Potassium-stabilized metastable carbides and chalcogenides via surface chemical potential modulation

Abstract

Metastable carbides and chalcogenides are attractive candidates for wide and promising applications. However, their inherent instability leads to synthetic difficulty and poor durability. Thus, the development of facile strategies for the controllable synthesis and stabilization of metastable carbides is still a great challenge. Here, taking metastable ɛ-Fe₂C as a case study, potassium ions (K⁺) are theoretically predicted and experimentally reported to control the synthesis of metastable ɛ-Fe₂C from an Fe₂N precursor by increasing the surface carbon chemical potential (μ_C). The controllable synthesis and improved stability are attributed to the better-matched denitriding and carburizing rates and the impeded spillover of carbon atoms in metastable ɛ-Fe₂C with high carbon contents due to the enhanced surface μ_C. In addition, this strategy is suitable for synthesizing metastable γ’-MoC, MoN, 1T-MoS₂, 1T-MoSe₂, 1T-MoSe_2xTe_2(1−x), and 1T-Mo_1−xW_xSe₂, highlighting the universality of the methodology. Impressively, gram-level scalable metastable ɛ-Fe₂C remains stable for more than 398 days in air. Furthermore, ɛ-Fe₂C exhibits remarkable olefin selectivity and durability for more than 36 h of continuous testing. This work not only demonstrates a facile, easily scalable, and general strategy for accessing various metastable carbides and chalcogenides but also addresses the synthetic difficulty and poor durability challenge of metastable materials.

Introduction

Metastable carbides and chalcogenides are widely demanded in various fields^1,2,3,4,5,6. For example, transition metal carbides (TMCs) with high carbon contents (e.g., ɛ-Fe₂C and γ’-MoC) are metastable carbide species that are promising candidates for CO_x hydrogenation, methanol reforming, and water‒gas shift reactions^7,8,9. However, the controllable synthesis of these metastable carbide species is still a great challenge^10,11. In addition, the inevitable restructuring that always occurs during the catalytic process also leads to undesirable on-stream stability¹², creating additional obstacles for the practical application of metastable carbide species.

Generally, synthetic difficulty and poor operating stability are associated with chemical equilibrium^13,14, which is influenced by the Gibbs free energy change (∆G) from a metastable state to a stable state. Under a specific system with the same temperature (T) and pressure, ∆G is related only to the chemical potentials (μ) of all the substances in the system, which are closely associated with the surface adsorbed species^15,16 (Fig. 1a). In terms of the reported strategies for constructing specific metastable carbides, μ is temporarily changed because of the extreme applied conditions (e.g., high pressure, rapid quenching, and narrow temperature range)^17,18. However, the chemical potential difference (∆μ) between the metastable and stable phases still drives structural evolution during catalytic operation. Specifically, for interstitial-type TMCs, the diffusion of carbon in the structure plays an important role in carbide phase conversion. For metastable carbides with high carbon concentrations, the lower surface carbon chemical potential (μ_C) under high-temperature catalytic conditions favors the diffusion of bulk carbon to form a relatively stable phase both thermodynamically and kinetically (e.g., ε-Fe₂C to χ-Fe₅C₂ or θ-Fe₃C, γ’-MoC to β-Mo₂C)^16,19,20. Therefore, exploring a strategy to modulate the surface μ_C during both carbide synthesis and operation to prevent carbon loss is pivotal for the controllable synthesis and stabilization of metastable TMCs.

Fig. 1: Calculation-guided design of metastable ɛ-Fe₂C.

a Schematic of the classical strategy for synthesizing metastable species and our proposed K⁺-modified strategy by modulating the chemical potential μ and reaction equilibrium. IP initial phase, R reactants, MP metastable phase, SP stable phase. b Schematic of the phase diagram of the Fe‒C system. c DFT calculation of the bulk phase evolution (∆μ_C), bulk over the Fe₃O₄ and Fe₂N precursors. d Free-energy diagram of the construction process over Fe₂N and K-Fe₂N under CO₂ and H₂ atmospheres. Purple: K, red: O, blue: N, pink: H, light brown: C, and brown: Fe. Visuals were produced with VESTA⁵⁰.

In this article, the synthesis of metastable ɛ-Fe₂C is chosen as a case study because of its predicted high activity but weak stability towards selective CO₂ hydrogenation to light olefins. Iron nitride (Fe₂N) was first selected as the precursor for subsequent carbonization through density functional theory (DFT) calculations because of its lowered carburizing barrier and similar crystal structure to that of the target ɛ-Fe₂C. DFT calculations were also employed to verify the feasibility of potassium ion (K⁺) modulation of the surface-adsorbed carbon-containing species for enhancing the surface μ_C (Fig. 1a), which is beneficial for the controllable reconstruction of Fe₂N to ɛ-Fe₂C. Then, ɛ-Fe₂C was successfully obtained through in situ reconstruction of the designed K⁺-modified Fe₂N loaded on carbon (K-Fe₂N@C). Moreover, the combined results of DFT calculations and operando characterizations demonstrate that the controllable construction of ɛ-Fe₂C is ascribed to the enhanced surface μ_C and the consequent better-matched denitriding and rates, which lowers the formation energy barrier of ɛ-Fe₂C and prevents carbon atom spillover. Furthermore, the K⁺-promoted μ modulation strategy has been proven to be universal for the controllable synthesis of other metastable TMCs, nitrides, and chalcogenides (e.g., metastable γ’-MoC, MoN, 1T-MoS₂, 1T-MoSe₂, 1T-MoSe_2xTe_2(1−x), and 1T-Mo_1−xW_xSe₂). Finally, as a proof-of-concept application, easily scalable ɛ-Fe₂C enables improved olefin selectivity (42.3%) and a prolonged lifetime of over 36 h.

