Stabilizing Co₂C with H₂O and K promoter for CO₂ hydrogenation to C₂₊ hydrocarbons

Abstract

The decomposition of cobalt carbide (Co₂C) to metallic cobalt in CO₂ hydrogenation results in a notable drop in the selectivity of valued C₂₊ products, and the stabilization of Co₂C remains a grand challenge. Here, we report an in situ synthesized K-Co₂C catalyst, and the selectivity of C₂₊ hydrocarbons in CO₂ hydrogenation achieves 67.3% at 300°C, 3.0 MPa. Experimental and theoretical results elucidate that CoO transforms to Co₂C in the reaction, while the stabilization of Co₂C is dependent on the reaction atmosphere and the K promoter. During the carburization, the K promoter and H₂O jointly assist in the formation of surface C^* species via the carboxylate intermediate, while the adsorption of C^* on CoO is enhanced by the K promoter. The lifetime of the K-Co₂C is further prolonged from 35 hours to over 200 hours by co-feeding H₂O. This work provides a fundamental understanding toward the role of H₂O in Co₂C chemistry, as well as the potential of extending its application in other reactions.

A Co₂C catalyst is in situ synthesized and stabilized by H₂O and K promoter for CO₂ hydrogenation to C₂₊ hydrocarbons.

INTRODUCTION

The continuously increasing greenhouse gas CO₂ levels originated from human activities have caused a series of environmental issues, including sea level rises, heat waves, and ocean acidification (1). Among the strategies considered, catalytic CO₂ reduction with green H₂, which is generated from water splitting by renewable energy, to valuable C₂₊ hydrocarbons (HCs) provides a potential technology to reduce CO₂ concentration and an alternative nonpetroleum-based production route toward light olefins (C₂ ~ C₄⁼) and liquid fuels (C₅₊) (2, 3). However, the chemical inertness of CO₂ molecules (C═O bond of 750 kJ mol⁻¹) and the high energy barrier of C─C coupling are detrimental to the formation of C₂₊ products (4, 5). Accordingly, it is highly attractive and challenging to develop effective catalysts for selective CO₂ conversion to C₂₊ HCs.

Much attention has been paid to using CO Fischer-Tropsch synthesis (CO-FTS) catalysts for CO₂ hydrogenation, and, in general, CO₂ is initially converted to CO intermediate via a reverse water gas shift (RWGS) reaction, followed by CO hydrogenation to C₂₊ HCs (6). For instance, iron carbide (FeC_x) exhibits high activity for CO_x hydrogenation to olefins in a high-temperature range (320° to 350°C) (7–9). Cobalt carbide (Co₂C) is also a promising catalyst for CO_x conversion owing to its adequate activation of C═O bonds and promoting effect on C─C coupling (10). Notably, it has distinct features in different reaction atmospheres. As for CO-FTS, nanoprism Co₂C with preferentially exposed (101) and (020) facets has been proven to be responsible for the synthesis of C₂ ~ C₄⁼, while electronic (alkali metal) and structural promoters (Mn and Ce) are instrumental in its morphological control (11–15). The synthesis of Co₂C is commonly divided into two steps, including an initial reduction and carburization in syngas (CO and H₂) or CO. Wavelet transform and linear combination fitting results of in situ x-ray absorption spectroscopy (XAS) indicate that cobalt oxide (CoO) is first reduced to metallic cobalt (Co⁰), and then carburized to Co₂C (16). Moreover, Co₂C is commonly used below 260°C and at near atmospheric pressure in CO-FTS to avoid its decomposition to Co⁰ and graphite (17). Despite numerous efforts, Co₂C is unstable and still underutilized for CO₂ hydrogenation (18, 19). Compared with CO-FTS, the conversion of equimolar CO₂ consumes more H₂, and the presence of excess H₂ and CO₂ accelerates the decomposition of presynthesized Co₂C (18). Besides, the activation of inert CO₂ often requires a high temperature, and it thus puts forward higher requirements for the thermal stability of Co₂C. As a result, the Co₂C catalyst tends to partially decompose to Co⁰ under CO₂ hydrogenation, which leads to a rapid side reaction toward CH₄ (20). In recent studies, SiO₂ support interacted with Co₂C was used for improving the stability, but a reconstruction to Co⁰ occurred while C₁ products dominate under reaction conditions (21). Co₂C with different morphologies was prepared from the ZIF-67 precursor by Zhang et al. (22) for RWGS reaction, but the atmospheric pressure condition only enables a reduction of CO₂ to CO and limits the selectivity to C₂₊ HCs. The facile synthesis and stabilization of Co₂C under a C-lean and H-rich atmosphere at higher temperatures and pressure are major bottleneck for its catalytic application in CO₂ hydrogenation to C₂₊ HCs.

H₂O is a common impurity for CO₂ capture from flue gas (23), and more H₂O is generated in CO₂-FTS compared with CO-FTS; therefore, its impact on the catalyst structure and performance needs to be pinpointed. As for Fe-based catalysts, H₂O-induced oxidation causes the evolution of FeC_x to iron oxides (FeO_x) and a decrease of the activity in both CO- and CO₂-FTS (8, 24). H₂O has also been proven as a poison owing to its competitive adsorption and oxidation of Co⁰ catalyst in CO-FTS (25–27). Thus, the rapid removal of H₂O boosts the syngas conversion to C₂₊ HCs (28). However, up to now, the effect of H₂O on Co₂C catalyst is still ambiguous in CO₂ hydrogenation, clarification, and optimization which is of importance to improving the practical application.

Here, we report the in situ synthesis and stabilization of Co₂C with H₂O from a K-modified Co₃O₄ precursor for CO₂ hydrogenation. In situ x-ray diffraction (XRD) and simulated results of chemical potential corroborate that the carburization route derived from CoO dominates under CO₂ hydrogenation conditions, while the reaction atmosphere in conjunction with the K promoter is important for the stabilization of Co₂C. Notably, co-feeding 2 volume % of H₂O accelerates the carburization by enhancing the formation of surface carboxylate, which prefers to further split for following C permeation. The K promoter also endows the adsorption of C atoms on the CoO surface, favoring the subsequent carburization. CO₂ converts to C₂₊ HCs via CO^* intermediate derived from carbonate splitting on K-Co₂C catalyst, and a selectivity to C₂₊ HCs up to 67.3% is achieved at 300°C, 3.0 MPa. Furthermore, H₂O (0.5 volume %) is applied to inhibit the decomposition of surface Co₂C and markedly prolongs the catalyst lifetime from 35 to over 200 hours. This work reveals a unique promoting role of H₂O in Co₂C-catalyzed CO₂ hydrogenation and provides a mechanistic understanding of the formation and evolution of Co₂C. It is expected to advance the application of Co₂C in CO₂ hydrogenation, and potentially other reactions.

RESULTS AND DISCUSSION

Synthesis, catalytic performance, and kinetics

As displayed in Fig. 1A, Co₃O₄ and K-modified Co₃O₄ precursors were prepared via a citric acid–induced sol-gel method and subsequent incipient wetness impregnation (IWI) using the solution of K₂CO₃ (0.98 wt %; details in Materials and Methods). XRD patterns and thermogravimetric analysis (TGA) results confirm that the foamy gels were completely decomposed to Co₃O₄ after the calcination at 450°C (fig. S1). The transmission electron microscopy (TEM) and high-resolution TEM (HRTEM) images on Co₃O₄ and K-Co₃O₄ illustrate that there is no substantial change in the average particle size (21 to 24 nm, beyond the sensitive size range in CO₂ hydrogenation) and morphology of Co₃O₄ with the K modification (fig. S2). The particle size of K-Co₃O₄ exhibits a narrower distribution owing to the secondary calcination. Energy-dispersive spectroscopy (EDS) mappings confirm that the K promoter is uniformly dispersed across the Co₃O₄ surface (fig. S3). These precursors were used in CO₂ hydrogenation reaction at 300°C, 3.0 MPa with a space velocity of 6000 ml g⁻¹ hour⁻¹. We collected the spent samples, which were tested for 3 hours of CO₂ hydrogenation, after careful passivation. The characteristic XRD peaks for spent samples derived from Co₃O₄ and K-Co₃O₄ are assigned to the phases of metallic Co (Co⁰), which consists of face center cubic–Co and hexagonal closest packed (hcp)–Co, and Co₂C, respectively (these two spent samples are denoted as Co⁰-t and K-Co₂C, respectively) (Fig. 1A). It indicates that these two phases were in situ generated under the reaction conditions.

