Capping Agent Effect on Pd-Supported Nanoparticles in the Hydrogenation of Furfural

: The catalytic performance of a series of 1 wt % Pd/C catalysts prepared by the sol-immobilization method has been studied in the liquid-phase hydrogenation of furfural. The temperature range studied was 25 – 75 °C, keeping the H 2 pressure constant at 5 bar. The effect of the catalyst preparation using different capping agents containing oxygen or nitrogen groups was assessed. polyvinylpyrrolidone (PVP), and poly (diallyldimethylammonium chloride) (PDDA) were chosen. The catalysts were characterized by ultraviolet-visible spectroscopy (UV-Vis), Fourier transform infrared spectroscopy (FTIR), transmission electron microscopy (TEM), and X-ray photoelectron spectroscopy (XPS). The characterization data suggest that the different capping agents affected the initial activity of the catalysts by adjusting the available Pd surface sites, without producing a significant change in the Pd particle size. The different activity of the three catalysts followed the trend: Pd PVA /C > Pd PDDA /C > Pd PVP /C. In terms of selectivity to furfuryl alcohol, the opposite trend has been observed: Pd PVP /C > Pd PDDA /C > Pd PVA /C. The different reactivity has been ascribed to the different shielding effect of the three ligands used; they influence the adsorption of the reactant on Pd active sites.


Introduction
Diminishing fossil fuel resources, a growing energy demand, and the increased environmental concerns caused by CO2 emissions have led to the search for new sustainable energy resources [1,2]. In this regard, waste lignocellulosic biomass, which mainly contains cellulose, hemicellulose, and lignin, is considered a promising sustainable resource. It is an abundant alternative carbon resource that can be used to produce chemicals and biofuels [3,4].
Among the value-added molecules, furfural is an important precursor in the generation of biofuels and chemical intermediates. It can be readily obtained from hemicellulose by acid-catalyzed cascade hydrolysis and dehydration [4,5]. Owing to its high unsaturation, the selective hydrogenation of furfural has attracted significant attention to produce a range of valuable C4 and C5 molecules including furfuryl alcohol (FA), tetrahydrofurfuryl alcohol (THFA), 2-methylfuran (2-MF), 2-methyltetrahydrofuran (2-MTHF), and others ( Figure 1). FA is widely used in the chemical industry, primarily for the production of foundry resins, polymers, synthetic fibers, and a chemical intermediate for the production of perfume and vitamin [6,7]. THFA is considered a green solvent and is usually used in printer ink, agricultural applications, and electronics cleaners [7,8]. 2-MF has applications in biorefineries as a consequence of its high octane number, and in chemical industries, e.g., the production of toluene [9]. The main issue with furfural hydrogenation is controlling its reaction route and hydrogenation degree through a selective catalyst [10]. Depending on the process conditions [11], the nature of the catalyst [12], and the type of the solvent [13], furfural hydrogenation proceeds through various pathways and with the formation of a significant number of products. A variety of heterogeneous catalysts mainly based on Pd [14,15], Pt [16], Ni [17], Cu [18], and Ru [19]. supported on carbon [20], titanium oxide [21], alumina [22], silica. or zeolite [22], have been reported for the hydrogenation of furfural.
In terms of supported metal nanoparticles, the capping agent can affect the particle size, size distribution, morphology, and stability [23,24]. Recently, a number of researchers revealed an additional benefit of capping agents, which act as promoters and/or selectivity modifiers by blocking specific surface sites in various liquid-phase reactions [25]. Medlin et al. used self-assembled monolayers (SAMs) to create a more favorable surface environment for specific product formation [26]. They used alkanethiolate SAM-modified Pd catalysts for the selective hydrogenation of furfural, increasing the selectivity to value-added compounds, i.e., FA and 2-MF, by selectively blocking facets, leaving only particle edges/corners exposed [26]. It was also observed that supported Pd nanoparticles (NPs) protected by PVA led to higher selectivity to benzaldehyde in the liquid phase oxidation of benzyl alcohol, despite the conversion being lower. This was mainly because of the preferential blocking of Pd (111) facets, which have been recognized to facilitate the decarbonylation process [27]. Indeed, Rogers et al. proposed that the interaction between the solvent and PVA alters the PVA binding on the metal surface. Consequently, this interaction contributes to the different Pd corner and edge sites available [21].
To further understand the role of capping agents in controlling the selectivity of furfural hydrogenation, we prepared carbon-supported Pd nanoparticles through controlled solimmobilization routes [28], using different capping agents including polyvinyl alcohol (PVA), polyvinylpyrrolidone (PVP), and poly(diallyldimethylammonium) chloride (PDDA). The aim of this work was to perform a more systematic study of the role of stabilizers under mild reaction conditions (25-75 °C, and 5 bar) and to enhance the understanding of this relatively under-reported effect.

