Esters exist in various natural products such as green olives and pears and have numbers of applications on the production of perfumes, flavorings, etc. [1]. Consequently, various methods have been reported to synthesize esters such as the esterification of carboxylic acids with alcohols [2], the condensation of acid chlorides and alcohols [3] as well as the alcoholysis of nitriles [4]. Compared with those methods, the functionalization of olefins in the presence of carbon monoxide (CO) and alcohols, which is named as hydroalkoxycarbonylation of olefins, has been considered to be one of the most attractive methods for synthesizing esters owing to the 100% of the atom economy and the using of readily available starting materials [5-15].

Controlling regioselectivity is a challenging task for chemists working in hydroalkoxycarbonylation of olefins [9, 15, 16]. Numerous researchers have attempted to control the regioselectivity through ligand design or modification in homogeneous catalytic systems. For example, monodentate N-phenylpyrrole phosphine ligand was designed by Beller's group [1] and applied in hydroalkoxycarbonylation of olefins with high regioselectivities of branched esters owing to the steric hindrance of the ligand (up to 91%), which had been verified through the density functional theory (DFT) calculation (Scheme 1a). Although it is well known that monodentate phosphine ligands are favorable for the production of branched esters, there are still some exceptions. Recently, a monodentate cage phosphaadamantane ligand in combination with Lewis acid such as SnCl₂ was reported to enhance the regioselectivities of the linear esters in the range of 70%–96% when PdCl₂ was used as the catalyst (Scheme 1b) [17]. Additionally, considering the fact that more attention has been turned to heterogeneous catalytic system owing to the difficulties in the separation as well as reusability of homogeneous catalysts [18], the investigation of factors affecting the regioselectivity attracts more and more attention in heterogeneous catalytic system. Recently, a heterogeneous catalytic system was reported and high regioselectivities of branched esters (up to 91%) were obtained with Pd-TPPTS-OTPPTS complex supported on acidic resin as the heterogeneous catalyst (Scheme 1c) [19, 20]. They found that the reaction regioselectivity was remarkably impacted by the pore size of the support owing to the steric confinement effect [20]. Therefore, the development of a greener heterogeneous catalytic system with high activity and excellent regioselectivity, especially of linear esters, as well as the in-depth investigating of the factors affecting regioselectivity is urgent for the reaction.

In our previous works, Ru-clusters/CeO₂ has been used in various carbonylation reactions [21-23]. For example, we reported the synthesis of methyl propionate from hydromethoxycarbonylation of ethylene over Ru-clusters/CeO₂ catalyst [21]. Although excellent catalytic performance was obtained, there is no regioselectivity for methyl propionate. Herein, Ru-clusters/CeO₂ was used as the catalyst for hydromethoxycarbonylation of a terminal olefin, that is, styrene, to investigate the factors influencing the regioselectivity. The conversion of styrene was > 99% and the regioselectivity for linear and branched esters was 83% and 12%, respectively (Scheme 1d). By using different supports (CeO₂ (nanoparticle), CeO₂-rod, and CeO₂-cube), three catalysts including Ru-clusters/CeO₂, Ru/CeO₂-rod, and Ru/CeO₂-cube were prepared and applied in this reaction. Structural characterizations demonstrate that the L/B ratio is related to the Ru size of supported Ru catalysts. Raman characterization and NH₃-TPD results show that the Ru size has a great relationship with the metal-support interaction and the concentration of oxygen vacancy of the catalyst. The highest regioselectivity of linear ester could be obtained using Ru-clusters/CeO₂ as catalyst owing to the smallest size of Ru clusters on the CeO₂ surface.

Initially, hydromethoxycarbonylation of styrene was conducted under 0.5 MPa of CO in methanol at 165 ℃ for 6 h. When Ru-clusters/CeO₂ was used as the catalyst, the conversion of styrene was 43% with 82% and 13% regioselectivity of linear and branched ester, respectively (Table 1, entry 1). The reaction did not take place in the absence of catalyst or only in the presence of CeO₂ support (Table 1, entries 2–3). In comparison, RuCl₃·nH₂O, the Ru precursor of Ru-clusters/CeO₂, gave very low regioselectivity of total ester (L/B = 36:6), albeit > 99% conversion of styrene (Table 1, entry 4). Other metals including Pd and Rh exhibited no activities at all (Table 1, entries 5–6). Both a low conversion of styrene (7%) and poor regioselectivity of 61:21 (L/B) were obtained when Ru/SiO₂ was used as the catalyst (Table 1, entries 7).

