Direct cleavage of C=O double bond in CO₂ by the subnano MoO_x surface on Mo₂N

Abstract

Compared to H₂-assisted activation mode, the direct dissociation of CO₂ into carbonyl (*CO) with a simplified reaction route is advantageous for CO₂-related synthetic processes and catalyst upgrading, while the stable C = O double bond makes it very challenging. Herein, we construct a subnano MoO₃ layer on the surface of Mo₂N, which provides a dynamically changing surface of MoO₃↔MoO_x (x < 3) for catalyzing CO₂ hydrogenation. Rich oxygen vacancies on the subnano MoO_x surface with a high electron donating capacity served as a scissor to directly shear the C = O double bond of CO₂ to form CO at a high rate. The O atoms leached in CO₂ dissociation are removed timely by H₂ to regenerate active oxygen vacancies. Owing to the greatly enhanced dissociative activation of CO₂, this MoO_x/Mo₂N catalyst without any supported active metals shows excellent performance for catalyzing CO₂ hydrogenation to CO. The construction of highly disordered defective surface on heterostructures paves a new pathway for molecule activation.

Introduction

The global warming caused by the increasing CO₂ emission has attracted extensive attention around the world. In addition to being a greenhouse gas, CO₂ can be considered as an available inert carbon source to synthesize a range of value-added chemicals and hydrocarbon fuels^1,2,3,4,5. For CO₂-related fine synthesis processes, carbonyl (*CO) is a pivotal platform species for the generation of the final products by coupling the further CO hydrogenation or carbonylation^6,7,8,9,10. The formation of *CO is closely linked to the activation of CO₂, which is always a great challenge due to its chemical inertness and the thermodynamic stability. In general, CO₂ activation pathways can be classified into two pathways, i.e., H₂-assisted CO₂ activation and the direct dissociation of CO₂^11,12. The major difference between the two is whether CO₂ reacts with the dissociated H species to form reaction intermediates, which then produce *CO or directly dissociates to *CO. Based on previous reports, both for metal complexes as homogeneous catalysts and supported metal catalysts in heterogeneous catalysis, the activation of CO₂ dominantly follows the H₂-assisted mode^13,14,15, in which the formation of active intermediates not only complicates the entire reaction route but also places redundant structural requirements on catalysts. The activation of CO₂ molecules by the high-rate cleavage of stable C = O double bond is a potential strategy and a crucial scientific issue to overcome the above obstacles, although it seems to be very difficult to achieve.

It has been established that due to the unique catalytic properties of oxygen vacancy, defect engineering based on reducible oxides has been widely used in CO₂-related catalytic reactions^16,17,18,19. On the one hand, the formation of oxygen vacancy can modulate the geometric and electronic structure of adjacent active metals^20,21, thus promoting the activation of CO₂ and the further formation of active intermediates at the active metal-oxygen vacancy interfaces. On the other hand, as one of typical coordination-unsaturated sites, oxygen vacancy itself can adsorb CO₂ molecules to participate in the catalytic reactions^16,22. However, for conventional reducible oxides, companied by the formation of oxygen vacancies, the exposed metal sites with unsaturated coordination donate poor electrons to the anti-bonding orbitals of CO₂, which largely limits the cleavage of the C = O bonds²³. If rich surface oxygen vacancies with strong electron donating capacity can be constructed, the above issues will be solved, while it has been rarely reported. According to recent reports, molybdenum nitride and molybdenum carbide-related catalysts exhibited an excellent catalytic efficiency in C1 chemistry^{24,25,26,27,28}. However, the high activity of MoC, Mo₂C and Mo₂N-based catalysts is always attributed to the MoC, Mo₂C and Mo₂N bulk structure^{11,25,26,27,28}, and the catalytic properties of MoO₃ passivation layer are ignored, especially for Mo₂N with well-defined MoO₃ surface structure^25,29. In our previous works, we pioneered the finding that the MoO₃ passivation layer with low formation energy of oxygen vacancies on the bulk Mo₂N was the intrinsic active surface of a Pt/Mo₂N catalyst to catalyze the water gas shift reaction²⁹. Subsequently, the synergistic interface between Pt clusters and oxygen vacancies on the defective MoO_x (x < 3) passivation layer was confirmed to largely enhance the H₂-assisted CO₂ activation³⁰. However, due to the superficial understanding of the intrinsic catalytic properties and the redox behavior of the defective MoO_x structure, the constructed active sites on a relatively ordered MoO_x surface in previous works was still unable to achieve the dissociative activation of CO₂. Nevertheless, compared to conventional reducible oxides, the special MoO₃/Mo₂N heterostructure with lower oxygen vacancy formation energy shows a huge potential to develop a new-type defective surface with strong electron donating capacity to directly cleave the solid C = O double bond over CO₂.

Herein, we clearly revealed the ultra-active redox property of subnano MoO₃ surface on bulk Mo₂N and constructed a dynamically changing surface to achieve the directly dissociative activation of CO₂. Under reductive atmospheres, companied with the reduction of Mo-O species, a deep-reduced subnano MoO_x (1<x < 3) surface on Mo₂N with high-density oxygen vacancies was created. This highly disordered subnano surface with abundant coordination-unsaturated Mo sites provided a favorable geometric and electronic microenvironment, which directly cleavage the stable C = O double bond in CO₂ (CO₂→CO + O_surface) via the electron transfer mediated by exposed Mo sites. The oxygen vacancies occupied by oxygen atoms leached from the CO₂ dissociation were easily regenerated by H₂ (H₂+O_surface→H₂O). This in situ cyclic consuming and generating process of oxygen vacancies accelerated the transfer of oxygen atoms from CO₂ to catalyst surface and then to final product, effectively catalyzing the CO₂ hydrogenation to CO, known as the reverse water gas shift (RWGS) reaction. Under harsh reaction conditions (600 °C, GHSV = 200,000 mL/g_cat/h), in the absence of any supported active metals, this MoO_x/Mo₂N catalyst exhibited a promoted CO₂ conversion (~48%), complete CO selectivity and durable cyclic stability (900 h), which even surpassed most of supported metal catalysts and demonstrated a great potential for industrial applications.

