1. Introduction
Carbon materials have attracted a lot of attention in heterogeneous catalysis because of their low cost and good chemical and mechanical stability [
1,
2,
3,
4]. Activated carbon (AC), carbon nanotubes (CNTs), carbon nanofibers (CNFs), carbon nanocoils (CNCs), ordered mesoporous carbons (OMCs), and graphene have been used as supports for metal nanoparticles (i.e., Au, Pd, Pt, Ru) [
5,
6,
7]. The utilization of carbon materials as catalyst supports for metal nanoparticles requires precise control over the surface chemistry in particular. This factor can strongly influence the metal active phase dispersion and electronic properties, which influence the overall catalytic performance [
1,
5]. A common strategy is the functionalization of the carbon surface with the incorporation of heteroatoms (commonly boron, nitrogen, oxygen, and phosphorus) [
8,
9,
10]. Several studies have shown that the introduction of oxygen and nitrogen groups can modify the acid/base properties of carbon materials, but also improve the metal dispersion, increase the stability, and modify the electronic properties of the metal nanoparticles [
11,
12,
13,
14,
15,
16,
17]. Beside O and N-doped carbon, recently, the effect of the introduction of P-groups on carbon surface has been an object of study in electrocatalysis [
18,
19,
20,
21,
22,
23,
24], acetylene hydrochlorination [
25], and liquid phase hydrogenation and oxidation [
26,
27,
28]. The beneficial effect of the functionalization of activated carbon with P groups on the activity and stability of Pd nanoparticles in the hydrogenation of nitroarenes has been reported [
26]. The authors demonstrated that this result has been attributed to the capability of P to modify the electronic properties of Pd as a result of the transfer of electrons from C atoms to P atoms, favoring the formation of stable electron-rich Pd–P species [
26]. Guo et al., reported the high efficiency of carbon supported Pd–P alloy (Pd–P/PCF) for the benzyl alcohol oxidation [
27]. They demonstrated that embedded P atoms influence the 3 d electron density of Pd and may weaken the binding of alcohol intermediates, thus promoting the catalytic performance of Pd–P/PCF. Pd deposited on nitrogen and phosphorus-codoped graphene sheets showed better activity than Pd on unfunctionalized carbon in the oxidation of 5-hydroxymethylfurfural (HMF) to furandicarboxylic acid (FDCA) [
28]. The authors showed that the functionalization of graphene with N- and P-groups increases the fraction of surface Pd
2+ species, which plays a determining role in reducing the activation energies of both HMF conversion and FDCA formation.
FDCA is an important high value product as it has been considered a potential replacement source of terephthalic acid, the monomer presently used for the production of polyethylene terephthalate (PET) as replacement of the oil-derived PET [
29,
30]. FDCA can be obtained by oxidation of HMF, which derived from the dehydration of fructose [
31,
32,
33]. The synthesis of FDCA from HMF has been reported using molecular oxygen as oxidant in the presence of Au, Pt, Pd, or Ru based heterogeneous catalysts, with water as the solvent and a base [
34,
35,
36,
37,
38,
39,
40,
41,
42]. Moreover, the combination of two metals in particular, Au and Pt, to form alloys has been reported to result in a catalytic system that is more effective than the corresponding monometallic counterparts [
43,
44].
We have recently used P-functionalized carbon nanofibers (P–CNFs) for the transformation of fructose into 5-hydroxymethylfurfural (HMF), showing that the incorporation of P groups on carbon nanofibers, with different graphitization degrees (pyrolytically striped carbon nanofibers (PS-CNFs), low-temperature heat treated carbon nanofibers (LHT-CNFs), and high-temperature heat treated carbon nanofibers (HHT-CNFs)), produces effective catalysts for this reaction [
45].
In this work, we have utilized these materials as support for Au, Pt, and Au–Pt nanoparticles as catalyst for the oxidation of HMF to FDCA. The aim is to investigate the role of P in enhancing the activity and the stability of the metal nanoparticles.
The effect of P amount has been evaluated in terms of activity, selectivity, and stability. Moreover, the effect of the reaction conditions (temperature, oxygen pressure, and NaOH amount) has been evaluated.