Fig. 1. Catalyst synthesis and catalytic properties.

(A) Scheme for the reaction-induced in situ synthesis and XRD patterns of K-Co₂C and Co⁰-t samples. (B) Catalytic performance on Co⁰-t and K-Co₂C at 300°C, 3.0 MPa, space velocity = 6000 ml g⁻¹ hour⁻¹, H₂/CO₂ = 3. (C) C₂₊ HCs space-time yield (STY) and catalytic performance for adjusted K loadings and alkali metal promoters at the same reaction conditions. (D) Catalytic performance at optimized reaction conditions, Co⁰-t: H₂/CO₂ = 2, 42,000 ml g⁻¹ hour⁻¹; K-Co₂C: H₂/CO₂ = 2, 4,500 ml g⁻¹ hour⁻¹. (E) Activation energies for CO₂ conversion and C₂₊ HCs or CH₄ formation on K-Co₂C. (F) Activation energies on Co⁰-t sample. (G) CO₂ and H₂ reaction orders evaluated at 260°C, 3.0 MPa. wt %, weight %; a.u., arbitrary units.

Some differences in the catalytic performance were observed on Co⁰-t and K-Co₂C. As shown in Fig. 1B, K-Co₂C offers a selectivity to C₂₊ HCs of 46.5%, whereas it is only 6.0% on Co⁰-t. High-value products C₂ ~ C₄⁼ and C₅₊ occupy 13.4 and 16.5% of K-Co₂C with an olefin/paraffin (O/P) ratio of 0.8 and a chain growth factor α of 0.53. In contrast, methane dominates (selectivity as 94.0%) on Co⁰-t, while ethane is the only C₂₊ product. The normalized space-time yields (STYs) of C₂₊ HCs based on catalyst mass and surface area reach 7.7 mmol g_cat⁻¹ hour⁻¹ and 0.31 mmol m⁻² hour⁻¹ on K-Co₂C sample, while those on Co⁰-t are 1.7 mmol g_cat⁻¹ hour⁻¹ and 0.08 mmol m⁻² hour⁻¹, respectively. We further modulated the K content and found that only K-Co₂C (K fraction as 0.98 wt %) delivered a moderate conversion (38.2%) and high selectivity and STY to C₂₊ HCs, while 0.49% K-Co₃O₄ and 1.96% K-Co₃O₄ samples only yield 2.2 and 1.8 mmol g_cat⁻¹ hour⁻¹, respectively (Fig. 1C and table S1). Further investigations on Na- and Cs-modified samples show decreased selectivity to C₂₊ HCs of 37.0 and 42.8%, while their STYs are 6.4 and 5.9 mmol g_cat⁻¹ hour⁻¹, respectively. In addition, the reaction temperature and pressure on K-Co₂C were optimized (table S2). We found that both CO₂ conversion and CH₄ selectivity increase with the temperature (260° to 340°C), indicating that a high-reaction temperature promotes the CO₂ activation but also boosts the deep hydrogenation to CH₄. The optimum yield to C₂₊ HCs and those to C₂ ~ C₄⁼ and C₅₊ (2.2 and 2.7 mmol g_cat⁻¹ hour⁻¹) were obtained at 300°C (fig. S4). Tests at various pressures show that the CO₂ conversion and C₂₊ HCs selectivity monotonically increase with the reaction pressure, whereas CO selectivity decreases (fig. S5A). Under pressurized conditions (0.8 to 3.0 MPa), more C₂₊ HCs were detected (above 34.5%), and all the spent samples show well-defined reflections of Co₂C. In contrast, the sample evaluated at atmospheric pressure (0.1 MPa) consists of CoO and Co₂C, while CO is the dominant product (61.8%), revealing that the reaction pressure affects the carburization and C─C coupling (fig. S5B). Considering that Co₂C tends to decompose to Co⁰ at a high temperature (above 260°C) and a high pressure (above 2 MPa) in CO-FTS (29), this in situ synthesized K-Co₂C catalyst operated at 300°C and 3.0 MPa extends its application in catalytic CO₂ conversion.

To inhibit the deep hydrogenation of the C₁ intermediate and increase the proportion of valuable C₂₊ HCs, we further optimized the H₂/CO₂ ratio to 2/1. At a comparable conversion (24.2 and 21.6%), Co⁰-t (42,000 ml g⁻¹ hour⁻¹) and K-Co₂C (4500 ml g⁻¹ hour⁻¹) exhibit the selectivity to C₂₊ HCs as 3.3 and 67.3%, respectively (Fig. 1D). The selectivity to C₂ ~ C₄⁼ and C₅₊ on K-Co₂C reach 31.6 and 28.7% with an O/P ratio of 4.5. The detailed distribution of C₂₊ HCs is shown in fig. S6 and table S3. The decreased space velocity and H₂/CO₂ ratio favor the carbon chain growth (α of 0.59) and inhibit the rehydrogenation of olefins, leading to a higher proportion of valuable C₂ ~ C₆⁼. Kinetic analysis was conducted at high space velocities to get insights on enhanced C₂₊ HCs generation on K-Co₂C. The apparent activation energies (E_a) for CH₄ and C₂₊ HCs generation were estimated as 96.7 ± 4.1 kJ mol⁻¹ and 50.1 ± 1.5 kJ mol⁻¹ on K-Co₂C versus 80.4 ± 4.6 kJ mol⁻¹ and 92.9 ± 4.8 kJ mol⁻¹ on Co⁰-t, revealing that C₂₊ HCs formation is more facile, whereas CH₄ formation is inhibited on K-Co₂C (Fig. 1, E and F). Apparent reaction orders on K-Co₂C (CO₂ α₁ = 0.48 ± 0.03 and H₂ β₁ = 0.98 ± 0.05) and Co⁰-t (CO₂ α₂ = 0.88 ± 0.01 and H₂ β₂ = 0.32 ± 0.02) further evidence that the enhanced CO₂ or suppressed H₂ activation profit the C₂₊ HCs formation instead of CH₄ on K-Co₂C (Fig. 1G) (30). Moreover, compared with K-Co₂C, the similar CO₂ reaction order (α₃ = 0.47 ± 0.01) and the evidently increased H₂ reaction order (β₃ = 1.34 ± 0.04) of 1.96% K-Co₃O₄ (fig. S7) suggest that the excessive K promoter inhibits the H₂ activation, instead resulting in the declined catalytic performance. We further investigated the specific properties of CO₂ and H₂ adsorption and activation on Co₂C and Co surfaces using density functional theory (DFT) calculations. Compared with Co, the CO₂ adsorption and dissociation energies on the Co₂C surface decreased from −0.08 to −0.26 eV, and −1.19 to −1.57 eV, respectively (fig. S8A). Thus, CO₂ adsorption and activation are enhanced on Co₂C. There is essentially no difference between the molecular adsorption energies of H₂ on the two surfaces (fig. S8B). However, the dissociation energy of H₂ on the Co surface (−1.22 eV) is exothermic, whereas the dissociation energy on Co₂C (0.04 eV) is energetically less favorable.