Further experiments optimizing of the reaction conditions to improve catalytic activity were conducted (Fig. 1). First, we investigated the influence of CO pressure on styrene hydromethoxycarbonylation catalyzed by Ru-clusters/CeO₂ (Fig. 1a). Obviously, when the CO pressure increased from 0.1 to 0.9 MPa, the conversion of styrene decreased from 87% to 43%. The regioselectivity of linear ester remained at 82% (L/B = 82:12). Therefore we chose 0.5 MPa as the optimizing pressure of CO, in which condition both a moderate conversion of styrene (51%) and high regioselectivity of 82:12 (L/B) could be obtained. Further optimization of the reaction temperature under 0.5 MPa of CO was conducted (Fig. 1b). The catalytic results at temperatures of 145–185 ℃ showed that the conversion of styrene increased rapidly from 4% at 145 ℃ to 51% at 165 ℃ and then to 80% at 185 ℃. The regioselectivity of 29:14 (L/B) obtained at 145 ℃ increased to 82:12 (L/B) at 165 ℃ and then decreased to 66:11 (L/B) at 185 ℃. Thus the temperature of 165 ℃ was considered to be the optimal temperature for the reaction. Further extending the reaction time to 14 h under 0.5 MPa of CO at 165 ℃, the conversion of styrene increased to > 99%, giving 83% and 12% regioselectivity of linear and branched ester, respectively (Fig. 1c).

Mechanism investigations for styrene hydromethoxycarbonylation over Ru-clusters/CeO₂ were then conducted (Fig. 2). Fig. 2a shows the plot of the logarithm of the rate constants log(k_X/k_H) plotted against Brown-Okamoto constant σ_p for each substituent on para-position in styrene (CH₃, H, F, or Cl). Obviously, there was no linear relationship between log(k_X/k_H) and σ_p, implying that the reaction is not sensitive to groups substituted on styrene [25]. Meanwhile, this result suggested the radical path of the reaction. Then further experiment using TEMPO (2, 2, 6, 6-tetramethylpiperidine-N-oxide) as the radical inhibitor was conducted and the conversion of styrene decreased from 43% to 9% when 0.2 mmol of TEMPO was added into the reaction system, proving the radical path of hydromethoxycarbonylation of styrene over Ru-clusters/CeO₂ (Fig. 2b) [26, 27].

The reaction kinetic of styrene hydromethoxycarbonylation was studied at temperatures of 145–175 ℃ (Fig. 3a). The concentration of methanol, CO, and catalyst remained constant during the reaction, maintaining the dependence of the rate on styrene to be isolated. The k values at 175, 165, 155, and 145 ℃ were 11.5 × 10^–6, 7.5 × 10^–6, 4.5 × 10^–6, and 5.6 × 10^–6, respectively, which were calculated from the slopes of the plots in Fig. 3a. Obviously, there was a linear Arrhenius plot of ln(k) versus 1000/T (Fig. 3b). And the activation energy (E_a) was calculated to be 48.50 kJ mol^–1, indicating the smoothly proceeding of styrene hydromethoxycarbonylation over heterogeneous Ru-clusters/CeO₂ catalyst.

In order to investigate the factors influencing the regioselectivity of styrene hydromethoxycarbonylation, three catalysts including Ru-clusters/CeO₂, Ru/CeO₂-rod, and Ru/CeO₂-cube were prepared by the same method [28, 31]. And then the three catalysts were applied in the hydromethoxycarbonylation of styrene under the same reaction conditions (Fig. 4a). When Ru/CeO₂-cube was used as the catalyst, the regioselectivity of 66:21 (L/B) was obtained. The reaction proceeded in higher regioselectivity of 75:20 (L/B) when Ru/CeO₂-rod was used as catalyst. A further higher regioselectivity of 82:13 (L/B) was achieved when Ru-clusters/CeO₂ was used as the catalyst. In order to show the difference of the regioselectivities over the three catalysts, L/B ratios were calculated and showed in Fig. 4b. The order of L/B ratios was in a subsequence of Ru-clusters/CeO₂ (6.92) > Ru/CeO₂-rod (3.75) > Ru/CeO₂-cube (3.14) (Fig. 4b), which means that the highest L/B ratio was obtained when Ru-clusters/CeO₂ was used as catalyst.