Results

Design of K⁺-modulated Fe₂N reconstruction for the synthesis of ɛ-Fe₂C

Metastable ɛ-Fe₂C is selected as the case study because it is theoretically predicted to be an appealing active species for CO₂ hydrogenation to light olefins, but still suffers from synthetic difficulty and instability concerns for most metastable materials. However, the instability of metastable materials makes their direct synthesis and stabilization highly challenging. Using the controllable synthesis of metastable ɛ-Fe₂C as a case study, carbonization kinetics are positively correlated with temperature (T), but the C atom adsorption energy of metastable ɛ-Fe₂C rapidly decreases with increasing T (Fig. 1b), resulting in difficulty in its synthesis and stabilization. Thus, two factors should be taken into consideration in the controllable synthesis and stabilization of ɛ-Fe₂C: (1) lowering the energy barrier of the carburization process to decrease T and (2) enhancing the surface μ_C (referred to as the negative shift in the μ_C value in this work) to increase carburization and prevent carbon loss from the bulk. Since we previously demonstrated that iron carbides can be a precursor, similar to Fe and/or Fe-based oxides, for the synthesis of iron carbides^7,21,22. However, the specific principles for precursor selection and the mechanism of stabilized ɛ-Fe₂C synthesis remain unclear. Here, ∆μ_{C, bulk} is defined as the energy barrier descriptor for the bulk carburization of different Fe-based precursors. The results of the DFT calculations indicated that the ∆μ_C of the bulk of Fe₂N (−10.02 eV) is much lower than that of its Fe₃O₄ and Fe counterparts (−9.85, −9.03 eV) (Fig. 1c, Supplementary Figs. 1 and 2, and Supplementary Note 1), indicating that Fe₂N is a suitable precursor for constructing metastable ɛ-Fe₂C^20,23. However, the formation trend of χ-Fe₅C₂ is still greater than that of the target ɛ-Fe₂C. Since the surface μ_C is highly related to the adsorbed carbon-containing intermediates, the K⁺ promoter is adopted to further increase carbonization and improve durability by modifying the adsorbates and enhancing the surface μ_C during the reconstruction and operation process^24,25. The feasibility of the K⁺-modulation strategy was subsequently verified by the energy change profiles of the transition states (TSs) of the carbonization process (Fig. 1d and Supplementary Figs. 3−13). Owing to the oxygen affinity exhibited by K atoms (insets corresponding to TS6 and TS7 in Fig. 1d), the dehydration process generating *CO and *C is more favorable over Fe₂N with K⁺ than that without K⁺, demonstrating the easy formation of surface carbon species, thus leading to a lower carburizing T and higher surface μ_C (Supplementary Figs. 3 and 4 and Supplementary Note 3). These calculated results reveal that the K⁺-modulated Fe₂N reconstruction strategy is promising for the controllable formation and subsequent stabilization of metastable ɛ-Fe₂C.

Synthesis and characterization of metastable ɛ-Fe₂C

To validate our calculation results, K-Fe₂N@C and its counterpart Fe₂N@C were synthesized via self-template impregnation-nitridation of the MIL-88 precursor with and without K₂CO₃ as the K⁺ source, respectively (Fig. 2a). The peaks in the X-ray diffraction (XRD) patterns of the obtained materials matched well with those of Fe₂N (ICDD-PDF-4 no. 01-072-2126), demonstrating the successful preparation of the Fe₂N precursors (Supplementary Figs. 14−18, Supplementary Notes 4 and 5, and Supplementary Table 1). Next, these iron nitrides were carburized under a 25% CO₂/H₂ atmosphere at 350 °C for 5 h to form iron carbides. As illustrated in Fig. 2b, the diffraction peaks of the counterpart without K⁺ are indexed to χ-Fe₅C₂ (ICDD-PDF-4 no. 01-080-9890) (denoted as Fe₅C₂@C), whereas those of ɛ-Fe₂C (ICDD-PDF-4 no. 00-036-1249), which features a similar crystal structure to that of the Fe₂N precursor, are obtained for the K⁺-containing sample (denoted as K-Fe₂C@C). The above XRD results primarily prove the direct construction of metastable ɛ-Fe₂C species. X-ray photoelectron spectroscopy (XPS) was subsequently conducted to further identify the two obtained iron carbide species. The characteristic peaks with lower binding energies in the XPS C 1s spectra and Fe 2p spectra are typical signals of Fe‒C bonds. The two groups of characteristic peaks over samples with and without the K⁺ promoter are typically attributed to χ-Fe₅C₂ and ɛ-Fe₂C species, respectively (Fig. 2c, d, Supplementary Figs. 19 and 20, and Supplementary Note 6)^26,27,28,29. Moreover, X-ray absorption spectroscopy (XAS) was employed to determine the local coordination environment of the as-prepared iron carbide samples. Fe K-edge XANES suggested that both samples exhibited similar oxidation states and had similar carbon atom neighbors with shorter Fe−C bonds around the central Fe atom (Supplementary Figs. 21 and 22). However, as evidenced by the extended X-ray absorption fine structure (EXAFS) spectra (Fig. 2e), the Fe−C coordination shells of K-Fe₂C@C (1.44 Å) are lower than those of Fe₅C₂@C (1.50 Å), whereas the average first Fe−Fe shell bond length narrows from 2.08 Å over Fe₅C₂@C to 2.21 Å over K-Fe₂C@C^30,31. These opposite trends confirm that a difference in the carbon content exists between these two carbide samples because the incorporation of more C atoms into the lattice shortens the Fe−C bonds and elongates the Fe−Fe bonds. Mössbauer spectroscopy, with high resolution for the chemical environment of the Fe atom, further proves the conversion from Fe₂N with K⁺-modification to metastable ɛ-Fe₂C, whereas the main carbide phase of the K⁺-free counterpart remains χ-Fe₅C₂ (Fig. 2f, g and Supplementary Table 2)³². These results together verify the successful construction of metastable ɛ-Fe₂C species and suggest that the reconstruction product of the Fe₂N precursor under a CO₂–H₂ atmosphere could be regulated from the stable χ-Fe₅C₂ phase to the metastable ɛ-Fe₂C phase via a simple K⁺-modulation strategy.

