Catalytic methoxycarbonylation reactionUnder the reaction conditions investigated, there exist three competing pathways for the catalytic reactions involving styrene, CO and methanol: hydrogenation, methoxycarbonylation and methoxylation. To ensure optimal carbonylation efficiency, there is clearly a need to identify a suitable catalyst that can selectively drive styrene toward carbonylation, while simultaneously inhibiting the undesirable hydrogenation and methoxylation pathways (Fig. 1c). Our initial investigations indicated that ruthenium is a highly effective active metal for this application (Fig. 1d, Supplementary Table 2). The commercially available carbon catalysts (Pt/C, Pd/C, Ir/C) showed major methoxylation products due to the strong acidity of activated carbon support, similar to acidic support catalysts (Ru/Hβ, Ru/WZrOx). The Ru/C catalyst exhibited a dominant hydrogenation pathway due to the strong metallicity of large-size Ru nanoparticles. The ceria-based Ru catalysts (Ru/CeO2, Ru/CeOx) showed distinct carbonylation selectivity (80–90%) despite poor conversion, due to their ability to form oxygen vacancy defects (Ov) on the surface. These oxygen defects can serve as active sites for anchoring and dispersing the Ru metal, as well as for methanol dissociation at the Ce-Ov site34,35. Our recent work has found that NbOx-based catalysts are effective at activating and dissociating the C-O bonds38,39,40, due to the abundant oxygen defects created through hydrogen treatment, despite Nb2O5 being typically less reducible than CeO2. The defective surface, along with its intrinsic acidic properties, makes NbOx-based catalysts potentially catalytically active for promoting the hydromethoxycarbonylation reaction.The synthesized Ru/NbOx catalyst exhibited a remarkable selectivity (up to 90%) to esters at 13% conversion, under 1 MPa of CO. Rather surprisingly, when the reaction pressure was reduced to 0.1 MPa a significant increase in conversion (55%) was observed, at no detriment to the ester selectivity (90%) (Fig. 1d). Encouraged by the good performance of the Ru/NbOx catalyst, additional experiments to optimize the reaction conditions were conducted (Supplementary Tables 4‒10). These studies revealed that higher temperatures and CO pressures were unfavorable for the reaction. The optimal temperature was found to be 170 °C, as higher temperatures promoted the competitive hydrogenation pathway. At 200 °C, the reaction showed a 40% decrease in the selectivity of carbonylation but doubled its selectivity of hydrogenation (up to 56%), in comparison to the reaction at 170 °C (Supplementary Table 4). In the absence of CO, hydrogenation became the primary pathway, with a comparatively low conversion of approximately 3%. Styrene conversion is optimal near ambient pressures of CO (0.06–0.1 MPa) and decreases rapidly with increasing pressure, as does the CO utilization, which was speculated to be limited by styrene concentration. Then we increased the styrene concentration by tenfold at a fixed CO pressure, and achieved a CO utilization of 62.4% (Supplementary Table 1). Interestingly, this did not appear to have a significant impact on the reaction selectivity, as the product distribution remained virtually unchanged (Fig. 1e). This decrease in activity (as pressure is increased) is likely to be attributed to an increased surface coverage of CO on the Ru (namely the occupation of more Ru surface), thus inhibiting the adsorption of other reactive species (e.g., styrene, CH3O* and H*). The progress of the reaction over time was subsequently investigated at 0.1 MPa CO (Fig. 1f). A high regioselectivity (ca. 75%) to linear esters was observed throughout the reaction and, thus, did not appear to be influenced by the conversion. After 16 h of reaction, the styrene conversion reached >95%, and the ester yield reached 72%, with a CO utilization of 23.7%. Importantly, the selectivity to linear esters, formed through an anti-Markovnikov addition, appeared to be independent of reaction time, temperature or pressure, accounting for approximately 75% of all carbonylation products.Kinetic experiments were conducted to establish apparent activation energy