Introduction

Arenes and heteroarenes are widely used in the chemical industry and are the most common structural motifs in small molecule pharmaceuticals, agrochemicals, and natural products1. They can be easily functionalized by various synthetic methods, making them versatile chemical intermediates for many products in our daily lives. Notably, the saturated (hetero)aromatic analogs are much less investigated compared with the well-explored (hetero)aromatic counterparts. To illustrate: Compared to aromatic acids/esters and the corresponding boronic acids/esters, cyclohexanecarboxylic acids/esters and cyclohexylboronic acids/esters are 1–3 orders of magnitude less commercially available (Supplementary Table 1). In contrast to their commercial availability, saturated (hetero)cycles have attracted rapidly growing attention in recent years in many fields such as the mass production of chemicals2, energy generation3, and storage4,5, drug discovery1, etc. Compared to the planar (hetero)aromatic compounds, their saturated (hetero)cyclic analogs have three-dimensional structures, especially when two or more substituents are present. This three-dimensionality of the structures offers the possibility of achieving significant positive changes in the compound properties. In the polymer industry, for example, the glass transition temperature (Tg) of a material can be adjusted by manipulating the ratio between cis and trans diastereomers2. A number of other examples can be found in current drug research. Here, there is a growing interest in the integration of more three-dimensional structural components and the use of sp3-C centers in the development of new drugs (“escape from the flatland”) in order to improve the diversity of bioactive compound libraries and expand the chemical space6,7,8. Indeed, beneficial correlations have been demonstrated between increased molecular complexity and the improved clinical success of compound transition from discovery to drugs9. As a consequence, 3D-saturated (hetero)cyclic scaffolds are featured in a large number of top-selling small molecule drugs in recent years. However, it remains challenging and step-intensive to synthesize the desired multi-substituted (hetero)cyclic 3D scaffolds10.

Among the numerous known synthetic methods for the preparation of aliphatic saturated (hetero)cyclic compounds, catalytic stereoselective hydrogenations are probably the most general, powerful, and atomically efficient method11. In this respect, diastereoselective hydrogenations offer a direct pathway to unlock the 3D chemical structures starting from abundant 2D (hetero)arenes with di- or multi-substituents. However, the high kinetic barrier to break aromaticity (e.g., the aromatic stabilization energy for benzene is 36 kcal·mol−1)12 and to achieve chemo- and stereoselectivity in the presence of other functional groups remain major challenges for any catalyst development. Nevertheless, sophisticated molecular catalysts including borenium-13, borane-14,15, iridium-16, rhodium-17,18,19,20, zirconium-21, and molybdenum-based complexes22 allowed to achieve such transformations. From a practical point of view, homogeneous systems have intrinsic disadvantages like more difficult catalyst recycling, sensitivity, often complex ligand synthesis, and/or the use of basic or acidic additives. Obviously, the development of a robust heterogeneous catalyst is more attractive for many applications, especially on a larger scale. Despite the many heterogeneous catalysts developed for arene hydrogenations3,4,23,24,25,26, there is only a very limited number of publications regarding the diastereoselective hydrogenation of multiple-substituted (hetero)arenes, and most known materials make use of chiral auxiliaries27,28 or Lewis acids29.

Following today’s vision of increasing sustainability in advanced chemical production, we herein present a unique heterogeneous Pt catalyst, which allows for the general diastereoselective hydrogenation of a variety of multi-substituted (up to six substituents!) and functionalized (hetero)arenes under mild reaction conditions. A combination of kinetic studies and modern characterization techniques identified a two-shell Pt structure as the most active site. To showcase the practicability of the catalyst and methodology, a phthalate-free plasticizer (cis−1,2-cyclohexanedicarboxylic acid, dioctyl ester) was prepared on a kg-scale under solvent-free conditions.

Results and discussion

Material synthesis and kinetic investigations of the benchmark system

As a model reaction, the diastereoselective hydrogenation of the industrial bulk product dimethyl phthalate (1a, DP for short) to provide dimethyl 1,2-cyclohexane dicarboxylate was used, as the related higher esters are of interest as an alternative for phthalate plasticizers. To achieve high stereoselectivity and to identify highly active materials, catalytic tests were performed under mild reaction conditions (25 °C, 10 bar H2 pressure).