2. Materials and Methods
CNFs functionalization: commercial Pyrograph®-III CNFs (code PR-24, average diameter of 100 ± 30 nm and a specific surface area of 40–45 m2/g) from Pyrograph Products, Inc. (Cedarville, OH, USA), an affiliate of Applied Science, Inc. (Grass Valley, CA, USA), were used as pristine carbon materials. In particular three samples thermally processed at different severity degrees were selected as carbonaceous substrates: vapor-grown carbon nanofibers treated at 700 °C (Pyrolytically Striped CNFs, hereafter shortened as PS-CNFs; average diameter of 100 ± 30 nm and a specific surface area of 45 m2/g), 1500 °C (Low-temperature Heat Treated CNFs, LHT-CNFs; average diameter of 100 ± 30 nm and a specific surface area of 43 m2/g), and 3000 °C (High-temperature Heat Treated CNFs, HHT-CNFs; average diameter of 100 ± 30 nm and a specific surface area of 41 m2/g).For the functionalization, 2 g of CNF sample were suspended in 250 mL of a HNO3–H3PO4 1:1 vol/vol mixture, stirred, and heated at 150 °C (2 h). The functionalized carbon samples were then filtered, thoroughly washed with distilled water till neutrality, and finally dried at 100 °C overnight. Functionalized samples were labeled as P–PS-CNFs, P–LHT-CNFs, and P–HHT-CNFs.
2.1. Catalyst Preparation
Au/CNF: Solid NaAuCl4·2H2O (ca. 20 mg, corresponding to Au = 0.051 mmol = 10 mg) and 0.5 mL of polyvinyl alcohol (PVA) 1 wt % solution (Au/PVA = 1/0.5 wt/wt ratio) were added to 100 mL of H2O. After 5 min, 2 mL of a fresh solution of NaBH4 (0.1 M) (Au/NaBH4 = 1/4 mol/mol) was added to the yellow solution under stirring. Within a few minutes of sol generation, the colloid was immobilized by adding the support (ca. 1 g) under vigorous stirring. The amount of support was calculated as having a total final metal loading of 1 wt %. After 2 h, the slurry was filtered, the catalyst was washed thoroughly with distilled water to remove PVA from the catalyst surface, and then dried at 80 °C for 4 h.
Pt/CNF: Solid K2PtCl4 (ca. 20 mg, corresponding to Pt = 0.051 mmol = 10 mg) and 0.5 mL of PVA (1 wt %) solution (Pt/PVA = 1/0.5 wt/wt) were added to 100 mL of H2O. After 5 min, H2 was bubbled (50 mL/min) under atmospheric pressure and room temperature for 2 h. The colloid was immobilized by adding the support (ca. 1 g) under vigorous stirring. The amount of support was calculated as having a total final metal loading of 1 wt %. After 2 h, the slurry was filtered, the catalyst was washed thoroughly with distilled water to remove PVA from the catalyst surface, and then dried at 80 °C for 4 h.
Au6Pt4/CNFs: NaAuCl4·2H2O (ca. 12 mg corresponding to Au = 0.031 mmol = 6 mg) was dissolved in 60 mL of H2O, and 0.3 mL PVA (1%, wt %) was added (Au/PVA = 1/0.5 wt/wt). The yellow solution was stirred for 3 min, after which 1.2 mL of a fresh solution of 0.1 M NaBH4 (Au/NaBH4 = 1/4 mol/mol) was added under vigorous magnetic stirring. Within a few minutes of sol generation, the gold sol was immobilized by adding the support (ca. 1 g) under vigorous stirring. The amount of support was calculated as having a gold loading of 0.60 wt %. After 2 h, the slurry was filtered and the catalyst washed thoroughly with distilled water (neutral mother liquors). The Au/support was dispersed in 40 mL of water, with K2PtCl4 (ca. 8 mg corresponding to Pt = 0.021 mmol = 4 mg) and 0.2 mL PVA solution (Pt/PVA = 1/0.5 wt/wt) added. H2 was bubbled (50 mL/min) under atmospheric pressure and room temperature for 2 h. The slurry was filtered, the catalyst was washed thoroughly with distilled water to remove PVA from the catalyst surface, and then dried at 80 °C for 4 h. The total nominal metal loading was 1 wt %.