Results
Pd colloids have been prepared in the presence of PVA, PVP, and PDDA as the stabilizer and NaBH4 as the reducing agent. PVA contains -OH groups, whereas PVP has a pyrrole-type N species in close proximity to a carbonyl group and PDDA contains dimethyl-ammonium groups ( Figure 2). The metal loading was confirmed by atomic absorption spectroscopy (AAS).

UV-Visible Characterization
UV-Vis spectroscopy was used to demonstrate the complete reduction of the Pd 2+ precursor to Pd 0 after the addition of NaBH4, before immobilization on the support ( Figure S1).
The peak at 236 nm, corresponds to the Pd-Cl metal charge transfer in [PdCl4] 2-, whereas the peak at 420 nm is related to d-d transitions [29,30]. The capping agent-H2O sols are transparent in the visible range; however, PVA has a band at 280 nm, which can be attributed to the π → π * transition of the carbonyl groups (C=O) associated with ethylene unsaturation [31]. The spectra of the Pd precursor+capping agent and Pd sols demonstrate the interaction of the metal with PVA and PVP. A shift at 286 nm (blue curve, Figure S1c) after adding PDDA to the Pd precursor, suggests a possible ligand-to-metal charge transfer interaction between PDDA and [PdCl4] 2-, as already evidenced in the case of Au [31]. The disappearance of the peaks related to Pd 2+ and the observed scattering suggest the complete reduction to Pd 0 and the formation of metal nanoparticles [32,33].

Infrared (IR) Spectroscopy
The interaction of the capping agents with Pd precursor was also investigated using Fourier transform infrared (FTIR) spectroscopy. The FTIR spectra of PVA and PVA-Pd precursor (Na2PdCl4) are shown in Figure 3a. The primary bands for the PVA films are: (i) 3431 cm −1 , O-H stretching vibrations (the broad nature of the -OH stretching vibration is characteristic of residual water, i.e., KBr is hygroscopic) and (ii) 2918 cm −1 , aliphatic C-H stretching vibrations (the peak at 2356 cm -1 is due to gas phase CO2). The peak at 1726 cm −1 can be assigned to C=O stretching deriving from the residual C=O due to the incomplete hydrolysis of the acetate group [34]. PVA is produced by the polymerization of vinyl acetate to poly(vinyl acetate), and its subsequent hydrolysis to PVA. The peaks between 1380 and 1247 cm -1 can be assigned to O-H deformation, whereas the peaks at 1200-1000 cm -1 can be attributed to the stretching vibrations of C-O-C linkage, which suggest the presence of cross-linked PVA molecules [35]. The slight modification observed for the peaks at 1380-1247 cm -1 in the spectrum of PVA+Na2PdCl4 suggests a weak interaction between the metal precursor and the -OH groups present in PVA. The prominent band related to PVP (Figure 3b) at 3335 cm −1 is due to the O-H stretching vibrations of adsorbed water at the surface of particles. The peak at 2951 cm −1 could be related to asymmetric CH2 stretching of the ring, while the one at 1671 cm −1 is assigned to the stretching vibration of C=O. The absorption peak at 1425 cm −1 is due to the С-H2 scissor. The peak at 1281 cm −1 can be assigned to the C-N stretching vibrations. Previous studies revealed that PVP can bind to Pd surfaces from the carbonyl group or nitrogen atom of the pyrrolidone units [36,37]. The decrease in intensity of the peak at 1671 cm -1 and at 1265 cm -1 , observed after the addition of Pd, confirms that both O and N groups present in PVP interact with the metal precursor [38,39].
The characteristic FTIR spectrum of PDDA is illustrated in Figure 3c. The peak at 1635 cm -1 is attributed to the bending vibration of the C-N group, and the one at 1466 cm -1 is assigned to C=C stretching, in a positively charged environment contributed by the positively charged nitrogen of the polycations [40]. After the addition of Pd, the observed decrease in the intensity of the peaks located at 1635 and 1466 cm -1 indicates that the activities of these vibrational modes are modified in the mixture, probably due to the PDDA-Pd interaction. With further inspection of Figure 3c, one can observe a loss in the peak at 2068 cm -1 after the addition of the Pd precursor. This result has been already observed in the literature, but the origin of this peak is not clear [40].

