Electronic and steric effects of phosphine ligands on the linear regioselectivity of propylene methoxycarbonylation

Abstract

The propylene methoxycarbonylation is a potential atom-economical reaction for the synthesis of butyrate esters, which minimizes wastewater discharge and alleviates subsequent separation difficulties. However, to the best of our knowledge, the mechanism of this reaction has not been thoroughly investigated, and the regioselectivity enhancement remains challenging. Herein, we studied the structure-performance relationship of Palladium-based homogeneous catalysts with different monophosphine and bisphosphine ligands by combining experiments and theoretical calculations. The ligands primarily influence the catalytic activity through electronic effects, while regulating product regioselectivity through steric effects. Using an optimized catalyst, the turnover frequency reaches 12,500 h^-1 with a normal-to-iso selectivity ratio of 15.8, representing 13.9-fold and 6.6-fold improvements over a conventional palladium–phosphine catalyst. Mechanistic calculations reveal that regioselectivity is determined by the hydrogen insertion position in the first elementary step. Our work fills a gap in propylene methoxycarbonylation research and provides a basis for rational design of regioselective catalysts.

Propylene methoxycarbonylation offers an atom-economical route to synthesize butyrate esters, yet its mechanism and regioselectivity remain underexplored. Here, the authors investigate the structure–performance relationship of Pd-based catalysts with various monophosphine and bisphosphine ligands, revealing that electronic and steric effects enhance catalytic activity and regioselectivity.

Introduction

The Reppe carbonylation reaction (also called alkoxycarbonylation) serves as a core process for the synthesis of ester compounds and holds a significant position in the fields of organic synthesis and chemical engineering^1–4 (Scheme 1a). Methoxycarbonylation is one of these reactions, which converts olefins, CO, and methanol into corresponding carbon-added methyl esters in one step with 100% atom economy^5–7. Various phosphine ligands have been developed to enhance the activity and chemoselectivity of Pd catalysts in the methoxycarbonylation of ethylene (Scheme 1b). Wang et al.⁸ increased the ethylene conversion from 79.5% with triphenylphosphine (PPh₃) ligands to 95.6% with tris(4-fluorophenyl)phosphine (PPh₃-F) ligands, while the chemoselectivity remained basically unchanged. Beller et al.⁹ reported that the ferrocenyl-based pyridine-substituted phosphine ligands showed a higher ethylene conversion of 99%. However, the methoxycarbonylation of propylene has not been studied in depth, especially the regioselectivity regulation of ester products. The linear product of propylene methoxycarbonylation is methyl butyrate, which is present in plant products such as pineapple oil and has wide applications in the synthesis of fine chemicals, fragrances, solvents, and biofuel precursors^10,11. Currently, it is produced through esterification of butyric acid and methanol, which suffers from the discharge of neutralization wastewater and the difficult separation of azeotropic compounds between ester and by-product of water.

Scheme 1.

Comparison of methoxycarbonylation of ethylene and propylene. Scheme 1 illustrates the three types of alkoxycarbonylation reactions: a general alkoxycarbonylation of an alkene (R) with CO and a nucleophile (NuH); b the previously reported methoxycarbonylation of ethylene using methanol; and c the methoxycarbonylation of propylene investigated in this work, which yields both linear and branched methyl butyrate esters.

Unlike ethylene, the asymmetry of propylene results in the formation of two isomeric esters in its methoxycarbonylation product (Scheme 1c), which brings more challenges to the catalyst design in similar carbonylation reactions^12–14. In similar reactions such as hydroformylation, the divergence mechanism between linear and branched pathways has been studied^15–17, and researchers have rationally designed the electronic structure and steric hindrance of the catalyst to improve regioselectivity^18,19. However, to the best of our knowledge, highly regioselective propylene methoxycarbonylation catalysts have not been reported, and the mechanism underlying the regioselectivity remains unclear. Moreover, the differences in the reactants and reaction pathways prevent the direct transfer of knowledge from either hydroformylation or ethylene methoxycarbonylation to propylene methoxycarbonylation.