Results

Redox and catalytic properties of the MoO_x/Mo₂N catalyst

The MoO₃/Mo₂N heterostructure was prepared by passivating the obtained Mo₂N synthesized from MoO₃ precursor²⁹. As shown in Fig. 1a, only the diffraction peaks of γ-Mo₂N were detected in the X-ray diffraction (XRD) pattern of the fresh and used samples, which indicated the bulk structure of this catalyst was γ-Mo₂N. In contrast to the XRD results, the Raman spectra only showed the signal of MoO₃ at 115, 128, 151, 197, 217, 242, 287, 339, 381, 667, 823 and 996 cm⁻¹, indicating that the surface MoO₃ was formed during the passivation process of γ-Mo₂N in 1% O₂/Ar or exposure to air (Fig. 1b)^31,32,33. And due to the weak Raman signal of γ-Mo₂N, only MoO₃ signal could be detected. Then, in situ Raman measurements were conducted to explore the intrinsic structure and the redox property of the surface MoO₃ in various atmospheres. As illustrated in Fig. 1c, after switching synthetic air to Ar flow at room temperature, the Raman signal of MoO₃ disappeared, and the peaks at 195, 222, 348, 485, 560, 722 cm⁻¹ of MoO_x (x < 3) were detected, which indicated that the MoO₃ transformed into defective MoO_x^31,32. This conversion indicated that oxygen vacancies were formed even under inert gas atmospheres at room temperature. When H₂ was introduced, the MoO_x signal could be detected at low temperatures. However, when the temperature was increased to 500 °C, the MoO_x signal was absent. The MoO_x Raman signal reappeared when the test temperature was returned to room temperature, indicating that the MoO_x passivation layer was still the intrinsic catalytic surface under the high-temperature reducing atmosphere. The absence of MoO_x Raman signal at 500 °C might be caused by the thermal emission that masked the weak vibrational information or the presence of laser that enhanced the reduction of surface oxygen atoms³⁴. After that, the MoO₃ signal was reproduced when the sample was exposed to synthetic air again. And based on the Raman spectra shown in Supplementary Fig. 1a, during above in situ Raman tests, no obvious MoO₂ and Mo₂N species could be detected. These in situ Raman results showed that the surface MoO₃ structure on bulk Mo₂N was highly reactive, which underwent dynamic structural transformations under different atmospheres. Subsequently, we measured temperature-programmed reduction by H₂ (H₂-TPR) to further explore the redox properties of the catalyst. As shown in Fig. 1d, the reduction peaks in H₂-TPR were attributed to the reduction of the Mo-O species on Mo₂N. It was noticed that even at 950 °C, the H₂ consumption was still found, demonstrating that part of the oxygen atoms was strongly coordinated with Mo atoms and were very hard to be removed by H₂. Besides, the observed signals of NH₃ and N₂ in the H₂-TPR could be contributed to the desorption of adsorbed NH₃ during the nitridation process and the denitriding of Mo₂N, respectively. The reduction of Mo-O species in this catalyst was also confirmed by the CO-TPR tests (Supplementary Fig. 1b). Interestingly, compared to MoO₃/Mo₂N, the reference MoO₃ sample was hardly reduced, which required a much higher reduction temperature (Supplementary Fig. 1c‒e). Based on our previous investigations, this improved reducibility of the surface MoO₃ in MoO₃/Mo₂N heterostructure was attributed to the stress between the MoO₃ passivation surface and the bulk Mo₂N²⁹.

Fig. 1: The structural characterization and catalytic property of MoO_x surface on Mo₂N.

a XRD pattern of fresh and used samples. b Raman spectra of fresh and used samples. c The in situ Raman spectra of MoO_x/Mo₂N with consecutive switch of air, Ar and H₂. d H₂-TPR pattern of MoO₃/Mo₂N. e Four-step surface reaction experiment on H₂-pretreated MoO_x/Mo₂N with consecutive switch of Ar, CO₂, Ar and H₂ at 600 °C. f In situ Raman of MoO_x/Mo₂N with consecutive switch of H₂, Ar and CO₂ at 500 °C. g The schematic diagram of proposed reaction pathway and corresponding structural evolution.

In addition, we also quantified the amount of reduced oxygen atoms during the H₂ reduction process through H₂-TPR experiments. Through the quantification of the H₂ consumption peaks from Supplementary Fig. 2a, we obtained that ~1905 µmol/g of oxygen atoms were reduced during the H₂ reduction from room temperature to 600 °C. After that, we utilized the H₂ reduction from room temperature to 950 °C to reduce all oxygen atoms in the catalyst (Supplementary Fig. 2b) and quantify the overall oxygen content of the entire catalyst. By quantifying the H₂ consumption in the H₂-TPR result (Supplementary Fig. 2c and d), the content of oxygen atoms in the sample was ~2844 µmol/g. Based on above results, after the H₂ reduction at 600 °C, ~67.0% of oxygen atoms could be reduced, and ~33.0% of O atoms in the MoO₃ passivation layer was inert. Due to only MoO₃ signals were detected in the Raman spectra of the as-prepared sample, we could assume that almost all Mo-O species on the surface of the as-prepared catalyst existed in the form of MoO₃. Therefore, we could obtain that the average value of x in MoO_x after the H₂ pretreatment was about 1.0. And during the H₂ pretreatment from RT to 600 °C, the range of x over the MoO_x active structure could be evaluated as 1<x < 3. With the increase of the H₂-pretreatment temperature to 600 °C, the value of x could reduce to 1.0.

The reduction of O atoms in the MoO₃ passivation layer is accompanied by the formation of oxygen vacancies. In general, oxygen vacancies can exhibit an enhanced catalytic activity to adsorb and activate oxygen-containing reactant molecules^16,32. CO₂ is a typical oxygen-containing molecule, while the stable C = O bond makes the catalytic conversion of CO₂ very challenging. Herein, the four-step surface reaction experiment with consecutive switch of Ar→CO₂→Ar→H₂ was carried out to explore the absorption and activation of CO₂ on the highly reduced MoO_x surface. Before the surface reaction measurement, the MoO₃/Mo₂N heterostructure was pretreated by 5% H₂/Ar at 600 °C for 1 h, and thus the defective MoO_x surface with rich oxygen vacancies was in situ constructed. As shown in Fig. 1e, for the H₂-pretreated sample, after the introduction of CO₂, CO signal appeared obviously, suggesting the direct conversion of CO₂ to gaseous CO. The gradually weakening CO₂ signal demonstrated that the sites to convert CO₂ into CO gradually decreased. Then, with the introduction of H₂, the observed H₂O signal indicated that CO₂ underwent the direct C = O cleavage on the highly reduced MoO_x surface and H₂ reacted with the oxygen atoms leached from the CO₂ dissociation to form H₂O. This four-step surface reaction experiment was conducted cyclically three times and gave repeatable results, strongly confirming that CO₂ was directly dissociated on the highly reduced MoO_x surface to produce gaseous CO. The accumulated oxygen atoms in the MoO_x surface were cleaned by H₂ to regenerate the consumed active sites. The in situ Raman of MoO_x/Mo₂N with a successive switch of H₂, Ar and CO₂ suggested the dynamical evolution of the oxidation state under H₂ and CO₂ atmospheres (Fig. 1f). Under H₂ flow at 500 °C, no Raman signal of Mo-O species was found at such a test temperature. However, the Raman signal of MoO_x formed with the following introduction of CO₂ flow under the same test temperature, which indicated CO₂ re-oxidized the reduced catalyst surface. This catalyst could reproduce such a structure evolution in three continuous rounds of Raman tests. By combing the results of the surface reaction experiment and in situ Raman, we speculated that the defective MoO_x surface with rich oxygen vacancies was an intrinsic active structure to directly dissociate CO₂ to generate CO. Further, the CO-TPD result indicated that there was no obvious desorbed CO signal from room temperature to 600 °C, suggesting a very weak CO adsorption on the MoO_x/Mo₂N surface (Supplementary Fig. 1f). The rapid desorption of CO prevented the further hydrogenation of the formed CO to CH₄, accelerating the equilibrium shift based on the Le Chatelier’s principle. As a result, the unique MoO_x structure on Mo₂N possessed a favorable microenvironment to catalyze the RWGS reaction. The discussion of the proposed catalytic pathway was concluded in Fig. 1g. First, during H₂ pretreatment, the MoO₃ passivation layer on the bulk Mo₂N was reduced to generate abundant vacancies, and then CO₂ was dissociated on the defective MoO_x surface to form gaseous CO. Subsequently, H₂ timely removed the leached O atoms from the CO₂ dissociation to promote the regeneration of active vacancy sites. Accompanied by the redox cycle of the MoO_x surface, the RWGS reaction proceeded rapidly through the redox mechanism.