Transmission Electron Microscopy (TEM)
After the immobilization of the Pd nanoparticles on the support, the morphological specification of the prepared catalysts was investigated by HRTEM ( Figure 4). It is noticeable that the capping agent has a great influence on the Pd NPs' size and distribution when using activated carbon (Camel (X40S)) as a support. All the catalysts showed an average particle size of 3-4 nm, with the presence of isolated larger particles, in particular for PdPDDA/C (see Figure 4).

X-Ray Photoelectron Spectroscopy (XPS)
X-ray photoelectron spectroscopy (XPS) of the catalysts was performed to investigate the surface chemistry of the carbon support, the chemical state, and the exposure of supported palladium species. XPS survey data indicated the presence of only Pd, C, N, and O species. No evidence of Na or B residues from NaBH4 was detected (Table 1). Depending on the capping agent used, significant differences were observed in the relative amount of Pd at the surface: PdPVA/C (1.30%) > PdPDDA/C (0.76%) > PdPVP/C (0.50) ( Table 1). The data revealed a different oxygen content in the samples. PdPVA/C shows the highest relative amount of O 1s (14.7%) compared to PdPDDA/C (9.70%) and PdPVP/C (9.10%), which show an oxygen content similar to bare carbon (9.17%) ( Table 1). The highest oxygen content on the surface of PdPVA/C catalyst is probably due to the presence of oxygen groups contained in PVA, confirming the presence of the capping agent on the surface of the catalyst.
PdPVP/C evidenced the highest amount of N (2.95%), superior to PdPDDA/C (1.86%), whereas for PdPVA/C nitrogen was not detected on the surface, as expected. The XPS spectra show two components in the Pd 3d3/2 region. The full width at half maximum (FWHM) was set between 1 and 3 eV. The measured spectra were deconvoluted with a few traces of the Gauss (80%)-Lorentz (20%) mixed function.
The XPS spectra show two components in the Pd 3d3/2 region. The first peak at a binding energy (BE) of ~335 eV is ascribed to Pd 0 , while a second peak centered at ~338 eV was ascribed to Pd δ+ species (Table 2 and Figure 4). Pd is mainly present in its oxidized form. However, the Pd 0 /Pd δ+ ratio varies when using different capping agents. A higher content of Pd 0 species was observed in the case of PdPVA/C and PdPVP/C catalysts (30.9% and 23.9%, respectively), compared to PdPDDA/C, which mainly contain oxidized Pd on the surface of the nanoparticles (Table 2, Figure 5c), in agreement with the increasing steric hindrance of the capping agent (PVA < PVP < PDDA).  In the C 1s region we can identify four peaks. The signal at 284.0-284.6 eV can be assigned to sp 2 -hybridised carbon, the one at 285.3-285.8 eV to sp 3 -hybridized carbon, the one at 286.6-287.9 eV to the C=O group, and the one at 288.8-292.2 eV to C-OH/C-O-C, as indicated in Table 3 and Figure  6. Four main oxygen groups were identified. Binding Energy (BE) of 529.5-531.4 eV corresponds to a carbon-oxygen double bond and the one at 532.1-532.8 eV to an ether-like single bond, whereas BE at 533.4-534.6 eV can be attributed to carbon-oxygen single bonds in hydroxyl groups [41] (Figure  7). Oxygen species around 536.0 eV refer to the presence of carboxylic groups [42]. Bare X40S carbon contains 30% C=O groups, 41.9% C-O-C, 22.1% C-OH groups, and 6.0% carboxylic groups (Table 3). After the addition of PdPVP nanoparticles, the oxygen species' relative ratio remains similar, whereas after the addition of PdPVA and PdPDDA the percentage of C-O-C increases with a decrease in C-OH species. For PdPVA, we can attribute the unexpected decrease in superficial -OH groups to the formation of C-O-C species between hydroxyl groups of the carbon and of PVA and the crosslinking of PVA chains, supporting the findings of FTIR spectroscopy

