Production of light olefins (C₂–C₄ olefins) via non-oil-based processes has motivated an increasing number of studies due to the growing demand for these key molecular building blocks and the concern regarding the diminishing oil reserves [1]. A promising route that first generates syngas (H₂ and CO) from alternative resources such as natural gas, coal, and renewable biomass and then converts the syngas to light olefins in direct or indirect processes has been widely investigated [1, 2]. Several indirect processes for converting syngas to light olefins have been developed and even commercialized, such as the methanol to olefins (MTO) process, but still have drawbacks of high equipment cost and high energy consumption due to the use of multiple steps in the process [1, 3, 4]. Recently, the direct Fischer-Tropsch to olefins (FTO) process using Fe- or Co-based catalysts has attracted significant interest [5-8]. However, due to the Anderson-Schulz-Flory (ASF) limit in Fischer-Tropsch synthesis, the maximum selectivity of C₂–C₄ olefins cannot exceed 60% [7, 8].

The limited selectivity for the process based on the Fe- or Co-based catalysts is mainly attributed to the uncontrollable coupling of the CH_x species at the same active sites where CO activation occurs, necessitating the development of novel strategies for the separation of CO activation and C–C coupling on different active sites. Bao et al. [9] reported an OX-ZEO (oxide-zeolite) process that separates CO activation on a reduced mixed metal oxide (ZnCrO_x) and C–C coupling in the zeolite (MSAPO) pores to reach a high C₂–C₄ olefins selectivity of 80% at a CO conversion rate of 17% at 673 K. Meanwhile, Wang et al. [10] developed a bifunctional catalyst composed of Zr–Zn oxide and SAPO-34 molecular sieve that offers a high C₂–C₄ olefins selectivity of 74% at a CO conversion rate of 11% at 673 K. Since these two pioneering works, intense studies of direct syngas conversion using the bifunctional OX-ZEO catalysts have been carried out. To date, various bifunctional catalysts have been developed by modifying both metal oxides and zeolites in order to improve the catalytic performance [11-17]. The increasing high selectivity is enabling the OX-ZEO process to become a potential competitor for industrial processes such as FTO and MTO [18]. In addition, further research effort has focused on the selective production of other hydrocarbons such as aromatics [19-25], and on the other hand, an increasing number of studies have been performed on the direct hydrogenation of CO₂ to light olefins using similar OX-ZEO catalysts [26-29].

Despite the significant progress achieved for the direct syngas conversion to light olefins over the OX-ZEO catalysts, some critical issues remain unresolved. For OX-ZEO catalysts, the mole ratio of the two metal elements in the mixed oxide, the role of the oxygen defects in the oxide, the density of the Brø nsted acid sites of the zeolites, and the proximity of the oxide and zeolite are the main factors controlling the CO conversion and light olefin selectivity that are still under study [13, 17]. Here, our research focuses on metal oxides since H₂ and CO are first converted to intermediates on the metal oxide surfaces [9, 10]. The two main issues related to our study are as follows. (1) The role played by the oxygen vacancies (V_Os) in metal oxides has not been fully elucidated even though intense experimental studies including the first two pioneering works mentioned above have indicated the importance of V_Os [9, 10]. While studies performed to date have demonstrated that oxygen vacancies are mainly responsible for CO activation [11, 13], detailed studies at the molecular level are needed to obtain deeper understanding of the role played by V_Os. (2) The debate about the reaction mechanism of CO hydrogenation for the formation of key intermediates such as ketene or methanol remains unresolved. Despite the partial evidence obtained by experimental studies, the two possible mechanisms proceeding via either the ketene or methanol intermediates are not fully understood [9, 10]. Hence, a mechanistic understanding of the generation of different intermediates at an atomic level is highly desirable.

To address the above-described issues, the selection of a suitable metal oxide as the research model is critical. Based on the experiments, the mixed metal oxides should be the first choice, but their crystal structures are too complex for use as a theoretical model [30]. Moreover, searching more complicated surface structures is still challenging due to the high computational cost. Interestingly, recent studies reported that the unary oxides such as MnO and ZnO [11, 16] also perform as well as the binary oxides such as ZnCrO_x, indicating that a unary oxide is sufficiently active for use in the OX-ZEO catalyzed direct syngas conversion to light olefins. Therefore, reducible unary oxides are considered in our studies because this approach avoids the need to treat the complex surface structures of mixed metal oxides.