Structural characterization

The specific surface areas of Co⁰-t and K-Co₂C are 20.9 and 24.8 m² g⁻¹ (table S4), and their average particle sizes are 23.2 ± 0.1 and 26.6 ± 0.2 nm as estimated from the XRD results, respectively. TEM and representative HRTEM images show that K-Co₂C represents a cylindrical morphology. The interplanar distances of 0.200, 0.218, and 0.243 nm match the (210), (020), and (101) planes of Co₂C, respectively (Fig. 2, A to C) (31). We verified that the K promoter is still uniformly dispersed on the Co₂C surface (fig. S9), suggesting that there is no obvious migration or agglomeration of K during the reaction. As for Co⁰-t, the (002) and (100) planes of hcp-Co were clearly observed (fig. S10). Previous DFT results by Zhang et al. (32) suggest that (101) and (020) facets of Co₂C compete for CH_x^* coupling, leading to the high selectivity to C₂₊ HCs on K-Co₂C. Ultraviolet (UV) Raman spectroscopy (Fig. 2D) with a 320-nm laser was used to probe the information of C species at about 10 nm in depth without the interference of fluorescence (33). Two Raman shifts observed on K-Co₂C of ~1453 and ~ 1606 cm⁻¹ are ascribed to the D and G bands of C species in Co₂C, which represents the A_1g vibration of disordered graphite and E_2g vibration of graphitic carbon, respectively (21, 34). The I_D/I_G value (calculated using the deconvoluted peak area) for K-Co₂C is ~1.38, suggesting the low disorder degree and surface energy (21). In contrast, no signal was detected on the metallic Co⁰-t sample.

Fig. 2. Structural characterizations of spent catalysts.

(A) TEM images and particle size distribution of K-Co₂C. (B and C) HRTEM images of K-Co₂C. (D) Ultraviolet Raman spectra of K-Co₂C and Co⁰-t. Quasi in situ x-ray photoelectron spectroscopy (XPS) spectra of K-Co₂C and Co⁰-t (E) Co 2p spectra and (F) C 1s spectra.

Considering the carbide is sensitive to air, the surface properties of spent catalysts were further investigated using quasi in situ x-ray photoelectron spectroscopy (XPS) without exposure to air. The K 2p features of K⁺ species in K₂CO₃ were observed at 296.8 and 294.2 eV on K-Co₂C (fig. S11) (35). The binding energies of Co 2p_1/2 and 2p_3/2 in Co₂C (793.5 and 778.3 eV on K-Co₂C) and Co⁰ (793.6 and 778.4 eV on Co⁰-t) are similar, except for the signals at 798.0 and 783.5 eV associated with the superposition of Co⁰ satellite peaks and Co²⁺ 2p peaks (Fig. 2E) (36). The Co₂C species on the surface of K-Co₂C was also identified according to the signal of the C 1s band at 283.1 eV, assigned to the Co─C bond (Fig. 2F). The characteristic peaks of the C═O group and the C─C bond originated from the surface adsorbed species appear at 290.0 and 284.9 eV, respectively (22). The surface C/Co molar ratio, which was estimated on the basis of the deconvoluted results, on K-Co₂C (0.27) is much higher than that on Co⁰-t (0.12), suggesting the enhanced carburization with K modification.

Adsorption properties and reaction mechanism

Temperature programmed desorption (TPD) of CO₂, CO, and H₂ combined with online mass spectrometry (MS) was performed to characterize the adsorption properties (fig. S12). Compared with Co⁰-t, K-Co₂C shows an enhanced medium-strong adsorption of CO₂ and CO (300° to 450°C), but a diminished adsorption of H₂. The medium-strong adsorbed CO₂ and CO species on K-Co₂C, which is more inclined to be activated and involved in the subsequent conversion, can create a relatively C-rich and H-lean environment, inhibiting the deep hydrogenation while promoting the C─C coupling (37). In situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) was carried out to gain insight into the reaction pathways, and the detailed peak assignments are listed in table S5. As shown in Fig. 3A, the spectra for CO₂ pre-adsorbed on K-Co₂C were collected during the pressurization of CO₂ to 1.2 MPa at 50°C and subsequent heating to 260°C. CO₂ was initially activated as monodentate carbonate (m-CO₃²⁻, at 1511, 1437, 1395, and 1278 cm⁻¹) and bidentate carbonate (b-CO₃²⁻, at 1676 and 1625 cm⁻¹) on the K-Co₂C surface (38–41). The features of adsorbed CO (CO_ads) on Co₂C (2078 cm⁻¹) and Co⁰ (2059 cm⁻¹) were detected once the pressure was increased to 0.6 MPa, and that of bicarbonate (HCO₃⁻, at 1606 cm⁻¹) emerges at 200°C, 1.2 MPa, indicating that the residual H species originated from the pretreatment or the hydroxyl group assists in the surface reaction (38). This CO_ads species reveals that, on the K-Co₂C surface, a small amount of CO₂ preferentially converted to CO at mild conditions (38). After switching to H₂ (Fig. 3B and fig. S13A), the features of bidentate formate (b-HCOO^*, the stretching and bending vibrations of C─H band at 2766, 2680, and 1400 cm⁻¹, and the vibration of OCO group at 1591 cm⁻¹) and that of generated gaseous CO (CO_gas, at 2178 and 2111 cm⁻¹) and CH₄ (3015 and 1306 cm⁻¹) and the vibrations of C─H bonds in C₂₊ HCs as CH₃ (2956 cm⁻¹) and CH₂ (2929 cm⁻¹) (the enlarged spectra are shown in fig. S13B) were observed (39, 42). The varying tendency of above species as determined by the featured peak (table S6) intensity with time is shown in Fig. 3C. It can be seen that the CO₃²⁻ rapidly reduces starting at 50°C, 1.2 MPa with an increasing intensity of CO_ads, indicating that CO₃²⁻ initially converts to CO_ads without the H assistance. Surface HCO₃⁻ has gradually accumulated and reaches a steady state after 260°C, 1.2 MPa. The introduction of H₂ results in the reduction of CO_ads and m-CO₃²⁻ with an increase of CH_x^* from 0 to 30 min and continuous generation of CO_gas and HCs. Moreover, the intensity of b-HCOO^* increases after H₂ addition but its further conversion is not observed. It is speculated that CO_ads derived from CO₃²⁻ is an important intermediate for CH_x^* generation, which is further coupled to C₂₊ HCs on K-Co₂C catalyst, whereas the HCO₃⁻ and b-HCOO^* seem acting as spectators. By comparison, as for Co⁰-t catalyst (figs. S14 and S15), the intensities of b-HCOO^* rapidly decreased after switching to H₂, while those of monodentate formate (m-HCOO^*) and product CH₄ simultaneously increased, indicating that the generation of CH₄ mainly undergoes the b-HCOO^*–mediated pathway. Moreover, the signal of CO_ads also emerged during CO₂ adsorption (200° to 260°C) and quickly vanished after the H₂ introduction. It reveals that the CO_ads species is also possibly involved in the CH₄ formation.

Fig. 3. Reaction mechanism studies.

In situ DRIFTS spectra on K-Co₂C for (A) CO₂ pre-adsorption. (B) Intermediates conversion with switching to H₂ at 260°C, 1.2 MPa. (C) Evolution of surface species according to the intensity of infrared featured bands.