Fig. 2: Synthesis and characterizations of metastable ɛ-Fe₂C.

a Scheme for the synthesis of ɛ-Fe₂C from K-Fe₂N (red arrow: K⁺-modified strategy). Red atoms: Fe, green atoms: C. Visuals of ɛ-Fe₂C and χ-Fe₅C₂ were produced with VESTA⁵⁰ and Blender. b XRD patterns of Fe₅C₂@C and K-Fe₂C@C. c, d XPS C1s spectra and Fe 2p spectra of Fe₅C₂@C and K-Fe₂C@C. e EXAFS spectra of Fe₅C₂@C and K-Fe₂C@C. f, g Mössbauer spectra of Fe₅C₂@C and K-Fe₂C@C. All “a.u.” represents arbitrary units.

Mechanistic studies on the K⁺-mediated controllable synthesis of ɛ-Fe₂C

To gain insight into the mechanism of K⁺-mediated Fe₂N selective reconstruction into metastable ɛ-Fe₂C, a series of in situ spectral techniques were used to monitor the evolution behaviors of Fe₂N@C and K-Fe₂N@C. For in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS), which detects changes in surface adsorbed species, the spectrum obtained when the temperature reached 350 °C under an Ar atmosphere was designated as the background, followed by time-dependent signal collection under a 25% CO₂/H₂ atmosphere (Fig. 3a–c). Although the intensities of the band at approximately 1448 cm⁻¹ (*COOH), which is related to the surface μ_C, increase markedly over both samples, more significant *COOH bands compared with the signal of inert *CO₃²⁻ over K-Fe₂N@C than that over Fe₂N@C suggest that more carbon-containing species exist after K⁺-modulation^33,34,35,36. The higher surface carbon-species concentration over K-Fe₂N@C is also substantiated by the CO-temperature-programmed desorption (TPD) and in situ Raman spectra (Supplementary Figs. 23 and 24 and Supplementary Notes 7 and 8). These results primarily indicate that the K⁺ promoter increases the surface μ_C and thermodynamically facilitates the synthesis of metastable ɛ-Fe₂C. In addition, the bands at 1608 cm⁻¹ (*N–H) and 1107 cm⁻¹ (*NH₃) were detected as inverted peaks during the acquisition time, indicating denitrification^37,38. The 50% attenuation point of the *N−H species was subsequently used as a descriptor for denitrification kinetics³⁹. As shown in Fig. 3c, it only takes 2.5 min for Fe₂N@C to reach the 50% attenuation point, whereas 3.8 min is needed for K-Fe₂N@C, at which the intensity percentages of *COOH, which is deemed the indicator of surface μ_C, are 38% and 47%, respectively. In other words, the K⁺ promoter delays the denitrification process, making it match better with a high surface μ_C. The change in surface μ_N after K⁺ modulation also supports this view (Supplementary Figs. 25 and 26 and Supplementary Note 9). In addition, the in situ XRD results provided further information on the crystal structure (Fig. 3d, e). For the K-Fe₂N@C precursor, the diffraction peaks of Fe₂N clearly shift to higher Bragg angles after 30 min at 350 °C because of the nitrogen spill, after which the predominant diffraction peak at 43.0° gradually vanishes as the profile shape of ɛ-Fe₂C becomes dominant. Conversely, the species construction over Fe₂N@C shifted to an earlier time and started after 17.2 min. The formation of χ-Fe₅C₂ species is accompanied by the typical diffraction of Fe (ICDD-PDF-4 no. 97-006-4795) at 44.5°, suggesting that the onset of denitrification is too premature to match carburization when K⁺ is lacking. The in situ XRD results also indicate that there is a matching problem. Accordingly, a CO₂-H₂-temperature-programmed surface reaction with mass spectrometry (TPSR‒MS) was conducted to further provide direct evidence of denitrification‒carbonization kinetics (Fig. 3f, g, Supplementary Figs. 27 and 28, and Supplementary Note 10). The signal of N-related escaped species appears ca. 210 °C over Fe₂N@C, whereas CO₂ remains unactivated until 315 °C, reaching a temperature gap (ΔT) as wide as 105 °C (Fig. 3f), making the reconstruction of Fe₂N mismatched with the high surface μ_C period. Then, the subsequent formation of carbides takes a time gap (Δt) of 5.8 min, which is indicated by the relaxation time of the N- and C-related escaped species. Notably, after K⁺ modulation, denitrification is delayed to 305 °C, which decreases ΔT to 45 °C, and the carburization process is accelerated since Δt is reduced to 4.7 min (Fig. 3g), which accords with the results obtained for the calculated surface μ_C (Supplementary Fig. 27). Thus, the more matched kinetics of the denitrification‒carbonization process enable the carbides to form at a high surface μ_C. To further determine the reason for the matched kinetics, the hydrogenation process of the N site over Fe₂N with and without the K⁺ promoter was calculated by DFT (the simulation details are displayed in the computational details section). The modification of K⁺ benefits the first N−H bond formation but increases the energy barrier of the second hydrogenation, impeding the entire denitrification process (Fig. 3h and Supplementary Figs. 29−35). In addition, the K⁺ promoter benefits surface carbon atom insertion, and the nitrogen atom escapes from the bulk phase (Fig. 3i and Supplementary Figs. 36−40). All these results are summarized in Fig. 3j and emphasize the importance of the K⁺ promoter in enhancing surface μ_C to affect denitrification-carbonization kinetics for the controllable construction and stabilization of ɛ-Fe₂C species under high surface μ_C conditions. Moreover, it should be noted that the direct construction of metastable ɛ-Fe₂C could be mediated not only by K⁺ but also by other alkalies such as Li⁺, Na⁺, Rb⁺, and Cs⁺ (Supplementary Fig. 41 and Supplementary Note 11). Additionally, the construction conditions are not confined to Fe₂N annealing in a CO₂-H₂ atmosphere. Fe₂O₃ can also act as an iron precursor and partially evolve into ɛ-Fe₂C with K⁺ modulation in a CO atmosphere (Supplementary Fig. 42 and Supplementary Note 12). Furthermore, under a wide range of temperatures, K⁺ and CO₂ concentrations (Supplementary Figs. 43–48 and Supplementary Notes 13–16), the K⁺-induced surface chemical potential modulation strategy is capable of inducing the formation of metastable ɛ-Fe₂C, further confirming its general applicability and potential.