Since the pioneering work by Stephan et al. [31, 32] in 2006 on the reversible metal-free hydrogen activation, the novel concept of frustrated Lewis pairs (FLPs) where the Lewis acid and Lewis base are sterically encumbered has attracted significant research interest [31] and homogenous FLPs have been demonstrated to activate syngas and generate formyl and alkoxy intermediates [32]. However, the utilization of the homogeneous molecular-based FLPs is hindered by the problematic and costly catalyst recycling. In recent years, the solid/surface frustrated Lewis pairs constructed based on the V_Os of reduced metal oxides have been shown to be able to circumvent the drawbacks of homogeneous FLPs, and have been reported in hydrogen dissociation, CO₂ hydrogenation, and methane dissociation [33-39]. Therefore, we expect that such kind of FLPs based on V_Os can also play an important role in the oxide-catalyzed syngas conversion. The present study focuses on the role of surface FLPs/oxides in the activation and conversion of syngas to intermediates (ketene or methanol) which in turn can reveal the role of V_Os in the OX-ZEO catalysis of syngas conversion as mentioned above. The surface FLPs are constructed on ceria due to the easy release/storage of oxygen by forming/filling oxygen vacancies [40]. As shown in Fig. 1(right), by removing two O atoms on CeO₂(110), the two adjacent surface Ce cations next to V_O acting as the Lewis acid and one neighboring O anion acting as the Lewis base construct the FLP sites. The distance between the acid site (coordinatively unsaturated Ce) and the base site (surface O) is ~4 Å, longer than that of the Ce−O bond (~2 Å) in the classical Lewis pairs (CLPs) on stoichiometric CeO₂(110) shown in Fig. 1(left). In addition, the two FLP sites in Fig. 1 are located close to each other. The structures of FLPs on reduced CeO₂(110) have been elucidated by both static density functional theory (DFT) calculations and ab initio molecular dynamics simulations under reaction conditions [37, 39]. Experimentally, it was shown that the addition of a trace amount of either Lewis base pyridine or Lewis acid pyrrole can completely quench the hydrogenation reactivity due to the blockage of the surface Lewis acidic or basic sites by these small molecules, respectively, indirectly suggesting the existence of FLPs [39]. More details about the FLPs on ceria surfaces are available in the previous reports [37-39]. Moreover, previous studies have reported that pure ceria was among the most promising catalysts for isosynthesis in which syngas is converted to hydrocarbons [41, 42]. In a recent study of syngas conversion, the pure ceria with HZSM-5 achieved high selectivity to C_2‒4 hydrocarbons and high CO conversion, indicating the possibility of syngas conversion on ceria surfaces [21]. Based on the catalytic features of solid FLPs and syngas conversion on pure ceria, it is desirable to study the direct syngas conversion to key intermediates at the FLP sites on the reduced ceria surface.

In this work, we first studied the activation of the H₂ and CO molecules at the FLP sites on the reduced CeO₂(110) surface. Then, four pathways for the conversion of H₂ and CO to ketene or methanol at a reaction temperature of 673 K were explored, and the dominant pathway was identified by calculating the Gibbs free energy of activation and the rate constants of the elementary steps.

All DFT calculations were performed using the Vienna Ab initio Simulation Package (VASP) [43-45] with the electron-ion interaction represented by the projector-augmented wave pseudopotentials [46]. The spin-polarized Perdew- Burke-Ernzerhof (PBE) version of the generalized gradient approximation (GGA) was used as the exchange-correlation functional [47], and the plane-wave kinetic energy cutoff was set to 400 eV. The DFT + U method with an effective U = 4.5 eV was used to treat the on-site Coulomb and exchange interaction of the strongly localized Ce 4f electrons [48-50]. The van der Waals dispersion forces were considered using Grimme's zero damping DFT-D3 method to account for the weak interactions between the adsorbates and the surfaces [51]. The first Brillouin zone k-point sampling utilized the Monkhorst-Pack scheme with a 7 × 7 × 7 mesh for the bulk and a 1 × 1 × 1 mesh for the CeO₂(110) surfaces [52]. The energy and force criteria for convergence of the energy and structure optimization were set to 10^–5 eV and 0.02 eV Å ^–1, respectively. The nudged elastic band combined with the minimum-mode following dimer method was used to search the transition state structures of the elementary steps [43, 53]. All of the transition states were identified using vibrational analysis. The reaction energy, ΔE, was defined as the energy difference between the product and the corresponding reactant. The activation energy, E_a, was defined as the energy difference between the transition state and the initial state, and the details regarding the calculation of the Gibbs free energy and rate constant are found in the Supporting Information. The calculations of the Gibbs free energy of reaction, ΔG, and the Gibbs free energy of activation, G_a, were similar to those of the reaction energy and activation energy, respectively. Atomic charges were computed using the atoms-in-molecules scheme proposed by Bader [54]. The crystal orbital Hamilton population (COHP) curves were generated with LOBSTER [55-58]. The relevant details about these methods can be found in our previous report [37].