Catalytic Results
The performance of the prepared catalysts was evaluated for the liquid-phase hydrogenation of furfural (furfural 0.3 M; furfural/metal ratio 500 mol/mol, 5 bar of H2, in the temperature range 25-75 °C), with 2-propanol as the solvent. As illustrated in Figure 1, the reaction route map is a cascade of reactions that forms various products such as FA, THFA, 2-MF, 2-MTHF, and ethers.
Hydrogenation of the carbonyl group (C=O) in furfural yields FA. The formation of THFA, on the other hand, requires the unselective hydrogenation of the furan double bonds along with the hydrogenation of the carbonyl group. Moreover, hydrogenolysis via ring activation is a primary pathway in the C−OH bond cleavage for the conversion of FA to 2-MF. By using alcohols as the solvent, several other byproducts can be formed, such as acetals [43] and ethers [44][45][46].
When the reaction was performed at 25 °C and 50 °C, PdPVA/C showed the highest initial activity (calculated after 5 min as 824 h -1 and 1775 h -1 at 25 °C and 50 °C, respectively), higher than PdPDDA (624 h -1 and 1422 h -1 at 25 °C and 50 °C, respectively) and PdPVP/C (304 h -1 and 573 h -1 at 25 °C and 50 °C, respectively) ( Table 4). At 75 °C all the catalysts were very active; it is not possible to precisely calculate the initial activity (conversion >55% after 5 min in all cases). We can exclude any effect of the size of Pd nanoparticles on the activity since all the catalysts had a similar particle size (3-4 nm, Table 4). However, it is possible to correlate the activity to the relative amount of Pd at the surface (Tables 2 and 4): PdPVA/C (1.3%) > PdPDDA/C (0.76%) > PdPVP/C (0.50%). Therefore, the initial activity can be affected by using different capping agents as they can change the available Pd surface sites regardless of any particle size effect. Indeed, as already reported in the literature, the N groups present in PVP blocks the Pd active sites more than in PDDA, whereas the oxygen groups in PVA interact weakly with the Pd surface [38]. By plotting the conversion versus time of reaction ( Figure  8), it is possible to observe an induction time in the first 15 min of the reaction, for all the tested catalysts, in particular for the reaction performed at 25 °C. It is worth noting that at 25 °C the activity of PdPDDA/C increased after 1 h, becoming the most active catalyst, whereas PdPVA/C and PdPVP/C showed a similar reaction profile. This behavior can be attributed to the easier removal of PDDA from Pd active sites during the reaction, and the reduction of the superficial oxidized Pd (Table 2). At 75 °C the catalysts show similar reactivity. We ascribed both effects to the partial removal of a capping agent during the reaction, thereby exposing a higher amount of active sites to the reactant [47].  