Biomass is an abundant renewable carbon resource and can be transformed into different kinds of valuable chemicals because they contain various oxygen-containing functional groups ^{1, 2}. Selective hydrogenation is a commonly used strategy to achieve the transformation of biomass and its derivatives ^{3, 4}. Since most biomass or its derivatives contain a variety of reducible functional groups, it is a great challenge to control the activity or selectivity in many cases.

5-(Hydroxymethyl)furfural (HMF), which can be prepared from cellulose, is an important biomass platform compound ^{5, 6}. There are different reducible chemical bonds (C＝O, C-OH, C＝C) in HMF. Various products (Scheme 1a) such as 2, 5-dihydroxymethylfuran (DHMF), 2, 5-tetrahydrofurandimethanol (DHTHF), 2, 5-dimethylfuran (DMF), 2, 5-hexanediol (HDO) have been obtained by the HMF selective hydrogenation ^7-14. 5-Methylfurfural (5-MF) is a very important chemical that can be used as a food additive. It is also a commonly used intermediate for synthesizing tobacco flavor, pyrethroid, prallethrin, agricultural chemicals, anti-cancer products, etc^15-18. In addition, it is an important chemical intermediate of synthesis high-quality diesel ^19-21. Preparation of MF from the selective hydrogenation of HMF has always been desirable. However, it is very difficult from both thermodynamic and kinetic aspects constraints. In our previous work, the direct hydrogenation of HMF to MF without additives has been realized by Pt₁/Nb₂O₅ catalyst ²².

In this work, the selectivity hydrogenation of HMF to MF has been controlled with a novel strategy, which elegantly combined the steric hindrance, the electron effect, and the hydrogen spillover. Based on this strategy, Pt@PVP/Nb₂O₅ was designed for selective hydrogenation of HMF to MF (Scheme 1b, this work). It was found that the catalyst could catalyze the reaction efficiently and the selectivity of MF could reach 92% when the HMF was converted completely.

Niobium oxalate hydrate (99%), polyvinylpyrrolidone (99%), hexadecylazaniumbromide (99%), H₂PtCl₆·6H₂O (99%) were supplied by Sinopharm Chemical Reagent Co. Ltd. Tetrahydrofuran (99%), 5-hydroxymethylfurfural (98%), 2-furylmethanol (98%), furfuryl alcohol (98%) and n-decane (98%) were obtained from J & K Ltd. 2, 5-Dimethylfuran (99%), 5-methylfurfural (99%) were provided by TCI (Shanghai) Development Co., Ltd. Pt black (1%, mass fraction) and TiO₂ (99%) were purchased from Sigma-Aldrich. Deionized water was used during the experimental process. H₂ (> 99.99%) was purchased from Beijing Analytical Instrument Company. All chemicals involved in the synthesis of catalysts were used directly.

Niobium pentaoxide nanoparticles were prepared according to the documented method ²². In the experiment, niobium oxalate hydrate (0.538 g) was dissolved in deionized water. 10 mL of CTAB (0.2 mol∙L⁻¹) surfactant aqueous solution was added. Then the mixture was loaded into a dried Teflon autoclave at room temperature, followed by hydrothermal treatment of the above solution at 180 ℃ for 10 h. And then the white precipitate was washed with distilled water and ethanol three times. The grinded powder was calcined in a muffle oven at 500 ℃ for 4 h.

Pt@PVP NPs were synthesized according to the method reported in other literature. 5.0 mL H₂PtCl₆ aqueous solution was added into 95 mL PVP solution. The solution was refluxed in a flask for 3 h, and then the PVP-stabilized Pt NP solution was obtained. The mole ratio of PVP and Pt was adjusted by changing the amount of H₂PtCl₆ and PVP added.

12 mL as-synthesized Pt nanoparticles solution was added into Nb₂O₅ aqueous solution and stirred for 2 h. Subsequently, the solvent was removed at 80 ℃. The obtained powder was washed with deionized water and dried at 80 ℃. Then, the grinded dark grey powder was calcined in a tube furnace (Anhui Kemi Machinery Technology Co., Ltd.; Model TFV-1200-50-Ⅰ-220) to obtain Pt@PVP/Nb₂O₅.