The calculated lattice parameter for bulk CeO₂ was 5.43 Å which is very close to the experimental value of 5.41 Å [59]. Since our previous studies concluded that the construction of FLPs by only regulating the surface V_Os on CeO₂(111) is unsuccessful and the CeO₂(110) surface is more stable than the CeO₂(100) surface as indicated by their respective surface energies [37, 38], the CeO₂(110) surface was selected. The surface was modeled using a periodic five-layer slab as shown in Fig. 2(a). Similar to previous work [37], the p(2 × 3) supercell was adopted to study syngas conversion. The bottom three layers of the slab were fixed at the bulk positions, and the top two layers with adsorbates were allowed to relax. Only the top-surface oxygen vacancies are included in the calculations described below [37]. Our previous study found that reduced CeO₂(110) with two V_Os in a p(2 × 3) supercell can form dynamic FLPs via thermal fluctuations and reactant adsorption [37]. Therefore, the calculations related to the FLPs are based on the surface structure shown in Fig. 2(b) with one O atom next to V_O at the original bulk position in the plane (P site) and another O atom next to V_O at the bridge site of the Ce atoms (B site). The optimized structure shown in Fig. 2(b) is the most stable structure with FLP sites. As shown in Fig. S1, our calculated results indicate that compared to the V_O formation through the generation of O₂ and H₂O, the surface oxygen vacancy formation through the generation of CO₂ from CO oxidation is the most thermodynamically favorable process at 673 K. In addition, the calculations related to CLPs are based on the surface structure shown in Fig. 2(a). All of the slab structures corresponding to the data in Figs. 3 and 5–8 are displayed in the Supporting Information with Ce³⁺ ions labeled.

First, the activation of H₂ and CO at the FLP sites on reduced CeO₂(110) was explored and compared to that at the CLP sites on stoichiometric CeO₂(110). As shown in Fig. 3(a), the molecular adsorption of H₂ on FLPs is weak (–0.27 eV), but is still stronger than that on CLPs (–0.14 eV). As indicated by the Bader charges of the H₂ molecule presented in Fig. 3(b) and Table S1, a larger electron transfer from the surface to H₂ on FLPs was found. As shown in Fig. 3(c), the length of the H–H bond increased significantly from 0.75 Å in the gas-phase to 0.80 Å on FLPs, in comparison to a slight increase of 0.01 Å on CLPs. Both the atomic charge and bond length suggest that H₂ is more effectively activated on FLPs than on CLPs. Further dissociation of H₂ at the FLP sites is almost barrierless (E_a = 0.01 eV) and thermodynamically favorable with reaction energy of –0.53 eV, in good agreement with the previous studies [38, 39]. However, both the activation energy (0.57 eV) and reaction energy (0.41 eV) are much higher on CLPs. Importantly, the adsorption of H₂ on FLPs is dissociative adsorption, producing one hydride (H_a in Fig. 3(b)) and one proton (H_b in Fig. 3(b)). Our previous study has found that the hydride can be stable at the FLP sites and serves as an active intermediate in the hydrogenation reactions, while the hydride at the CLP sites can easily transfer to the nearby O atoms and form strong O–H bond [38]. In addition, the formation of hydride on reduced ceria has also been evidenced experimentally [60, 61].