We commenced the study testing different commercially available and homemade supported metal catalysts (Table 1). Except for Pt-based catalysts, which showed limited conversions, no activity was observed for all the other tested materials (Pd/C, Ru/C, Ni/C, Co/C, and Cu/C) under these conditions (Table 1, entries 1–7). Variation of the support showed Pt/TiO2 as the optimal catalyst composition (Table 1, entries 6–10). For all the tested Pt/TiO2 materials, an excellent diastereoselective ratio (d.r.) of 99:1 towards the cis-product (2a) was obtained in a wide temperature range from 25 °C to 120 °C (Table 1, entries 11–13), which is different compared to other Pt-based catalysts30. The ester group also remained intact, and no over-hydrogenation occurred31. With an increased reaction time, excellent yield and d.r. can be observed at ambient conditions (25 °C, 1 bar H2 pressure, Table 1, entry 14). We also noted that, commercial Pt/C from different suppliers exhibited different activities (Table 1, entries 6 and 7). It is thus interesting and significant to identify the real catalytically active site to allow reproducible material preparation and the rational design of a highly efficient catalyst. Hence, we prepared a series of Pt/TiO2 catalysts with different Pt loadings (denoted as xPt/TiO2, where x represents the weight percentage of Pt) using an impregnation method as reported32. The measured Pt loading of all the prepared samples is very close to the designed value, as evidenced by the ICP-OES results (Supplementary Table 2). As shown in Fig. 1a, turnover frequency (TOF) shows an increasing trend when the Pt loading increases from 0.5 wt% to 4 wt%, followed by a slight decrease when further increasing the Pt loading to 6 wt%. To illustrate this phenomenon, kinetic studies were performed. The effect of mass transfer was eliminated (Supplementary Fig. 1) to ensure that the data were collected in a kinetically controlled region. The reaction orders for substrate 1a and hydrogen are almost identical for all samples tested (Fig. 1b): an order close to 0 for 1a and a positive order in the range of 0.7 ± 0.1 for hydrogen (Fig. 1c, data summarized in Supplementary Table 4). These reaction order values are consistent with those reported in the literature for the hydrogenation of benzene33,34, toluene35, and benzoic acid36 over Pt-based catalysts. A zero-order reaction for 1a indicates that the substrate 1a is involved in the reaction in a strongly adsorbed state. In addition, the reaction order values measured here are supposed to reflect a reaction mechanism in which the hydrogen addition to the intermediate on the catalyst surface is most likely the rate-determining step of this reaction. While we should note that it is also possible that during the reaction, the catalyst surface is saturated with substrate 1a, resulting in the hydrogen chemisorption as the kinetically limiting step. The effect of temperature on the reaction rates of all catalyst samples can be seen in the Arrhenius diagram (Fig. 1d). The dependencies in the Arrhenius diagram were linear in the investigated temperature range of 298–313 K for all investigated samples. Hence, the apparent activation energy (Ea) can be determined within this region. For all the samples, the Ea values are quite close, within the range of 33–42 kJ·mol−1, which is in agreement with reported data34,37. The measured kinetic parameters, including the reaction orders for the substrate 1a and H2, together with the apparent activation energies, did not differ significantly for all the tested samples, suggesting no obvious change in the reaction mechanism. Consequently, we would consider that the difference in reactivity originates from the number of active sites. In other words, since the total number of Pt centers involved in the reaction is the same (0.5 mol%) for all samples, it is reasonable to assume that some specific centers are the most active sites for the reaction, while others are less active or even inactive.

Fig. 1: Activity tests and kinetic studies of Pt/TiO2-catalyzed hydrogenation of 1a.
Fig. 1: Activity tests and kinetic studies of Pt/TiO2-catalyzed hydrogenation of 1a.
Full size image

a) The effect of Pt loading on the activity. b Reaction orders with respect to the substrate. c Reaction orders with respect to H2. d Arrhenius plots. The error bars represent the standard deviation obtained from three independent performance tests. Reaction conditions: 0.25–1 mmol substrate, 0.5 mol% Pt, 5–20 bar H2, 2 mL cyclohexane, 25–40 °C.