The effect of CO₂ on the bulk Mo₂N structure was also clarified in this work. Under CO₂ flow, for H₂-pretreated MoO_x/Mo₂N, with the increase of test temperature, CO signal was detected companied by the CO₂ consumption (Supplementary Fig. 3a). We also found that in the absence of H₂, CO₂ conversion decreased gradually with the extension of the test time at 600 °C, 700 °C and 800 °C, respectively (Supplementary Fig. 3a‒c). For MoO₃/Mo₂N without H₂ pretreatment, CO₂ could also be converted into CO, but at a slower rate (Supplementary Fig 3d). Combined with the in situ XRD pattern of MoO₃/Mo₂N under 2% CO₂/Ar, it could be speculated that CO₂ could oxidize the bulk Mo₂N to MoO₂, and thus forming CO (Supplementary Fig. 3e). However, with the conversion of bulk Mo₂N to MoO₂, the active MoO_x surface could not be regenerated, which explained why the initial intensity of the H₂O signal in Fig. 1e continued to decrease with the increase of the number of the treatment cycles. In brief, during the CO₂ hydrogenation, H₂ was necessary to timely remove oxygen atoms generated from the CO₂ dissociation, thus preventing the over-oxidation of internal Mo₂N to MoO₂. In addition, compared to product CO, the formed H₂O during the reaction was also able to oxidize the bulk Mo₂N structure (Supplementary Fig. 3f and g).

In order to compare the catalytic property of this highly reduced MoO_x with that of conventional reducible metal oxides, we carried out the CO₂ dissociation experiments over the H₂-reduced CeO₂ sample. As a typical reducible metal oxide, CeO₂ has been widely used in various important catalytic reactions^{35,36,37,38,39}. The H₂-TPR result of CeO₂ showed that before 600 °C, CeO₂ underwent obvious reduction (Supplementary Fig. 4a). And the amount of reduced O atoms from room temperature to 600 °C over CeO₂ was quantitatively calculated to be ~748 µmol/g. Then, we used CO₂ flow to treat the CeO₂ after the H₂ pretreatment at 600 °C. Interestingly, based on CO₂ treatment experiments and in situ diffuse reflectance infrared Fourier transform spectroscopy tests (DRIFTS), we found that for H₂-pretreated CeO₂, CO₂ did not dissociate to generate CO, but formed stable carbonates (Supplementary Fig. 4b‒d). Due to the deep black color of MoO₃/Mo₂N, we could not find any adsorption signals of carbon/oxygen-containing intermediates from the DRIFTS (Supplementary Fig. 4e). Nevertheless, the temperature-programed desorption of CO₂ (CO₂-TPD) suggested that compared to H₂-pretreated CeO₂, it was difficult for CO₂ to form carbonate intermediate on the MoO_x/Mo₂N surface (Supplementary Fig. 4f). Therefore, we inferred that the significant difference in the catalytic activity of oxygen vacancies on MoO_x/Mo₂N and CeO₂ leaded to two distinct catalytic mechanisms toward the RWGS reaction. According to whether CO₂ can be dissociated directly into CO, the mechanisms of the RWGS reaction can be divided into redox pathway and associative pathway^12,40. Because the direct cleavage of stable C = O double bond in CO₂ happened on highly reduced MoO_x surface, the catalytic process catalyzed by MoO_x/Mo₂N followed the redox mechanism. On the contrary, for the H₂-reduced CeO₂, the strong adsorption rather than the direct dissociation of CO₂ on CeO₂ was confirmed, suggesting the reaction catalyzed by CeO₂ could only follow the associative mechanism.

Catalytic performance of the MoO_x/Mo₂N catalyst in the RWGS reaction

For the RWGS reaction with harsh reaction conditions, supported catalysts with active metals were susceptible to suffering from deactivation due to the sintering of active metals, although they might exhibit a high initial activity. Owing to the unique redox properties and the absence of supported active metal, the MoO_x/Mo₂N catalyst is expected to achieve both high activity and solid stability for the high-temperature RWGS reaction. As shown in Fig. 2a, under a high gaseous hourly space velocity (GHSV) of 200,000 mL/g_cat/h, the MoO_x/Mo₂N catalyst exhibited very high activity and complete CO selectivity. At 600 °C, the CO₂ conversion was close to the thermodynamic equilibrium limitation. And in six rounds of cyclic stability tests, the catalytic activity was reproduced well, suggesting the durability of this catalyst under the start-up cool-down conditions. In addition, CO selectivity of this catalyst was 100%, no CH₄ was detected throughout the activity evaluation, therefore, this catalyst effectively catalyzed the RWGS reaction rather than methanation (Supplementary Fig. 5a). Compared to H₂-pretreated CeO₂, MoO_x/Mo₂N with the unique redox properties demonstrated a much higher activity. Besides, the apparent activation energy (E_a) of MoO_x/Mo₂N was only 45.0 kJ/mol (Fig. 2b), which was lower than that of CeO₂ (102.1 kJ/mol) and Pt-CeO₂ (55.6 kJ/mol). The low E_a value suggested the RWGS reaction could easily take place on MoO_x/Mo₂N with the redox mechanism. Importantly, even compared to most reference-supported metal catalysts with CeO₂ and Al₂O₃ as supports, MoO_x/Mo₂N showed largely improved catalytic performance (Fig. 2c). And compared to other Mo-based catalysts including MoO₃, Mo₂C and MoS₂, MoO_x/Mo₂N exhibited better catalytic performance (Supplementary Fig. 5c and d). We would like to emphasize that in this work, in order to more accurately compare the activity between the MoO_x/Mo₂N catalyst and other reference catalysts, all the reference catalysts in Fig. 2 and Supplementary Fig. 5 were prepared in-house and their activity tests were measured by using the same reactor under the same test conditions. Among all reference catalysts, Pt/CeO₂ had a higher CO₂ conversion than MoO_x/Mo₂N, while it had a poorer CO selectivity (Supplementary Fig. 5b). It was worth noting that with the weakening of external diffusion limitation, the CO formation yield of MoO_x/Mo₂N could reach 159.6×10⁻⁵mol/g_cat/s at a very high GHSV of 3,800,000 mL/g_cat/h (Supplementary Fig. 5e), exceeding almost all reported catalysts in previous literatures (Supplementary Table 1).