Here, we carried out an analysis of the crystal orbital Hamilton population in order to further study the activation of H₂ on FLPs [55-58]. The negative projected COHP (pCOHP) curves presented in Fig. 4(a) show that the lowest unoccupied H–H orbital (the partial charge density from 2.0 to 3.0 eV in Fig. 4(b)) on FLPs is closer to the Fermi level than that on CLPs, suggesting an increased ability of H₂ on FLPs to accept electrons. More importantly, in the enlarged local pCOHP shown in Fig. 4(a), a slightly negative peak in the range from –3.0 to –1.0 eV corresponding to the occupied antibonding state is observed on FLPs, suggesting the H–H bond is weakened by electron transfer from the surface to the H–H antibonding orbital. The electrons in the energy range from –3.0 to –1.0 eV are mainly located around the hydride (H_a) close to the Ce ions as shown in Fig. 4(b), due to the local electric field of the metal oxide surface. Overall, the above analysis indicates that the electron transfer from the surface to the antibonding molecular orbital of H₂ and the local electric field both contribute to the activity of Lewis pairs.

Previous studies have reported that chemical adsorption of CO₂ on metal oxide surfaces is strong and can form CO₃^2– group [62-64]. Similarly, CO on CeO₂(110) can also bind with two surface O atoms to form a CO₃^2– group. The strongest adsorption on CLPs releases heat of 3.50 eV as shown in Figs. 3 and S2. The formation of a stable CO₃^2– group with a large adsorption energy suggests that it is difficult for the subsequent hydrogenation steps to form methanol or ketene to occur. Fortunately, CO can form a bent CO₂ group when inserting into FLPs as shown in Figs. 3(b) and S3, and as supported by a previous Fourier-transform infrared spectroscopy study of CO adsorption on CeO₂ [65]. The adsorption energy (absolute value) decreases to –1.68 eV, and the atomic charges of the CO₂ group at FLPs is –1.38 e (Fig. 3(b) and Table S2), close to that of the CO₃^2– group at CLPs (–1.58 e), while the number of the surface Ce³⁺ ions remains the same after CO adsorption (Fig. S3), suggesting the formation of a CO₂^2– group.

The electronic structures of the CO₂^2– group on FLPs and the CO₃^2– group on CLPs are analyzed (Fig. 4). The pCOHP curves for the three C–O bonds in the CO₃^2– group reveal that the bonding orbitals (–pCOHP > 0) are below the Fermi level and the antibonding states (–pCOHP < 0) are almost above the Fermi level, indicative of strong C–O bonds. For the two curves of the CO₂^2– group, two sharp peaks close to the Fermi level are observed at –0.8 eV and at 2.3 eV. The occupied antibonding orbitals ranging from –2.0 to 0.0 eV reflect the weaker C–O bonds on FLPs. Therefore, the COHP analysis indicates that two C–O bonds in the CO₂^2– group are weakened and the chemically adsorbed CO* on FLPs can be more active in the subsequent reactions.

By calculating the adsorption energies and analyzing the electronic structures of H₂ and CO, we found that by forming surface FLPs, the oxygen vacancy can serve as catalytic centers for the dissociation of hydrogen to hydride and proton and for binding CO to form active CO₂^2– groups. Moreover, as shown in Fig. 3(a) the adsorption of ethylene which is a product of syngas conversion is weaker on FLPs (–0.49 eV) than both the dissociative adsorption of H₂ (–0.80 eV) and adsorption of CO (–1.68 eV) on FLPs, indicating that the adsorption of H₂ and CO is advantageous compared to that of C₂H₄ on the reduced CeO₂(110) surface.

In the previous section, we only considered the activation of a single molecule at a single FLP site, namely either the transformation of H₂ into H⁺ and H^– or the transformation of CO into CO₂^2–. To reveal the hydrogenation mechanism, four pathways classified by the adsorption states and the interactions between H₂ and CO at the beginning of the reactions were proposed. In pathway Ⅰ, H₂ and CO first occupy two neighboring FLP sites (see the FLPs in Fig. 1) and then the surface reactions proceed. In pathway Ⅱ, two CO occupy two neighboring FLP sites and then C–C coupling occurs. In pathway Ⅲ, the H₂ first dissociates into a proton and hydride and then CO attacks the adsorbed hydride. In pathway Ⅳ, the CO first adsorbs at the FLP sites and then the H₂ directly dissociates at the Lewis pairs formed by CO₂^2– and surface ions. In the following, these four syngas conversion pathways at the reaction temperature of 673 K will be discussed in detail.