Comparing the selectivity at low temperature (25 °C and 50 °C), PdPVP/C exhibits better selectivity towards furfuryl alcohol (64% and 56.7% at 25 °C and 50 °C, respectively) compared to the other catalysts (45.2% and 49.6% at 25 °C and 50 °C, respectively, for PdPVA/C and 52.8% and 49.0% at 25 °C and 50 °C, respectively, for PdPDDA/C) ( Table 4). We already reported that the presence of a higher amount of protective agent can enhance the selectivity to furfuryl alcohol, directing the geometry of the adsorption of the furfural on the surface of the catalyst [21,48]. The evolution of the selectivity with the time on stream for the reaction performed at 25 °C is reported in Figure S2. For all catalysts, with a low reaction time, the main product is the acetal. When increasing the conversion rate, the production of acetal decreases, with the formation of furfuryl alcohol as the main product and furfuryl ether. According to the literature, Pd/C can catalyze the formation of ether even in the absence of an acid support [44]. Indeed, the Pd hydride can catalyze the formation of acetal, followed by hydrogenolysis to form the corresponding ether. Figure S2 shows the initial formation of acetals, which are converted to ether at a higher conversion rate, confirming the proposed mechanism. At 75 °C the selectivity is similar due to the removal of the protective agent.
The stability of the most active catalyst (PdPVA/C) has been studied. Five successive reactions were conducted by simply filtering the catalyst and reusing it either without any treatment or by washing with acetone. As shown in Figure 9a, the conversion starts to decrease immediately after the first run. Washing with acetone was performed to prove if the deactivation is caused by the irreversible adsorption of the products. Pretreating the catalyst after each run, the catalyst maintained almost the same activity and selectivity (Figure 9b), confirming our assumption. We performed XPS of the used catalysts (Table 1, Figures S4-S6). The catalyst used without any treatment showed a high O/C ratio (1.10) (10 times higher than in the fresh one (0.10)), and a low surface Pd content (0.32 and 1.30 at % for the used and fresh PdPVA/C, respectively; see Table 1), attributed to the adsorption of furfural and the hydrogenation products (Table 1). After washing, the O/C ratio decreased to 0.22, and Pd exposure increased from 0.32 to 0.93 at %, confirming the partial removal of the adsorbed species. The Pd3d spectrum of the used PdPVA/C catalyst shows that a reduction of Pd δ+ to Pd 0 occurred during the reaction (Pd 0 of 30.9% to 72.7% before and after reaction) ( Table 2). This result can also explain why the catalysts become more active after 15 min of reaction. The slight deactivation observed in the recycling tests performed after washing the catalyst after each run can be attributed to the formation of palladium carbide, as already demonstrated by Rogers et al. [21]. Isopropyl-furfuryl ether and isopropyl-tetrahydrofurfuryl ether were obtained by reacting isopropanol with furfuryl alcohol and tetrahydrofurfuryl alcohol, respectively, using ZSM5 as a catalyst at 150 °C and as a standard.