Fig. 3: Mechanistic studies.

a, b In situ DRIFTS spectra of Fe₂N@C (a) and K-Fe₂N@C (b) in a 25% CO₂/H₂ atmosphere at 0.1 MPa. c Time course of the relative percentage of the infrared bands of Fe₂N@C and K-Fe₂N@C. Solid lines were obtained by fitting signals according to the Hill equation^51,52. d, e In situ XRD patterns of Fe₂N@C and K-Fe₂N@C recorded in 25% CO₂/H₂ at 350 °C and 0.1 MPa. f TPSR−MS profile of Fe₂N@C and g K-Fe₂N@C in a 25% CO₂/H₂ atmosphere at 0.1 MPa. “a.u.” represents arbitrary units. h, i Free-energy diagram of the N site hydrogenation process and carbon‒nitrogen atom exchange process over Fe₂N with or without K⁺. j Scheme showing how the K⁺ promoter affects the construction of the Fe₂C phase. Visuals were produced with VESTA⁵⁰ and Blender.

Methodology universality

These aforementioned results of metastable ɛ-Fe₂C synthesis imply that the modulation of the surface μ_C is an intrinsic factor and is of critical importance for metastable carbide generation via other alkali ion-modulation strategies, such as the use of K+ ions. Thus, to further explore the universality of the proposed strategy and mechanism, molybdenum (Mo)-based metastable species (e.g., γ’-MoC, MoN, and 1T chalcogenides) with wide applications were chosen as the target products^40,41,42,43. First, the surface μ_X values of these interstitial atoms (X = O, N, C, S) over Mo with and without the K⁺ promoter were further studied via DFT calculations (Supplementary Fig. 49). The DFT results reveal that the addition of K⁺ enhances the surface μ_X and then the reaction equilibrium of the corresponding Mo-based materials (Fig. 4a and Supplementary Table 3). The values of ∆μ_{X, O} with and without K⁺ as descriptors of the energy barriers for MoX formation from the MoO_x precursor are displayed in Fig. 4b. The ∆μ_{X, O} values of X = C and N negatively shift after K⁺ modulation, which indicates that the energy barriers for carburization and nitridation from MoO_x are lowered by the K⁺ promoter, benefiting the formation and stabilization of metastable Mo-based carbides and nitride species with relatively high carbon or nitrogen contents. In addition, although the changes in the ∆μ_S, O values are not obvious, the ∆μ_S after K⁺ modulation exhibits the same trend as the ∆μ_C and ∆μ_N, implying that the sulfuration process could also increase thermodynamically, which may also induce the formation of metastable chalcogenide species (Supplementary Table 3). Then, as a proof of concept, MoO₃ and K-MoO₃ were treated under carbonization conditions (Supplementary Fig. 50 and Supplementary Notes 2 and 17). As shown in Fig. 4c, metastable γ’-MoC is successfully obtained from K-MoO₃, whereas when MoO₃ is used as a precursor without the K⁺ promoter, only β-Mo₂C can be synthesized (Supplementary Figs. 51 and 52 and Supplementary Notes 18 and 19). Similarly, the construction of metastable MoN, 1T-MoS₂, 1T-MoSe₂, 1T-MoSe_2xTe_2(1−x), and 1T-Mo_1−xW_xSe₂ was also achieved over K⁺-modified MoO₃ under identical conditions (Fig. 4d−m, Supplementary Figs. 53−63, Supplementary Notes 20−30, and Supplementary Table 4)^44,45,46,47, demonstrating the universality of the proposed strategy. Such mechanistic insight might provide a supplementary interpretation for the phase engineering strategy for constructing the 1T-Mo(W)S(Se, Te)₂ materials proposed by Zhang et al.^2,3.