As shown in Fig. 5(a), the activations of both CO and H₂ are facile with an exothermic adsorption free energy of –0.61 eV for CO adsorption and a small G_a of 0.22 eV for hydrogen dissociation. Then, the active hydride attacks the C atom to form CHO* (from A3 to A4); this step is also facile with a low G_a of 0.59 eV and an exothermic ΔG of –1.65 eV. The second CO occupies the vacant position left by the hydride (state A5) and then binds a proton to form another CHO* (state A6). However, the second hydrogenation (from A5 to A6) needs to overcome a G_a of 1.25 eV. After a rotation of CHO*, the two adsorbed CHO* occupy two neighboring FLP sites, respectively, forming the most stable intermediate (state A7) in pathway Ⅰ. To continue the reaction, H₂ dissociates at the nearby CLP sites (from A8 to A9) with a G_a of 0.90 eV. The subsequent hydrogenation of CHO* by a hydride has a G_a of 1.15 eV. The surface C–C coupling occurs between CHOH* and CHO* (from A10 to A11) with a moderate G_a of 0.84 eV and an exothermic ΔG of –0.38 eV. Then, the newly formed OCHCHOH* successively undergoes two facile steps, namely breaking of the C–O bond (from A11 to A12) and hydrogen transfer between two C atoms (from A12 to A13), to form CH₂CO* (ketene). To release the ketene into the gas phase, two C–O bonds must be broken which is difficult due to a G_a that is as high as 1.68 eV. After ketene desorption, the generation of water into the gas phase (from A15 to A18) proceeds through a three-step pathway with a G_a as high as 1.27 eV. For pathway Ⅰ, the rate-determining step is the cleavage of the C–O bond for the release of CH₂CO* into the gas phase.

Pathway Ⅱ starts from two CO molecules adsorbed at two neighboring FLP sites with an adsorption free energy of –0.45 eV (state B2) as shown in Fig. 6. As the two adsorbed CO* are activated by FLPs, the coupling of two CO* to form OCCO* only needs to overcome a small G_a of 0.28 eV (from B2 to B3). Similarly, the coupling of two adsorbed CO molecules has been reported in the studies of the electrocatalytic reduction of CO and CO₂ on Cu(100) using density functional theory investigations [66] and Fourier transform infrared spectroscopy [67]. Since the FLPs are occupied by OCCO*, the dissociation of H₂ (from B4 to B5) occurs at the nearby CLP sites with a G_a of 1.18 eV. The hydride first binds the nearby C_a atom (from B6 to B7), and then transfers to another C_b atom (from B7 to B8). Since OCHCO* mainly occupies a single FLP site, the surface OH group can diffuse above the two Ce atoms (from B8 to B9) to form a hydrogen bond with the O atom in OCHCO*. After the H atom returns back to C_a atom from the C_b atom (from B9 to B10), the proton almost spontaneously transfers from the surface OH group to the O atom in OCCHO* (from B10 to B11). The addition of the first H₂ to OCCO* (from B3 to B11) is facile because the highest G_a is only 1.05 eV. For the OCCHOH* in state B11, the bond length between the C_a atom and the O atom of hydroxyl is 1.48 Å, longer than the C–O bond in methanol (1.43 Å), indicating that the C_a–O(H) bond is weak. Hence, we consider the H₂ dissociation between the O atom of hydroxyl and the C_b atom, which is a C_b···O FLP with a distance of 2.44 Å. The reaction (from B11 to B12) following the Eley-Rideal mechanism needs to overcome a high G_a of 1.68 eV and leads to the formation of one adsorbed H₂O and addition of one H atom to C_b atom. After the desorption of water (from B12 to B13) and transfer of the H atom (from B13 to B14), the adsorbed CH₂CO* forms. The subsequent cleavage of two C–O bonds to release the ketene into the gas phase needs to overcome a G_a that is as high as 1.21 eV. Unlike for pathway Ⅰ, the rate-determining step in pathway Ⅱ is the simultaneous dissociation of the second hydrogen molecule and the cleavage of one C–O bond to form adsorbed water.