Fig. 2: The catalytic performance of MoO_x/Mo₂N for CO₂ hydrogenation to CO.

a CO₂ conversion over MoO_x/Mo₂N in the RWGS reaction for six start-up cool-down cycles (Test conditions: 0.1 MPa, 300 °C‒600 °C, GHSV = 200,000 mL/g_cat/h). b Apparent activation energy (E_a) of MoO_x/Mo₂N, Pt/CeO₂ and CeO₂ catalysts. c Comparison of CO₂ conversion over MoO_x/Mo₂N with other reference catalysts. The reference catalysts were synthesized by ourselves in our lab, and the corresponding catalytic tests were carried out on same reactor. d Three rounds of stability tests over MoO_x/Mo₂N under harsh reaction conditions with 600 °C and a high GHSV of 200,000 mL/g/h (the time interval between two rounds of testing was 55 days).

In order to assess the durability of the MoO_x/Mo₂N catalyst in the high-temperature RWGS reaction, three rounds of stability tests were measured at 600 °C. The duration of each test was 300 h, and the time interval between two rounds of testing was 55 days. As shown in Fig. 2d, this catalyst maintained stable CO₂ conversion and CO selectivity over 900 h with a CO₂ conversion more than 45% and a complete CO selectivity. Importantly, even after 55 days of exposure to air, the active sites of the used sample were not deactivated by air corrosion, suggesting a superior recyclability. Meanwhile, compared to other reported catalysts, this MoO_x/Mo₂N catalyst without any supported active metals could effectively avoid the deactivation caused by the sintering of active metals, suggesting a great potential in practical application (Supplementary Fig. 5f). In addition to the atmospheric reaction condition, this catalyst maintained its activity well in a 20-hour stability test at enhanced reaction pressures of 2.0 Mpa (400 °C) and 1.0 Mpa (600 °C) (Supplementary Fig. 5g and h).

In order to investigate the relationship between the catalytic activity and the H₂/CO₂ ratio, we further supplemented the activity tests of the catalyst under other H₂/CO₂ ratios (2:1 and 1:1). As shown in Supplementary Fig. 6a–d, under the gas flow with the H₂/CO₂ ratio of 1:1 and 2:1, the CO₂ conversion of the catalyst was much lower than the equilibrium line, which suggested that the catalytic activity was poor at a low H₂ proportion. The XRD results of the used catalysts showed distinct diffraction peaks of MoO₂ but no Mo₂N (Supplementary Fig. 6e), indicating that during the reaction, bulk Mo₂N was oxidized by CO₂. Compared to the XRD result of the used catalyst under a H₂/CO₂ ratio of 3:1 (Fig. 1a), above XRD results confirmed that the catalytic performance and the redox behavior of MoO_x was largely determined by the H₂/CO₂ ratio. A high H₂ proportion could guaranteed the reversible redox of MoO_x during the reaction, but the low H₂ proportion could not protect the bulk Mo₂N from the over-oxidation by CO₂, which led to the irreversible redox of the MoO_x structure and the decrease of the catalytic performance.

Surface and bulk structure of the MoO_x/Mo₂N catalyst

As shown in the high-resolution transmission electron microscopy (HRTEM) images (Supplementary Fig. 7a‒c), the fresh catalyst showed a sheet-like morphology consisting of Mo₂N particles, and no information of MoO₃ passivation surface could be detected. After the high-temperature RWGS reaction, this catalyst could remain the sheet-like morphology with that of the fresh sample (Supplementary Fig. 7d). And as illustrated from the high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images of the fresh and used catalysts (Fig. 3a‒c and Supplementary Fig. 7e and g), the information of MoO₃ component was still absent. However, after the RWGS reaction, a disordered surface layer with a thickness of less than 1 nm was clearly observed, which could be regarded as the pivotal surface structure to form oxygen vacancy-rich subnano MoO_x layer to activate CO₂ during CO₂ hydrogenation. Meanwhile, the surface Mo-O species could be verified by the elemental mapping, with O signal was highly dispersed on Mo₂N surface (Fig. 3d and Supplementary Fig. 7f and h). Besides, although O element signal was weak, STEM-EDS line scanning of the fresh and used samples indicated a decrease in the atomic fraction of O from the surface to the bulk, which also demonstrated the presence of surface Mo-O species (Fig. 3e, f and Supplementary Fig. 8a‒f). Scanning Electron Microscope (SEM) images also suggested that the fresh catalyst showed a stacked sheet-like structure, and the nanosheets were made up of numerous nanoparticles (Supplementary Fig. 8g). After the RWGS reaction, the morphology of the catalyst maintained stable (Supplementary Fig. 8h). The X-ray photoelectron spectroscopy (XPS) over the fresh and used catalysts in Mo 3d, O 1 s, and N 1 s regions clearly illustrated the reduction of surface MoO₃ and the denitrification of Mo₂N. As shown in Supplementary Fig. 9a and Supplementary Table 2, the Mo 3d XPS spectra contained mixed components with contributions from Mo⁶⁺, Mo⁵⁺, Mo⁴⁺ and Mo-N²⁵. The appearance of Mo⁶⁺ confirmed the presence of the surface MoO₃. Furthermore, the Mo⁶⁺ signal over the used sample was much weaker than that of the fresh sample, indicating the reduction of MoO₃ during the RWGS reaction. By combining the in situ Raman spectra in Fig. 1c, it could be confirmed that when the used sample was exposed to air, a part of reduced Mo atoms could be re-oxidized to MoO₃. The reduction of MoO₃ was also confirmed by the O 1 s XPS spectra with much-decreased lattice oxygen after the reaction (Supplementary Fig. 9b). And the significant increase of surface oxygen species for the used catalyst demonstrated the created oxygen vacancies during the surface MoO₃ reduction favored the adsorption of oxygen-contained molecules in air, such as H₂O and CO₂ (Supplementary Fig. 9b and f).