Table 1 Diastereoselective hydrogenation of dimethyl phthalate (1a)
Full size table

Identification of catalytically active sites

The catalyst structure essentially comprises the electronic and the geometric structure38. To uncover the most active site of the catalyst, we first performed X-ray Photoelectron Spectroscopy (XPS) to explore the effect of the electronic structure on the reactivity. As shown in Supplementary Fig. 2, both Pt0 and PtO were observed for all the samples. The fractions of these two species are quite close (around 62–66% for Pt0 and 34–38% for PtO, as shown in Supplementary Table 5), indicating that the electronic structure is not the decisive factor for the difference in the catalytic reactivity. Therefore, we assume a determinant role of the geometric structure, although controversial conclusions have been drawn for the hydrogenation of benzene and toluene. Such reactions have been designated as structure-insensitive reactions by groups of Verykios39 and Figueiredo40, while Somorjai and co-workers37 observed a moderate structure sensitivity for these hydrogenation reactions. To investigate the size distributions of supported Pt clusters and nanoparticles (NPs), we used scanning transmission electron microscopy (STEM). Some representative high-angle annular dark-field (HAADF) images are shown in Fig. 2a–e, which are categorized according to the physical model with different numbers of shells (definition is given in Supplementary Figs. 3 and 4 and Table 2)32. A significant amount of one-shell structured Pt was observed over 0.5Pt/TiO2 and 1Pt/TiO2. Because these two samples exhibited the least reactivity (Fig. 1a), we conclude that this type of Pt structure is not the most active site. When the Pt loading increases, the number of clusters (particle size < 0.5 nm) becomes negligible, while that of two-shell structured Pt and larger Pt NPs increases to some extent. Here, we summarize the relative populations of supported Pt with different structures. When the Pt NP are present in a three-shell structure or even larger size, its property becomes similar to the Pt(111) surface, which generally has a weak affinity to the substrate. For this reason, we have grouped these NPs together. As shown in Fig. 2f and the inset, a good correlation has been observed between the fractions of two-shell structured Pt and TOF, suggesting that such a type of Pt structure may be the most active Pt site for the reaction. To prove this hypothesis, we synthesized another two types of Pt/TiO2 catalysts: (1) supported Pt clusters with almost all the Pt size below 1 nm (denoted as PtS/TiO2, Fig. 3a) and (2) supported Pt NPs with Pt size > 2 nm (denoted as PtL/TiO2, Fig. 3b). Both catalysts exhibited limited activity, with reaction TOF even lower than that of 0.5Pt/TiO2 (Fig. 3c). These results strongly suggest that the two-shell structured Pt is most likely the most active site for this reaction.

Fig. 2: Representative HAADF–STEM images and the derived particle size distributions of the series of Pt/TiO2 catalysts with different Pt loadings, categorized according to the physical model.
Fig. 2: Representative HAADF–STEM images and the derived particle size distributions of the series of Pt/TiO2 catalysts with different Pt loadings, categorized according to the physical model.
Full size image

a 0.5Pt/TiO2. b 1Pt/TiO2. c 2Pt/TiO2. d 4Pt/TiO2. e 6Pt/TiO2. f Relative populations of supported Pt clusters/NPs with different structures; Inset: the correlation between the frequency of two-shell structured Pt and TOF.

Fig. 3: Representative HAADF–STEM images, the derived particle size distributions, and activity tests of the PtS/TiO2 and PtL/TiO2 catalysts.
Fig. 3: Representative HAADF–STEM images, the derived particle size distributions, and activity tests of the PtS/TiO2 and PtL/TiO2 catalysts.
Full size image

a PtS/TiO2. b PtL/TiO2. c Comparison of TOF with 0.5Pt/TiO2 and 4Pt/TiO2. The error bars represent the standard deviation obtained from three independent performance tests. Reaction conditions: 0.5 mmol substrate, 0.5 mol% Pt, 10 bar H2, 2 mL cyclohexane, 25 °C, conversions and yields were determined by GC-FID using n-hexadecane as an internal standard.