In situ synthesis and stabilization of Co₂C

We resorted to in situ XRD and theoretical calculations to shed light on the structural evolution of Co-species during the reaction. The reduction properties were first determined as discerned by the in situ XRD patterns of as-prepared Co₃O₄ and K-Co₃O₄ in H₂ (fig. S16). We found that K addition increased the complete reduction temperature from 300° to 360°C, suggesting that, to some extent, it inhibits the reduction of cobalt oxide toward the Co⁰. Under the CO₂ hydrogenation atmosphere (Fig. 4A), the Co₃O₄ modified with K was first reduced to CoO and then carburized to Co₂C at 260°C within 2 hours, and the reflection of Co⁰ was absent throughout the test. In combination with the semiquantitative XPS analysis (Fig. 2, E and F) that the surface C/Co ratio (0.27) on K-Co₂C is relatively lower than the standard stoichiometric coefficient in Co₂C (0.50), we speculate that, with K modification, the Co₂C was in situ generated, followed by its partial decomposition to Co⁰ on the catalytic surface. Regarding that Co⁰ species generally participates in deep hydrogenation to CH₄, a strategy against this decomposition is still required. In comparison, the Co₃O₄ without K was reduced to CoO and then directly to Co⁰ at 260°C within 50 min (fig. S17 and green arrows in Fig. 4B). Ab initio thermodynamics methods were applied to derive the phase diagram and depict the structural evolution of the catalyst. The chemical potentials for C (μ_C) and O (μ_O) are determined according to the gas-phase composition and reaction conditions (43). As displayed in Fig. 4B, the initial reaction condition is located at the red circle (μ_O = −7.02 eV, μ_C = −9.69 eV) which is near the computed boundary of the CoO phase. This μ_O value is below the Co₃O₄ phase. Therefore, once Co₃O₄ is exposed to the environment, a reduction in the CoO will occur. During the reaction, μ_O decreases, while μ_C increases due to the CO generated from CO₂ reduction. As a result, the final chemical potentials migrate to the black circle (μ_O = −7.79 eV, μ_C = −8.16 eV), which is inside the domain of Co₂C, suggesting that its formation from CoO is thermodynamically favorable under reaction conditions. Considering that the Gibbs free energy change (ΔG) is more negative for the carburization from CoO than that from Co⁰ (fig. S18), and no bulk Co⁰ was observed during the evolution (Fig. 4A), we speculate that the thermodynamic advantages directly propel CoO into carburization without the reduction to Co⁰. We then investigated the structural evolution in CO₂ hydrogenation for the K-Co⁰ sample, which was obtained from the decomposition of K-Co₂C. The in situ carburization of metallic cobalt toward Co₂C was also observed, exhibiting the gradually increased C₂₊ HCs selectivity and reduced CH₄ selectivity along with the reaction (fig. S19). The CH₄ dominates in the initial products of K-Co⁰ [time on stream (TOS) = 1 hour] or Co⁰-t samples, suggesting that the introduction of the K promoter does not substantially influence the product selectivity. The temperatures for the emergence of Co₂C reflections and the disappearance of Co⁰ reflections increase to 270° and 330°C, respectively (Fig. 4C), suggesting that the synthesis of Co₂C from Co⁰ precursor requires a high temperature for the C permeation into the Co lattice. We are thus confident that the in situ synthesis of Co₂C in CO₂ hydrogenation starts from K-modified CoO (blue arrows in Fig. 4B) instead of Co⁰. By comparison, the syngas preferentially induces the reduction to Co⁰, while the following carburization originates from Co⁰ (16). The replacement of CO with CO₂ results in an increased μ_O, favoring the unique CoO route for the synthesis of Co₂C. Moreover, this oxide-mediated carburization is different from that of iron catalysts although their carbides have many similarities. Metallic iron is more advantageous than its oxide for the formation of FeC_x, and this key difference is helpful to understand the distinct evolution and deactivation pathways for cobalt- and iron-based catalysts in CO_x conversion (44, 45).

Fig. 4. Structural evolutions and theory calculations.

(A) In situ XRD patterns of K-Co₃O₄ in CO₂ hydrogenation at 0.8 MPa. (B) Phase diagram of Co-O-C trinary system derived from the DFT energies of bulk crystals. (C) In situ XRD patterns of K-Co⁰ sample in CO₂ hydrogenation at 0.8 MPa; the K-Co⁰ sample was synthesized from the decomposition of K-Co₂C in H₂ at 340°C, 0.8 MPa. In situ XRD patterns of K-Co₂C (D) in pure CO₂ at 0.8 MPa, (E) in 25 volume % H₂O/N₂ at 0.1 MPa. Potential energy profiles. (F) CO₂ dissociation to CO^* and O^* on Co₂C (101) surface with or without K₂O.

Structural stability is a key factor for catalysts. As evidenced by the in situ XRD patterns in N₂ at 0.8 MPa (fig. S20), K-Co₂C is thermally stable below 340°C without bulk decomposition. Its stability in the presence of reactants and products is also important but still uncertain. Co₂C is considered a metastable phase under the H-rich environment, and, from the in situ XRD results of K-Co₂C in pure H₂, it can be clearly seen that the transition of Co₂C to Co⁰ occurs at 290°C (fig. S21). A recent report has proven that CO₂ adsorbs on the surface as carboxylate (CO₂^δ−) and then splits to C^* in the presence of H₂, enabling the following C permeation (22). We performed the in situ XRD investigations on K-Co₂C in pure CO₂ and found that excessive CO₂ also removes the C from Co₂C at 300°C (Fig. 4D). Previous results have shown that H₂O can oxidize hcp-Co to CoO or cobalt hydroxide but its effect on Co₂C is unclear (46). We investigated the evolution of K-Co₂C in 25 volume % (the theoretical value of volume fraction and the same as below) H₂O/N₂ and found that excessive H₂O causes the transition of Co₂C to hcp-Co in the range of 280° to 310°C (Fig. 4E). There is no obvious oxidation in bulk below 340°C, possibly originated from the fast O^* removal with K-promoting effect (47). When the H₂O content was decreased to 5%, this transition intensively occurred in the temperature region of 330° to 340°C (fig. S22), which is higher than that with co-feeding 25% H₂O, implying that the H₂O concentration also has an impact on the decomposition of Co₂C. Inspired by the above findings, we believed that the single reactant or product (H₂O) of high concentrations could destroy the K-Co₂C but a unique balance between them in CO₂ hydrogenation presents an opportunity for its stabilization. We further inquired into the effect of the K promoter. We prepared a Co₂C sample without K (denoted as Co₂C-t) through a reduction-carburization procedure from Co₃O₄. However, it rapidly decomposed to hcp-Co at 260°C in CO₂ hydrogenation within 20 min as confirmed by the in situ XRD results (fig. S23). Moreover, the DFT results on the Co₂C (101) surface reveal that with the modification using the K₂O promoter, the direct splitting of CO₂ to CO is more facile, considering that the energy barrier and energy of reaction (E_reaction) reduce from 1.10 to 0.56 eV and −0.36 to −1.04 eV, respectively (Fig. 4F). Since CO₂ is adverse to the stabilization of Co₂C whereas CO enhances the carburization, the promoting effect on CO₂ direct dissociation here is also important for stabilizing the Co₂C. Hence, we reveal that the reaction atmosphere in conjunction with the K promoter is crucial in the stabilization of Co₂C.

H₂O promoted carburization and stabilization

Since H₂O is a common impurity for CO₂ capture and more H₂O is generated in CO₂-FTS compared with CO-FTS, we further delve into its influences on the carburization and the catalytic reaction. We simulated the carburization from K-CoO in 10% CO/N₂ and, as shown in Fig. 5A, in situ XRD results demonstrate that the transformation of CoO to Co₂C started at 260°C and finished at 280°C as evidenced by the vanishing CoO feature at 36.5°. While 2% H₂O was co-fed (at 0.1 MPa, Fig. 5B), despite the constant temperature for the beginning of carburization at 260°C, the transition of CoO to Co₂C was accomplished at 270°C. Once the H₂O content was increased to 5%, the onset and end temperatures for the carburization shifted to 290° and 300°C, respectively (fig. S24). As stated above, a moderate amount of H₂O facilitates the carburization of K-CoO and reduces the carburizing temperature, while its content is crucial. Most previous studies reported the negative effect of H₂O on the active centers, including the induced oxidation or sintering of Co⁰ nanoparticles (48–50). So far, the promoting effect of H₂O on the carburization to Co₂C has never been reported to the best of our knowledge.

Fig. 5. Facilitating carburization of K-CoO sample with a moderate amount of H₂O.

In situ XRD patterns recorded (A) in 10 volume % CO/N₂ at 0.8 MPa and (B) in 10 volume % CO + 2 volume % H₂O/N₂ at 0.1 MPa. Quasi in situ XPS spectra in the near-ambient pressure (NAP) XPS system recorded after reaction at 0.1 MPa (C) in 10 volume % CO/N₂ and (D) in 10 volume % CO + 2 volume % H₂O/N₂.