Fig. 3: The identification of surface and bulk structure over MoO_x/Mo₂N.

a‒c HAADF-STEM of spent MoO_x/Mo₂N after the RWGS reaction. d STEM and element mapping of Mo, N and O of spent MoO_x/Mo₂N after the RWGS reaction. e HAADF-STEM image of the used sample and f, STEM-EDS elemental line scanning of O element over the highlighted area in e. g Ex situ and quasi in situ Mo 3d XPS for fresh MoO_x/Mo₂N and MoO_x/Mo₂N after H₂ pretreatment and high-temperature RWGS reaction at 600 °C. h In situ Raman spectra of MoO_x/Mo₂N under the RWGS reaction atmosphere at various temperatures. i In situ XRD pattern of MoO_x/Mo₂N under 5% H₂/Ar. j The scheme of surface evolution over MoO₃/Mo₂N under different atmospheres and temperatures.

Then, in order to identify the oxidation states of surface Mo species in the H₂ pretreatment and the RWGS reaction, the quasi in situ XPS was measured. As shown in Fig. 3g, high-temperature H₂ pretreatment led to the reduction of MoO₃, and although the RWGS reaction gas contained 23% CO₂, the oxidation states of Mo remained relative durable in the actual reaction process. Combined with the results of four-step surface reaction experiment and in situ Raman (Fig. 1c, d), the stable oxidation states of Mo were owing to the timely removal of the generated oxygen atoms from the CO₂ dissociation by H₂. Compared to the used sample with abundant adsorbed oxygen species (Supplementary Fig. 9b), the limited surface oxygen species of samples after the in situ H₂ pretreatment and the RWGS reaction was related to the quick CO₂ dissociation, CO desorption and H₂O formation (Supplementary Fig. 9d and f). Meanwhile, the catalysts that underwent the H₂ pretreatment and the reaction process possessed a weaker N 1 s signal compared to the fresh sample²⁵, suggesting the denitrification of Mo₂N and the formation of nitrogen vacancies (Supplementary Fig. 9c and e). We speculated that in the reaction process, nitrogen vacancies were likely to be occupied by oxygen atoms from CO₂, and then converted to oxygen vacancies by the further H₂ reduction. After that, the surface and subsurface structure information over the catalyst after the long-term RWGS reaction (70 h) was also investigated. As shown in Supplementary Fig. 10a, the catalyst suffered from partial carbonization during the RWGS reaction, with the observation of the Mo-C signal. While, the largely increased Mo nitride signal after the Ar ion etching indicated the subsurface of the catalyst had more Mo-N species^24,25, suggesting that the bulk Mo-N structure was relative stable under the CO₂ hydrogenation process (Supplementary Fig. 10b). Besides, the Mo⁶⁺ content on the catalyst surface was higher than that in the catalyst subsurface, again confirming the presence of the MoO₃ passivation layer on bulk Mo₂N (Supplementary Fig. 10c).

The in situ Raman was carried out to further explore the intrinsic surface structure during the actual reaction process. Under the reaction atmosphere, the MoO_x (x < 3) signal appeared at room temperature and maintained until 500 °C (Fig. 3h). Based on the quasi in situ XPS spectra of Mo 3d and O 1 s, as well as the Raman spectra after the sample was cooled down to room temperature, the absence of MoO_x signal in the Raman spectra at 500 °C did not imply the complete removal of surface Mo-O species, but indicated the masked vibrational information caused by the high test temperatures. Importantly, although the MoO_x surface structure possessed extremely active redox properties, the bulk Mo₂N exhibited a solid stability under reductive atmosphere. As illustrated in Fig. 3i and Supplementary Fig. 9g, under 5% H₂/Ar and the RWGS reaction flow at 600 °C for 8 and 6 h, respectively, only diffraction peaks of Mo₂N could be detected, suggesting a high stability of the bulk-phase Mo₂N. The surface structural evolution over MoO₃/Mo₂N under different atmospheres and temperatures was illustrated in Fig. 3j. In summary, for the MoO₃/Mo₂N heterostructure, the subnano MoO₃ passivation layer inhibited the overoxidation of the internal Mo₂N in air, guaranteeing the excellent recyclability of this catalyst. In addition, under the reductive reaction atmosphere, the bulk Mo₂N not only endowed with subnano MoO₃ surface active redox properties but also prevented the collapse of the active surface.

CO₂ activation mechanism and RWGS reaction pathway

By utilizing a series of in situ characterization techniques, the presence of rich and efficient oxygen vacancies on the highly reduced subnano MoO_x surface during the reaction process had been confirmed. The density functional theory (DFT) calculations were further performed to explore the crucial role of surface oxygen vacancies in CO₂ activation. Based on characterization results, a model of the MoO₃/Mo₂N heterostructure was constructed (Supplementary Fig. 11a, b). In addition, the models of MoO_x/Mo₂N with different oxygen defect concentrations were shown in Supplementary Fig. 11c. We thoroughly examined both physical and chemical adsorption configurations of CO₂ on these models. For the 2Ov and 4Ov models, CO₂ molecules adsorb at isolated oxygen vacancies, primarily through weaker physical adsorption, with adsorption energies less than −0.1 eV. Chemical adsorption at these sites was energetically unfavorable, showing positive adsorption energies. Conversely, for the 6Ov, 8Ov, and 14Ov models, CO₂ chemically adsorbs at sites with adjacent vacancies, resulting in significantly more negative adsorption energies, indicating stronger bonding (Fig. 4a and Supplementary Fig. 12). In this process, oxygen and carbon atoms from CO₂ bonded with coordination-unsaturated Mo sites. An increase in oxygen vacancy density on the catalyst surface could effectively enhance the adsorption energy of CO₂. According to H₂-TPR, CO-TPR and in situ Raman results, we obtained MoO_x/Mo₂N catalysts with varying oxygen vacancy concentrations by regulating H₂-pretreated temperatures. By analysing the O 1 s XPS spectra of the related samples, we found that the oxygen vacancy concentration was positively correlated with the pretreatment temperatures under 5% H₂/Ar (Supplementary Fig. 13a, b). And by combining these findings with activity evaluations, we speculated that the increased oxygen vacancy concentration led to improved catalytic activity (Fig. 4b).

Fig. 4: Role of the surface oxygen vacancy on MoO_x surface in catalyzing the CO₂ hydrogenation to CO.

a The adsorption energy of CO₂ on MoO₃/Mo₂N and MoO_x/Mo₂N with various vacancy densities. b The CO₂ conversion of MoO_x/Mo₂N at 400 °C after different H₂-pretreated temperatures. c Energy diagram for CO₂ hydrogenation and corresponding configurations of optimized intermediates and transition states.