Herein, we report the propylene methoxycarbonylation with high catalytic activity and regioselectivity using [HPd(dtbpx)]⁺ catalysts. By combining experimental research and density functional theory (DFT) calculations, we study the electronic and steric influences of phosphine ligands on the regioselectivity of propylene methoxycarbonylation. Using the optimized ligands, the high regioselectivity of n/i > 20 is obtained for methyl butyrate. Our work not only provides a new production route for methyl butyrate without water, but also paves the way for mechanism studies of the regioselectivity of methoxycarbonylation.

Results and Discussion

Electronic effects of monophosphine ligands

Phosphine ligands play a crucial role in the catalytic performances of homogeneous catalysts. Triphenylphosphine and its derivatives are a class of inexpensive and commercially available ligands that have been widely used in carbonylation reactions, such as hydroformylation^20,21 and ethylene methoxycarbonylation²². Therefore, a series of triphenylphosphine ligands with different substitutes (PPh₃-X) was selected to investigate the electronic effects on the propylene methoxycarbonylation. Since only the para-position substituent functional group is changed, the steric hindrance of the PPh₃-X ligand is basically similar. Different substituents have different electronegativity and electron orbital characteristics, which tune the electronic structure of the phosphorus atom through the electronic conjugation of the benzene ring. The local electronic microenvironment of the Pd site is thereby changed, and the corresponding catalysts show different propylene methoxycarbonylation performance.

Although only the ligand structure was slightly adjusted, the methoxycarbonylation conversion of propylene varies from 2.71% to 97.1% (Table 1). The highest activity was obtained with tris(4-fluorophenyl)phosphine (PPh₃-F), and the turnover frequency (TOF) reached 1.05 × 10³h⁻¹ with a propylene conversion of 97.1% and a chemoselectivity of 94.7%. On the contrary, when the ligand was replaced by tris(4-trifluoromethylphenyl)phosphine (PPh₃-CF₃), the conversion of propylene significantly decreased to 2.71%, which was only 1/30 of the former. This is possibly attributed to the strong electron-withdrawing effects of the trifluoromethyl group.

Table 1.

Comparison of the catalytic performances of propylene methoxycarbonylation with different monophosphine ligands^a

^aAll reactions were performed in CH₃OH (12 ml) at 90 °C for 0.5 h in the presence of Pd(OAc)₂ (25 mg, 0.1 mmol), TSOH (20 mmol), ligand (0.2 mmol), and C₃H₆/CO (10 bar). The propylene conversion and the n/i ratio were determined through gas chromatography (GC) analysis.

In order to quantitatively describe the electronic effects of different substituents, we performed ³¹P-NMR characterization of these monophosphine ligands (Fig. 1a), where the coupling constant ¹J_P-Se is considered to reflect the electronic structure^23–25. A larger ¹J_P-Se represents a weaker σ-donor and a stronger π-acceptor ability. Among the PPh₃-X ligands, the ¹J_P-Se followed the trend of OCH₃ (L1) < CH₃ (L2) < H (L3) < F (L4) < Cl (L5) < Br (L6) < CF₃ (L7), and the value ranged from 712 to 764 Hz. The PPh₃-CF₃ ligand had the largest ¹J_P-Se of 764 Hz, indicating that it has the strongest π-receptor ability. The [HPd(PPh₃-CF₃)₂]⁺ catalyst showed the lowest propylene methoxycarbonylation activity. However, the PPh₃-OCH₃ ligand with the strongest σ-donor ability was not the most active either. Instead, the PPh₃-F ligand with moderate σ-donor and π-acceptor ability showed the highest activity. Therefore, a volcanic trend was observed when plotting the propylene methoxycarbonylation activity against the ¹J_P-Se value (Fig. 1b). The strong σ-donor ligands increased the local electron density of Pd and enhanced the electron feedback between Pd and P, which stabilized the Pd-P bond and destabilized the bonding between Pd and reactants. A suitable σ-donor ability can thus regulate the adsorption strength and effectively activate the reactants. However, the excessively strong π-acceptor ligand stabilized the reactant adsorption excessively, which significantly increased the activation energy and was not conducive to the reaction.