and the reaction orders of CO and styrene, as illustrated in Supplementary Figs. 1–3. The reaction order for styrene was determined as 0.88, with the corresponding rate constant k of 2.9 × 10−5 at 170 °C. Within the investigated CO pressure range (0.1–1 bar), the reaction order of CO is 0.76, with the rate constant k of 5.6 × 10−10 (Supplementary Fig. 1). In a typical methoxycarbonylation reaction, the reaction was performed at a stirring rate of 600 rpm. Notably, the reaction rate was obviously low at 400 rpm, whereas the reaction rate at 600 rpm closely resembled that at 800 rpm, suggesting the elimination of potential mass transfer effects (Supplementary Fig. 2). The apparent activation energy of methoxycarbonylation reactions was determined as 60 ± 11 kJ/mol (Supplementary Fig. 3), basically aligning with the theoretical results (0.96 eV)35.Next, the effect of the size-dependent Ru speciation was investigated. The single Ru atom catalyst (Ru1/NbOx) and 4 wt.% Ru/NbOx catalyst with a diameter of 3.1 nm (Ru/NbOx-3.1 nm) were synthesized (Supplementary Fig. 4 and Supplementary Table 11). Evaluated by the methoxycarbonylation of styrene, the activity of carbonylation presented a volcano curve with the size of Ru clusters (or nanoparticles). The 2 wt.% Ru/NbOx catalyst, featuring a diameter of 2.2 nm by CO- chemisorption, exhibited a better activity and carbonylation yield (62%), compared to the others tested (Supplementary Table 12). The Ru1/NbOx showed that only <1% yield of hydrogenation was observed and no ester products were detected, most likely due to the absence of the accommodating reactive sites for multiple reactants. The Ru/NbOx-3.1 nm exhibited a low carbonylation yield of 14% but a higher hydrogenation yield of 28%. The larger-size Ru nanoparticles favored the side-reaction hydrogenation of alkene substrates, most likely due to the enhanced metallicity.Catalyst structural characterizationTo uncover the origin of the 2 wt.% Ru/NbOx catalysts’ enhanced performance, its physicochemical properties were probed. High-resolution transmission electron microscopy (HRTEM) confirmed that this material is comprised of nanosheets with layers of exposed Nb2O5 (100) facets (Fig. 2a‒c, Supplementary Fig. 5). To gain a more direct visualization of the metal distribution of 2 wt.% Ru/NbOx, the aberration-corrected high-angle annular dark field scanning transmission electron microscopy (AC-HAADF-STEM) was performed. The images clearly demonstrated that ultrafine Ru nanoclusters were uniformly dispersed over the support (Fig. 3d–h). Quantitative electron paramagnetic resonance (EPR) was employed to identify (and quantify) defect sites. Analysis of the spectra showed that NbOx materials exhibited significantly higher concentrations of oxygen vacancy (Ov) sites compared to the analog Nb(OH)5 and Nb2O5 (Fig. 2e‒f). Interestingly, the Ru/NbOx catalyst possessed the highest concentration of Ov sites, reaching up to 1.74 × 1015 spins/g (Supplementary Fig. 6b). Furthermore, X-ray photoelectron spectroscopy (XPS) analysis of the O1s region confirmed that the Nb2O5 material exhibited a higher proportion of lattice oxygen (OL), while NbOx had a greater concentration of defective Ov compared to the other two (Fig. 2d). Thus, the findings from XPS corroborate our EPR analysis. Notably, the Ru/NbOx material exhibited a two-fold higher yield of esters (12%) than that of an analogous Ru/Nb2O5 catalyst (6%), thus accentuating the significance of Ov sites in the carbonylation reaction (Fig. 2d).Fig. 2: Characterizations of Ru/NbOx and its analogs.a‒c HRTEM of 2 wt.% Ru/NbOx and crystal plane (001) model of Nb2O5. d O1s XPS. e, f Oxygen vacancy concentrations and EPR spectra. g Quasi-in situ Ru 3d XPS. h CO-probe FTIR of 2 wt.% Ru/NbOx.Fig. 3: Structural characterizations of 2 wt.% Ru/NbOx catalyst.a The normalized X-ray absorption near-edge spectra (XANES) at the Ru K-edge. b The k2-weighted Fourier transform extended X-ray absorption fine structure spectra (EXAFS) in r-space. c Wavelet transform of Ru foil, Ru/NbOx, and RuO2. d, e AC-HAADF-STEM images and f–h the corresponding element map.The electronic and coordinative structures of 2 wt.