In addition, the CO₂ activation also relies on the electron transfer between the chemically adsorbed CO₂ molecule and the catalyst surface^2,15,41. The injection of electrons into the anti-bonding orbitals of the CO₂ molecule directly determines the activation of C = O bonds. Projected electronic density of states (Supplementary Fig. 14) demonstrated that CO₂ had effective interactions with the defective MoO_x catalytic surface. Moreover, as shown in Supplementary Fig. 15a of the differential charge density, due to the high electron-donating capacity of the oxygen-rich surface, CO₂ molecule acquired electrons from the surrounding coordination-unsaturated Mo atoms, suggesting the effective CO₂ activation on catalyst surface. And the integration of Δρ further confirmed the strong electronic interaction between the adsorbed CO₂ and the MoO_x surface with rich oxygen vacancies (Supplementary Fig. 15b). Based on DFT simulations, the differences in CO₂ activation mechanisms on MoO_x/Mo₂N and the traditional reducible oxide of CeO₂ were also investigated. For MoO_x/Mo₂N, we chose MoO_x (8O_v)/Mo₂N as the initial catalyst model. As illustrated in Supplementary Fig. 16, on the H₂-reduced CeO₂ surface, the chemisorbed CO₂ molecule on oxygen vacancies tended to react with another lattice oxygen atom to form carbonates rather than directly dissociating into CO. This was consistent with the observation of carbonate signals and the absence of gaseous CO in the in situ DRIFTS (Supplementary Fig. 4c, d). However, compared to CeO₂, the adsorption configuration of CO₂ on MoO_x/Mo₂N surface was quite different, in which two oxygen atoms and one carbon atom from CO₂ could be activated simultaneously by forming chemical bonds with two adjacent coordination-unsaturated Mo atoms on the catalyst surface. Therefore, CO₂ could be effectively activated through the electron transfer mediated by more than one exposed Mo site. This efficient activation of CO₂ facilitated the direct breaking of the stable C = O double bond to form *CO and *O with a much lower energy barrier of 1.50 eV (Supplementary Fig. 16a), which could be easily overcome at reaction temperature. Conversely, the formation of *CO₃ on the MoO_x surface was more difficult, with a relatively higher energy barrier of 1.77 eV. These results indicated that, compared to the reduced CeO₂ surface, the subnano MoO_x passivation surface with high-density oxygen vacancies possessed much improved ability to directly dissociate CO₂ (Supplementary Fig. 16b, c).

To understand the underlying reaction pathway of CO₂ hydrogenation to CO on the MoO_x surface, we conducted the further DFT calculation (Fig. 4c). CO₂ was first captured by an oxygen vacancy, which was a prerequisite for CO₂ hydrogenation. The subsequent dissociation of CO₂ to *CO and *O was the critical step for the catalyst to catalyze CO₂ hydrogenation to CO, which had an energy barrier of 1.5 eV. Then the formed *CO left the catalyst surface to generate gaseous CO. Subsequently, the *O left on the catalyst surface reacted with a dissociated H atom to form *OH, which followed by the reaction with another dissociated H atom to produce H₂O. We calculated the energy barriers for the transition states of dissociative H₂ adsorption (H₂ + 2* → 2*H) and H₂O formation (*H + *OH → H₂O + 2*), which were 0.97 eV and 1.23 eV at 0 K, respectively. These barriers were comparatively lower than the CO₂ dissociation barrier (*CO₂ → *CO + *O), which is 1.5 eV, suggesting that CO₂ dissociation was indeed the rate-limiting step in this reaction over MoO_x/Mo₂N catalysts. With H₂O desorption, the consumed reactive oxygen vacancy was regenerated, and the catalyst structure returned to its initial state. Our calculations also showed that raising reaction temperature from 0 K to 600 K could effectively accelerate the desorption of the formed intermediates, especially for *CO, preventing the poisoning of active oxygen vacancies.

Discussion

The catalytic conversion of CO₂ is largely determined by the CO₂ activation. In this work, by utilizing the active redox property of MoO₃ passivation layer on Mo₂N, we constructed high-performance MoO_x surface with rich and efficient oxygen vacancies, which achieved the direct cleavage of C = O double bonds over CO₂. The high-density oxygen vacancies with abundant coordination-unsaturated Mo sites provided a favorable spatial and electronic environment, largely promoting the CO₂ activation via the electron transfer mediated by exposed Mo sites. The reversible switch of different Mo oxidation states on MoO_x surface allowed the cyclic consumed and regenerated oxygen vacancies to serve as independent active sites to complete the whole CO₂ hydrogenation process. In addition, by using comprehensive in situ characterizations, the intrinsic active structure of Mo₂N-based catalytic materials was deeply uncovered. The MoO₃ passivation surface on Mo₂N exhibited unusual catalytic properties, although it has been ignored in almost all other reports. These findings in this work showed that as a heterostructure with unique surface redox properties, MoO_x/Mo₂N possessed huge application potential in various heterogeneous reactions. This work also provided a well reference to develop effective defect engineering in designing high-performance catalysts without relying on supported active metals.

Methods

Synthesis of MoO₃

The MoO₃ precursor was prepared by hydrothermal method. Firstly, 1.0 g of triblock copolymer (P123) was dispersed in deionized water (40 mL) by stirring for 1 h. Then, 0.9 g of Na₂MoO₄·9H₂O was dissolved in deionized water (5 mL) and transferred into the obtained P123 solution. After stirring for 30 min, 3 mL of HCl (37 wt.%) was added. Subsequently, the obtained stock solution was transferred into a Teflon bottle and then tightly sealed in a stainless-steel autoclave. The hydrothermal reaction was carried out at 100 °C for 12 h in the oven. After that, the produced solid was washed by deionized water and ethanol followed by drying at 70 °C for 10 h. Finally, the above solid was calcined in a tube furnace at 400 °C for 4 h.

Synthesis of MoO₃/Mo₂N

400 mg of MoO₃ was loaded in a quartz tube and placed in a horizontal tube furnace. Then, the MoO₃ was heated under pure NH₃ with a gas flow of 40 mL/min at 650 °C (5 °C/min) for 4 h. And then, after complete cooling to room temperature, the 1% O₂/Ar (30 mL/min) was switched to passivate the obtained black powder for 2 h.

In this work, the as-prepared MoO₃/Mo₂N catalyst with MoO₃ passivation layer was denoted as MoO₃/Mo₂N. And due to the MoO₃ passivation layer on Mo₂N could be easily reduced under H₂ and RWGS reaction flow, the catalyst underwent Ar, H₂, or RWGS reaction flow treatment was denoted as MoO_x/Mo₂N catalyst (1<x < 3).