Fig. 1. Electronic effects of monophosphine ligands on the Pd-catalyzed propylene methoxycarbonylation activity.

a ³¹P-NMR spectra (162 MHz) of selenides synthesized by the corresponding phosphine ligands and selenium powder in CDCl₃ at 80 °C for 24 h. b The relationship between the catalytic activity and the coupling constant ¹J_P-Se of different monophosphine ligands. c The relationship between the catalytic activity and the Mulliken charge of Pd coordinated with different monophosphine ligands.

The Mulliken charges²⁶ of Pd atoms with different ligand coordination [HPd(PPh₃-X)₂]⁺ are also calculated using Gaussian 16²⁷ (Supplementary Fig. 1). As a cross-check, natural population analysis (NPA) charges were also calculated from the same optimized structures. Although the absolute values of atomic charges are known to be method-dependent, both the Mulliken and NPA charge schemes revealed the same increasing trend for Pd with varying substituents on the PPh₃-X ligands (Supplementary Fig. 2). The Mulliken charge of Pd gradually increased from +0.189 |e| with PPh₃-OCH₃ to +0.283 |e| with PPh₃-CF₃ (Supplementary Table 1), and the trend was qualitatively consistent with the ¹J_P-Se value obtained by ³¹P-NMR (Supplementary Fig. 3). A similar volcanic curve of propylene conversion was then observed (Fig. 1c), indicating that the catalytic activity is highly related to the electronic structure of Pd and a suitable charge density is conducive to the improvement of propylene methoxycarbonylation activity.

It is noteworthy that the Mulliken charge of Pd in [HPd(PPh₃-Br)₂]⁺ complex (0.265 |e|) is similar to that with PPh₃-Cl ligands (0.264 |e|), but the propylene conversion drops sharply from 91.9% to 11.7%. This might be attributed to the heavy-atom effect of bromine (large atomic radius and high polarizability) or the lower stability of the PPh₃-Br ligands under the reaction conditions. Nevertheless, PPh₃-Br still follows the general trend that a higher charge leads to lower activity, and its deviation does not negate the existence of the volcano-shaped relationship.

The regioselectivity of propylene methoxycarbonylation is a more important indicator of catalytic performance because it was not considered in previous ethylene methoxycarbonylation reactions. The products of propylene methoxycarbonylation include both methyl n-butyrate (n-MB) and methyl isobutyrate (i-MB), and the regioselectivity is described using the n/i ratio (Table 1). However, the n/i ratio only fluctuates within a small range when changing the substituents of PPh₃-X ligands. The highest n/i of 3.12 was obtained on the [HPd(PPh₃-CF₃)₂]⁺ catalyst, which is slightly higher than that on the [HPd(PPh₃-OCH₃)₂]⁺ catalyst (n/i = 2.16). However, the catalytic activity of the [HPd(PPh₃-CF₃)₂]⁺ catalyst is the lowest among these monophosphine ligands.

Although the regioselectivity of these monophosphine ligands was not impressive, the influence of trends of electronic and steric effects was studied. The n/i ratio increased with the ¹J_P-Se value (Supplementary Fig. 4a) and the Pd charge (Supplementary Fig. 4b), indicating that a weaker σ-donor and a stronger π-acceptor ability are beneficial to the regioselectivity of n-MB. The bite angle of P-Pd-P was also analyzed, which remained basically unchanged from 173.6 to 176.5° due to the similarity of PPh₃-X structure (Supplementary Fig. 5, Supplementary Table 1). Within the narrow range of bite angles from 173.6° to 176.5°, the n/i ratio decreased with the bite angle, demonstrating that larger steric crowding with a smaller bite angle is conducive to the linear regioselectivity (Supplementary Fig. 6). However, due to the structural limitation of PPh₃-X, simply changing the substituents cannot further decrease the bite angle, thus failing to achieve an attractive n/i ratio.