Table 2 Calculated van der Waals diameter of dimethyl phthalate and model catalysts
Full size table

Density functional theory (DFT) calculations were performed to illustrate the size compatibility between the catalyst and substrates. The van der Waals diameter of 1a, calculated to be 1.03 nm (Table 2 and Supplementary Table 6), is larger than the size of a one-shell structured Pt nanoparticle (0.88 nm). Since H2 molecules should also dissociatively adsorb on Pt sites, it is natural to conclude that the size of this one-shell Pt nanoparticle (or even smaller Pt cluster) is too small to accommodate both 1a and H2 concurrently. The surface site will be over-covered by 1a based on the rather strong adsorption energy (−2.28 eV, Supplementary Table 7) on Pt13/TiO2. This might explain the very limited reactivity observed over the small-sized Pt catalyst. However, once the Pt size is too large, the adsorption of 1a on the Pt surface becomes much weaker (−0.63 eV for three-shell Pt159 and −0.21 eV for Pt(111) surface, see Supplementary Table 7), and this is contrary to our kinetic analysis, which shows that 1a was involved in the reaction in a strongly adsorbed state. Consequently, the two-shell structured Pt nanoparticle, which has a size of 1.40 nm (Table 2), is large enough to accommodate both 1a and H2 concurrently, and the adsorption energy of 1a (−2.07 eV., Supplementary Table 7) indicates a strong adsorption, as found in our kinetic analysis. Therefore, both factors make the two-shell structured Pt the most active site for the diastereoselective hydrogenation of dimethyl phthalate.

Synthetic applications

After identifying the optimal Pt/TiO2 catalyst in the benchmark reaction, we tested the hydrogenation of diverse substituted arenes and heteroarenes to evaluate the reaction scope and limitations (Fig. 4). Obviously, a broad substrate scope and functional group tolerance are prerequisites for advanced fine chemical applications. Substrates with substituents in the different positions of the arene ring were efficiently hydrogenated and gave very good to excellent yields and diastereoselectivities. By varying the position of the substituents in phthalates from ortho, meta to para, a slight decrease in diastereoselectivity is observed. While ortho-phthalates gave >99/1 diastereoselectivity, generating only a single diastereomer in quantitative yield (2a), the meta- and para-phthalates provided quantitative yields but lower selectivity (2b, 2c). In all cases, the corresponding cis-isomers were produced preferentially. Notably, product 2d was synthesized from n-octyl phthalate41,42, a commercial plasticizer, using standard reaction conditions with >99% diastereoselectivity. Similar to phthalates, substituted benzoic acid esters with electron-donating or -withdrawing moieties, showed comparable reactivity and diastereoselectivity (2e–2h). Irrespective of the electronics of the substituents, ortho-substituted benzoates gave excellent diastereoselectivities. Furthermore, methyl 3,4-dimethyl benzoate provided 2i in 90% isolated yield and a high d.r. value (92/8). Hydrogenation of ibuprofen methyl ester was performed at 60 °C using cyclohexane as solvent, which provided the corresponding hydrogenated product in 94% isolated yield (2v). In this case, the diastereoselectivity was comparable to other ester-containing arenes like 2g, 2h, and 2j.

Fig. 4
Fig. 4
Full size image

Synthetic applications and kg-scale synthesis of non-phthalate-based plasticizers.

The diastereoselective hydrogenation of non-functionalized substrates and substrates with weakly coordinated functional groups is considered an extremely challenging problem43,44,45. In contrast to starting materials that bind to the catalyst (center) via a strong interaction, the fundamental difficulty with such transformations is that a large number of transition states are feasible. Exemplarily, catalytic hydrogenation of xylenes was performed. Interestingly, in all the cases a similar d.r. Ratio (around 3:1) is obtained irrespective of the position of the second methyl group (2k–2m). In these cases, product yields were calculated based on 1H NMR due to the lower boiling point of the corresponding 1,n-dimethylcyclohexanes or based on isolated products after filtering of the catalyst using neat conditions.