In a typical procedure, an appropriate amount of H₂PtCl₆ aqueous solution was slowly added into the 100 mg Nb₂O₅ solution with magnetic stirring for 5 h. Then, under the ice bath temperature, the NaBH₄ (freshly prepared) was added dropwise into the solution. After 2 h, the obtained grey granules were washed and dried at 80 ℃ for 12 h.

The powder X-ray diffraction (XRD) patterns were obtained on Rigaku D/max 2500 equipped with nickel-filtered Cu-K_α (D/MAX-RC, Japan). The microstructure and size of Pt/Nb₂O₅, Pt@PVP NPs, and Pt@PVP/Nb₂O₅ were obtained by using transmission electron microscopy (TEM) on JEM-2100F at 120 kV (JEOL, Japan). The Pt contents of different catalysts were analyzed by ICP-AES (Thermo, America). XAFS experiments were operated at the 1W2B beamline at Beijing Synchrotron Radiation Facility (BSRF). The X-ray photoelectron spectroscopy (XPS) was collected on the Thermo Scientific ESCA Lab 250Xi electron spectrometer (Thermo, America).

In a typical experiment, appropriate amount of catalyst was dispersed in 2 mL THF and then 0.3 mmol HMF was added to the mixture in a Teflon-lined stainless-steel reactor of 20 mL. The reactor was purged with H₂ to remove the excess air. The reaction was carried out at desired temperature. After completion of the hydrogenation reaction, n-decane was added to the reactor. The solution was centrifuged and then analyzed by gas chromatography-mass spectrometry (GC-MS, Agilent 7890B-5977A) with a flame ionization detector (FID) (Agilent, America).

The catalyst Pt@PVP/Nb₂O₅ (10/1) was synthesized by the reduction-deposition method and the overall synthesis strategy was illustrated in Fig. 1a, in which 10/1 is the mole ratio of Pt and PVP. Detailed preparation processes for catalysts were described in the Experimental Section. Briefly, Nb₂O₅ was fabricated from niobic acid roasting at 500 ℃ in air ²³. Known amounts of PVP and H₂PtCl₆ were added into ethanol under stirring and Pt@PVP was obtained. Then, the Pt@PVP nanoparticle suspension was added to Nb₂O₅ aqueous solution under stirring (Fig. 1a).

The XRD patterns of the Nb₂O₅ and Pt@PVP/Nb₂O₅ are given in Fig. 1b. The diffraction peaks at 2θ = 22.6°, 28.5°, and 36.6° are assigned to the pseudohexagonal Nb₂O₅ crystal. The TEM images of Nb₂O₅ and Pt@PVP/Nb₂O₅ are shown in Fig. 1c and Fig. 1d. Uniform Pt nanoparticles (NPs) with average diameter of 2.8 nm were immobilized on Nb₂O₅. The high-resolution TEM (HRTEM) image (Fig. 1e) showed that the interplanar distance of Pt NPs on Nb₂O₅ is 0.23 nm, which corresponds to the standard lattice fringe of Pt (200) surface. High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) and energy-dispersive X-ray (EDS) elemental mapping (Fig. S1) revealed that all elements were uniformly distributed throughout the Pt@PVP/Nb₂O₅. The content of Pt in Pt@PVP/Nb₂O₅ was 0.75% (mass fraction) (Table S1).

The catalytic performance of Pt@PVP/Nb₂O₅ was studied. Interestingly, the selectivity of MF could reach 92% with > 99% conversion of HMF in 24 h (Table 1, entry 1). To get some information about the roles of the components in the catalyst, the catalytic performances of Nb₂O₅, Pt/Nb₂O₅, and Pt@PVP were also checked (Table 1). Although the conversion of HMF over Nb₂O₅ was only 7%, the selectivity of MF could reach 63% (Table 1, entry 2). The by-product was from the C-C coupling reaction during the process of dehydroxylation because of the strong acidity of Nb₂O₅ and no DHMF was detected which was from the hydrogenation of C＝O. It means that Nb₂O₅ was active for the dehydroxylation and inactive for the hydrogenation of C＝O in HMF. The low activity of the Nb₂O₅ may be attributed to its low ability to split H₂. The cat alytic activity of Pt/Nb₂O₅ was much higher than pure Nb₂O₅ because Pt nanoparticles can efficiently activate H₂. However, DHMF, which was generated from hydrogenation of the C＝O group of HMF, was the main product over Pt/Nb₂O₅ (Table 1, entry 3). The conversion of HMF and the selectivity of MF was 16% and 37% over Pt@PVP in 12 h, respectively (Table 1, entry 4). PVP also improved the selectivity of MF over Pt/C. No MF was obtained over commercial Pt/C catalyst (Table 1, entry 5) and the selectivity increased to 32% over Pt@PVP/C catalyst (Table 1, entry 6). The Pt@PVP/TiO₂ had low catalytic activity for the target product (Table 1, entry 7). The results indicate that Pt, PVP, and Nb₂O₅ were indispensable for the highly selective hydrogenation of HMF to MF.