To obtain more evidence on the catalytic surface, we collected the quasi in situ XPS spectra in the NAP-XPS system during the carburization of K-CoO. As shown in the Co 2p spectra (Fig. 5, C and D), the characteristic peaks at 780.3 and 796.4 eV and their satellite features at 786.3 and 802.9 eV are attributed to the divalent Co species in the octahedral site of CoO (51). In 10% CO/N₂, the features of CoO kept almost unchanged, while only a faint signal of Co─C bond was first detected at 300°C, inferring a slight surface carburization of K-CoO (Fig. 5C). However, by co-feeding 2% H₂O, the peak of Co─C bond (283.0 eV) initially emerged at 250°C, and then gradually developed with the increasing temperature (Fig. 5D). Meanwhile, the featured Co 2p peaks completely shifted from CoO (780.0 eV) to Co₂C (778.1 eV) at 300°C, illustrating the full carburization of the surface. These results explicitly verify the promoting effect of H₂O in the carburization of K-CoO. Furthermore, for the fresh K-CoO sample, the K 2p_1/2 and 2p_3/2 peaks at 295.2 and 292.7 eV are associated with K⁺ species in K₂O (52). With the increasing temperature in CO or CO + H₂O flow, these characteristic peaks gradually shifted to high binding energies and lastly reached at 295.7 and 293.0 eV at 300°C, revealing that the surface K₂O species reacted with the CO to form carbonate.

To gain insight into this promotion, we contrasted the interaction between CO and pristine CoO surface with and without H₂O. As shown in the quasi in situ XPS spectra, the surface CoO (780.2 eV) was fully reduced to Co⁰ (778.5 eV) at 300°C in 10% CO, while co-feeding 2% H₂O prevents this reduction (Fig. 6A). For C 1s spectra (Fig. 6B), a shoulder peak at ~288.0 eV assigned to surface C─O species, including the possible CO₃²⁻, HCO₃⁻, and CO₂^δ−, appears with H₂O addition. These adsorbed species were further determined by in situ DRIFTS. The spectra collected at 1 and 30 min are used for probing the initial and final adsorbed species, respectively, while the changing processes are recorded in figs. S25 and S26. As shown in Fig. 6C, for the CoO-t sample in 10% CO, the initial peaks at 1609, 1509, and 1268 cm⁻¹ correspond to b-CO₃²⁻ adsorbed on the CoO surface and those at 1366 and 1318 cm⁻¹ correspond to polydentate carbonate (p-CO₃²⁻) species (40). Along with the surface reduction (fig. S25), some adsorbed species gradually shift to HCO₃⁻ (1604 cm⁻¹, H is from surface hydroxyl) and m-CO₃²⁻ (1463, 1377 cm⁻¹) on Co⁰. While 2% H₂O was added, a band at 1438 cm⁻¹ and the shoulder peak at 1425 cm⁻¹, which are attributed to the stretching vibration of the C═O band and the bending vibration of the C─H band in formyl (HCO^*) species, were observed on the CoO surface (53). The b-CO₃²⁻ and monodentate formate (m-HCOO^*, which is identified by the stretching vibrations of the OCO group at 1542 and 1358 cm⁻¹, and the bending vibration of C─H band at 1392 cm⁻¹) were also observed (54). Without K modification, these species mostly convert to adsorbed m-CO₃²⁻, which is inactive for carburization. As for the K-CoO sample, a signal at 1745 cm⁻¹ associated with adsorbed CO on three- or fourfold hollow sites (55) was both observed with and without co-feeding H₂O, suggesting that the K promoter provides additional sites for enhanced CO adsorption, which is in accordance with the CO-TPD-MS results (fig. S12B). In 10% CO, b-HCOO^* at 1565 and 1364 cm⁻¹, and m-CO₃²⁻ are initially adsorbed and accumulated, while additional HCO₃⁻ species (1605 cm⁻¹) was observed at 30 min on the K-CoO surface. Notably, with the addition of 2% H₂O, the bands at 1556, 1415, and 1336 cm⁻¹, which are assigned to the CO₂^δ−, were detected (22, 56, 57). With an extension of the treatment time, these species have accumulated in 1 to 10 min and then are gradually consumed along with the surface carburization (fig. S26). The final adsorbed species on a fully carburized surface is HCOO^* while physically adsorbed H₂O (1660 cm⁻¹) was also detected. In literature results, compared with HCOO^* and CO₃²⁻, CO₂^δ− is easier to split and form C^* species for carburization (22). According to the above XPS and infrared (IR) results, we speculate that the promotion originates from the H₂O-facilitated formation of surface C─O species, including the HCO^* or CO₂^δ− species. However, without the K promoter, the HCO^* converts to CO₃²⁻; while only the K promoter and H₂O jointly act on the CoO surface, the active CO₂^δ− species can be generated and involved in the subsequent carburization.

Fig. 6. Effect of co-feeding H₂O on structural properties and catalytic performance.

Quasi in situ XPS spectra collected on CoO-t sample after treating in 10 volume % CO with or without co-feeding 2 volume % H₂O (A) Co 2p spectra and (B) C 1s spectra. (C) In situ DRIFTS spectra collected at 1 and 30 min on CoO-t and K-CoO samples under different treated conditions. (D) Catalytic stability test with co-feeding 0.5 volume % H₂O within 210 hours. Reaction conditions: 300°C, 3.0 MPa, space velocity = 6000 ml g⁻¹ hour⁻¹, H₂/CO₂ = 3. Quasi in situ XPS spectra of K-Co₂C sample that was used in CO₂ hydrogenation with co-feeding 0.5 volume % H₂O (E) Co 2p spectra and (F) C 1s spectra.

We calculated the dissociation of CO₂^δ− (CO₂^*) toward C^* species on the CoO (200) surface. As demonstrated in fig. S27, with the presence of K₂O promoter and H₂O, the CO₂^* dissociation energy (reaction I: CO₂^* → CO^* + O^*) declines about 0.7 to 2.07 eV. Notably, the reaction energies for C^* formation of reactions II (CO^* → C^* + O^*) and III (CO^* + CO^* → C^* + CO₂^*) reduce to 0.67 and −0.54 eV, which are lower than those on clean CoO (6.47 and 3.19 eV) or K₂O-CoO (5.45 and 2.47 eV) surfaces. The results show that K₂O and H₂O jointly promote the dissociation of surface CO₂^δ− to C^* species. Besides, after the dissociation of the C─O species to C^*, the adsorption of surface C atoms is important for carburization, and we further investigated this process on CoO. The average number of electrons transferred between C atoms and the CoO surface (described by Bader charge) and the average formation energies of adsorbed C atoms (ΔE_form) calculated with and without K₂O co-adsorption are summarized in table S7. Two or three C atoms (marked as 2C and 3C) were deposited on the CoO surface to restore the situation of multicarbon co-adsorption. The Bader charges are 2.13 (2C) and 2.16 (3C) on the clean CoO surface. They increase to 2.28 (2C) and 2.22 (3C) in the presence of K₂O, indicating that more electrons are transferred from the C atoms to the CoO surface, and that K₂O enhances the electronic interaction between C^* and the catalyst surface. As a result, with the K₂O addition, the ΔE_form decreases from 3.69 (2C) and 3.63 eV (3C) to 3.25 (2C) and 3.20 eV (3C), respectively. This reduction in formation energies (exceed, 0.4 eV) demonstrates that the adsorption of C^* species on the CoO surface is easier with K₂O promotion, which favors C accumulation and permeation to form a bulk carbide. These calculation results agree with the experimental observations that K₂O and H₂O promote carbide formation.