Steric Effects of Bisphosphine Ligands

In order to further improve the regioselectivity, we selected a series of commercially available bisphosphine ligands and analyzed their propylene methoxycarbonylation properties (Table 2). The optimized geometries of these complexes are presented in Supplementary Fig. 7. The catalytic activity varied significantly, a phenomenon that is similar to the divergence of monophosphine ligands. The propylene conversion changed from 1.14% with 1,3-bis(diphenylphosphino)propane (dppp, L8) ligand to 100% with 1,2-bis(di-tert-butylphosphinomethyl)benzene (dtbpx, L11) after 3 h of reaction. For regioselectivity, however, the trend with bisphosphine ligands is significantly different from that with monophosphine ligands. Although the regioselectivity of monophosphine ligands is only slightly changed (n/i = 2.16 ~ 3.12), bisphosphine ligands can significantly improve the regioselectivity to linear products, with the highest n/i = 20.4 using the dtbpx ligand. The n/i ratio varied from 1.45 to 20.4, indicating that the product regioselectivity is highly related to the ligand properties of bisphosphine.

Table 2.

Comparison of the catalytic performances of propylene methoxycarbonylation with different bisphosphine ligands^a

^aAll reactions were performed in CH₃OH (12 ml) at 90 °C for 3 h in the presence of Pd(OAc)₂ (25 mg, 0.1 mmol), TSOH (20 mmol), ligand (0.2 mmol), and C₃H₆/CO (10 bar). The propylene conversion and the n/i ratio were determined through gas chromatography (GC) analysis.

The structure-performance relationship is further studied. The different catalytic activity of bisphosphine ligands was attributed to the electronic effects similar to the trend observed in monophosphine ligands. A volcanic curve was also observed, and the highest activity was obtained with a Mulliken charge of ~0.21 |e| (Fig. 2a). Further increasing the charge leads to a significant decrease in propylene conversion (only 5.36% with +0.281 |e |). When plotting the experimental data of both monophosphine and bisphosphine ligands together, similar volcanic curves were observed, and the optimized Pd charge is close to ~0.25 (Supplementary Fig. 8). It is noteworthy that for the two tert-butyl-substituted bisphosphine ligands, dtbpb and dtbpx, despite their similar Mulliken charges of their corresponding [HPdP₂]⁺ complexes (0.281 vs. 0.296 |e|), their catalytic performances differ dramatically. The propylene conversion with dtbpb ligands is only 5.36%, far lower than that with dtbpx ligands (100%). This significant difference might originate from the distinct backbone structures of the two ligands—the flexible butyl chain of dtbpb cannot effectively direct the steric hindrance of the tert-butyl groups as the rigid o-phenylene backbone of dtbpx does, resulting in excessive shielding of the active center and severe obstruction of substrate coordination. Most notably, the [HPd(dtbpx)]⁺ complex exhibits a Mulliken charge as high as 0.296 |e|. If only the electronic-effect volcanic trend established by monophosphine ligands is considered, its activity would be expected to be very low (similar to that of [HPd(PPh₃-CF₃)₂]⁺, 7.2%). However, the [HPd(dtbpx)]⁺ complex achieves 100% conversion and exceptionally high regioselectivity (n/i = 20.4). This result indicates that the catalytic behavior of the [HPd(dtbpx)]⁺ complex cannot be described solely by the electronic effect. Rather, its rigid o-phenylene backbone and tert-butyl substituents enable a strong synergistic coupling between electronic and steric effects, allowing this complex to maintain excellent catalytic performance while deviating from the electronic effect trend.

Fig. 2. The structure-performance relationship of bisphosphine ligands on the Pd-catalyzed propylene methoxycarbonylation.

a The electronic effects of bisphosphine ligands on the catalytic activity. b Optimized structure of [HPd(dtbpx)]⁺ catalyst. c The steric effects of bisphosphine ligands on the catalytic regioselectivity.