The effect of the mole ratio of Pt and PVP was checked for the Pt@PVP/Nb₂O₅ catalyst (Fig. S2). The TEM images of the modified Pt nanoparticles with different Pt/PVP ratios are shown in Fig. S3. The activity of the catalyst gradually decreased with the increase of the amount of PVP due to the covering of the metal surface, while the selectivity of MF increased. When the ratio of Pt/PVP was greater than 10/1, the selectivity of MF could reach above 90%. The selectivity of MF was constant when further increase the amount of PVP. When the mole ratio of PVP/Pt was less than 10/1, HMF could still be adsorbed by Pt nanoparticles, and the selectivity of MF was low. When the Pt nanoparticles were coated with plenty of PVP, a large amount of PVP inhibits the adsorption of HMF by Pt nanoparticles, and the selectivity of MF could remain at 90%. These results indicate that PVP could improve the selectivity of MF significantly and the effect was kept unchanged after the amount of PVP adsorbed on the Pt particles was large enough. The effect of reaction temperature on the Pt@PVP/Nb₂O₅ catalytic performances was researched (Fig. S4). When we increase the temperature from 100 ℃ to 160 ℃, the conversion of HMF increases. Although the selectivity of MF was the highest at 120 ℃ (> 99%), the conversion was very low. The most suitable reaction temperature was 140 ℃. To verify the durability of the Pt@PVP/Nb₂O₅ catalyst, the reusability of the catalyst was also tested for the hydrogenation of HMF to MF. After the reaction, the catalyst was washed with THF. Next, the catalyst was dried in vacuum oven at 80 ℃ overnight before being reused in the next run. The results in Fig. S5 clearly showed that the Pt@PVP/Nb₂O₅ catalysts could be reused at least 4 times without obvious decrease in activity and product selectivity.

To further study the reaction pathway, some control experiments were conducted. The effect of the reaction time on the catalytic performance over Pt@PVP/Nb₂O₅ was studied and the result was demonstrated in Fig. S6. Fortunately, the selectivity of MF could reach 92% and the selectivity was independent of conversion. Even if the reaction time was prolonged to 30 h, MF was not further hydrogenated to DMF, indicating that the by-product was not from the hydrogenation of C＝O. In addition, MF was also used as the substrate to check the catalytic performance of Pt/Nb₂O₅ and Pt@PVP/Nb₂O₅. The Pt/Nb₂O₅ could efficiently catalyze the hydrogenation of MF to DMF with almost complete conversion. However, no MF was converted over the Pt@PVP/Nb₂O₅ (Table S2). The results indicate that Pt@PVP/Nb₂O₅ is inert for the hydrogenation of C＝O, which explains the high selectivity of MF. The variation trend of the catalytic performance with the reaction time over the Pt/Nb₂O₅ was also checked and the result was shown in Fig. S7. The conversion of HMF was 80% with 12% selectivity of MF in 3 h. However, MF was converted to DMF with the prolongation of the reaction time, demonstrating that PVP was crucial for the high selectivity.

It is well known that the electronic state often affects the performance of catalysts significantly. The electron interaction among the three interfaces in Pt@PVP/Nb₂O₅ was studied by X-ray photoelectron spectroscopy (XPS), and the results are presented in Fig. 2. The binding energy of Nb in Pt/Nb₂O₅, Pt@PVP/Nb₂O₅ shifts toward lower binding energy compared with that of Nb in pure Nb₂O₅, indicating that the electrons transferred from Pt to Nb₂O₅ and then Nb in Pt@PVP/Nb₂O₅ get more electrons. The binding energy of Pt 4f in Pt@PVP/Nb₂O₅ displayed a slight shift toward higher binding energy compared with Pt@PVP and Pt/Nb₂O₅, demonstrating the lowest electron density of Pt in Pt@PVP/Nb₂O₅ compared with Pt@PVP and Pt/Nb₂O₅ (Fig. 2b). All the results above show that Pt was more electron-deficient and Nb₂O₅ more electron-rich in Pt@PVP/Nb₂O₅.