2.2. Catalytic Tests
The reactions were performed in a stainless steel reactor equipped with a magnetic stirrer. HMF 0.15 M and the catalyst (substrate/total metal = 50/500 mol/mol) were mixed in distilled water (total volume 10 mL), in presence of NaOH (1–4 equiv) or under base free conditions. The reactor was pressurized at 1–5 atm of oxygen and set to 30–80 °C. Once the temperature was reached, the gas supply was switched to oxygen and the monitoring of the reaction started. The reaction was initiated by stirring.
Samples were removed periodically and analyzed by high performance liquid chromatography (HPLC, Agilent Technologies, 1220 Infinity, Santa Clara, CA, USA) using a column (L × i.d.: 300 mm × 7.8 mm, Alltech OA-10308, Fisher Scientific, Hampton, NH, USA) with ultraviolet (UV) and refractive index (RI) detection to analyze the mixture of the samples. Aqueous H3PO4 solution (0.1 wt %) was used as the eluent. Products were identified and quantified by comparison with the original standard samples. The selectivity to products was calculated based on the number of carbon atoms. The difference in concentration from the carbon mass balance was attributed to unknown products detected on HPLC chromatograms. The catalyst was recycled in the subsequent run after filtration without any further treatment. The recovery of the catalyst was always >98%.
2.3. Characterization of Catalysts
Raman spectroscopy was performed with a Renishaw inVia Raman microscope (Wotton-under-Edge, Gloucestershire, UK). Bare supports and fresh catalysts were analyzed. Typically, a sample of approximately 0.01 g Awas placed on a metal slide inside the spectrometer. The powder was analyzed under an IR class laser (514 nm) with a laser intensity of 50%. The sample was scanned at an attenuation time of 22 s and 10 scans were carried out to give a spectrum.
Particle size distributions and mean particle size were obtained by means of TEM using a JEOL JEM 2100 TEM operating at 200 kV (Peabody, MA, USA). Samples for examination were prepared by dispersing the catalyst in high purity ethanol. A drop of the suspension was allowed to evaporate on a holey carbon film supported by a 300-mesh copper TEM grid. Samples were subjected to bright field diffraction contrast imaging experiments. Mean particle sizes and particle size distributions were determined by measuring the size of over 200 particles from different selected areas.
The actual metal loading was verified by atomic absorption spectroscopy (AAS) analysis, on a Perkin Elmer 3100 instrument (Waltham, MA, USA). After filtration and washing of as-prepared catalysts, residual metal content in the filtrate was evaluated by the method of standard additions.
X-ray photoelectron spectroscopy (XPS) was performed on a Thermo Scientific K-alpha spectrometer (Waltham, MA, USA). Samples were analyzed using a monochromatic Al X-ray source operating at 72 W (6 mA × 12 kV), with the signal averaged over an oval-shaped area of approximately 600 × 400 microns. Data was recorded at pass energies of 150 eV for survey scans and 40 eV for high resolution scan with a 1 eV and 0.1 eV step size, respectively. Charge neutralization of the sample was achieved using a combination of both low energy electrons and argon ions (less than 1 eV), which gave a binding energy of 284.8 eV for the sp3 component of C 1s.
All data were analyzed using CasaXPS (v2.3.17 PR1.1, Teignmouth, Devon, UK) using Scofield sensitivity factors and an energy exponent of −0.6. The regions of C 1s and Au 4f and Pt 4f have been fitted using a Shirley type background.
3. Results
Carbon nanofibers (CNFs) treated at 700 °C (pyrolytically striped, PS-CNFs), 1500 °C (LHT-CNFs), and 3000 °C (HHT-CNFs) were used as supports for Au–Pt nanoparticles. The heat treatment modifies the graphitization degree of CNFs, with HHT-CNFs being the more graphitized according to Raman spectra analysis (
Table 1). A portion of these materials have been functionalized, introducing phosphorus-containing groups on the surface. We have previously shown that the surface chemistry of CNFs, in particular the graphitization degree, could influence the amount of introduced P [
45]. XPS analysis has shown the presence of 1.19% of P on the surface of P–HHT-CNFs, whereas a lower amount has been observed for P–PS-CNFs and P–LHT-CNFs (0.23% and 0.24%, respectively,
Table 2). Raman analysis has shown that the functionalization has a different impact on the resulting graphitization degree on the three different CNFs. In the case of PS-CNFs, the acid functionalization increased the graphitization (
ID/
IG of 0.69 and 0.37 for PS-CNFs and P–PS-CNFs, respectively,
Table 1), probably removing the amorphous carbon present on PS-CNFs surface [
46]. On the contrary, the functionalization negatively impacts on the graphitization of the LHT-CNFs (
ID/
IG of 0.93 and 1.18 for LHT-CNFs and P–LHT-CNFs, respectively,
Table 1), whereas no significant modifications were observed for HHT-CNFs (
ID/
IG of 0.12 and 0.14 for HHT-CNFs and P–HHT-CNFs, respectively,
Table 1).