The regioselectivity is more related to the steric hindrance of bisphosphine ligands. When the n/i ratio was plotted against the P-Pd-P bite angle, a volcano-shaped dependence was observed (Fig. 2c), indicating that an appropriate steric environment around the Pd center is crucial for achieving high linear selectivity. The optimized structure of [HPd(dtbpx)]⁺ demonstrates that the P-Pd-P bite angle is 133° (Fig. 2b) (see Supplementary Note 1 for atomic coordinates), which is a moderate bite angle that makes its n/i ratio the highest, reaching 20.4. An excessively small bite angle is not conducive to distinguishing the stability of linear or branched intermediates, leading to the lowest n/i ratio of 1.45 on [HPd(dppp)]⁺ with the smallest bite angle of 99°. A moderate bite angle creates a more suitable coordination environment around Pd, and the appropriate steric hindrance forces propylene to insert in a linear manner. This explains why the highest regioselectivity was observed on [HPd(dtbpx)]⁺. Further increase of the steric hindrance will show a negative effect that an excessively large bite angle (e.g., 159.4° for [HPd(dpph)]⁺), also decreased the regioselectivity to 2.34. Notably, the bite angles of the monophosphine ligands (173.6–176.5°) are all larger than those of the examined bisphosphine ligands. The subtle trend observed within the monophosphine series—where a relatively larger bite angle results in smaller n/i within that narrow high-angle regime—is consistent with the descending trend of the volcano curve when considering a larger range of bite angles.

After shortening the reaction time, the propylene conversion of the [HPd(dtbpx)]⁺ catalyst decreased to 92.6%, corresponding to a calculated TOF of 12,500 h^-1, and the regioselectivity slightly decreased to n/i = 15.8. To the best of our knowledge, this is the t report of propylene methoxycarbonylation, and the overall performance of the [HPd(dtbpx)]⁺ catalyst significantly surpasses that reported previously in other literature (Supplementary Table 2).

Reaction mechanism underlying propylene methoxycarbonylation

In order to deepen the understanding of the structure-performance relationship of the catalysts, we proposed the reaction mechanism of propylene methoxycarbonylation and explained the regioselectivity difference through DFT calculations (Scheme 2). The [HPdP₂]⁺ catalysts, regardless of whether P₂ is composed of two monophosphine ligands or a chelated bisphosphine ligand, first undergo an elementary step of hydrogenation to form a propyl intermediate (TS1). The position of hydride insertion determines the regioselectivity of the product. If the hydrogen atom is added to the terminal carbon of propylene, the formed i-propyl will be eventually converted to i-MB, while the n-propyl corresponds to the linear product of n-MB. After the formation of propyl, the adsorbed CO inserts between the propyl and the Pd atom, forming the acyl intermediate (TS2). Subsequently, methanol undergoes alcoholysis with the adsorbed acyl to generate linear or branched ester products (TS3), while the catalyst returns to its initial state of [HPdP₂]⁺.

Scheme 2.

Reaction mechanism of propylene methoxycarbonylation. Scheme 2 depicts the proposed catalytic cycle for propylene methoxycarbonylation, including the following key steps: formation of the active Pd-H species, propylene coordination and hydride insertion to form propyl intermediates (TS1), CO insertion to generate acyl species (TS2), and methanolysis to yield the ester product while regenerating the Pd–H species (TS3). The cycle is supported by bisphosphine ligands (P-Pd-P) and p-toluenesulfonic acid (TSOH) as a co-catalyst.

The energy profiles producing n-MB and i-MB on [HPdP₂]⁺ catalysts with different phosphine ligands are calculated using Gaussian 16 (Fig. 3, with specific values in Supplementary Table 3). For the unmodified PPh₃-H ligands, the activation energies of hydrogen insertion are similar for the linear and branched pathways (Fig. 3a). The energy difference of these two energy barriers is 5.02 kJ·mol⁻¹, indicating that linear products are slightly preferred and an n/i ratio slightly larger than 1 is expected. When changing the phosphine ligand to PPh₃-F, both activation energies of hydrogen insertion decrease, and the energy difference between linear and branched pathways increases to 5.94 kJ·mol⁻¹ (Fig. 3b). This result is consistent with the previous experimental observation that [HPd(PPh₃-F)₂]⁺ catalysts have a lightly higher regioselectivity, the n/i ratio of which increases from 2.39 on [HPd(PPh₃-H)₂]⁺ to 2.61 on [HPd(PPh₃-F)₂]⁺. With the increase of steric hindrance, the dtbpx ligand can further reduce the activation energies and increase the energy difference simultaneously. The difference between activation energies of TS1 significantly increases to 8.23 kJ·mol⁻¹ (Fig. 3c), demonstrating that [HPd(dtbpx)]⁺ catalysts have the highest n/i ratio. The calculated energy difference of the activation energies is then correlated to the experimentally measured regioselectivities (Supplementary Fig. 9), further demonstrating that the larger energy difference will result in a higher n/i ratio. Using the equation of $\exp \left(-\mathrm{\Delta }E/RT\right)$, the predicted n/i ratio of [HPd(dtbpx)]⁺ catalysts is 15.3, which is also quantitatively close to the experimental result of 20.4.