% Ru/NbOx catalysts were further characterized by the X-ray absorption spectra (XAS) technique41,42. Figure 3a displayed the X-ray absorption near-edge spectra (XANES) at the Ru K-edge of the Ru catalyst and references. The adsorption threshold E0 for Ru/NbOx is higher than Ru foil but lower than RuCl3, suggesting the most likely coexistence of metallic and ionic Ru speciation, consistent with the XPS characterization results. Further, the coordination environment of Ru atoms was determined by the extended X-ray absorption fine structure spectra (EXAFS). As shown in the Fourier-transformed k2- weighted EXAFS spectra at the Ru K-edge (Fig. 3b), in contrast to the reference samples of Ru foil and RuO2, the Ru catalyst showed three major peaks, which could be ascribed to Ru-O, Ru-Ru, and Ru-O-Nb/Ru coordination. To further resolve Ru-O, Ru-Ru, and Ru-O-Nb/Ru coordination, wavelet transform (WT) of Ru K-edge EXAFS oscillations was carried out owing to its more powerful resolutions in both k and r spaces (Fig. 3c)41. The three hills at (1.45 Å, 5.00 Å−1), (2.38 Å, 8.50 Å−1), and (2.80 Å, 9.50 Å−1), associated with Ru-O, Ru-Ru, and Ru-O-Nb/Ru contribution, respectively, was clearly observed from the WT contour plots of Ru/NbOx catalyst. Consistently, the best-fitted EXAFS result revealed Ru-O at 1.94 Å with coordination number (CN) of 3.6, Ru-Ru at 2.67 Å with coordination number (CN) of 5.1, and Ru-O-Nb/Ru shell at 3.08 Å with coordination number (CN) of 1.3, respectively, in Ru/NbOx (Supplementary Table 3 and Supplementary Fig. 7). The interfacial Ru-O-Nb-Ov sites consist of Ru sites and intimate O or Ov sites on the support. These sites are proposed to be synergistically active for the adsorptive dissociation of CH3OH on Ov sites (Fig. 4a), CO adsorption exclusively on Ru sites (Supplementary Figs. 8–11), and alkene adsorption on both Ru and acidic Nb-Ov sites (Figs. 4, 5 and Supplementary Figs. 9, 11), therefore promoting the subsequent hydromethoxycarbonylation reaction of three substrates. We will discuss the role of the active sites in detail later.Fig. 4: DFT calculations on methanol dissociation and migration, and the methoxycarbonyl (COOMe) pathway to yield esters.a Energy profiles for adsorptive dissociation of methanol on Nb2O5 and NbOx, respectively. b Energy profiles for migration of *OCH3 and *H species from NbOx support to Ru surface, respectively. c Energy profiles for the methoxycarbonyl pathway to yield linear- or branched-selective esters. The x-axis shows the reaction intermediates and transition states (TS); the y-axis shows the relative energy of each state. Nb, O, C and H atoms are shown in green, red, gray and white, respectively.Fig. 5: The holistic elementary reaction networks of potential routes for the hydromethoxycarbonylation reaction are based on DFT calculations.Four potential reaction routes to produce esters: ①styrene carbonylation (styrene + CO), ② styrene methoxylation (styrene + *OCH3), ③ styrene hydrogenation (styrene + *H), and ④ methoxycarbonyl pathway (CO + *OCH3). The energy barriers and enthalpy changes of each elementary step have been calculated and marked. The structures of the key intermediates are shown in Supplementary Fig. 14.Potential reaction routesThe field of heterogeneous methoxycarbonylation is in its earliest infancy, with vast important unknowns yet to be explored. Besides developing an efficient catalyst, understanding the reaction pathway and mechanism is of high significance. First, the adsorption behaviors of methanol on the surface of Ru metal, the NbOx support and an Nb2O5 material (as a control), were investigated using density functional theory (DFT) calculations. The results demonstrated that methanol preferentially adsorbs onto Ov sites on the NbOx surface (Gads = −0.96 eV), compared to Nb2O5 (Gads = −0.58 eV) or Ru (Gads = −0.61 eV). Following this, the adsorbed CH3OH on Ov sites can undergo dissociation to *OCH3 and *H, with a remarkably low energy barrier of 0.1 eV, which is significantly lower than the 0.49 eV barrier observed over Nb2O5 (Fig. 4a). Methanol-probe Fourier transform infrared (FTIR) experiments support these findings, showing distinct O-CH3 vibrations associated with *OCH3 