Au–Pt catalysts have been prepared using a two-step procedure developed by our group that ensures the formation of homogeneously alloyed Au–Pt nanoparticles [
48]. Gold was first deposited on the support as PVA-protected nanoparticles. In the second step, Pt precursor solution was added to the solid Au/support material and slowly reduced by H
2. Monometallic Au and Pt have been prepared for comparison. The introduction of Au–Pt nanoparticles had an unexpected effect on the graphitization of the materials (
Table 1). Indeed, the
ID/
IG ratio did not change for P–PS-CNFs, whereas the graphitization decreased for P–LHT-CNFs and increased for P–HHT-CNFs (
Table 1). The reason for this trend remains unclear. Transmission electron microscopy (TEM) has been utilized to determine the morphology of Au–Pt nanoparticles. Similar mean metal particle size in the range of 4–5 nm has been observed for the catalysts, (
Table S1,
Figure 1). In all the sample, metal nanoparticles were homogeneously dispersed on the external CNF surface.
Scanning transmission electron microscopy (STEM) images showed a non-optimal dispersion of metal nanoparticles with the presence of few agglomerates regardless the support used (
Figure 2).
The catalytic performance of the prepared Au–Pt nanoparticles supported on functionalized carbon nanofibers has been evaluated in the oxidation of HMF using the reactions conditions reported in our previous studies (HMF 0.15 M, Alcohol/metal: 200 mol/mol, T = 60 °C, pO
2 = 3 atm, 2 equiv. NaOH) (
Table 3) [
44]. The influence of catalyst amount and stirring rate has been investigated to be sure to work under kinetic regime (
Figures S1 and S2). The main products detected were 5-hydroxymethyl-2-furancarboxylic acid (HFCA), deriving from the oxidation of the carbonyl group of HMF, and 2,5-furandicarboxylic acid (FDCA), the desired product, derived from the oxidation of both carbonyl and alcoholic group (
Scheme 1).
Table 3 reports the activity, the conversion after 2 h, and the selectivity at iso-conversion (80%). The activity has been calculated as mol of converted HMF per mol of metal per hour and was calculated after 15 min of reaction. The comparison between Au–Pt and the corresponding monometallic Au and Pt was initially considered using PS-CNF as a reference support. The beneficial synergistic effect of the bimetallic systems was envisaged with an activity of Au–Pt (136 h
−1) higher than Au (92 h
−1) and Pt (58 h
−1). Considering the bimetallic systems on pristine CNFs, it can be observed that Au–Pt/PS-CNF and Au–Pt/LHT-CNF showed almost the same activity (136 h
−1 and 140 h
−1, respectively) and selectivity as FDCA (51% and 54% at 90% conversion, respectively). On the contrary, Au–Pt/HHT-CNF showed a better performance with an activity of 185 h
−1 and a selectivity to FDCA of 68%. These results cannot be explained in terms of particle size being similar for the three catalysts (
Table S1), but probably to the higher graphitization degree of the functionalized CNF support. Similar findings have been reported for glycerol oxidation under similar reaction conditions, where the higher activity has been correlated with a lower amount of structural defects, leading to Au particles strongly anchored to the orderly exposed graphite edges [
49,
50]. The introduction of a small amount (~0.2%) of phosphorous in P–PS-CNF and P–LHT-CNF slightly increased the activity of Au–Pt catalysts compared with undoped Au/PS-CNF and Au–Pt/LHT-CNF (
Table 3). The higher P content in P–HHT-CNF (1.19%) has a higher impact on the activity of Au–Pt (368 h
−1 and 185 h
−1, for Au–Pt/P–HHT-CNF and Au–Pt/HHT-CNF, respectively) with almost full conversion and 96% of selectivity to FDCA after only 1 h.