Given that the surface of K-Co₂C decomposes in the reaction (Fig. 2F), there is a possibility to exploit the beneficial effect of co-feeding H₂O for inhibiting this decomposition and improving the catalytic stability. We first evaluated the effect of co-fed H₂O content on the performance (fig. S28; the data are collected at TOS = 3 hours, while there is no obvious deactivation) and found that when it increases to 0.25 and 0.5%, the CO₂ conversion slightly decreases from 37.9 to 35.6% and 35.3%, respectively. However, more H₂O addition exacerbates the activity declination, while the conversion drops to 24.2% with co-feeding 2.5% H₂O. While the added H₂O content was increased to 25%, the overall CO₂ conversion further dropped to 17.5% and the CH₄ (selectivity as 86.8%) dominates in the products. The generated H₂O in the actual reaction or its extra addition gives rise to the competitive adsorption with other reactants or intermediates and increases the rate of reverse reaction, which jointly causes a gradually reduced CO₂ conversion (28). Moreover, a volcanic curve was observed for C₂₊ HCs selectivity and yield with H₂O content, and at the optimum content of 0.5%, the C₂₊ HCs selectivity increases to 53.1% and its STY of C₂₊ HCs is 8.0 mmol g⁻¹ hour⁻¹. As can be seen from the C₂₊ HCs composition (fig. S29), co-feeding 0.5% H₂O leads to a more centralized product distribution (α decreases from 0.53 to 0.48) and an enhanced generation of light olefins (the C₂ ~ C₄⁼ selectivity increases to 19.9% and O/P ratio increases to 1.0), indicating that co-feeding 0.5% H₂O here protects the surface Co₂C from decomposition. Moreover, in CO-FTS, H₂O sometimes promotes the formation of CH_x^* species via an H₂O-assisted methylidyne mechanism and increases the C₅₊ selectivity (58). However, a slightly decreasing C₅₊ selectivity was observed with H₂O addition here. It is speculated that the low concentration of H₂O (0.5% ~ 2.5%) does not substantially influence the reaction pathway but is enough to modulate the surface compositions. The promoting effect of H₂O on stabilizing the structure of Co₂C and its catalytic features is expected to enable its practical application in the CO₂ hydrogenation reaction with high temperature and pressure.

The catalytic stability of K-Co₂C in CO₂ hydrogenation was examined, as shown in figs. S30 and S31, we found that the H₂O content has an impact on the catalyst deactivation. Within the initial 35 hours, without co-feeding H₂O, the CO₂ conversion and C₂₊ HCs selectivity decreased from 38.2 to 31.6% and 46.5 to 38.8%, respectively. The addition of H₂O effectively inhibited this activity decline, and when co-feeding 0.5% or more H₂O, the catalytic performance basically kept stable in the initial 35 hours. We further compared the changes in catalytic performance with various H₂O contents (fig. S32). The drops of C₂₊ HCs STY can be clearly observed when the H₂O content is 0 or 0.25%. As for the H₂O content of 0.5 ~ 2.5%, the ratios of CO₂ conversion collected at 3 and 35 hours (ratio A) and those of C₂₊ HCs STY (ratio B) are all in the range of 96 to 98%, which are higher than those recorded at 0 or 0.25% H₂O. It reveals that co-feeding H₂O is beneficial to mitigating the deactivation, and when the H₂O content exceeded 0.5%, the features of deactivation are quite similar. Furthermore, without co-feeding H₂O, the decrease of the activity (31.6 to 24.2%) and selectivity alterations were clearly observed in 35 to 50 hours, while final CH₄ selectivity increased to 74.7% and the STY of C₂₊ HCs dropped from 7.6 to 2.6 mmol g⁻¹ hour⁻¹ within 50 hours on stream (fig. S33). By comparison, with co-feeding 0.5% H₂O (Fig. 6D), the K-Co₂C catalyst showed higher stability throughout a 210-hour test. The final selectivity to C₂₊ HCs remained at 49.2%, exceeding 90% of the initial values (53.0%, at 15 hours), and the STY of C₂₊ HCs still retained 7.2 mmol g⁻¹ hour⁻¹ at 210 hours. The XRD patterns of spent samples (fig. S34) show that co-feeding 25% H₂O caused the complete decomposition of Co₂C, which is in accordance with the in situ XRD results (Fig. 4E). However, no substantial difference was observed in spent samples with co-feeding 0.5 or 2.5% H₂O, indicating that their evolution mainly occurs on the catalytic surface. The quasi in situ XPS spectra of the spent K-Co₂C with 0.5% H₂O are shown in Fig. 6 (E and F). The binding energies of Co 2p_1/2 and Co2p_3/2 at 793.5 and 778.3 eV reveal that the catalytic surface is Co₂C. Notably, the addition of 0.5% H₂O enhances the signal of the Co─C bond at 283.1 eV (C 1s spectra) in comparison to the results of the K-Co₂C sample which was used without co-feeding H₂O (Fig. 2F). The estimated surface C/Co ratio increased from 0.27 (without H₂O) to 0.46 with co-feeding 0.5% H₂O, particularly substantiating that the decomposition of Co₂C has been inhibited by adding moderate H₂O in the reaction.

In conclusion, we provide an in situ synthesized and stable K-Co₂C catalyst for CO₂ hydrogenation which shows outstanding activity and selectivity to C₂₊ HCs up to 67.3% at 300°C, 3.0 MPa. Compared with metallic cobalt catalysts, K-Co₂C is more competitive in accelerating the formation of C₂₊ HCs, and a C-rich and H-lean surface environment suppresses the side reaction toward CH₄. This K-Co₂C catalyst is promising to be combined with the zeolites or membrane reactors to further optimize the product composition and alleviate the downstream separation needs for improving its practical applications (3, 59). Adsorbed CO derived from carbonate splitting is an important intermediate for coupling to C₂₊ HCs as confirmed by in situ DRIFTS. Multispectral studies, including in situ XRD, quasi in situ XPS, and DFT calculations, reveal that Co₂C is directly generated from CoO instead of undergoing the reduction to Co⁰. The reactants (H₂ and CO₂) or the product (H₂O) of high concentrations all can cause the decomposition of Co₂C to Co⁰, while the delicate balance of the reaction atmosphere and the K promoter plays vital roles in the stabilization of Co₂C. Co-feeding a small amount of H₂O instead facilitates the carburization and stabilization of Co₂C. During the carburization, 2% H₂O and the K promoter jointly act on the CoO surface and stimulate the formation of key CO₂^δ− species, followed by its dissociation to active C^* species. The K promoter also enables the subsequent C^* adsorption on the CoO surface, favoring the C accumulation and permeation. Inspired by this finding, 0.5% H₂O was co-fed in the feed gas to stabilize the surface structure of K-Co₂C and it markedly prolongs the catalytic stability from 35 to over 200 hours. Given that H₂O is an inevitable impurity in the flue gas (a key source for CO₂ capture) and generally considered a poison in CO₂ hydrogenation reaction, our study highlights the unique promoting effect of H₂O in feed gas on Co₂C catalyst and it is expected to improve its practical application in CO₂ conversion. However, considering that the excessive H₂O addition is conversely detrimental to Co₂C, its separation before the catalytic conversion or the surface hydrophobic or hydrophilic modifications to control its content within the ideal range is still needed. The fundamental understanding of the formation and stabilization of Co₂C also provides opportunities for its potential application in other reactions.

MATERIALS AND METHODS

Catalyst preparation

Synthesis of Co₃O₄ and modified Co₃O₄ precursor

Co(NO₃)₂·6H₂O was purchased from Aladdin Biochemical Technology Co., Ltd. Citric acid, Na₂CO₃, K₂CO₃, and Cs₂CO₃ were purchased from Bodi Chemical Trade Co., Ltd. Co₃O₄ precursor was prepared using a citric acid–induced sol-gel method. Typically, citric acid and Co(NO₃)₂·6H₂O (molar ratio = 13/20) were dissolved in ethanol and the mixed solution was aged at 30°C for hydrolysis and complexation. Then, the ethanol was fully evaporated at 80°C. The obtained foamy gel was dried at 80°C for 12 hours and further calcined in air at 450°C for 4 hours to get the Co₃O₄ precursor.

Alkali metal–modified Co₃O₄ precursors were synthesized by an IWI method using an aqueous solution of their carbonates. The molar amount of Na₂CO₃ or Cs₂CO₃ was fixed as 0.025 mmol, whereas that of K₂CO₃ was tried as 0.013, 0.025, and 0.050 mmol (the corresponding mass fractions of K are 0.49, 0.98, and 1.96%). As an example, 0.0035 g (0.025 mmol) of K₂CO₃ was dissolved in 0.12 g of H₂O through sonication, and then the solution was dropped on 0.2 g of Co₃O₄ under stirring at room temperature. The resulting power was dried at 80°C for 12 hours and calcined in air at 400°C for 2 hours to get the 0.98% K-Co₃O₄ precursor.