Considering the difference of binding energy of Pt in Pt/Nb₂O₅ and Pt@PVP/Nb₂O₅ was very small, Pt L3-edge X-ray absorption near edge structures (XANES) analysis was performed to further characterize the electronic state of Pt, and the results are given in Fig. 3. XANES analysis showed an increase in the Pt d-band vacancy for Pt@PVP/Nb₂O₅ compared with Pt/Nb₂O₅ (Fig. 3a). Namely, the electron density of Pt was lower in Pt@PVP/Nb₂O₅ than that in Pt/Nb₂O₅. This is consistent with the results of XPS characterization that there exist strong electronic interactions between Pt and PVP or Nb₂O₅. Fourier transformation of extended X-ray absorption fine structure (FT-EXAFS) further showed weakening of Pt-Pt coordination shells while emerging Pt-N shells ²⁴ (Fig. 3b and Fig. S8). It means that exists interaction between PVP and Pt. Wavelet transform (WT) oscillations further display visually the Pt-N throughout the whole catalyst. In Fig. 3c, WT plots of Pt/Nb₂O₅ showed intensity maximum near 2.5 Å (1 Å = 0.1 nm), which is associated with the Pt-Pt. However, as shown in Fig. 3d, another intensity maximum can be observed at 1.9 Å, which can be attributed to the existence of Pt-N coordination. These results further confirmed the formation of Pt-N in Pt@PVP/Nb₂O₅.

The electronic states of Pt were further characterized by in situ CO-diffuse reflectance infrared Fourier transform (CO-DRIFT) spectra measurements. It can be known from in situ CO-DRIFT spectra in Fig. S9 that the band for CO adsorbed on metallic Pt (CO-Pt0) in the prepared Pt@PVP/Nb₂O₅ was observed at 2083 cm⁻¹. By comparison, the Pt/Nb₂O₅ catalyst showed substantial CO chemisorption at 2072 cm⁻¹, and the band for CO adsorbed on metallic Pt (CO-Pt0) in the synthesized Pt@PVP was observed at 2066 cm⁻¹. These results showed that electron transfer from Pt nanoparticles to Nb₂O₅ support. The characterization results are consistent with the above XPS.

The adsorption mode of the substrate on the solid surface of the catalyst is a key factor to affect the selectivity of the reaction. The chemisorption of C-OH, and C＝O can be studied by FT-IR spectroscopy using n-propanal and methanol as model molecules ^{25, 26}. The bands of the methoxy species observed at 1211 cm⁻¹ and 1153 cm⁻¹ were assigned to the ν(C-O) bands of the on-top and bridged sites, respectively (Fig. S10). This is the direct evidence of chemical adsorption of C-OH by the surface of Nb₂O₅. However, Nb₂O₅ could not adsorb C＝O groups chemically. It means that Nb₂O₅ could activate C-OH selectively. It has been reported that Nb₂O₅ can promote the deprotonation of the phenol molecules and activate C_aromatic-OH ^{27, 28}. So, the phenomena are consistent with those reported in the literature.