To elucidate the reasons for observed different L/B ratios obtained in hydromethoxycarbonylation of styrene over the three catalysts, we first characterized their structures using electron microscopy. Both high-angle annular dark-field imaging-transmission electron microscope (HADDF-STEM) and energy-dispersive X-ray spectroscopy (EDX) using the Ru L line confirmed that the highly-dispersed Ru clusters existed on the defected CeO₂ surface for Ru-clusters/CeO₂ (Fig. 5a and 5b). However, a small amount of Ru nanoparticles with most of Ru clusters could be observed on the surface of Ru/CeO₂-rod catalyst based on the characterizations from HADDF-STEM and EDX mappings using the Ru L line (Fig. 5c and 5d). The TEM and HR-TEM characterizations showed that a large number of Ru nanoparticles unevenly distributed on CeO₂-cube surface (Fig. 5e).

By combining the different Ru sizes and the L/B ratios obtained in styrene hydromethoxycarbonylation over the three catalysts, we found that the L/B ratio is related to the Ru size of supported Ru catalysts. The Ru-clusters/CeO₂, who has the smallest Ru size on CeO₂ surface, gave the highest L/B ratio in hydromethoxycarbonylation of styrene.

Considering the fact that metal-support interaction between metal and their oxide supports has profound effects on the dispersion and size of metal species in heterogeneous catalysts, the metal-support interaction for the three catalysts was then investigated via Raman characterization (Fig. 6a) [32-38]. The results are given in Fig. 6a and show that the Ru-clusters/CeO₂, Ru/CeO₂-rod, and Ru/CeO₂-cube exhibited a strong peak at 453 cm^–1 and a relative weaker peak at 595 cm^–1, corresponding to the vibration model of octahedral local symmetry around CeO₂ lattice and the defect-induced modes of CeO₂, respectively [39-43]. Besides the two peaks of CeO₂, two peaks at 695 and 980 cm^–1 were also observed for Ru-clusters/CeO₂ and Ru/CeO₂-rod, which could be assigned to Ru–O–Ce bond resulting from the stronger interaction between Ru species and the different supports (CeO₂ and CeO₂-rod) [44]. However, no peaks at 695 and 980 cm^–1 were observed in Ru/CeO₂-cube, indicating a weak interaction exists between Ru species and CeO₂-cube. Thus, the Raman characterization showed that there was a stronger metal-support interaction in Ru-clusters/CeO₂ and Ru/CeO₂-rod than that in Ru/CeO₂-cube. Combining the Raman characterization and the structural characterizations, we found that the metal support interaction has a great influence on the Ru size formed on different supports.

In the literature reports, the formation of metal size has a great relationship with the concentration of oxygen vacancy, which presents a linear relationship with the concentration of acid site [40, 45]. Thus, we turned to characterize Ru-clusters/CeO₂, Ru/CeO₂-rod, and Ru/CeO₂-cube using NH₃-TPD to investigate the concentration of acid site of the catalyst (Fig. 6b). Obviously, the order of acid concentration of the three catalysts was in the subsequence of Ru-clusters/CeO₂ > Ru/CeO₂-rod > Ru/CeO₂-cube (Table S1). Thus, the order of the concentration of oxygen vacancy was in the subsequence of Ru-clusters/CeO₂ > Ru/CeO₂-rod > Ru/CeO₂-cube (Fig. 6c). Therefore, the smallest size of the Ru with the highest concentration of oxygen vacancy, having the largest steric hindrance, is favor to the linear adsorption mode of styrene, which is benefit to generation of linear product. Similar results are also reported before [46].

In summary, Ru-clusters/CeO₂ is used as an active catalyst for the hydromethoxycarbonylation of styrene with > 99% conversion of styrene, giving 83% and 12% regioselectivity of linear and branched ester, respectively. Structural characterizations demonstrate that the L/B ratio is related to the Ru size of supported Ru catalysts. Further Raman characterization and NH₃-TPD demonstrate that the metal-support interaction and the concentration of oxygen vacancy of the catalyst have a great influence on Ru size. Mechanism investigation proves the radical path of hydromethoxycarbonylation of styrene over Ru-clusters/CeO₂. Considering the complication varying of CeO₂ shape, further research work, in-depth investigating of the factors affecting regioselectivity, is ongoing.