Catalyst characterization

X-ray diffraction (XRD)

The XRD pattern was measured on PANalytical X’pert3 powder diffractometer (40 kV, 40 mA) using Cu Ka radiation. Appropriate mass of powder sample was placed in a quartz-glass holder for the ex situ XRD measurement. For in situ XRD, appropriate mass of powder sample was sealed in the in situ chamber. And in situ XRD tests of the following two programs were carried out.

(1) The oxidation of the bulk Mo₂N over MoO₃/Mo₂N under CO₂ flow. The sample was purged by 2% CO₂/Ar at room temperature to remove the residual air and then heated to 600 °C with a heating rate of 10 °C/min. The XRD pattern was measured at room temperature (25 °C), 100 °C, 300 °C and 600 °C after stabilizing at the set temperature for 30 min. Then, XRD test was carried out every 1.5 h at 600 °C.

(2) The stable bulk Mo₂N over MoO₃/Mo₂N under H₂ flow. The sample was purged by 5% H₂/Ar at room temperature to remove the air in sample chamber and then heated to 600 °C (10 °C/min). The XRD pattern was obtained every 1.5 h at 600 °C under 5% H₂/Ar. For RWGS reaction conditions, the sample was purged by 30% H₂, 10% CO₂/N₂ at room temperature and then heated to 600 °C (10 °C/min).

Raman

All ex situ and in situ Raman spectra were collected by a LabRAM HR800 spectrometer (HORIBA JY) with excited laser at 473 nm. Approximately 10 mg of catalyst was used in the in situ Raman testing. The Raman reaction cell (Xiamen TOPS) equipped with a quartz window and a heating module that allowed samples to be heated controllably. A manual six-way valve was used to rapidly switch the gases without dead volume. In addition, recycled ethanol was used to cool down the cell during the high-temperature testing. For MoO₃ sample, 5% H₂/Ar was injected at room temperature, and then the sample was heated from room temperature to 500 °C. The corresponding Raman spectra were measured at 100 °C, 300 °C and 500 °C, respectively. For MoO_x/Mo₂N, the in situ Raman spectra were obtained by using the following programs.

(1) The Raman spectra were collected under air. Then, pure Ar was injected at room temperature to remove the air, and Raman spectra were measured. After that, the gas was switched into 5% H₂/Ar, and the sample was heated to 500 °C from room temperature with a ramping rate of 10 °C/min. Raman spectra were collected under Ar flow after the H₂ treatment at room temperature, 100 °C, 200 °C, 300 °C, 400 °C, 500 °C and cooling down to room temperature, respectively. Then, after cooling to room temperature, Raman spectrum was measured after the exposure of air.

(2) The MoO_x/Mo₂N sample was treated by 23% CO₂/69% H₂/Ar from room temperature to 550 °C (10 °C/min). Raman spectra were collected in Ar flow after the treatment under the RWGS reaction flow at room temperature, 100 °C, 200 °C, 300 °C, 400 °C, 500 °C, 550 °C, respectively. Then after cooling down to room temperature, the Raman spectra was measured again.

(3) The MoO_x/Mo₂N sample was heated from room temperature to 500 °C under 5% H₂/Ar. At 500 °C, after H₂ treatment for 15 min, pure Ar was injected for another 15 min, and Raman spectra were collected under Ar flow. Then, the gas was switched from Ar to 2% CO₂/Ar for 15 min, and Raman spectra were measured. Above test program was cycled for three times.

X-ray photoelectron spectroscopy (XPS)

All XPS experiments were measured on a Thermo Scientific ESCALAB Xi+ XPS instrument. The C 1 s line located at 284.8 eV was used to calibrate each spectrum for accurate binding energies. For the ex situ XPS measurement of the used catalysts after different pretreatment temperatures, the sample was pretreated at 400 °C, 500 °C and 600 °C under 5% H₂/Ar for 1 h, respectively, and then followed one-hour RWGS reaction at 400 °C. The reaction at 400 °C was chosen to avoid that at higher temperature, the RWGS reaction atmosphere would further reduce Mo-O species on the catalyst surface, which could mask the effect of the pretreatment temperature on oxygen vacancy concentration. For the XPS etching experience, the ion energy of Ar ion was set to 2000 eV. For the quasi in situ XPS measurements, the samples after the treatments at predetermined temperature and gas flow were transferred from the pretreatment chamber to the test room under the vacuum without exposure to air. The quasi in situ XPS measurements were carried out by the following three programs.

(1) The spectrums of C 1 s, O 1 s, N 1 s, Mo 3d were measured after the sample was treated by 5% H₂/Ar at 600 °C for 1 h.

(2) The spectrums of C 1 s, O 1 s, N 1 s, Mo 3d were obtained after the sample was treated by 23% CO₂/69% H₂/N₂ at 600 °C for 5 h.

Transmission electron microscopy (TEM)

The TEM was conducted on a JEOL JEM-1011 instrument operating at 100 kV. The powder sample was dispersed in ethanol under ultrasound. And then, the sample was dropped at the carbon-coated Cu grid before test. The high resolution transmission electron microscopy (HRTEM) was measured at JEM-2100. The High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images were acquired at Spectra 300 (Thermofisher). The STEM-EDS elemental line scanning and STEM-EDS elemental area mapping element mapping were obtained from the same equipment under the STEM mode.

Temperature program reduction by hydrogen (H₂-TPR)

For MoO₃ sample, H₂-TPR was carried out on Builder PCSA-1000 instrument with a thermal conductivity detector (TCD) to test the H₂ consumption. The sample was first pretreated by pure air at 300 °C for 30 min with a ramping rate of 10 °C /min. After cooling at room temperature, the flow was switched to 5% H₂/Ar, and the H₂ consumption was measured under 5% H₂/Ar with a ramping rate of 10 °C/min from room temperature to 950 °C. For MoO₃/Mo₂N sample, the H₂-TPR was obtained on a mass spectrometer. In order to prevent the bulk Mo₂N from oxidation by air, the sample was pretreated by pure Ar at 300 °C for 30 min. And after cooling to room temperature, the H₂-TPR was acquired from room temperature to 950 °C (10 °C/min). For the quantification of reduced oxygen atoms during the H₂ pretreatment process, 50 mg of MoO₃/Mo₂N was pretreated at 600 °C (10 °C/min) under pure Ar flow for 20 min to exclude the disruption of the desorption signals of NH₃ and N₂, and after cooling to room temperature, H₂-TPR was acquired from room temperature to 600 °C (10 °C/min) and maintained at 600 °C for 1500 s. For the quantification of all oxygen atoms in MoO₃/Mo₂N, 150 mg of MoO₃/Mo₂N was pretreated at 950 °C (10 °C/min) under pure Ar flow for 20 min to exclude the disruption of the desorption signals of NH₃ and N₂, and after cooling to room temperature, H₂-TPR was acquired from room temperature to 950 °C (10 °C/min) and maintained at 950 °C for 1500 s. And during this process, in order to exclude the influence of denitriding and the formation of NH₃ at high temperature on the TCD signal, the products generated from the H₂-TPR was further detected by mass spectrometer.