Synthesis of testing samples

We used the in situ XRD reactor chamber to prepare the samples which were denoted as K-CoO, K-Co⁰, CoO-t, and Co₂C-t, respectively. The K-CoO sample was prepared from the reduction of 0.98% K-Co₃O₄ in H₂ at 240°C, 0.8 MPa. The K-Co⁰ sample was obtained from the decomposition of K-Co₂C in H₂ at 340°C, 0.8 MPa. The CoO-t sample was prepared from the reduction of Co₃O₄ precursor in H₂ at 220°C, 0.1 MPa. The Co₂C-t sample was prepared from two-step treatments from Co₃O₄ precursor, including the initial reduction at 260°C, and the following carburization in CO, as recorded in fig. S23A. Before the ex situ structural characterizations, the Co⁰-t and K-Co₂C samples were carefully passivated in the flow of 1% O₂/N₂ (20 ml min⁻¹) at ~20°C for 1 hour and then were transferred to the glove box for preservation.

Catalyst characterization

XRD patterns were recorded on a Rigaku SmartLab 9-kW diffractometer with Cu Kα radiation (λ = 1.5406 Å) and with a scanning rate of 8°/min. In situ XRD measurements were performed in an XRK 900 reactor chamber and the heating process was controlled by a TCU 750 temperature control unit. The patterns were collected for each 10 min until the structure is stable for at least 1 hour at the given conditions. As for the structural evolution in CO₂ hydrogenation, the sample was directly heated to the target temperature in the reactive gas at 0.8 MPa. For the structural evolution in other atmospheres (H₂/CO₂/CO/H₂O), the specific conditions were introduced in the corresponding results.

Quasi in situ XPS was performed on a spectrometer equipped with an Al K x-ray source at 300 W. Before the test, the samples, including the K-Co₂C and Co⁰-t (without the passivation), were carefully transferred from the glove box to the XPS analysis chamber without the exposure to air. Besides, the carburization processes for CoO-t and K-CoO samples were simulated in 10% CO/N₂ or 10% CO + 2% H₂O/N₂ in the reaction chamber equipped with the NAP-XPS system. Here, these quasi in situ XPS spectra were recorded after reaching the setting temperatures for 20 min. The thermal couples were placed at the side of the powder sample for in situ XRD (set temperature, ±30°C) and on the upper surface of the sample piece for quasi in stu XPS, respectively (fig. S35).

In situ DRIFTS experiments were performed using the Thermo Scientific Nicolet iS50 spectrometer with a mercury cadmium telluride detector, and the background spectrum was collected after N₂ purge for at least 1 hour. The flow rate of reactive gas (CO₂/H₂/N₂ = 21:63:16) or N₂ is 30 ml min⁻¹, whereas that of CO₂, CO, or H₂ is 10 ml min⁻¹. As for mechanism studies on K-Co₂C and Co⁰-t samples, the testing temperature is not above 260°C for avoiding the destroy from CO₂ or H₂ on catalyst structure. As shown in fig. S36, these samples were in situ synthesized in the IR cell at 300°C, 1.2 MPa for 3 hours, and then was purged (300°C) and cooled (50°C) in N₂, atmospheric pressure before the adsorption of CO₂ and the following switching to H₂. The co-fed H₂O at atmospheric pressure for in situ XRD or DRIFTS and quasi in situ XPS test is stored in a glass wash bottle, which is placed before the reactor chamber. Its content is dependent on the controlled temperature of wash bottle. The theoretical H₂O content was calculated according to the following equation and was used for the expression. The actual H₂O content was determined using the NaOH and silica gel as the absorbents (fig. S37), while the above results are listed in table S8.

$${\text{Theoretical Content}}_{{\mathrm{H}}_2\mathrm{O}}=\frac{p_{{\mathrm{H}}_2\mathrm{O}}^{\ast }}{p_0}\times 100\mathrm{\%}$$

TPD tests were carried out on a Quantachrome ChemBET Pulsar analyzer, while the desorbed species was detected by the Pfeiffer GSD-350 online MS. Typically, the samples were loaded into a quartz tube and flushed with N₂ at 300°C for 1 hour to remove undesired adsorbates, and then cooled to room temperature. The samples were first retreated in the reactive gas at 300°C for 1 hour to remove the external passivation layer, and then flushed with N₂ for 1 hour. These samples were cooled to 30°C to adsorb CO₂ or CO in 1 hour, followed by switching to N₂ to purge for 30 min. Then, the desorption and analysis program were conducted with a heating rate of 10°C min⁻¹ to 700°C.

TEM and HRTEM images were obtained on a Tecnai F30 HRTEM instrument (FEI Corp) with a voltage of 300 kV. EDS elemental mapping images were obtained on a JEM ARM200F thermal-field emission microscope equipped with a probe spherical aberration (Cs) corrector with a voltage of 200 kV. UV Raman spectra were collected on a homemade triple-stage UV Raman spectrometer with a resolution of 2 cm⁻¹. The wavelength of UV laser line from a double-frequency 514-nm laser was set at 320 nm. The textural properties were determined by Ar absorption-desorption on a Quantachrome AUTO-SORB-1-MP sorption at 87 K. The surface area was calculated using the Brunauer-Emmett-Teller (BET) method. The concentrations of alkali metals and Co were determined by inductively coupled plasma optical emission spectrometry on a PerkinElmer AVIO 500 instrument and the results are listed in table S9. TGA was conducted on a TGA-SDTA851e thermobalance in the air flow (25 ml min⁻¹) with a heating rate of 5°C min⁻¹ from 50° to 800°C.

Catalytic activity test

The CO₂ hydrogenation was conducted in a stainless steel fixed-bed flow reactor with an 8-mm inner diameter. A total of 150 mg of Co₃O₄ or alkali metal–modified Co₃O₄ precursor (20 to 40 meshes) was diluted with 450 mg of quartz sand (20 to 40 meshes), and then loaded into the middle of the reactor. Catalytic performance was tested in the reactive gas (CO₂/H₂/Ar = 1:3:1.5, space velocity = 6000 ml g⁻¹ hour⁻¹, P = 3.0 MPa unless otherwise noted) at 260°, 300°, and 340°C. The products were collected at about 3 hours after reaching the steady state on steam. As shown in fig. S38, all the products, including the liquid oxygenates and C₅₊ HCs, were heated by an oven and heating belt at 100°C for full vaporization and accessed into the online chromatography (Agilent, 7890B) for the analysis by the thermal conductivity detector and flame ionization detector, while the Ar was used as an internal standard. CO₂ conversion and CO selectivity were calculated on a carbon-atom basis according to the following equations

$${\mathrm{CO}}_2\mathrm{Conv}.(\mathrm{\%})=\frac{n_{{\mathrm{CO}}_2,\mathrm{in}}-n_{{\mathrm{CO}}_2,\mathrm{out}}}{n_{{\mathrm{CO}}_2,\mathrm{in}}}\times 100\mathrm{\%}$$

$$\mathrm{CO}\mathrm{Sel}.(\mathrm{\%})=\frac{n_{\mathrm{CO},\mathrm{out}}}{n_{{\mathrm{CO}}_2,\mathrm{in}}-n_{{\mathrm{CO}}_2,\mathrm{out}}}\times 100\mathrm{\%}$$

where n_CO2,in and n_CO2,out represent the concertation of CO₂ at the inlet and outlet. n_CO,out represents the CO concertation at the outlet. The total selectivity of alcohols and dimethyl ether is below 1.5% and therefore was not reported here.