Fig. 4: Methodology universality.

a, b Calculation of the chemical potential change of different atoms on the MoO₃ surface with and without K⁺. c–l XRD patterns and Raman spectra of the carbonization, nitrogenization, sulfuration, selenylation, and antimonization of MoO₃ and K-MoO₃. m Scheme for the K⁺ promoter-mediated construction of the metastable phase. All “a.u.” represents arbitrary units.

Proof-of-concept application of the proposed K⁺-modulation strategy

The gram-scale synthesis of ɛ-Fe₂C, MoN, and γ’-MoC could be realized by using a tubular furnace with a diameter of 6 cm (Fig. 5a, Supplementary Fig. 64, and Supplementary Note 31), suggesting its advantages for easy scaling up and considerable application potential. Therefore, to evaluate the stability of metastable ɛ-Fe₂C, we first studied ∆μ_O, C theoretically and found that there was a prominent increase in ∆μ_O, C after K⁺ modulation, suggesting that the oxidation resistance of the as-prepared metastable ɛ-Fe₂C was markedly increased through our strategy (Fig. 5b, c, Supplementary Fig. 65, and Supplementary Table 5). As predicted, the synthesized K-Fe₂C@C exhibited outstanding stability, with the characteristic peaks of ɛ-Fe₂C remaining distinguishable even when exposed to air for more than 398 days (Fig. 5d, Supplementary Fig. 66, and Supplementary Note 32).

Fig. 5: Scalable synthesis and a proof-of-concept application.

a Verification of the gram-level synthesis. b. Calculation of the chemical potential change of O and C on the Fe₂O₃ surface with and without K⁺. c XRD pattern and XPS Fe 2p spectra of K-Fe₂C@C exposed to air for 398 days. d Hydrocarbon selectivity and CO₂ conversion rate over K-Fe₂C@C and K-Fe₂C@C-I₂ (the sample in which I₂-acetonitrile was used to remove K from K-Fe₂C@C). “a.u.” represents arbitrary units. e–g CO-free hydrocarbon selectivity and CO₂ conversion over K-Fe₂C@C and K-Fe₂C@C-I₂. Reaction conditions: 300 °C, 1.0 MPa, space velocity = 20,000 mL g⁻¹ h⁻¹, H₂/CO₂ = 3. The inset image in g is the XRD pattern of the tested K-Fe₂C@C-I₂ after 6 h. “C₂⁼–C₄⁼” represents light olefins, “C₂^o–C₄^o” represents light paraffins, “Con.” represents conversion, “Sel.” represents selectivity, and O/P indicates the olefin/paraffin ratio of C₂–C₄ production. K-Fe₂C@C-I₂ is the sample that uses I₂-acetonitrile to remove K⁺ from K-Fe₂C@C.

CO₂ hydrogenation was used as a model reaction to evaluate the catalytic potential of the as-prepared metastable ɛ-Fe₂C materials. As predicted in early reports, K-Fe₂C@C exhibited better CO₂ hydrogenation activity and lower olefin hydrogenation activity than Fe₅C₂@C during transient surface reactions (Supplementary Fig. 67 and Supplementary Note 33). The quantification performance comparison is displayed in Fig. 5e, Supplementary Figs. 68 and 69, and Supplementary Note 34, and Supplementary Table 6. The equilibrium CO₂ conversion rate reached 34.9%, with outstanding C₂⁼–C₄⁼ (light olefins) selectivity and O/P ratios of 42.3% and 8.7, respectively, achieved over K-Fe₂C@C, which are obviously greater than those achieved over Fe₅C₂@C (Con. rate, 28.2%; C₂⁼–C₄⁼ Sel., 20.4%; O/P, 1.2). Moreover, the performance and morphology of K-Fe₂C@C remained unchanged after the 36 h durability test (Fig. 5f, Supplementary Fig. 70, and Supplementary Note 35), demonstrating the considerable operating stability of the metastable species obtained from the K⁺-modulation strategy. The dynamic mutual transformation of ɛ-Fe₂C and χ-Fe₅C₂ after removing/re-adding the K⁺ promoter during the reaction further verify its crucial effect on metastable construction and stabilization (Fig. 5g, Supplementary Fig. 71, and Supplementary Note 36). Notably, this phenomenon implies the important role of the K⁺ promoter in the evolution of metastable active species, which are often overlooked, in the CO₂ hydrogenation process rather than just in electronic structure modification.

Discussion

In summary, modulating the surface carbon chemical potential is theoretically reported to be the crucial factor and descriptor for the synthesis of metastable carbides. A K⁺-modulation strategy is subsequently developed to convert iron nitride precursors to metastable ɛ-Fe₂C. An increased surface carbon chemical potential leads to a better-matched denitriding–carburizing rate and impedes carbon spillover because of the thermodynamically lower formation energy barrier, which kinetically guarantees enhanced stability. Moreover, this K⁺-induced surface chemical potential modulation is suitable for the controllable preparation of metastable γ’-MoC, MoN, 1T-MoS₂, 1T-MoSe₂, 1T-MoSe_2xTe_2(1−x), and 1T-Mo_1−xW_xSe₂ from K⁺-modified MoO₃ and easy scaling to gram-level synthesis, highlighting the universality and applicability of the methodology. Impressively, metastable ɛ-Fe₂C can be stable for more than 398 days under an air atmosphere. Furthermore, as a proof of concept, a considerable olefin selectivity of 42.3%, with a lifetime of over 36 h, is achieved during the process of CO₂ hydrogenation over in situ-formed ɛ-Fe₂C, demonstrating its remarkable operating durability. Our work not only demonstrates a simple surface chemical potential modulation method for constructing highly stable metastable carbides and chalcogenides but also rethinks the real role of alkaline promoters in synthetic and catalytic processes.