Temperature program reduction by CO (CO-TPR)

The mass spectrometer was used to analyze the outlet gases for the CO-TPR. After pretreatment of MoO_x/Mo₂N by pure Ar at 300 °C for 30 min, the CO-TPR was measured under 2% CO/Ar from room temperature to 950 °C.

Temperature program desorption by CO (CO-TPD)

The mass spectrometer was used to analyze the outlet gases for the CO-TPD. Firstly, 50 mg of MoO₃/Mo₂N was pretreated under 5% H₂/Ar at 600 °C for 1 h. After cooling to room temperature, the sample was treated by 2% CO/Ar (30 mL/min) for 30 min followed by purging with Ar flow (30 mL/min) for 30 min to remove all physically adsorbed CO molecules on catalyst surface. And then, the CO-TPD measurement was carried out from room temperature to 600 °C with a ramping rate of 10 °C/min. The desorption of CO was detected online by a mass spectrum.

CO₂ treatment experiment

The CO₂ treatment experiment was performed on a lab-made reactor and the outlet gases were further analyzed by mass spectrum and an infrared gas analyzer (Gasboard-3100, Wuhan Sifang Corp.). The equipped infrared gas analyzer was used to obtain the CO₂ conversion rates. For MoO₃/Mo₂N and CeO₂, before CO₂ treatment, the samples were pretreated by 5% H₂/Ar at 600 °C for 1 h. After cooling to room temperature, the samples were purged by pure Ar to clear the residual H₂. Then, 2% CO₂/Ar was switched to reactor and then the samples were heated from room temperature to 800 °C. The test temperatures were maintained at 600 °C, 700 °C and 800 °C for 1 h, respectively. The CO₂ treatment experiment of non-pretreated MoO₃/Mo₂N was also operated by using the same test program. For 10 h of CO₂ treatment over H₂-pretreated MoO_x/Mo₂N, the 2% CO₂/Ar was switched to 5% H₂/Ar at 600 °C and 700 °C, respectively, and the outlet components were analyzed by a mass spectrum and an infrared gas analyzer.

Four-step surface reaction experiment

A mass spectrometer was used to analyze the gaseous products online during the surface reaction. 100 mg of sample was pretreated at 600 °C for 1 h under 5% H₂/Ar (30 mL/min) with a ramping rate of 10 °C/min. After that, the sample was purged by Ar, followed by the consecutive switch of 2% CO₂/Ar, Ar, 5% H₂/Ar and Ar for three times. All the spectrometer signal was measured at the temperature of 600 °C.

In situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) spectra

The in situ DRIFTS spectra of the CeO₂ sample were collected by using a Bruker Vertex 70 FTIR spectrometer. An acquisition time of 30 s at a resolution of 4 cm⁻¹ was used for the spectrum collection. The DRIFTS cell (Harrick) was equipped with CaF₂ windows and a heating cartridge that allowed samples to be heated. A quick switching system was used to rapidly change the gases without a dead volume through a six-way valve. Prior to the in situ DRIFTS measurement, ~20 mg sample was pretreated at 600 °C for 60 min under 5% H₂/Ar mixed gas. And then, the background spectra were collected under Ar flow at 600 °C. After the collection of the background spectrum, the mixed gas of 2% CO₂/Ar was introduced into the chamber. Continuous recording of the IR profiles was maintained for 3 min. All DRIFTS results were analyzed by using OPUS software.

Activity evaluation and kinetics measurement

A fixed-bed reactor was used for all RWGS reaction tests. For activity tests, 15 mg catalyst (20–40 mesh) was loaded in quartz tube and pretreated at appropriate temperature under 5% H₂/Ar for 1 h. After the sample was cooled to room temperature, 23% CO₂/69% H₂/Ar at 50 mL/min was switched to the reactor. The catalytic activity was measured at 300 °C, 400 °C, 500 °C and 600 °C, respectively. At each test temperature, the products were analyzed after 1 h of steady-state reaction by an on-line gas chromatograph (GC-2014C, Shimadzu) equipped with a TCD detector. For stability test, for the first round test, the sample was pretreated under 5% H₂/Ar at 600 °C for 1 h. After one round of stability test, the reaction tube containing the sample was removed from the heating furnace. Subsequently, after the sample has been exposed to air for 55 days, the reaction tube with sample was placed in the reactor again, and other stability tests were performed without any pretreatment. CO₂ conversion and CO selectivity were calculated using the following equations:

$${X}_{{\text{CO}}_{2}}=\frac{{n}_{{\text{CO}}_{2}}^{\text{in}}-{n}_{{\text{CO}}_{2}}^{\text{out}}}{{n}_{{\text{CO}}_{2}}^{\text{in}}}\times 100\%\,$$

(1)

$${S}_{\text{CO}}=\frac{{\text{n}}_{\text{CO}}^{\text{out}}}{{n}_{\text{CO}+}^{\text{out}}{n}_{{\text{CH}}_{4}}^{\text{out}}}\times 100\,$$

(2)

where ${n}_{{\text{CO}}_{2}}^{\text{in}}$ is the CO₂ concentration at inlet and ${n}_{{\text{CO}}_{2}}^{\text{out}}$, ${n}_{\text{CO}}^{\text{out}}$, ${n}_{{\text{CH}}_{4}}^{\text{out}}$ are the concentrations of CO, CO₂, CH₄ in the outlet.

In the tests of apparent activation energy (E_a) over measured catalysts, the CO₂ conversion was controlled within 15% by regulating the mass of catalysts, test temperature and flow rates. The CO yields at 600 °C under different weight hourly space velocities were measured by using 3 mg of catalyst, and the corresponding detailed parameters were given in Supplementary Table 1. The reaction rates of measured catalysts were calculated by using the following equation:

$$\,r=F\times {{Conv}\cdot }_{\text{CO}2}{\div}W\,$$

(3)

where $r$ is the RWGS reaction rate by means of CO₂ (mol_CO2/g_cat/s). F is the total flow rate of CO₂ in reaction stream (mol/s). ${{Conv}.}_{\text{CO}2}$ is the concentration of converted CO₂ (%). $W$ is the mass of the catalyst (g). During the RWGS reaction from 300 °C to 600 °C, the carbon balance exceeded 97.5%.