Catalysts Preparation
Supported Pd nanoparticles were synthesized by the sol-immobilization method [27]. Two aqueous solutions, one containing the metal precursor (Na2PdCl4, 1 mL of a 1.26 × 10 −4 mol L -1 solution) and one containing the desired capping agent (1 mL of a 1 wt % solution, capping agents/Pd (w/w) = 1), were added to 100 mL of deionized water. The third solution of NaBH4 (3.7 mL of a 0.1 mol L -1 solution; NaBH4/Pd (mol/mol) = 8) was then added to the yellow-brown solution under stirring. After the reduction of Pd species (30 min), the colloidal solution was immobilized on activated carbon (Camel X40S; Surface Area = 1200 m 2 g -1 ) under vigorous stirring. The amount of support material required was calculated to give a final metal loading of 1 wt %. The suspension was acidified to pH 2 by sulfuric acid before being stirred for 60 min to accomplish full immobilization of the metal NPs onto the support [49]. The slurry was filtered, washed completely with distilled water, and dried at 80 °C for 2 h.

Catalytic Tests
The liquid phase hydrogenation of furfural was carried out in a stainless-steel autoclave equipped with a heater and a magnetic stirrer. For a typical experiment, 0.063 g of catalyst (1:500 metal/substrate) was loaded in the reactor and tested at 5 bar of H2 and different temperatures (25 °C, 50 °C, and 75 °C) for 2 h. Before starting the reaction, the reactor was charged with 10 mL of 0.3 M furfural solution using 2-propanol as the solvent; the reactor was then purged with nitrogen and finally charged at 5 bar of H2. The reactor was heated to the desired temperature and the reaction commenced as soon as the stirring speed was 1000 rpm. Samples were collected at regular intervals for gas chromatography analysis (Agilent 6890, equipped with a Zebron ZB5 60 m x 0.32 mm x 1 μm column and a FID detector, Agilent, Santa Clara, CA, USA). Response factors were determined using a known concentration of standard solutions of the pure compounds. To identify unknown products, gas chromatography-mass spectrometry (GC-MS) was used (Thermo Scientific ISQ QD, equipped with an Agilent VF-5ms column, 60 m × 0.32 mm I.D × 1 µm film thickness, Thermo Scientific, Waltham, MA, USA).

Characterization
The ultraviolet-visible (UV-Vis) absorption spectra of capping agents and the corresponding Pd sols were collected using a UV-Vis spectrophotometer (Shimadzu UV3600, Kyoto, Japan) with quartz cuvettes with 1 cm optical path length, in the wavelength range 150-600 nm.
The FTIR spectra of catalysts were recorded with a JASCO spectrometer (Jasco FT/IR-4100; Jasco Corporation, Tokyo, Japan) in the range of 4000-400 cm -1 . For the measurements, 1 mL of a solution of capping agent (1 wt %) was mixed thoroughly with 1 mL of a metal precursor solution (10 mg Pd mL -1 ); 0.15 g of KBr were then added, and the solution was mixed until all the solid was completely dissolved. The obtained solution was dried in a rotary evaporator and the resulting solid was crushed in a mortar and pressed to obtain a thin film.
The morphological characteristics of the samples were determined using a high-resolution transmission electron microscope (LIBRA 200FE Zeiss at 200 kV equipped with a high-angle annular dark-field detector (HAADF), Zeiss, Oberkochen, Germany). The samples were suspended in isopropanol and sonicated; afterwards, each suspension was dropped onto a carbon-coated copper grid (300 mesh), and the solvent was evaporated. Particle size distribution curves were obtained by counting onto the micrographs at least 300 particles.
X-ray photoelectron spectra (XPS) were acquired in an M-probe apparatus (Surface Science Instruments, Warrington, UK) equipped with an atmospheric reaction chamber. The XPS lines of Pd 3d, O 1s, and C 1s were recorded by applying a Al Kα characteristic X-ray line and (hv = 1486.6 eV) pass energy.
The metal content was checked by atomic absorption spectroscopy (AAS) analysis of the filtrate on a Perkin Elmer 3100 instrument (Perkin Elmer, Waltham, MA, USA). After the catalyst filtration, a 10 mL sample of the filtered solution was collected and analyzed

Conclusions
A sol-immobilization methodology was utilized for preparing a series of catalysts with different capping agents (PVA, PVP, and PDDA), using carbon as the support (Camel X40s). UV-Vis and FTIR analyses were applied for studying the interaction between the Pd precursor and the capping agents, and XPS and HRTEM techniques were used to identify the metal surface exposure and particle size/distribution, respectively.
The results showed differences in the relative amount of Pd at the surface of catalysts. Moreover, the use of different capping agents affected the initial activity of the catalysts (PdPVA/C > PdPDDA/C > PdPVP/C), depending on the Pd exposure on the catalyst surface. However, in terms of particle size and distribution, the capping agents were not as effective as expected, since all the catalysts revealed a roughly similar range of particle size and distribution when using carbon as the support. We also found that the use of different capping agents affects the product distribution at low temperatures (25 °C and 50 °C).