Methods

Materials

Iron(III) chloride hexahydrate (FeCl₃·6H₂O), potassium carbonate (K₂CO₃), dimethylformamide (DMF), and terephthalic acid were purchased from Aladdin Ltd. (Shanghai, China). The mixed gases of carbon dioxide and hydrogen (CO₂/H₂ with a volume ratio of 1/3) and ammonia (NH₃) were purchased from Lian Bo (Tianjin) Co., Ltd. Deionized water (DIW) was used in all the experimental processes. All chemicals were of analytical grade and were used without further purification.

Synthesis of MIL-88 precursors

According to the literature⁴⁸, a typical procedure is as follows: First, 0.81 g of FeCl₃·6H₂O and 0.50 g of terephthalic acid were dissolved in 50 mL of DMF. The obtained solution was then vigorously stirred for 15 min. Afterward, the mixture was transferred to a Teflon-lined stainless steel autoclave and crystallized for 24 h at 110 °C. Finally, the dark yellow product was harvested by centrifugation, washed with DIW and ethanol several times, and dried in a vacuum oven overnight.

Synthesis of carbon-wrapped iron nitride with (K-Fe₂N@C) and without (Fe₂N@C) K⁺ modulation

K-Fe₂N@C was obtained by facile impregnation of K species and subsequent calculations under an NH₃ atmosphere. First, K₂CO₃ was added to a 1 g mL⁻¹ solution. Owing to the water absorption of the MIL-88 precursor, 3 μL of potassium solution was added dropwise to 0.1 g of MIL-88 precursor with vigorous stirring until it became sticky. The mixture was subsequently dried at 80 °C, and the obtained powder was placed in a 6 mm diameter quartz tube and put into the tube furnace. The temperature was subsequently increased to 500 °C at a heating rate of 5 °C min⁻¹ under an NH₃ atmosphere for 5 h to obtain K-Fe₂N@C. After cooling to room temperature, the sample was sealed under an Ar atmosphere to prevent oxidation. Fe₂N@C was synthesized via a process similar to that described above without the addition of K species.

Synthesis of carbon-wrapped iron carbide with (K-Fe₂C@C) and without (Fe₅C₂@C) K⁺ modulation

K-Fe₂C@C was obtained by in situ carburization under a 25% CO₂/H₂ atmosphere. A 0.3 g sample of sieved (40–80 mesh) K-Fe₂N@C precursor was put into a stainless-steel tube reactor with an inner diameter of 8 mm. Then, the temperature was increased to 350 °C at a heating rate of 5 °C min⁻¹ under a 25% CO₂/H₂ atmosphere for approximately 2 h. After cooling to room temperature, the samples could be directly used for subsequent CO₂ hydrogenation or removed for subsequent characterization. In addition, Fe₅C₂@C could be synthesized via a similar process using Fe₂N@C as a precursor.

General characterizations

Powder X−ray diffraction (XRD) was performed on a Bruker D8 Focus Diffraction System (Germany) using a Cu Kα source (λ = 0.154178 nm). Scanning electron microscopy (SEM) was conducted with an FEI Apreo S LoVac microscope (10 kV). Transmission electron microscopy (TEM) and high-resolution transmission electron microscopy (HRTEM) images were obtained with a JEM-2100F system equipped with an EDAX Genesis XM2. X−ray photoelectron spectroscopy (XPS) measurements were conducted with a PHI-1600 X-ray photoelectron spectrometer equipped with Al Kα radiation. All the peaks were attributed to the C 1s peak at a binding energy of 284.8 eV. Raman spectroscopy was carried out with a Horiba Labram HR Evolution Raman microscope under excitation with a 532 nm laser at a power of 20 mW. Fourier transform infrared spectroscopy (FTIR) was performed on a FEI iS50 instrument. Before analysis, the samples were degassed at 120 °C under vacuum overnight. X−ray absorption fine structure spectroscopy (XAFS) was conducted at the 1W1B station of the Beijing Electron–Positron Collider II (BSRF). The room temperature 57Fe Mössbauer spectra of Fe₅C₂@C and K-Fe₂C@C were recorded under various simulated conditions by using a Topologic 500 A Mössbauer spectrometer with a 57Co (Rh) radiation source and proportional counter detector. The speed was calibrated with α-Fe foil, and the absorption thickness of the sample was adjusted to 10 mg Fe cm⁻². The analysis of the Mössbauer spectrum was completed on the basis of the Lorentz absorption curve via a computer and MossWinn 3.0i software. Qualitative analysis of the various phases in the catalyst was carried out by fitting the parameters of the isomorphic energy shift (IS), quadrupole splitting (QS), absorption linewidth (LW), and hyperfine field (H_hf). The relative content of various iron phases in the catalyst was determined by integrating the absorption peak area of each phase. Nano secondary ion mass spectroscopy (NanoSIMS) was performed on a CAMECA nanoSIMS 50L (France) using a Cs⁺ source.