Next, we investigated some tri-substituted arenes under standard reaction conditions. Indeed, the products 2n–2p formed with similarly good diastereoselectivities compared to di-substituted arenes. Even tetra- and penta-substituted benzenes can be diastereoselectively (i.e., up to 80:20) hydrogenated to 2q–2r with the Pt/TiO2 catalyst system. As an example of fused-ring arenes, tetrahydronaphthalene was hydrogenated to form octahydronaphthalene (2s) in 98% GC yield and 80/20 d.r. value. Finally, bicyclic strained arenes were hydrogenated, providing 2t–2u with excellent diastereoselectivities, too.

The stereoselective hydrogenation of heteroarenes offers an effective strategy for the direct access to saturated heterocycles. In recent years, several homogeneous46,47,48 and heterogeneous49,50,51 catalyst systems have been developed for such reductions with high levels of chemo- and stereoselectivity. However, using heterogeneous catalysts for this task remains a challenge. Hence, we investigated the hydrogenation of pyridine (1ab) and phthalimides (1w–1z) in the presence of water as solvent at 60 °C. While dimethyl pyridine-2,3-dicarboxylate was hydrogenated with moderate diastereoselectivity, in the case of phthalimides, the reaction was highly selective to generate one diastereomer in high isolated yields. Also, 6-membered heterocyclic 9H-xanthene provided both arene hydrogenated product (2ac) in very good d.r. The heterocyclic product 2ad was also easily isolated with excellent diastereoselectivity after performing the hydrogenation reaction at 60 °C in water. 2ad is an important building block in the synthesis of (4aS,7aS)−6-benzyloctahydro-1H-pyrrolo[3,4-b] pyridine, which is commonly used as a crucial intermediate in the preparation of the drug moxifloxacin52,53,54. Similarly, hydrogenation of isoindoline-1,3-dione derivatives, which constitute an important group of medicinal substances55,56, provided the corresponding saturated heterocycles in excellent diastereoselectivities and with very good, isolated yields (2ae–2af).

Next, two examples of pinacol borane esters containing arenes with additional one and two substituents (2ag–2ah) were tested. While at room temperature, no conversion was observed, the corresponding hydrogenated products were provided with very good diastereoselectivity at 80 °C. Interestingly, substituted benzamides (2ai–2ap) formed the cyclohexanecarboxamides in good d.r. values, too. Compared to secondary amides, primary amides gave higher diastereoselectivity, while tertiary amides didn’t show any reactivity even at higher reaction temperatures.

Reusability, stability, and easier separation of product and catalyst are intrinsic advantages of heterogeneous catalysts, as described at the beginning. For the model reaction, the Pt/TiO2 catalyst shows a stable conversion (ca. 40%) and d.r. value (99/1) during 5 runs (Supplementary Fig. 5). STEM characterization of the recycled catalyst reveal that the proposed Pt structure basically holds, with a slight decrease in the fraction of one-shell NPs and a tiny increase in the proportion of larger particles (>2 nm) (Supplementary Fig. 6). In order to demonstrate the practicability of this catalyst system and the process described herein, the diastereoselective hydrogenation of di-n-octyl phthalate was performed on kg-scale. This reaction was selected due to the concerns about the toxicity of phthalates and the interest in developing alternative non-phthalate-based plasticizers in industry2. To examine its stability and recycling, the Pt/TiO2 catalyst was recycled four times under neat conditions. The desired product was collected simply using filtration, without any further purification operations. 876 g of the cis-product was obtained with a d.r. value > 99:1 (Fig. 4).

In conclusion, we have developed a heterogeneous Pt catalyst allowing for the general diastereoselective hydrogenation of a wide range of (hetero)arenes under comparably mild conditions (10 bar H2; 25–120 °C). Detailed characterizations and kinetic experiments revealed a two-shell Pt structure as the most active site, which opens up the possibility for further rational design of catalysts for diastereoselective hydrogenations. The kg-scale synthesis of a phthalate-free plasticizer demonstrates the potential of such transformations for industrial applications.