It has been reported the activated hydrogen species on metal surfaces can migrate to the supports to take part in the hydrogenation reaction, which seemed to be hydrogen spillover ^29-31. The hydrogen spillover can be studied by mixing catalyst with a support ³². The physically mixed commercial Pt/C catalyst and Nb₂O₅ were used to study the hydrogen spillover and the results are given in Table 2. Although the HMF conversion was very low, the selectivity of MF can reach 61% in a short time (Table 2, entry 1). The conversion of HMF was 65% while no MF was produced over commercial Pt/C catalyst (Table 2, entry 2). However, MF was produced over the physically mixed commercial Pt/C catalyst and Nb₂O₅. With the decrease of the amount of Pt/C catalyst, the selectivity of MF increased gradually (Table 2, entries 3−5). When the dosage of Nb₂O₅ was 40 mg and the Pt/C was 3 mg, the selectivity of MF could reach 43% with 8% conversion of HMF (Table 2, entry 5). When only 3 mg Pt/C was used as the catalyst, no MF was detected (Table 2, entry 6). The result shows that hydrogen spillover exists in the reaction system. The hydrogenation of C＝O occurred on the surface of Pt NPs and the formation of MF can be attributed to the surface of Nb₂O₅. In addition, the physical mixture of WO₃ and Pt@PVP/Nb₂O₅ catalyst showed the transformation from WO₃ with bright yellow color to WO_3−x with light blue color at room temperature, confirming the occurrence of hydrogen spillover (Fig. S11). The hydrogen spillover can also be proved by pulse adsorption of hydrogen and the results are shown in Table S3. The metal dispersion of Pt@PVP was only 2.33% while it increased to 46.63% for Pt@PVP/Nb₂O₅. Due to the presence of hydrogen spillover, the increase in the number of active sites for H₂ activation. The metal dispersion of Pt/Nb₂O₅ was 71.85%, suggesting that some active sites of H₂ were blocked by PVP in Pt@PVP/Nb₂O₅. The higher activity of Pt@PVP/Nb₂O₅ over Pt@PVP (Table 1, entries 1 and 4) suggests that the reaction proceeded mainly via the hydrogen spillover pathway, namely the activated hydrogen on Pt NPs migrated to the surface of Nb₂O₅ and reacted with C-OH activated by Nb₂O₅.

The hydrogen spillover and the chemical adsorption of C-OH on the surface of Nb₂O₅ have been confirmed in our experiments. In this work, we also studied the hydrogenation process of HMF to MF on the surface of Nb₂O₅ in the presence of activated hydrogen with density functional theory (DFT). The calculation was conducted through the Vienna Ab-initio Simulation Package (VASP) in this work. The results (Fig. 4) show that the reaction contains three primary steps. Firstly, the C-OH bond of HMF is ruptured because of the strong interaction between Nb and OH to form methylene groups and the free energy barrier is 0.96 eV. Secondly, the Nb-OH reacts with activated hydrogen atoms of H₂ and forms Nb-OH₂. Finally, one of the hydrogen atoms in Nb-OH₂ transfers to the methylene groups and generates MF, and meanwhile another active hydrogen atom of H₂ combines with the Nb-OH to form H₂O. The free energy barrier of the whole process for the selectivity hydrogenation of HMF to MF on the surface of Nb₂O₅ was 0.96 eV, which can be overcome at 140 ℃. The DFT calculation also shows that the hydrogenation of HMF to MF can proceed on Nb₂O₅ surface in the presence of activated hydrogen migrated from the surface of Pt.

According to the experimental and DFT calculation results, we can discuss the interesting phenomenon that HMF could be hydrogenated to MF selectively over Pt@PVP/Nb₂O₅, which is shown in Scheme 2. The activation of HMF on Pt was inhibited by the PVP layer, while Pt could activate hydrogen because it can penetrate the PVP layer due to its small size. The C-OH groups activated by Nb₂O₅ were hydrogenated by the active hydrogen transferred to Nb₂O₅ from Pt via hydrogen spillover. The selectivity was very high because the acidic surface of Nb₂O₅ made it easier to adsorb C-OH, rather than C＝O. In addition, the electron effect was also beneficial to the formation of MF. As discussed above, the electron interaction at the three-phase interface in the Pt@PVP/Nb₂O₅ made the Pt particles more electron-deficient and Nb₂O₅ more electron-rich. The electron-rich Nb₂O₅ decreased the C-C coupling reaction, and electron-deficient Pt was favorable to hydroxyl adsorption and unfavorable to C＝O adsorption, which further enhanced the selectivity of MF ³³.

In summary, Pt@PVP/Nb₂O₅ could catalyze the selective hydrogenation of HMF to MF with a selectivity of up to 92%. The steric effect, hydrogen spillover, and electron transfer at the interface of the three-phase catalyst system contribute simultaneously to the reaction. Although PVP blocked the access of HMF to the surface of Pt, H₂ can be activated on the surface of Pt, which will migrate further to the Nb₂O₅ surface through the hydrogen spillover pathway. When the HMF is adsorbed on Nb₂O₅, the C-OH will be hydrogenated with high selectivity. This work represents a new paradigm to control the hydrogenation selectivity, and we believe that the route to produce MF from renewable HMF has great potential for the green production of such important chemical intermediate.

Supporting Information: available free of charge via the internet at http://www.whxb.pku.edu.cn.