The beneficial effect of P cannot be attributed to possible changes in Au–Pt aggregation state, as the nanoparticle size is similar to Au–Pt/HHT-CNF (4.5 and 3.9 nm for Au–Pt/HHT-CNF and Au–Pt/P–HHT-CNF, respectively), but rather to the capability of P to modify the electronic properties of the metal active sites, as already envisaged in previous studies for Pd [
26,
27,
28].
To verify this hypothesis, XPS analysis was performed to unravel eventual electronic modifications induced on metal NPs by P functionalities of doped CNF surface (
Table 4). XPS analysis confirmed the presence of P on the catalyst surface. Single P peaks with BE in the range of 133.5–134.0 eV revealed that P is prevalently present as phosphate groups in all the catalysts, as reported in our previous work [
45]. Compared with the reference XPS data from bulk metal materials (84.0 eV and 71.0 eV for Au 4f
7/2 and Pt 4f
7/2, respectively), slight shifts toward higher binding energy were observed in both Au and Pt XPS peaks. The relatively more pronounced shift in Au peak was revealed for the Au–Pt/P–HHT-CNF sample (84.12 eV). Concerning the Pt XPS signals, the peaks centered at 71.28 and ca. 72.5 eV can be assigned to the Pt 4f
7/2 of the Pt metallic state (Pt
0) and to oxidized Pt species (Pt
2+), respectively. It is noted that the position of Pt
0 signal is not affected by the nature of the support, being exactly the same (71.28 eV) for Au–Pt/P–PS-CNF, Au–Pt/P–LHT-CNF, and Au–Pt/P–HHT-CNF. Differently, the binding energy for Pt
2+ signal increases according to the following order: Au–Pt/P–HHT-CNF (72.40 eV) < Au–Pt/P–LHT-CNF (72.49 eV) < Au–Pt/P–PS-CNF (72.60 eV). Quantitative analysis of the Pt
0/Pt
2+ atomic surface ratio showed that in all catalysts, the Pt metallic state was predominant and this is more evident in Au–Pt/P–HHT-CNF. In particular, the Pt
0/Pt
2+ atomic surface ratio increases in the order Au–Pt/P–PS-CNF (4.7) < Au–Pt/P–LHT-CNF (6.4) < Au–Pt/P–HHT-CNF (7.6), reflecting the increase of P content (
Table 2). Consequently, it seems that a support P-rich surface induces a decrease in the electron density of Au atoms and the stabilization of Pt
0 species, which are less prone to be oxidized. Both these pieces of evidence could explain the observed trends in catalytic data reported in
Table 3, as discussed below (
Section 4).
To better understand the different catalytic behavior, in particular, the different selectivity to FDCA at iso-conversion of Au–Pt/P–PS-CNF (56%), Au–Pt/P–LHT-CNF (61%), and Au–Pt/P–HHT-CNF (96%), the products distribution as a function of time of reaction has been analyzed.
Figure 3 showed a different rate in the conversion of HFCA to FDCA using the three different catalysts. The conversion rate of HFCA to FDCA increased in the following order: Au–Pt/P–PS-CNF < Au–Pt/P–LHT-CNF < Au–Pt/P–HHT-CNF. In other words, Au–Pt/P–PS-CNF and Au–Pt/P–LHT-CNF are able to easily oxidize the carbonyl group, resulting in the formation of HFCA but not in the oxidation of the alcoholic group, whereas Au–Pt/P–HHT-CNF is effective in the oxidation of both groups.
The effect of reaction parameters, in particular, NaOH amount, temperature, and O2 pressure, have been investigated for the most active catalyst, Au–Pt/P–HHT-CNF.