at 1030 cm−1 on NbOx, but absent on Nb2O5 (Supplementary Figs. 8, 9). The *OCH3 and *H species formed can migrate from the NbOx support to the Ru surface with energy barriers of 0.28 eV and 0.77 eV, respectively (Fig. 4b). Complementary FTIR experiments were used to support this and showed increased O‒CH3 vibration signals over the Ru/NbOx (and Ru/Nb2O5) material(s), compared to the Ru-free support (Supplementary Figs. 9–11). These results evidenced that the NbOx support is responsible for the dissociation of CH3OH into *OCH3 and *H species, which likely (at least partially) transferred to the Ru surface (or interface). The acidic sites of the Ru/NbOx catalyst were identified by pyridine-probe infrared spectra (Supplementary Fig. 12). The acidic NbOx support can facilitate the adsorption of alkenes, as evidenced by in situ infrared spectra (Supplementary Fig. 9). Additionally, CO exclusively adsorbs on the Ru surface with stronger adsorption energy (−1.65 eV) compared to NbOx (−0.23 eV), as confirmed by FTIR analysis (Supplementary Figs. 9, 11). Furthermore, styrene adsorption onto Ru surfaces is significantly more favorable (Gads = −3.46 eV) than on NbOx (Gads = −0.98 eV) (Supplementary Fig. 13).These findings imply that the reactions involving styrene, CO, *OCH3 and *H species most likely occur on the Ru surface. We subsequently explored four potential routes among styrene, CO, *OCH3 and *H. They include (1) styrene carbonylation (styrene + CO), (2) styrene methoxylation (styrene + *OCH3), (3) styrene hydrogenation (styrene + *H), and (4) methoxycarbonyl pathway (CO + *OCH3). Each of the elementary reactions that could yield esters (both linear and branched) were considered. As shown in Fig. 5, (1) styrene carbonylation and (2) styrene methoxylation face significant barriers of approximately 1.20‒1.60 eV, making the initial binding of styrene with CO or *OCH3 unfavorable. For route (3) styrene hydrogenation, the energy barriers required for a *H species to attack an internal or terminal carbon in the side chain of styrene are low (0.59 eV and 0.86 eV, respectively). This suggests that *H species are more likely to interact with styrene (PhCHCH2), particularly by hydrogenating its internal carbon atom to form *PhCH2CH2. However, when CO or *OCH3 species attempt to react with this hydrogenated intermediate, they encounter large energy barriers that exceed 1.80 eV.In the case of the route (4) methoxycarbonyl pathway, the first step involves the coupling of CO with a *OCH3, resulting in the formation of *COOMe intermediate with a comparatively lower barrier of 1.06 eV (ΔG = 0.32 eV). The crucial *COOMe species can selectively attack the terminal carbon in the side chain of styrene to form linear C8H8COOMe, facilitated by a lower energy barrier of 1.04 eV (TS2), while the production of branched C7H6(COOMe)CH2 requires a higher barrier of 1.42 eV (TS2’). The terminal carbon atom of styrene is positioned 2.118 Å away from *COOMe in TS2 (Fig. 4c). Finally, the linear PhCHCH2COOMe intermediate undergoes hydrogenation to complete the cycle (a barrier of only 0.41 eV). Additionally, the formed *COOMe species can react with the hydrogenated *PhCH2CH2 or *PhCHCH3 intermediates from styrene hydrogenation, but this step encounters large energy barriers of 1.99 eV and 1.32 eV, respectively (Fig. 5). Based on these calculations, we propose that the reaction proceeds via the methoxycarbonyl pathway, as shown in Fig. 4c. This involves the reaction of *CO and *OCH3, leading to the formation of a *COOMe surface species, which can attack the terminal carbon in the side chain of styrene, leading to the formation of linear ester. From the perspectives of adsorption and reaction, when *COOMe attacks the terminal C atoms of styrene (yielding linear esters), it experiences less steric hindrance, as well as a significantly lower energy barrier (1.04 eV). Besides, the secondary *C intermediate formed via the terminal attack is more stable than the primary *C species formed via the internal attack. In contrast, attacking the internal C atoms would encounter a much higher barrier (1.42 eV) to yield branched esters. These factors provide a rational explanation for