In the initial stages of pathways Ⅰ and Ⅱ, the H₂ and CO molecules chemically adsorb on FLPs and then the surface reaction occurs. In pathway Ⅲ, the CO initially physically adsorbs on the top site of Ce atom and then binds with the hydride on FLPs (from C3 to C4) to form CHO*. The hydrogenation of CHO* by a proton (from C4 to C5) generates CH₂O* and the further rotation stabilizes the CH₂O* by forming the second C–O bond with the O atom (state C6) of FLPs. The second H₂ dissociates at another FLP site with a low G_a of 0.20 eV. The G_a of the elementary steps from state C0 to state C8 (CH₂O* and 2H* at two neighboring FLP sites) are no greater than 0.50 eV. However, the subsequent transfer of the hydride to the O atom of CH₂O* (from C8 to C9) must overcome a high G_a of 1.70 eV. Then, the transfer of the H atom from the O atom to the C atom to form CH₃O* (from C9 to C10) is favorable both kinetically (0.21 eV) and thermodynamically (–0.40 eV). After two steps of the surface diffusion of OH* and CH₃O*, the breaking of the C–O bond and formation of the O–H bond to generate and release CH₃OH into the gas phase need to overcome a ΔG of 1.00 eV. Overall, the rate-determining step is the hydrogenation of the adsorbed CH₂O* with a G_a of 1.70 eV.

Similar to the direct attack of the dissociated H₂ by CO in the initial stage of pathway Ⅲ, the dissociation of intact H₂ by the chemically adsorbed CO was studied in pathway Ⅳ. The dissociation process is assisted by the Lewis pairs between the O atom in CO and the surface Ce ion, indicated by the transition state structure shown in Fig. S11. The process results in the formation of CHOH* with a high G_a of 1.97 eV. The dissociation of the second H₂ occurs between the C atom and a Ce cation (from D4 to D5) to form CH₂OH* that has a low G_a of 0.62 eV. The transfer of the H atom from the O atom to the C atom to form CH₃O* is also facile with a low G_a of 0.46 eV. The surface transfer of the H atom (from D7 to D8) has a high G_a of 1.61 eV. Similar to pathway Ⅲ, the generation and release of CH₃OH into the gas phase has a ΔG of 1.20 eV. Overall, the dissociation of the first H₂ with a G_a of 1.97 eV is still the rate-determining step in pathway Ⅳ.

The four syngas conversion pathways discussed above present two routes for ketene production and two routes for methanol generation as shown in Fig. 9. Comparing the G_a values for the elementary steps shows that pathway Ⅳ is distinctly more difficult than the other three routes due to the high G_a of its rate-determining step (1.97 eV). The G_a values of the rate-determining steps of pathways Ⅰ, Ⅱ and Ⅲ (1.68, 1.68 and 1.70 eV, respectively) are very close to each other. Since the transition state structures for the endothermic desorption of CH₂CO, H₂O, and CH₃OH cannot be found, the desorption free energies (ΔG) are used as the reaction barriers and are compared to the G_a of the surface reactions; this overestimates the kinetics of the desorption steps. Moreover, the partial pressure of the reactants and products can also affect the rates of adsorption, desorption, and surface reactions in the Eley-Rideal mechanism. Therefore, to comprehensively evaluate the reaction activity, the rate constants at 673 K under 2.5 MPa of syngas with a H₂/CO ratio of 1.5 for all the elementary steps, including surface reactions, adsorption, and desorption were calculated and are presented in Tables S3-6. Table 1 lists the elementary steps that have a forward rate constant of < 10³ s^–1 in the four pathways. Using the rate constant of 1.0 s^–1 as a benchmark to evaluate the reactivity of the elementary steps on the active sites [68], pathways Ⅰ, Ⅱ and Ⅲ are promising, whereas pathway Ⅳ is difficult due to its low forward rate constant of 2.48 × 10^–2 s^–1. Notably, pathway Ⅱ for the formation of ketene is dominant as it is nearly 5 times faster than pathway Ⅰ and 7 times faster than pathway Ⅲ.