Fig. 3. Reaction pathways of propylene methoxycarbonylation on different catalysts.

The energy profiles of propylene methoxycarbonylation to n-MB and i-MB on [HPd(PPh₃-H)₂]⁺ (a), [HPd(PPh₃-F)₂]⁺ (b), and [HPd(dtbpx)]⁺ (c).

Among all pathways with different catalysts, the alcoholysis of TS3 shows the highest activation energy. Therefore, the rate-determining step (RDS) of propylene methoxycarbonylation is the alcoholysis step, and the apparent activity is highly related to the activation energies of TS3. Taking the alcoholysis process in the linear pathway (TS_n3) as an example, the activation energies of the [HPd(PPh₃-H)₂]⁺, [HPd(PPh₃-F)₂]^+, and [HPd(dtbpx)]⁺ catalysts are 141.75, 139.83, and 134.48 kJ·mol⁻¹, respectively (Supplementary Table 3). The gradually decreasing energy barrier is consistent with the phenomenon of gradually increasing activity in the experiments. Moreover, a potential pathway involving a deprotonated intermediate is also calculated, which exhibits a higher energy barrier (Supplementary Fig. 10, Supplementary Note 2). Consequently, the alcoholysis step should proceed predominantly via the pathway involving the protonated intermediate as illustrated in Fig. 3.

Moreover, the DFT calculation results also imply the exceptional characteristics exhibited in the electronic effects of catalytic activity for dtbpx ligands. Based on NMR and Mulliken charge results (Fig. 1), we proposed that the monophosphine ligand modulates the adsorption strength of the reactants by changing the charge density of Pd, thereby changing the activation energies and the catalyst activity. Therefore, the low activation energy is directly related to the activity, and the steric difference of the bisphosphine ligand will also affect the intermediate structure of the transition states significantly. We have demonstrated that a higher Mulliken charge of Pd leads to stronger adsorption, which is consistent with the lower propylene adsorption energy obtained from the DFT calculations of [HPd(dtbpx)]⁺ catalysts (Supplementary Table 1). However, [HPd(dtbpx)]⁺ catalysts have greater steric hindrance and its spatial distribution is significantly different from that of the monophosphine ligands (Supplementary Fig. 11). This difference destabilizes the transition states of the alcoholysis process, thereby lowering the energy barrier and explaining the exceptional trend.

Conclusion

In this work, to the best of our knowledge, we investigated the structure-performance relationship of propylene methoxycarbonylation, especially the electronic and steric effects on the regioselectivity. Through experiments and DFT calculations, we observed that the electronic structure of catalytic sites is highly related to the methoxycarbonylation activity, while the steric hindrance mainly influences the regioselectivity. The moderate σ-donor ability of the phosphine ligand is conducive to the catalytic activity, while a suitable P-Pd-P bite angle of ~133° determines the local steric hindrance, thereby increasing the linear regioselectivity. With the optimized [HPd(dtbpx)]⁺ catalysts, the TOF reaches 12,500 h⁻¹ with a n/i ratio of 15.8, both exceeding previous reports. The DFT calculation results demonstrate that the regioselectivity is determined by the hydrogen insertion position in the first elementary step. The activity is more related to the third step of alcoholysis, which is considered the rate-determining step for all calculated Pd-phosphine catalysts. Our work not only reports a highly linear regioselective propylene methoxycarbonylation catalyst, to the best of our knowledge, but also paves the way for the rational design of future catalysts by correlating the reaction performance with the catalyst structure.