the predominant linear regioselectivity observed in the methoxycarbonylation reaction over the Ru/NbOx catalyst.Identification of heterogeneous methoxycarbonyl mechanismIn situ Fourier-transform infrared (FTIR) surface reactions were subsequently conducted to experimentally evidence the methoxycarbonyl pathway. The illustration scheme of the in situ infrared spectroscopic system applied in this work was depicted in Supplementary Fig. 15. Temperature-resolved FTIR experiments were conducted by saturating the catalyst with CO at 30 °C, followed by evacuating the cell. Subsequently, a methanol/styrene saturated vapor was introduced into the cell under vacuum until the system remained steady, and the infrared spectra were recorded. Stretching vibrational bands that are indicative of O‒H, C=O, and CAr‒CAr bonds in methanol, CO, and styrene are observed at 3700, 2080, and 1450 cm−1, respectively (Fig. 6a). As the temperature jumps from 30 to 60 °C, a new band attributed to C=O stretching vibrations emerged at 1750 cm−1, exhibiting an increasing intensity over the temperature, indicating the gradual formation of *COOMe species43,44. The stretching vibration frequency of the C=O bond of *COOMe is further confirmed by the simulation using the Vienna Ab initio Simulation Package, at 1720 cm−1 (Supplementary Fig. 16). The O‒H vibrations of methanol at 3700 cm−1 decreased by nearly 50% at 150 °C, due to both methanol consumption (for *COOMe formation) and heat-induced desorption. In contrast to the Ru/Nb2O5 analog, the Ru/NbOx catalyst showed a distinct peak at 1750 cm−1, confirming its efficiency in generating *COOMe (Supplementary Figs. 10, 11). The time-resolved FTIR experiments were conducted at 150 °C to further investigate the surface species. The procedures were similar to the above experiment except that the experimental temperature was fixed at 150 °C. Following CO adsorption and evacuation in a vacuum, a methanol/styrene saturated vapor was introduced into the cell, and the infrared spectra were recorded. Immediately the vibrations at 1750 cm−1 appeared and increased (by orders of magnitude) with time (Fig. 6b).Fig. 6: Experimental evidence for the methoxycarbonyl species formation.In situ FTIR surface reactions of CO, styrene and methanol over Ru/NbOx: a temperature-resolved reaction after chemisorption of CO and methanol/styrene vapor. b time-resolved process at 150 °C. c adsorption of styrene vapor after CO desorption at 150 °C. d adsorption of methanol vapor after CO desorption at 150 °C.The Ru 3d XPS analysis of the Ru/NbOx catalyst revealed that the majority of the Ru species present on the material are in a metallic state (Ru0), with only a minor proportion existing as Run+ (Fig. 2g). In CO-probed FTIR spectra, the corresponding assignments of IR bands of Ru carbonyl species were assigned (Fig. 2h), based on the well-documented literature (Supplementary Table 14)45,46,47,48,49. Tricarbonyl or dicarbonyl Ruδ+(CO)n species typically presented two high frequencies at ~2130 ± 15 and ~2070 ± 15 cm−1, the monocarbonyl Ru0(CO) species shows a frequency at 2030 ± 30 cm−1. Two in situ FTIR experiments were conducted to gain further insights into the active Ru sites of the catalyst. As shown in Fig. 6c, Ru/NbOx was saturated with CO at 150 °C, followed by the introduction of styrene vapor through argon bubbling. No vibration at 1750 cm−1 corresponding to *COOMe was detected. Notably, the signals assigned to Run+(CO)n at 2138 cm−1 and 2072 cm−1 decreased as styrene molecules were introduced. This decrease can be attributed to the partial occupation of the Ru surface by styrene. In the second experiment, Ru/NbOx was first saturated with CO before methanol vapor was introduced through bubbling with argon. The peaks at 1750 cm−1 appeared and then continued to intensify over time, indicating that *COOMe was being formed through the combination of MeO* and *CO (Fig. 6d). Based on these findings, we consider that metallic Ru0 and ionic Run+ sites synergistically contribute to the catalytic process under hydrogen atmosphere (Supplementary Fig. 17). This was strongly supported by a significant fall in the carbonylation yield over the NbOx supported