO/He. The quantity of CO adsorbed was detected by TCD by assuming a stoichiometric ratio of one CO molecule per surface Ni atom.

CO₂ hydrogenation was evaluated in a packed bed reactor, in which 100 mg catalyst powder mixed with 300 mg pre-calcinated SiO₂ were loaded into a quartz tube with an inner diameter of 6 mm. The reactant was composed of 12% CO₂/48%H₂/15%Ar in N₂ with a total flow rate of 50 mL (STP)/min. Ar was used as the internal standard. The activity test was conducted at 0.1 MPa in the temperature range of 548 to 623 K with a GHSV of 30,000 mL g⁻¹ h⁻¹. The products were analyzed online using an Agilent GC7890B equipped thermal conductivity (TCD) and flame ionization detector (FID), respectively. For the Mo₂C-Ni/SiO₂ samples, the standard pre-treatment was conducted at 590 °C for 2 h in 15%CH₄/H₂ with a flow rate of 50 mL (STP)/min; whereas, H₂ was only used for Ni/SiO₂ references. In the quantification, the conversion of CO₂ is defined as:

$$X_{CO2}\left(\%\right)=\frac{n_{CO2}^{in}-n_{CO2}^{out}}{n_{CO2}^{in}}*100\%$$ 2

The observed rate (r_observed) is corrected using the Temkin-Pyzhev formalism as follows:

$$r_{corrected}=\frac{r_{observed}}{\left(1-\eta \right)}$$ 3

The carbon balance is defined as:

$$B_C\left(\%\right)=\frac{n_{C{O}_2}^{out}+n_{CO}^{out}+n_{C{H}_4}^{out}}{n_{C{O}_2}^{in}}\times 100\%$$ 4

In all tests, the carbon balance of the system is above 94%.

Computational details

The spin-polarized first-principle calculations are performed with Vienna Ab Initio Simulation Package (VASP 5.4.4). Perdew-Burke-Ernzerhof (PBE) functional was used to solve the Kohn-Sham equations within periodic boundary conditions. The electron-nucleus interactions are described using PAW pseudopotentials. A 4 × 4 × 4 Monkhorst-Pack k-point grid was used for bulk β-Mo₂C and Ni, and a 2 × 2 × 1 k-point grid was used for the layered SiO₂ structure derived from a recently reported silicate. Γ-centered grids with equivalent k-point density were then used in calculations of clean and Ni-deposited surface structures on β-Mo₂C and SiO₂. A planewave cutoff of 400 eV was used throughout this work. Second-order Methfessel-Paxton smearing with a width of 0.10 eV was used to handle the partial occupancies of the bands. The convergence criterion for energy is 1 × 10⁻⁵eV and that for ionic relaxation is 0.01 eV/Å. With the above setup, the calculated lattice parameters a, b and c of bulk β-Mo₂C are 6.06, 6.07 and 4.07 Å, respectively, while that for bulk face-center-cubic Ni was 3.51 Å. The layered SiO₂ was found plausible in an orthorhombic lattice of 10.94 × 10.30 Ų. After test calculations, (101) surface slab with thickness of 4 Mo layer with mixed C and Mo termination was selected to mimic (101) surface of β-Mo₂C, as the adsorption energy of a single Ni atom on (1 × 1) slabs in the configuration bridging 2 nearest neighboring surface C already converges within 0.03 eV. The stability of deposited Ni_n structures was investigated on a (2 × 2) supercell of (101) β-Mo₂C and a (1 × 1) orthorhombic slab of SiO₂. The freestanding Ni_n clusters were investigated in a 15.00 × 15.10 × 15.20 ų orthorhombic cell. The relative stabilities of Ni_n structures ($\mathrm{\Delta }{G}_{{\mathrm{f},\mathrm{Ni}}_{\mathrm{n}}/\mathrm{Support}}$) were determined with the normalized formation free energy, calculated as:

$$\mathrm{\Delta }{G}_{{\mathrm{f},\mathrm{Ni}}_{\mathrm{n}}/\mathrm{Support}}=\left(G_{{\mathrm{Ni}}_{\mathrm{n}}/\mathrm{Support}}-n\times E_{\mathrm{Ni},\mathrm{atom}}-G_{\mathrm{Support}}\right)/n$$ 5

where $G_{{\mathrm{Ni}}_{\mathrm{n}}/\mathrm{Support}}$, $n$, ${\mathrm{E}}_{\mathrm{Ni},\mathrm{atom}}$ and $G_{\mathrm{Support}}$ are the free energy of the Ni_n/support composite, the number of Ni atoms in the Ni_n/support composite, electronic energy of a single Ni atom and the free energy of the support, respectively. First-principles thermodynamics scheme was adapted to calculate free energy of Ni_n/support composites and the support at finite T. Free energy is approximated from Helmholtz free energy by ignoring pV term, and was calculated as:

$$G\left(\mathrm{T}\right)=E_{\mathrm{DFT}}+E_{\mathrm{vib}}-T{S}_{\mathrm{vib}}$$ 6

Where ${\mathrm{E}}_{vib}$ and ${\mathrm{S}}_{vib}$ are vibrational energy and entropy, respectively. ${\mathrm{S}}_{vib}$ was calculated with partition functions of structures in the temperature range of interest. For reference, the formation energies of these Ni_n structures ($\mathrm{\Delta }{E}_{{\mathrm{f},\mathrm{Ni}}_{\mathrm{n}}/\mathrm{Support}}$) were also calculated as:

$$\mathrm{\Delta }{E}_{{\mathrm{f},\mathrm{Ni}}_{\mathrm{n}}/\mathrm{Support}}=\left(E_{{\mathrm{Ni}}_{\mathrm{n}}/\mathrm{Support}}-n\times E_{\mathrm{Ni},\mathrm{atom}}-E_{\mathrm{Support}}\right)/n$$ 7

where $E_{{\mathrm{Ni}}_{\mathrm{n}}/\mathrm{Support}}$, $n$, $E_{\mathrm{Ni},\mathrm{atom}}$ and $E_{\mathrm{Support}}$ are the total energy of the Ni_n/support composite, the number of Ni atoms in the Ni_n/support composite, electronic energy of a single Ni atom and the total energy of the support, respectively. With the above setup and definition, the calculated $\mathrm{\Delta }{E}_{\mathrm{f},\mathrm{Bulk\; Ni}}$ is −4.79 eV.

The pathways for CO₂ hydrogenation over Mo₂C(101), Ni₁/Mo₂C(101) and Ni₄/SiO₂ were investigated with extensive spin-polarized first-principles-based calculations using Perdew-Burke-Ernzerhof (PBE) functional within the formalism of generalized gradient approximation (GGA) as implemented in the DMol3 package. DFT semi-core pseudopotentials (DSPP) and the double numerical plus polarization (DNP) basis sets were used to describe the core and valence electronic states. Test calculations on bulk and surface structures of Mo₂C, Mo₂C(101), Ni/Mo₂C(101) and Ni₄/SiO₂, as well as energetics and structural of CO, CO₂ and H₂ adsorption over these model structures yield essentially the same results as those obtained with VASP.

Author contributions

H.W.: conceptualization, investigation, methodology, formal analysis, data curation, writing—original draft, software, writing—review & editing, visualization, funding acquisition. X.Q.: XAS analysis, methodology, formal analysis, data curation, software, writing—review and editing. Z.G.: TEM operation and analysis, methodology, formal analysis, data curation, software, writing—review and editing. Y.D., S.L. (Shenghua Liu), M.P., S.L. (Shida Liu): data curation, writing—review and editing. S.H.: conceptualization, Supervision. X.L.: conceptualization, project administration. X.G., D.M., and C.S.: conceptualization, supervision, funding acquisition, project administration.

Peer review

Peer review information

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

Data availability

The data generated in this study are available within the paper, its supplementary information files, and the repository. Source data are provided with this paper (Figshare 10.6084/m9.figshare.29924090). Data are available from the corresponding authors upon request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-64030-9.