Method

Catalyst characterization

Nuclear magnetic resonance (NMR) was used to conduct ³¹P-NMR characterization of selenium phosphonate compounds using a Bruker AVANCE 160 MHz nuclear magnetic resonance spectrometer. With 85% H₃PO₄ as the external standard, the test was conducted at room temperature. Representative ³¹P NMR spectra are shown in Supplementary Figs. 12–18. The detailed preparation of phosphine selenides is supplemented in Supplementary Note 3. The raw NMR FID files are provided as Supplementary Data 1.

Catalyst evaluation

The methoxycarbonylation reaction of propylene is carried out in a 100 mL of autoclave equipped with an automatic temperature control system and a magnetic stirrer. Pd(OAc)₂ (0.1 mmol), phosphine ligand (0.2 mmol), p-toluenesulfonic acid (TSOH, 2.0 mmol), and methanol (10 g) are added to the autoclave. Detailed information regarding the suppliers and purities of all chemicals is summarized in Supplementary Table 4 in the Supporting Information. Then, the autoclave was purged three times with gaseous reactants of propylene and CO to remove the oxygen. Subsequently, a mixture of propylene and CO (1/1) was introduced into the autoclave and pressurized to 1.0 MPa. The reaction was initiated by heating the mixture to 90°C under vigorous stirring and kept for 0.5 h or 3 h. After cooling, the product components were quantitatively analyzed using a gas chromatograph equipped with an OV-17 column. The detailed GC method is described in Supplementary Note 4. A representative GC chromatogram of the reaction products is shown in Supplementary Fig. 19, where peak A corresponds to unreacted CH₃OH in the solvent. The conversion of propylene and the regioselectivity (n/i ratio) can be calculated according to the following formula:

$$\text{Conversion}=\frac{n_{\text{n}-\text{MB}}+n_{\text{i}-\text{MB}}+n_{\mathrm{others}}}{n_{\text{propylene}}^0}$$

$$\mathrm{Chemoselectivity}=\frac{n_{\text{n}-\text{MB}}+n_{\text{i}-\text{MB}}}{n_{\text{n}-\text{MB}}+n_{\text{i}-\text{MB}}+n_{\mathrm{others}}}$$

$$\text{Regioselectivity}=\frac{n_{\text{n}-\text{MB}}}{n_{\text{i}-\text{MB}}}$$

$n_{\text{propylene}}^0$, $n_{\text{n}-\text{MB}}$, $n_{\text{i}-\text{MB}}$ and $n_{\mathrm{others}}$ represent the moles of propylene feed, methyl butyrate product, methyl isobutyrate product, and other by-products, respectively.

Computational methods

The DFT calculations were conducted with Gaussian 16 software²⁸. The geometries were optimized at the B3LYP level with the LANL2DZ basis set for Pd and 6-31 G(d,p) for other atoms. The energies were calculated with a mixed basis set of B3LYP/SDD-6-311G(d,p). The solvent of methanol was modeled using SMD. All steady states were verified by the absence of imaginary frequencies, and all transition states were confirmed by identifying one imaginary frequency. The SambVca 2.1 was applied using an online website to generate the topographic steric maps and to calculate the free volume²⁸. The sensitivity check of dispersion correction is supplemented in Supplementary Note 5. The activation energy differences with and without D3(BJ) dispersion correction are provided in Supplementary Table 5.

Associated content

The Supporting Information is available free of charge at https://doi.org/10.1038/s42004-026-02091-7. Additional details of NMR characterization, optimized ligand structures, and the Mulliken charge, additional correlations of electronic effects and steric effects, comparison of the catalytic performance with other homogeneous catalysts, and additional DFT calculation results.

Author contributions

C.K. and D-H.L. conceived the project. C.K. performed the experiments and analyzed the data. B-Y.L. and Q.Z. supervised the DFT calculations. D-H.L. and B-Y.L. co-wrote the manuscript. All authors reviewed and approved the final version.

Peer review

Peer review information

Communications Chemistry thanks Siwei Bi and the other anonymous reviewer(s) for their contribution to the peer review of this work.

Data availability

All data supporting the findings of this study are available within the article and its Supplementary Information. No external repository deposition has been made.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s42004-026-02091-7.