Methods

Catalyst preparations

TiO2-supported Pt catalysts with different Pt loadings (0.5 wt%, 1 wt%, 2 wt%, 4 wt%, 6 wt%) were synthesized using the wet-impregnation method. For a typical catalyst preparation (0.5 g), the desired amount of aqueous hydrogen hexachloroplatinic acid solution (7.5 mg Pt·mL−1, Alfa Aesar) was diluted to 16 mL using deionized water, and heated in an oil bath to 60 °C with vigorous stirring (800 r.p.m.) for 15 min, followed by the addition of TiO2 support (P25, Sigma-Aldrich). The temperature was kept at 60 °C for 16 h. After being cooled down, water was removed using rotatory evaporation. The resultant material was then put into the oven at 110 °C overnight. The sample was ground thoroughly and then calcined at 450 °C for 4 h with a heating rate of 10 °C·min−1. Finally, the sample was reduced in a flow-tube furnace under a steady flow of 5% H2/Ar at 250 °C for 1 h with a heating rate of 5 °C·min−1. This series of catalysts are denoted as xPt/TiO2, where x represents the weight percentage of Pt. For the catalyst synthesis of kg-scale reaction, a similar procedure was used for catalyst preparation except that concentrated aqueous hydrogen hexachloroplatinic acid solution (37.7 mg Pt·mL−1, Alfa Aesar) was used.

Preparation of PtS/TiO2

A two-step impregnation method was used to synthesize PtS/TiO2. For a typical catalyst preparation (1.0 g), desired amount of aqueous hydrogen hexachloroplatinic acid solution (7.5 mg Pt·mL−1, Alfa Aesar) and ethylenediamine solution (Pt: ethylenediamine = 1:2 in mol) was diluted to 16 mL in deionized water, and heated in an oil bath to 60 °C with vigorous stirring (800 r.p.m.) for 15 min, followed by the addition of TiO2 support (P25, Sigma-Aldrich). The temperature was kept at 60 °C for 2 h. After being cooled down, water was removed using rotatory evaporation. The sample was ground thoroughly and then calcined at 300 °C for 3 h with a heating rate of 10 °C·min−1. Then the catalyst was washed with 1 M NH4Cl, and the resultant material was put into an oven at 110 °C overnight. Finally, the sample was reduced in a flow-tube furnace under a steady flow of 5% H2/Ar at 300 °C for 3 h with a heating rate of 5 °C·min−1.

Preparation of PtL/TiO2

Sol-immobilization method is used to prepare PtL/TiO2. To initiate the synthesis process, a sol containing Pt NPs was prepared first. The desired amount of H2PtCl6 solution was pipetted into the beaker. Subsequently, 28 mL of water was added to the beaker, and the solution was stirred for 5 min. The PVA solution, with a PVA: Pt weight ratio of 1:1, was then added to the beaker, followed by stirring for 20–30 min to form Pt-PVA colloid. To reduce the Pt particle and control its size, a freshly prepared sodium borohydride (NaBH4) solution was added to the mixture. The NaBH4: Pt mole ratio was set at 4:1. The stirrer speed was then increased to 800 r.p.m., and the solution was left to stir for 30 min. Then H2SO4 was added to adjust the pH to 3. Subsequently, the desired amount of TiO2 support was gradually added to the beaker, ensuring the dissolution of any powder on the liquid surface before adding more. The solution was left to stir for 40 min. The catalyst solution was poured into the funnel, with the beaker rinsed using deionized water to ensure complete transfer. The water level in the funnel was maintained during the washing process by adding additional water as needed. After all the water had passed through the funnel, the funnel was placed in an oven (110 °C) overnight.

Diastereoselective hydrogenation of dimethyl phthalate

Unless otherwise stated, reactions were performed in autoclaves. Solvents were used directly without further purification. Gas chromatography (GC) was performed on an HP 6890 with a HP5 column (Agilent). All the materials were tested as catalysts without any further treatment. For a typical small-scale reaction (2 mL) procedure, the desired amount of catalyst (0.5 mol% of Pt), 0.5 mmol of substrate, and 2 mL of solvent were added in a 4 mL vial with a magnetic stirring bar and septum cap. Then, a needle was inserted in the septum, allowing H2 to enter the vial. The vials (up to eight) were set in an alloy plate and placed in a 300 mL steel Parr autoclave. The autoclave was flushed with H2 for times at 10 bar and finally pressurized to 10 bar. Then, it was placed into an aluminum block and kept at 25 °C for 16 h. When the reaction is completed, the autoclave is vented. Finally, the samples were removed from the autoclave with the addition of n-hexadecane as the internal standard and ethyl acetate to the crude mixture, followed by filtration to separate the solid catalysts using a celite pad. The organic layer was analyzed at the GC-FID (Agilent 7890 A) and by NMR.