In situ XRD experiments

In situ XRD measurements were conducted on a Smartlab8KW Diffraction System using a Cu Kα source (λ = 0.154178 nm) with an XRK 900 heater. The experiments were carried out under conditions close to those of the carbonization process, i.e., mixed gas with H₂/CO₂ volume ratios of 3 and 350 °C at ambient pressure for 70 min. Diffraction patterns were recorded within a 2θ range of 30–50° because the most intense diffraction peaks of the relevant phases, i.e., Fe₂N, Fe₂C, and Fe₅C₂, mainly fall in this range. The scanning rate was 20°/min.

In situ DRTIFTS measurements

In situ DRIFTS measurements were conducted in continuous flow mode under an atmosphere of mixed H₂ and CO₂ gas with a volume ratio of 3. Specifically, the catalysts and anhydrous KBr were ground together at a ratio of 1/5 and degassed at 450 °C under a He atmosphere for 2 h. Then, T was decreased to 350 °C, and the reaction gases (H₂ and CO₂ with a ratio of 4/1) were introduced into the reactor. Next, the background spectrum was collected under the same conditions, and the spectrum was recorded as a function of time to monitor the intermediates adsorbed on the catalyst surface during the on-stream reaction.

In situ Raman analysis

In situ Raman spectroscopy was performed via a Raman microscopy system (LabRAM HR Evolution, Horiba Jobin Yvon) with a 532 nm Ar ion laser beam. The powdered sample was added to the reactor, which was ramped up to 450 °C with N₂ for 2 h to degas. Afterward, T was decreased to 350 °C, and a 25% CO₂/H₂ atmosphere was introduced into the reactor for spectrum acquisition.

TPD/TPSR-MS experiments

TPD/TPSR-MS experiments were carried out in a TP-5076 characterization system (Xianquan Co., Ltd., Tianjin, China), and a QAS mass spectrometer (Linglu Instruments, Shanghai, China) was used to analyze the reactants and products online. The m/z values detected were as follows: 2 for H₂, 14 for the released N species, 28 for ethylene, 30 for ethane, and 44 for CO₂. For CO-TPD, approximately 100 mg of catalyst was first treated under an Ar flow (20 mL/min) at 60 °C for 30 min, degassed at 300 °C for 2 h using a 20 mL/min Ar flow, and then cooled to room temperature. The sample was purged with 10% CO/Ar mixed gas (20 mL/min) for 1 h under ambient conditions. After adsorption saturation, the sample was heated to 450 °C under an Ar flow (20 mL/min) at a rate of 10 °C min⁻¹. The amount of CO flowing out of the sample was monitored by a thermal conductivity detector (TCD). For the TPSR-MS analysis, the same amount of catalyst was first degassed for the CO-TPD measurements. Then, the temperature was decreased to 60 °C, and the sample was exposed to 25% CO₂/H₂ (20 mL/min) at 60 °C for approximately 15 mins. After that, the catalyst was heated to 350 °C at a rate of 10 °C min⁻¹. The tail gas was monitored by MS.

Computational details

In this work, all DFT calculations were performed via the Vienna Ab initio Simulation Package (VASP). The projector augmented wave (PAW) pseudopotential with the PBE generalized gradient approximation (GGA) exchange correlation function was utilized in the computations. The cut-off energy of the plane wave basis set was 500 eV, and a Monkhorst-Pack mesh of 3 × 3 × 1 was used for K-sampling in the adsorption energy calculation. The long-range dispersion interaction was described via the DFT-D3 method. All the atoms were fully relaxed with an energy convergence tolerance of 10⁻⁵ eV per atom, and the final force on each atom was <0.01 eV Å⁻¹. The TS searches were performed via the Dimer method in the VTST package. The final force on each atom was <0.05 eV Å⁻¹. The TS search is conducted by using the climbing-image nudged elastic band (CI-NEB) method to generate initial guess geometries, followed by the dimer method to converge to the saddle points.

The adsorption energy of the reaction intermediates can be computed via Eqs. (1 and 2):

$$\Delta E={E}_{\left({ads}/{slab}\right)}{-E}_{{ads}}-{E}_{{slab}}$$

(1)

$$\Delta G=\Delta E+{\Delta E}_{{ZPE}}-T\Delta S$$

(2)

where ∆E_ZPE is the zero-point energy change and ∆S is the entropy change. In this work, the values of ∆E_ZPE and ∆S are obtained via vibration frequency calculations via vaspkit⁴⁹. Visuals were produced with VESTA⁵⁰ and Blender.

To simplify the calculation process, we did not consider the adsorption and desorption processes of molecules such as CO₂ and H₂ here, but focused on the surface reaction process. H₂ is believed to decompose into *H and adsorb on the surface before the reaction. Owing to the presence of 50% unfilled N vacancies on the Fe₂N surface, CO₂ can be reduced to C in the unfilled vacancies, while the N sites can be further hydrogenated to NH₃ and escape from Fe₂N precursors. Thus, the whole conversion process of iron nitrides to the corresponding carbides is decoupled into two processes for convenient simulation: the surface N sites of Fe₂N are gradually hydrogenated to NH₃ and escape, and then the *C formed through CO₂ permeates into the sublayer for the formation of iron carbides. The energy of the reaction can be calculated via Eqs. (3−15):

$$ \ast+{{{\rm{CO}}}}_{2}+4*{{\rm{H}}}\to*{{{\rm{CO}}}}_{2}+4*{{\rm{H}}}$$

(3)

$$*{{{\rm{CO}}}}_{2}+4*{{\rm{H}}}\to*{{\rm{COOH}}}+3*{{\rm{H}}}\,({{\rm{TS}}}1)$$

