Formic Acid‐Assisted Selective Hydrogenolysis of 5‐Hydroxymethylfurfural to 2,5‐Dimethylfuran over Bifunctional Pd Nanoparticles Supported on N‐Doped Mesoporous Carbon

Abstract Biomass‐derived 5‐hydroxymethylfurfural (HMF) is regarded as one of the most promising platform chemicals to produce 2,5‐dimethylfuran (DMF) as a potential liquid transportation fuel. Pd nanoparticles supported on N‐containing and N‐free mesoporous carbon materials were prepared, characterized, and applied in the hydrogenolysis of HMF to DMF under mild reaction conditions. Quantitative conversion of HMF to DMF was achieved in the presence of formic acid (FA) and H2 over Pd/NMC within 2 h. The reaction mechanism, especially the multiple roles of FA, was explored through a detailed comparative study by varying hydrogen source, additive, and substrate as well as by applying in situ ATR‐IR spectroscopy. The major role of FA is to shift the dominant reaction pathway from the hydrogenation of the aldehyde group to the hydrogenolysis of the hydroxymethyl group via the protonation by FA at the C‐OH group, lowering the activation barrier of the C−O bond cleavage and thus significantly enhancing the reaction rate. XPS results and DFT calculations revealed that Pd2+ species interacting with pyridine‐like N atoms significantly enhance the selective hydrogenolysis of the C−OH bond in the presence of FA due to their high ability for the activation of FA and the stabilization of H−.


List of abbreviations
Hamburg (Germany). 3 For the measurements at the Pd K-edge, a Si(311) C-type double crystal monochromator was used. The beam current was 100 mA with a ring energy of 6.08 GeV. The samples were measured in glass capillaries without dilution. All spectra were recorded as continuous scans in fluorescence mode at ambient temperature and pressure in the range of -150 eV to 1000 eV around the edge within 180 sec. For calibration, a palladium foil was measured as a reference simultaneously with the samples.
The data treatment was performed using the Demeter software package. 4 In order to compensate for the oversampling of the continuous scan mode, the data points of the obtained spectra were reduced with the help of the 'rebin'-function of the Athena software (edge region: -50 to +50 eV; pre-edge grid: 5 eV; XANES grid: 0.5 eV; EXAFS grid: 0.05 Å-1). For data evaluation, a Victoreen-type polynomial was subtracted from the spectrum to remove the background using the Athena software. The first inflection point was taken as edge energy E0. No phase shift corrections have been applied. The EXAFS analysis was performed using the Artemis software.
Prior to the fitting procedure, the amplitude reduction factor S02 and the Debye-Waller factor σ² (Pd-Pd) were determined for a Pd reference foil and used as a fixed parameter for all materials.

Catalytic tests
Hydrogenolysis of HMF with H2. The catalytic performance of the catalysts for the hydrogenolysis of HMF was tested in a stainless-steel autoclave (Parr Autoclave 4560, 160 mL). Typically, 1.5 mmol HMF and 50 mg catalyst were added into the vessel precharged with 30 mL tetrahydrofuran. After purging with H2, the reaction was performed at 160 ℃, 5 bar initial pressure with a stirring speed of 600 rpm for 5 h. Liquid samples of 0.5 mL were taken via a sampling line after 1, 2, 3, 4, and 5 h. The liquid samples were filtered using membrane filters and then analyzed by gas chromatography (GC). GC analysis was performed using an Agilent 7820A GC system equipped with a DB-XLB column (30 nm × 0.18 nm × 0.18 μm) and S5 a FID detector. All analyses were performed three times each. Biphenyl was used as the internal standard and the carbon balance based on furan was in the range of 95 to 103 %. For easy comparison, quantification was achieved by using a normalization method. The errors of the measurements were calculated to be 3-4%.
Hydrogenolysis of HMF with formic acid. A similar procedure was applied to the conversion of HMF to DMF with formic acid as the hydrogen donor. Briefly, 1.5 mmol HMF, 50 mg catalyst, and 45 mmol formic acid (30 equiv.) were dissolved in 30 mL THF. After purging and pressurizing with 5 bar N2, the reaction was conducted at 160 ℃ and 600 rpm. Liquid samples were taken periodically.  Figure S15 confirms the successful synthesis of FMF.

Hydrogenolysis of HMF with formic acid in
Hydrogenolysis of FMF to DMF with formic acid. In a typical experiment, 1.5 mmol FMF, 50 mg Pd/NMC, and 45 mmol formic acid (30 equiv.) were dissolved in 30 mL THF. After purging and pressurizing with 5 bar N2, the reaction was conducted at 160 ℃ and 600 rpm.
Liquid samples were taken periodically.