Calculation of TOF

The reaction was performed at three different times (in all cases conversions < 20%). Then we plotted the concentration of the reactant as a function of time. At such a low conversion, the coverage of the product on the catalyst surface was negligible, and we could assume that all the active sites were involved in the reaction. A linear correlation could be obtained by curve fitting. The initial rate (r) was calculated from the slope57:

$$r=-\frac{{{{\rm{d}}}}[{{{\rm{R}}}}]}{{{{\rm{dt}}}}}=-{slope}$$
(1)

Where r is the initial rate, mol·L−1·h−1; [R] is the concentration of the reactant, mol·L−1; t is the reaction time, h.

The TOF is calculated as follow:

$${TOF}=\frac{{rV}}{{n}_{{Pt}}}$$
(2)

Where V is the volume of the reaction mixture, L; nPt is the amount of surface Pt, mol, which is calculated as follow:

$${n}_{{Pt}}={n}_{{total\; Pt}}\times d$$
(3)

Where ntotal Pt is the total amount of Pt in one reaction, mol; d is the dispersion of Pt, determined by CO chemisorption, as shown in Supplementary Table 3.

Each measurement was repeated three times to plot the error bar.

Catalyst recycling

In order to exclude the effect of catalyst loss during the operation procedure, we used two batches of catalysts for recycling tests. In batch A0, we used the standard amount of catalyst. In the second batch, B0, an excess amount of catalyst (ca. ten times that of A0) was used. The reaction condition and operation procedure were identical to the hydrogenation reaction. After one hydrogenation reaction, the reaction mixture in batch A0 was analyzed by GC-FID, and the result was recorded as the activity for the fresh catalyst. On the other hand, the catalyst in batch B0 was separated by centrifugation and washed with cyclohexane three times, followed by being dried in an oven at 80 °C for 2 h. Then the catalyst was collected and ground thoroughly, after which it was put into the vacuum line overnight for the complete removal of adsorbed species. For the second run, the standard amount of catalyst was taken out from batch B0 and denoted as batch A1, and the rest was denoted as batch B1. Both batches were set up for reaction. The result of batch A1 was regarded as the activity for the second run. The catalyst in batch B1 was regenerated using the same method above. The procedure was repeated up to five runs of the catalyst.

Kg-scale reaction

Desired amount of catalyst (0.5 mol% of Pt) and ca. 250 mL of di-n-octyl phthalate were added into a 450 mL steel Parr autoclave. The reaction proceeded under mechanical stirring. The autoclave was flushed with H2 three times at 10 bar and finally pressurized to 40 bar. Then, it was placed into an aluminum block and kept at 80 °C. The pressure of H2 was continuously monitored during the reaction process. When the H2 pressure decreased below 20 bar, it was recharged to ca. 40 bar. When the H2 pressure remained stable, the reaction was supposed to be completed. The autoclave was vented, and then the solid catalyst and product were separated by filtration. The liquid product was directly collected without purification. The catalyst was washed using cyclohexane, followed by drying in an oven at 80 °C for 2 h. Then the catalyst was collected and ground thoroughly, followed by being put into a vacuum line overnight for the complete removal of adsorbed species. The recycled catalyst was directly used for the next run without any further regeneration procedure. Four cycles were performed in total.