Initially, the effect of the base amount has been considered. As reported in previous works [
34,
35,
36,
37], the presence and the amount of base is a key parameter in HMF oxidation to FDCA. In the absence of base, very low HMF conversion was observed; moreover, FDCA and other byproducts are insoluble in acidic media and tend to precipitate on the catalyst surface, thus contributing to fast catalyst deactivation. The addition of a base stabilizes FDCA in its anionic form and prevents the degradation of HMF. The reported optimal value for gold catalysts is four equivalents. When lower concentrations of base are used, a negligible amount of FDCA is produced, making the intermediate oxidation product (HFCA) the main product. On the other hand, when more than four equivalents of NaOH are used, no significant improvements in the catalytic performances have been observed. In our experiments, the NaOH amount varied from 0 to 4 equiv., keeping constant the temperature (60 °C) and the oxygen pressure (3 atm) (
Table 5). It should be noted that 2 equiv. corresponds to the stoichiometric amount needed to neutralize the two acid groups of FDCA. Increasing NaOH amount to 4 equiv., the activity increases, but the selectivity to FDCA decreased from 96 to 92 because of the formation of unknown products. When a lower amount than the stoichiometric one is used, the catalyst possesses lower activity, in particular, in the transformation of HFCA to FDCA, confirming previous evidence from the literature [
34,
35,
36,
37]. In particular, running the reaction in the absence of a base (pH = 5), Au–Pt/P–HHT-CNF is still active, but only 15% of conversion has been obtained after 1 h. Moreover, almost no FDCA has been observed with the formation of 40% of unknown products, probably because of the decomposition of HFCA under acidic conditions.
Increasing the temperature from 30 to 80 °C has a positive effect on the activity and in the transformation of HFCA to FDCA (
Table 6). However, it can be observed that at higher temperature, for example, at 80 °C, the amount of undesired products increases (11%), whereas at lower temperature, the concentration of by-products is <2%.
Varying the oxygen pressure from 1 to 5 bar, it can be observed that the HMF conversion maintains almost complete after 1 h in the range of examined pressures (
Table 7). However, large differences were observed in the FDCA production. Indeed, O
2 pressures higher than 3 atm were necessary to achieve high selectivity (>96% at 80% conversion) to FDCA.
Combining the results attained and varying the reaction parameters, it is possible to identify the optimal operative conditions for the selective oxidation of HMF to FDCA in the presence Au–Pt/P–HHT-CNF; working at 60 °C, with 2 equiv. NaOH and 5 bar of pressure, allows to reach the maximum yield to FDCA. Therefore, in order to study the durability of the catalyst, stability tests have been performed under the optimized reaction conditions. Recycling tests have been performed by filtering the catalyst after each run and reusing it without any pretreatment. The results have been reported as yield to FDCA.
Figure 4 showed that the catalyst is stable after six runs. Atomic absorption spectroscopy (AAS) performed on the solution did not reveal any leaching of metal components (Au–Pt) or phosphorus during the reaction.
4. Discussion
Three commercial carbon nanofibers (CNFs) with different graphitization degree, PS-CNFs, LHT-CNFs, and HHT-CNFs, have been selected as pristine carbon materials to be doped with P-containing functionalities and then used as support for bimetallic Au–Pt NPs. The incorporation of phosphorus on these materials has been previously reported to proceed successfully [
45]. Moreover, as-obtained P-doped CNFs have demonstrated to be effective metal-free catalysts for the selective fructose dehydration to 5-hydroxymethyl furfural (HMF) [
44]. On the basis of these promising results, we explored the potentiality of P–CNFs as supports for Au–Pt-based catalysts in the selective oxidation of HMF to furan dicarboxylic acid (FDCA). The functionalization was performed by treating pristine CNFs with a HNO
3–H
3PO
4 1:1 vol/vol mixture at 250 °C. Beside introducing O- and P-functionalities, this treatment altered the defectiveness of the materials as evidenced by evaluating the
ID/
IG ratio from Raman spectra. In particular, an enhancement of graphitization was obtained for PS-CNFs, while an opposite trend was revealed for LHT-CNFs. No significant variations were observed for HHT-CNFs, which still result in the more graphitic materials.
The immobilization of Au–Pt NPs on P–CNFs produced catalysts with similar mean metal particle sizes (3–4 nm) as evidenced by TEM analysis. Although surface doping is usually known to provide effective anchoring sites able to stabilize very small metal NPs, in this case, the introduction of P-functionalities did not improve the metal dispersion compared with undoped CNFs. The presence of large aggregates was disclosed by STEM-EDX analysis in all the samples, independent of the nature of the support. Au–Pt-based materials were tested as catalysts in HMF oxidation reaction and the catalytic performances were compared with the ones of monometallic Au or Pt counterparts. As expected, the addition of Au to Pt resulted in more active catalysts compared with monometallic systems, as a consequence of the insurgence of synergistic effects. Working with same particle size of metal nanoparticles allowed us to rule out metal particle size effect, and thus to unveil the role of the support in determining the catalytic behavior of studied materials (
Table 8).