(4)

$$*{{\rm{COOH}}}+3*{{\rm{H}}}\to*{{\rm{CO}}}+\ast {{{\rm{H}}}}_{2}{{\rm{O}}}+2*{{\rm{H}}}\,({{\rm{TS}}}2)$$

(5)

$$*{{\rm{CO}}}+\ast {{{\rm{H}}}}_{2}{{\rm{O}}}+2*{{\rm{H}}}\to*{{\rm{COH}}}+\ast {{{\rm{H}}}}_{2}{{\rm{O}}}+\ast {{\rm{H}}}\,({{\rm{TS}}}3)$$

(6)

$$*{{\rm{COH}}}+\ast {{{\rm{H}}}}_{2}{{\rm{O}}}+\ast {{\rm{H}}}\to*{{\rm{C}}}+\ast {{\rm{OH}}}+\ast {{{\rm{H}}}}_{2}{{\rm{O}}}\,({{\rm{TS}}}4)$$

(7)

$$*{{\rm{C}}}+\ast {{\rm{OH}}}+\ast {{{\rm{H}}}}_{2}{{\rm{O}}}\to*{{\rm{C}}}+2*{{{\rm{H}}}}_{2}{{\rm{O}}}\,({{\rm{TS}}}5)$$

(8)

$$*{{{\rm{CO}}}}_{2}+4*{{\rm{H}}}\to*{{\rm{CO}}}+\ast {{{\rm{H}}}}_{2}{{\rm{O}}}+2*{{\rm{H}}}\,({{\rm{TS}}}6)$$

(9)

$$*{{\rm{CO}}}+\ast {{{\rm{H}}}}_{2}{{\rm{O}}}+2*{{\rm{H}}}\to*{{\rm{C}}}+2*{{{\rm{H}}}}_{2}{{\rm{O}}}\,({{\rm{TS}}}7)$$

(10)

$$*{{\rm{N}}}+3*{{\rm{H}}}\to*{{\rm{NH}}}+2*{{\rm{H}}}\,({{\rm{TS}}}8)$$

(11)

$$*{{\rm{NH}}}+2*{{\rm{H}}}\to*{{{\rm{NH}}}}_{2}+\ast {{\rm{H}}}\,({{\rm{TS}}}9)$$

(12)

$$*{{{\rm{NH}}}}_{2}+\ast {{\rm{H}}}\to*{{{\rm{NH}}}}_{3}\,({{\rm{TS}}}10)$$

(13)

$$*{{{\rm{N}}}}_{{{\rm{bulk}}}}\to*{{{\rm{N}}}}_{{{\rm{surface}}}}\,({{\rm{TS}}}11)$$

(14)

$$*{{{\rm{C}}}}_{{{\rm{surface}}}}\to*{{{\rm{C}}}}_{{{\rm{bulk}}}}\,({{\rm{TS}}}12)$$

(15)

Performance evaluation

Flow reactor studies of CO₂ hydrogenation were performed in a stainless steel tube reactor with an inner diameter of 8 mm under a pressure of 1 MPa. For the steady-state experiments, 300 mg of sieved catalyst (40–80 mesh) was loaded into a reactor tube and held in place by quartz wool. The temperature was increased to 300 °C at a rate of 10 °C/min, and the samples were pretreated in the presence of reactants (36 mL min⁻¹ CO₂ + 144 mL min⁻¹ H₂). The gas was then changed to pure N₂ for degassing. After that, the reactants were introduced again. Finally, the products were subjected to a GC-2014 gas chromatograph equipped with a PLOT-Q column (with Ar as the carrier gas) and a flame ionization detector (FID) for analysis of the C₁–C₈ hydrocarbons. For CO, CO₂, and N₂, a thermal conductivity detector (TCD) with a 5 A molecular sieve column was used. The evolution rates of different products and CO₂ conversion rates were determined via Eqs. (16) and (17). All the experiments were repeated three times.

$$ {{\rm{Evolution\; Rate}}}({{\rm{\mu }}}{{\rm{mol}}}/{{\rm{g}}}/{{\rm{s}}})=({{\rm{peak}}}\; {{\rm{area}}}\; {{\rm{of}}}\; {{\rm{X}}}\times {{\rm{C}}})/\\ ({{\rm{peak}}}\; {{\rm{area}}}\; {{\rm{of}}}\; {{\rm{standard}}}\; {{\rm{gas}}}\times {{\rm{m}}})\times {F}_{X}$$

(16)

$$ {{{\rm{CO}}}}_{2}{{\rm{Conversion}}}(\%)=({{\rm{peak}}}\; {{\rm{area}}}\; {{\rm{of}}}\; {{\rm{B}}}-{{\rm{peak}}}\; {{\rm{area}}}\; {{\rm{of}}}\; {{\rm{A}}})/\\ ({{\rm{peak}}}\; {{\rm{area}}}\; {{\rm{of}}}\; {{\rm{B}}})$$

(17)

X: Different products, including H₂, CO, and C₁–C₈ hydrocarbons.C: The concentration of X in standard gas. m: The mass of the catalyst used. A: The area of the CO₂ outlet; B: The area of the CO₂ inlet. F_X: The flow rate of X.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.