Characterizations

CO chemisorption. CO chemisorption measurements were conducted using a 3-Flex Micromeritics instrument. A 200 mg sample was placed into a U-shaped quartz reactor and heated from room temperature (RT) to 450 °C at a rate of 10 K/min under a flow of 5% O2/He (50 ml/min), with the temperature maintained for 60 min. The sample was then cooled under an Ar flow (50 ml/min). Following this, the sample was reduced in 5% H2/Ar (50 ml/min), heated to 200 °C at 5 K/min for 60 min, and further heated to 300 °C (10 K/min) under Ar. After cooling to RT, the sample loop was filled with 20% CO/He, and CO pulses were introduced into the sample using He as the carrier gas (50 ml/min). The resulting peaks were recorded and integrated using a thermal conductivity detector (TCD). The chemisorption stoichiometry was assumed to be Ptsurf: CO =  1:1, and a spherical metal particle shape was considered for calculations.

XPS

The XPS measurements were performed on an ESCALAB 220iXL (Thermo Fisher Scientific) with monochromated Al Kα radiation (E = 1486.6 eV). Samples are prepared on a stainless-steel holder with conductive double-sided adhesive carbon tape. The measurements are performed with charge compensation using a flood electron system combining low-energy electrons and Ar+ ions (pAr = 1 × 10−7 mbar). The electron binding energies are referenced to the C 1 s core level of carbon at 284.8 eV (C-C and C-H bonds). For quantitative analysis, the peaks were deconvoluted with Gaussian-Lorentzian curves using the software Unifit 2023. The peak areas were normalized by the transmission function of the spectrometer and the element-specific sensitivity factor of Scofield.

STEM

The morphology and particle size/structure of Pt-loaded catalysts were characterized using a Thermo Fisher Spectra 300 scanning transmission electron microscope equipped with a high-angle annular dark field (HAADF) detector and a Cs probe corrector (CEOS), operating at 300 kV. The TEM samples were prepared by dispersing the catalyst powders into a methanol solution, followed by sonication and drop-casting the solution onto a copper grid coated with an amorphous holey carbon film. The particle size distribution was analyzed from HAADF electron micrographs using the ImageJ software package58. To make the analysis representative, we counted ca. 500 particles for each sample.

Nuclear magnetic resonance (NMR)

NMR spectra were recorded on Bruker AV 300 spectrometers. All chemical shifts (δ) are reported in parts per million (ppm) and coupling constants (J) in Hz. All chemical shifts are reported relative to the peaks deuterated solvent (e.g., with respect to CDCl3, δ (ppm) = 7.26 for 1H NMR and δ (ppm) = 77.16 for 13C NMR, respectively).

DFT calculations

All spin-polarized calculations were performed using the Vienna Ab initio Simulation Package (VASP, Version 6.4.2)59, and the generalized gradient approximation with Perdew–Burke–Ernzerhof (GGA-PBE) functional60,61 was employed to describe the exchange and correlation energies of electrons. The interaction between ion and valence electrons was described using projector-augmented wave (PAW)62 potential. The energy cutoff for plane wave was set to 400 eV with the convergence criteria of 1 × 10−4 eV/atom and 0.03 eV/Å for energy and max force in all structure optimizations and frequency calculations. The Brillouin zone was sampled with the Monkhorst–Pack method63.

To understand the activity for catalysts of different sizes at the atomic level, supported Pt13 and Pt55 nanoclusters on the (3 × 3) and (3 × 4) anatase-TiO2(101) surfaces were used to simulate the one-shell and two-shell structures, respectively. Besides, the Pt159 nanoparticle and the plain Pt(111) facet with a (8 × 8) supercell were constructed to simulate the structure of a three-shell and even larger surface size. All TiO2(101) surfaces were constructed with a six-atomic-layer thickness, where the bottom three layers were fixed in their bulk positions, and the others were relaxed. A vacuum region of at least 15 Å was set for all surface models to avoid the perpendicular interactions with adjacent slabs in the Z-direction, and an effective Hubbard parameter (Ueff)64 of 4.0 eV was set for Ti-3d states. The k-points for Pt13/TiO2(101), Pt55/TiO2(101), Pt159, Pt(111) models were setting to 3 × 3 × 1, 2 × 2 × 1, 2 × 2 × 1 and 2 × 2 × 1, respectively.

The adsorption energy (Eads) for the reactant was computed according to the equation, Eads = EX/slab ‒ EX ‒ Eslab, where the EX/slab, EX, and Eslab are the total optimized energy for the adsorbed configurations, isolated species, and bare slab, respectively.