sole Ru0 catalysts (Supplementary Table 15 and Fig. 2g). These infrared experiments strongly support the methoxycarbonyl pathway, and the results are consistent with the DFT calculations. According to this mechanism, methanol initially dissociates into *OMe and *H species on NbOx, which then migrate to the Ru surface. CO adsorbed on the Ru surface first binds with *OMe to form the *COOMe intermediate, which subsequently attacks the terminal carbon atom of the styrene side chain, followed by hydrogenation and desorption, as illustrated in Fig. 7.Fig. 7: Heterogeneous alkoxycarbonylation reactions.a The reaction route. b The methoxycarbonyl mechanism.The substrate scopes and regioselectivity controlTo assess the generality of the catalysts, a series of other substrates were also investigated under the optimized reaction conditions (Supplementary Tables 16‒20). The Ru/NbOx catalyst also exhibited activity for styrene derivatives with para-position substituents such as acetoxyl, bromo, methoxy, hydroxyl, and tertiary butyl, resulting in an ester yield of ca. 50% (Supplementary Table 16). All three α-olefins with vinyl, propenyl, and butenyl side chains achieved conversion of >96%, while styrene showed the highest carbonylation product yield of 71%, due to its inherent lower steric hindrance (Supplementary Table 17). To explore the influence of C=C positions in the substrate on carbonylation, the reactivity of allyl and propenyl benzene was assessed. Interestingly, similar product distributions were obtained regardless of the C=C positions, primarily due to preferential isomerization (Supplementary Tables 18, 19). Among the four alcohols examined (methanol, ethanol, propanol, and butanol), methanol demonstrated the highest selectivity to alkoxycarbonylation products (Supplementary Table 20). Controlling regioselectivity (toward Markovnikov or anti-Markovnikov) is evidently crucial for the efficient carbonylation of alkenes. We speculate that the acid/alkalinity of the environment affects the stabilization of primary/secondary intermediate and, thus, influences the regioselectivity. To probe this, some additional experiments were conducted where trace quantities of base or acid were spiked into the styrene methoxycarbonylation reaction. The results show that the presence of basic CsCO3 or Me3EtOK species, results in exclusive regioselectivity to branched esters via Markovnikov addition. On the contrary, anti-Markovnikov additions resulting in the formation of linear esters accounted for 75% of the products in both acidic and neutral environments (Supplementary Fig. 18).Despite the advantages of heterogeneous Ru/NbOx catalyst, there remains an issue of cycling stability (Supplementary Table 21). Comprehensive characterization of the fresh and cycled catalysts from the same batch was conducted (Supplementary Fig. 19). Pyridine-probe FTIR revealed a significant decline (ca. 50% or more) in Brönsted acid sites for the used catalyst compared to the fresh one, while the amount of Lewis acid sites remained nearly unchanged. CO-probe FTIR indicated a substantial decrease in exposed Ru species for the used catalyst, suggesting a potential structural change of Ru surface during the reaction (Supplementary Fig. 19b). The AC-HAADF-STEM images of the used catalyst exhibited prevalent particulate Ru aggregates, while no highly dispersed Ru clusters were observed in the investigated regions (Supplementary Fig. 20). ICP-OES results also showed the loading of Ru metal slightly declined from 1.89% to 1.28%. Furthermore, EPR analysis of fresh and used catalysts was carefully measured and confirmed that the number of oxygen vacancies in the used catalysts did not decline after the reaction (Supplementary Fig. 19c). Ru 3d XPS results showed that Ru/NbOx catalyst, both before and after the reaction, contained both Ruδ+ and Ru0 components with a slightly changed proportion (Supplementary Fig. 19d). Based on these analyses, we can conclude that the agglomeration of highly dispersed Ru clusters over support is most likely responsible for the observed deactivation. Given the significance of this issue, we aim to address it to improve recyclability in the near future.