The selectivity to HC C_nH_m was calculated for representing the HCs distribution

$${\mathrm{C}}_{\mathrm{n}}{\mathrm{H}}_{\mathrm{m}}\mathrm{Sel}.(\mathrm{\%})=\frac{\mathrm{n}{\mathrm{C}}_{\mathrm{n}}{\mathrm{H}}_{\mathrm{m},\mathrm{out}}}{\text{Σ}\mathrm{i}{\mathrm{C}}_{\mathrm{i}}{\mathrm{H}}_{\mathrm{m},\mathrm{out}}}\times 100\mathrm{\%}$$

where C_nH_m,out represents moles of detected individual HCs product.

The STY, the O/P ratio, and the chain growth factor α were calculated as shown below

$${\mathrm{STY}}^{\mathrm{a}}(\mathrm{mmol}{\mathrm{g}}^{-1}{\mathrm{h}}^{-1})=\frac{{\mathrm{C}\mathrm{O}}_2\mathrm{Conv}.\times {\mathrm{C}}_{\mathrm{n}}{\mathrm{H}}_{\mathrm{m}}\mathrm{Sel}.\times (1-\mathrm{CO}\mathrm{Sel}.)\times 6000}{22.4\times 5.5}$$

$${\mathrm{STY}}^{\mathrm{b}}(\mathrm{mmol}{\mathrm{m}}^{-2}{\mathrm{h}}^{-1})=\frac{{\mathrm{STY}}^{\mathrm{a}}}{{\mathrm{S}}_{\mathrm{BET}}}$$

$$\mathrm{O}/\mathrm{P}\mathrm{ratio}=\frac{\mathrm{Sel}.({\mathrm{C}}_2\sim {\mathrm{C}}_4^{=})}{\mathrm{Sel}.({\mathrm{C}}_2\sim {\mathrm{C}}_4^0)}$$

$$\ln \frac{W_n}{n}=n\cdot \ln \mathrm{\alpha }+n\frac{{(1-\mathrm{\alpha })}^2}{\mathrm{\alpha }}$$

where STY^a and STY^b represent the STY values normalized based on catalyst mass and surface area (S_BET; table S4). W_n and n represent the mass fraction and the carbon number of HCs products. The carbon balance was in the range of 95 to 105%.

The contents (0.25 to 25%) of H₂O mentioned in this work represent those of extra-added H₂O in the carburization gas or feed gas. The H₂O was added using a steel gas wash bottle, which can be pressurized to 3.0 MPa, and the setting temperatures are listed in table S8. For the kinetics tests (activation energy and reaction order), the conversion was decreased below 10% by varying the space velocity.

Theory computational methods

The basic thermodynamic calculations (ΔG, ΔH, and ΔS) were carried out by the Aspen Plus V11 software. All spin-polarized DFT calculations were performed with the Vienna Ab initio Simulation Package. The exchange-correlation energies were calculated by the generalized gradient approximation approach with the Perdew-Burke-Ernzerhof functional (60). Core electrons were frozen and treated with the projector-augmented wave theory (61). Valence electrons were taken as Co (4s²3d⁷), C (2s²2p²), and O (2s²2p⁴). The plane-wave basis set was truncated at 500 eV for gas molecules and surface calculations and was increased to 650 eV for bulk crystals to mitigate the effect from Pulay stress. Gaussian smearing with 0.1-eV width was applied to oxides and molecules while first order of Methfessel-Paxton smearing was applied to metal and carbides (62). Brillouin zones were treated with the Monkhorst-Pack k-points mesh (63). The k-points, lattice parameters, space groups, and magnetics are summarized in table S10. The van der Waals interaction correction was applied for all calculations with the DFT-D3 method (64). The climbing image nudged elastic band method was used to find transition states, combined with the dimer method. It was verified that each transition state had only one imaginary vibrational frequency along the reaction coordinate direction. A three-layer slab model was built for the CoO (200) surface and the bottom two layers were fixed during structural optimization. The criteria of force convergence for surface calculations were set to 0.03 eV/Å on all atoms. The Hubbard U-correction was applied to oxides for localized d-state electrons (65). A value of 3.3 eV for Co was taken from the work of Wang et al. (66). Bader charge analysis was performed to investigate the charge transfer of the adsorbate-surface system (67). The adsorption and activation of CO₂ and H₂ were investigated on Co₂C (101) and (020) surfaces, and the Co (002) surface. All surface calculations were performed in the Monkhorst-Pack scheme using k-points of 2 × 2 × 1 with the addition of a vacuum layer of 15 Å to avoid interactions between repeating periodic elements. In the structural optimization of Co₂C surfaces, the bottom 1/3 Co and C atoms are fixed in their original atomic positions, while the remaining top 2/3 Co and C atoms together with the adsorbate are completely relaxed. We constructed three kinds of CoO (200) surfaces: clean surface, surface containing only K₂O, and surface containing both K₂O and H₂O, and calculated the reaction energy of the stepwise decomposition of CO₂^δ− to C^* species on these three surfaces. The optimized structures involved are displayed in figs. S39 and S40, respectively.

Ab initio thermodynamics (68) was applied to construct the bulk phase diagram, locate the reaction conditions on the resulting phase diagram, and calculate formation energies. The explanation pertaining to the equilibrium assumptions inherent to this thermodynamics method is shown in the Supplementary Materials. In the bulk phase diagram, the phase with lowest formation energy is identified as the stable phase. The chemical potentials of C and O are determined by the free energies of the gas species with the assumption that the gas phase is equilibrated with the catalysts. The free energy of a molecule is calculated according to the following equations

$${\mathrm{\mu }}_{\mathrm{O}}=G_{{\mathrm{CO}}_2}-G_{\mathrm{CO}}$$

$${\mathrm{\mu }}_{\mathrm{C}}=G_{\mathrm{C}\mathrm{O}}-{\mathrm{\mu }}_{\mathrm{O}}$$

$$G_{\mathrm{gas}}=E_{\mathrm{gas}}^{\mathrm{DFT}}+{\mathrm{ZPE}}_{\mathrm{gas}}+\text{Δ}{H}_{\mathrm{gas}}^{0\mathrm{K}\to 533.15\mathrm{K}}-T{S}_{\mathrm{gas}}^{533.15\mathrm{K}}+RT\cdot \ln \frac{p_{\mathrm{gas}}}{p_{\mathrm{ref}}}$$

where G_CO2 and G_CO are the free energy of CO₂ and CO. G_gas is the free energy of the gas molecule, $E_{\mathrm{gas}}^{\mathrm{DFT}}$ is the DFT energy of the gas molecule, ZPE_gas is the zero-point vibrational energy of the gas molecule, $\text{Δ}{H}_{\mathrm{gas}}^{0\mathrm{K}\to 533.15\mathrm{K}}$ is the enthalpy change of the gas molecule from 0 to 533.15 K, T is set as 533.15 K according to the initial carburization temperature for K-Co₃O₄ sample, and $S_{\mathrm{gas}}^{533.15\mathrm{K}}$ is the entropy of the gas molecule at 533.15 K, R is the gas constant, p_gas is the partial pressure of the gas, and p_ref is the reference pressure, which is set to 1 bar. The following temperature and partial pressures were used in determining the chemical potential of O and C in Fig. 4B: red dot (533.15 K, 1.45 bar CO₂, 1 × 10⁻¹⁰ bar CO), black dot (533.15 K, 1.45 bar CO₂, 1.645 × 10⁻³ CO). The pressure of CO was set to 1 × 10⁻¹⁰ bar to represent that it does not exist in the initial reaction condition, while all other partial pressures were measured from our experiments.

The average formation energy ∆E_form of each C atom adsorption on the CoO (200) surface with or without the co-adsorption of K₂O is defined with the following equation

$$\text{Δ}{E}_{\mathrm{form}}=(E_{\mathrm{total}}-E_{\mathrm{sur}}-z\cdot {\mathrm{\mu }}_{\mathrm{C}}^0)/z$$

where E_total is the total DFT energy of a surface adsorption structure with z C atoms, E_sur is the DFT energy of the clean CoO surface (either with or without co-adsorbed K₂O). z is set as 2 and 3, while ${\mathrm{\mu }}_{\mathrm{C}}^0$ is set as −8.16 eV on the basis of the above calculation result of the chemical potential of carbon.

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S40

Tables S1 to S10