Reusability study
For catalytic reusability tests, the hydrogenolysis of HMF to DMF was performed with formic S6 acid in the presence of H2 over Pd/NMC. After 3 h reaction, the catalyst was recycled from the reaction mixture by centrifugation, washing with THF and acetonitrile, and drying overnight at 80 ℃. The recovered Pd/NMC catalyst was subsequently reused for the hydrogenolysis of HMF.

In situ ATR-IR spectroscopy
In situ ATR-IR spectroscopy was employed to monitor the reaction progress and to investigate the reaction pathways. The hydrogenolysis of HMF to DMF over Pd/NMC was carried out in a 300 mL stainless-steel autoclave (Berghof BR-300), and the in situ ATR-IR spectra were recorded every 2 min using a Mettler Toledo ReactIR TM 15 spectrometer equipped with a 6.35 mm diameter Dicomp probe. Each spectrum was collected with a resolution of 4 cm -1 and 256 scans in the range of 650 to 4000 cm -1 .
In a typical run, 45 mmol HMF, 500 mg Pd/NMC catalyst, and the required amount of FA were dissolved in 120 mL THF. After purging with H2, the reactor was pressurized with 10 bar H2.
The reaction was conducted at 160 ℃ and was monitored by in situ ATR-IR for 6 h. The reference spectra of the standard compounds (i.e., HMF, BHMF, 5-MF, and DMF) were recorded prior to the reaction separately. Figure S1. N2 physisorption isotherms of Pd/NMC and Pd/CMC.         Figure S10. H2 TPR profiles of as-prepared Pd/NMC (black line) and Pd/CMC (red line).

Characterization
The H2 TPR profiles of the Pd/NMC and Pd/CMC are shown in Figure S10. For Pd/NMC, a reduction peak at 265 K is observed, owing to the reduction of Pd δ+ and Pd 2+ to Pd 0 . By comparison, this reduction peak is not observed for Pd/CMC, probably because the Pd NPs in Pd/CMC are mostly in the reduced metallic state (see Table S1) and the Pd loading is low. The large and broad peaks with onset temperature of 300 K for Pd/NMC and 375 K for Pd/CMC are the reduction of the surface functional groups of NMC and CMC, respectively. There is one negative peak at 315 K, which originates from the decomposition of palladium hydride. This palladium hydride peak is not observed for Pd/NMC, probably due to the overlapping with the broad reduction peak of the surface functional groups.   XAS spectra were recorded to confirm the oxidation state and to investigate the local chemical environment around the Pd centers. The XANES spectra of both samples were significantly different from metallic palladium and showed a significant whiteline indicating that a major amount of palladium atoms was present in an oxidic state ( Figure S11a). Furthermore, the edge position shifted to higher energies by approximately 2.5 eV (inset Figure S11a), which also suggested the presence of oxidic Pd centers. However, a larger shift of the edge (≈ 5 eV) would be expected for a completely oxidized material containing only Pd 2+ . 6 Therefore, both materials contain a mixture of Pd 2+ /Pd δ+ and Pd 0 . In accordance with the XPS results, the slightly higher intensity of the whiteline and the slightly larger shift of the edge indicate that a larger fraction of Pd is oxidized in Pd/NMC compared with Pd/CMC. For Pd/CMC, the Fourier-transformed EXAFS spectra ( Figure S11b) showed the presence of two shells of backscatterers at 1.54 Å and close to 2.6 Å (note that no phase shift corrections have been applied and the real distances are expected to be ≈ 0.4 Å larger). The comparison to the Pd foil suggested that the shell at 2.6 Å corresponds to a shell of Pd backscatterers in the metallic state. In contrast, the shell at 1.54 Å is expected to be formed by light backscatterers like carbon or oxygen, which would be expected for Pd species that are anchored to the support.
For Pd/NMC, also a shell at 1.54 Å from presumably light backscatterers was clearly present.
However, the presence of a shell resulting from metallic Pd backscatterers cannot clearly be confirmed solely by comparison of the spectra. A comparison to a Fourier-transformed EXAFS spectrum of PdO 7 confirmed that no PdO clusters were present in either sample.
In addition to the qualitative evaluation of the Fourier-transformed EXAFS spectra, we performed an EXAFS structure fitting to derive quantitative information about the local chemical environment around the Pd centers. For both materials, the best fits were obtained assuming two different shells of backscatterers (

Computational Methodology
The quantum chemical study was performed using density functional theory (DFT) with the TPSS 13 functional and def2-SVP 14 basis set, which includes Stuttgart-Cologne effective core potentials (def2-ecp) 15 for the Pd atoms. All calculations were carried out using the TURBOMOLE 16