In pathway Ⅱ, the C–C coupling can occur easily between two chemically adsorbed CO* with a small G_a of 0.28 eV, which is attributed to the existence of the bent CO₂^2– active intermediate shown in Fig. 4. In pathway Ⅰ, the C–C coupling between CHO* and CHOH* also has a moderate G_a of 0.84 eV, suggesting the reactivity of CHO*. In Fig. 10(a), the electronic structure analysis of the CHO* actually forming a HCOO group with the O atom of FLPs is presented. The presence of the HCOO group (formate species) has been previously confirmed by in situ FT-IR studies of syngas conversion over the ZnO-ZrO₂ catalyst [13]. The pCOHP curves of the C–O bonds show that the unoccupied antibonding orbitals are close to the Fermi level. The projected density of states (PDOS) of C 2p shows a large contribution of the C atom to the antibonding orbitals that is indicative of a tendency of the C atom in CHO* to accept electrons. The rate-determining step in pathway Ⅲ is the hydrogenation of CH₂O* with a large G_a of 1.70 eV, suggesting the stability of CH₂O*. Fig. 10(b) shows that no large peaks are observed close to the Fermi level in both pCOHP and PDOS of CH₂O*, indicating that the activation of CH₂O* is more difficult than that of CHO*. Similarly, CH₂O* binds with the surface O atom of FLPs to form an OCH₂O group in which the C atom is located at the center of the tetrahedron constituted by the four atoms in the OCH₂O group. Both electronic structure and geometric structure reflect the stability of CH₂O* on the reduced CeO₂(110) surface. Even though the H atom can first bind with an O atom to break the stable structure and then easily move to the C atom to form CH₃O*, our calculations show that the hydrogenation step is still challenging. Overall, the C atoms in CO*, CHO* and CH₂O* on FLPs all bind with two O atoms, one from CO molecule itself and another from the CeO₂(110) surface. Moreover, the C atoms in the former two are unsaturated while the C atom in the latter is saturated. Therefore, the easier C–C coupling of CO* or CHO* compared to the hydrogenation of CH₂O* is the underlying reason for the preferential formation of ketene on the FLPs of reduced metal oxides.

In addition, previous studies have proposed that CO is first activated through a disproportionation reaction to CO₂ and adsorbed C, i.e., the Boudouard reaction, followed by a combination between the CH₂ species (originating from the hydrogenation of the adsorbed C) and CO to generate CH₂CO [9, 11]. This mechanism is supported by the formation of CO₂ during the interaction of CO with metal oxides such as ZrCrO_x and MnO [9, 11]. In our mechanism at the FLP sites on the reduced CeO₂(110) surface, only water is produced together with ketene. However, the water produced in syngas conversion can react with CO to form CO₂ via a water gas shift reaction. In addition, after the formation of ketene through the C–C coupling of CO* or CHO*, CO can react with the surface O atom to generate CO₂, similar to the removal of the O atom by water production. Therefore, our proposed mechanism can also explain the formation of CO₂ during the generation of ketene over ZrCrO_x and MnO.

In this study, we investigated the role of reduced metal oxides, and particularly the frustrated Lewis pairs, in syngas conversion to ketene/methanol, the key intermediate step in the conversion of syngas to olefins via OX-ZEO catalysis. The main conclusions are as follows. (1) The frustrated Lewis pairs constructed via the oxygen vacancies of the metal oxide play a key role in syngas (H₂ and CO) activation. The H₂ is activated via a heterolytic dissociative pathway forming a hydride and a proton, and CO is activated in the form of CO₂^2– by combining with the basic site (O atom) of the FLPs. The hydride stabilized at the FLP site is the critical active species for the hydrogenation of CO. (2) Four pathways for the conversion of syngas were explored on FLPs, with two pathways for the formation of ketene and two pathways for the production of methanol. Rate constant calculations showed that the predominant route is the formation of ketene via pathway Ⅱ which is attributed to the easier C–C coupling of CO* or CHO* compared to the hydrogenation of CH₂O*. Moreover, the neighboring FLP sites were also found to play a crucial role in the surface C–C coupling to produce ketene. Overall, our study provides novel insights into the role of oxygen vacancies in metal oxides and the reaction mechanism of syngas conversion over OX-ZEO catalysts.

Acknowledgments

The calculations were performed by using the HPC Platform at Xi'an Jiaotong University and National Supercomputing Center in Guangzhou.