As general trend, the incorporation of P on the support slightly improved the activity of Au–Pt-based catalysts. From the evaluation of activity, conversion, and selectivity at iso-conversion, Au–Pt/P–HHT-CNF emerged as the best catalyst (activity 368 h
−1, 98% conversion after 2 h, and 96% selectivity to FDCA at 80% conversion). Analyzing the product distribution as a function of the reaction time for the studied catalysts, it appears clear that Au–Pt/P–HHT-CNF is the only catalyst able to quantitatively oxidize HFCA to FDCA (
Figure 3 and
Table 8). This evidence could be related to the interaction between Au–Pt NPs and P functionalities of the support. Chen et al. [
28] reported that Pd
2+ species are stabilized by P functionalities in Pd NPs deposited on highly porous phosphorus-doped graphene sheet. The Pd
2+ species are involved in accelerating the oxidation of hydroxyl group of HFCA into aldehyde group of 5-formyl-2-furancarboxylic acid (FFCA), which is the rate-determining step in FDCA production from HMF. A linear dependence of the activation energy for FDCA production on the fraction of Pd
2+ has been observed. In our case, XPS analysis revealed a shift to higher binding energy for Au 4f
7/2 signal of Au–Pt NPs supported on P-doped CNFs (
Table 4 and
Table 8). This shift could be ascribed to a decreased electron density on gold centers and it is more pronounced for the Au–Pt/P–HHT-CNF catalysts, which contains the highest P-content and showed the best catalytic performances in HMF oxidation to FDCA. Consequently, we might guess an effect similar to the one disclosed by Chen et al. for Pd catalysts; the P functionalities of the support are able to stabilize Au
δ+ species, which might be responsible for the high FDCA production. This evidence has not been reported previously for Au-based catalysts in HMF oxidation, but it is in agreement with the results of Wang et al. [
25], who studied phosphorus-doped carbon as supports for gold-based catalysts in acetylene hydrochlorination. The authors ascribed the enhanced catalytic performances of the studied catalysts to the stabilization of the catalytic active gold species Au
3+ through specific interaction with phosphate groups on support surface.
The presence of P on carbon surface also influences the oxidation state of Pt and in particular, the higher the content of P in the support, the higher the Pt
0/Pt
2+ ratio. This experimental outcome might reveal an opposite trend for Pt compared with Pd and Au. Indeed, it seems that a higher P content is associated with a greater stabilization of Pt
0 than Pt
2+ species. This apparent contradiction might find a possible explanation in the particular synthetic route used for the preparation of Au–Pt catalysts. Indeed, in our two-step procedure, Pt addition occurred in the second step, when Au NPs have been already immobilized on the support. Therefore, it may reasonably be supposed that when Pt is introduced on the catalyst surface, no phosphate groups are available to interact with Pt
2+ ions, as they are involved in strong interactions with Au centers. Consequently, Pt
2+ ions undergo reduction in the presence of H
2, thus forming metal Pt NPs. However, in this hypothesis Pt
0/Pt
2+ ratio should be insensitive to the P-content, in contrast to experimental evidences. Again, the interactions between Au and phosphate groups could be invoked to justify the Pt
0/Pt
2+ ratio dependence on P-content. Indeed, it is known that because of the difference in Fermi levels, electrons tend to migrate from Pt to Au, when the two metals are contacted. In our case, the strong interaction between phosphate groups and Au NPs would make Au
δ+ centers less prone to be reduced by Pt, thus minimizing the electron flow from Pt to Au and the consequent formation of Pt
2+ species (
Scheme 2).
Considering that Pt
0 has been reported to play a key role in HMF oxidation [
51,
52], we cannot rule out a concomitant influence of the Pt
0/Pt
2+ ratio on the catalytic performances. In fact, the best catalytic performances have been attained with Au–Pt/P–HHT-CNF, which also possesses the highest Pt
0/Pt
2+ ratio.
Finally, the experimental conditions have been optimized in order to maximize the FDCA yield. It has been found that at 60 °C, with 2 equiv. NaOH and 5 bar of pressure, it is possible to quantitatively and selectively oxidize HMF to FDCA. Under these conditions, the catalyst was stable up to six runs (
Figure 4).