Palladium catalyzed hydrogenation of biomass derived halogenated furfurals

Abstract

The formation of valuable products and especially fuel candidates from lignocellulosic biomass is highly desirable. Lignocellulose derived halogenated furfurals formation was reported as highly efficient in comparison to the current existing methods of lignocellulose transformation. However, halogenated furfurals are platform chemicals and not end use chemicals, certainly not for fuels. Therefore, transformation methods of halogenated furfurals into fuels, fuels additives, or other valuable compounds are desirable. In this work we present the hydrogenation of halogenated furfurals over carbon supported palladium catalysts. Palladium catalysts showed better performance in the formation of 5-methyl furfural (MF) from halogenated furfurals compared to other catalysts. The reaction products were identified using GC-MS, FT-IR and NMR, and they were quantified using GC analysis. Catalysts were characterized with SEM, BET and pH meter. The role of catalysts properties and reaction parameters in MF preparation, and their effect on MF yields and selectivity were examined. In addition, the catalysts recovery and reuse in subsequent cycles was examined together with the recovery of hydrochloric acid or hydrobromic acid, formed as by products in halogenated furfurals hydrogenation.

---

1. Introduction

Environmental and political concerns combined with diminishing petroleum resources have driven our society to look for renewable and sustainable replacements for petroleum based commodities. For this reason, the bio-refinery concept has attracted much attention in the last decade. The bio-refinery facility aims to produce chemicals and fuels from biomass components.¹ Biofuel production is a key process for the profitability of the bio-refinery facility. Today, the transportation sector is strongly dependent on petroleum; in the U.S. about 92% of the energy used by the transportation sector is sourced from petroleum, while natural gas and biofuels contribute only about 3% and 5%, respectively.² Biomass is attractive for the production of transportation fuels since it can supply organic carbon in liquid form with physical properties similar to petroleum-derived fuels.^3,4 Currently, biodiesel, produced primarily from vegetable oils, and bio-ethanol, produced from sugar canes and starch, are commercially available and are used mainly as blends with traditional petro-fuel in existing engines.^5–8 The production of fuels from edible resources is limited by agricultural lands shortage.^9,10 Therefore, non-edible resources must be considered for biofuel production. The most abundant non-edible material in nature is lignocellulose. It is estimated that 170 billion tons a year of globally produced biomass is roughly consist of 75% carbohydrates, 20% lignin and only 5% of other chemicals (oils, fats, proteins, terpenes, alkaloids, etc.).^11,12 Lignocellulosic biomass consists of cellulose (40–50%), hemicellulose (20–40%) and lignin (20–30%)^13,14. While lignin is a three-dimensional aromatic polymer, cellulose and hemicellulose are polysaccharides structured from carbohydrates monomers. Since carbohydrates are the major ingredients of lignocellulose, they are the most interesting lignocellulosic components for biofuels and value added chemicals production. Cellulose, the major component in lignocellulose, has a high degree of crystallinity and its bonds are not accessible to a great extant as a result of its glycoside bonds arrangement, formed by an extensive hydrogen bonding.^15,16 Hence, the conversion of cellulose into valuable products is limited by the high degree of its crystallinity. Promising compounds that are possible from carbohydrates, and specifically from lignocellulose, in high yields are from the furfural family.^17,18

In 2004 the U.S. department of energy (DOE) released a report reviewing the most promising carbohydrates derived chemicals as a platform for bio-refinery processing.¹⁹ Taking into consideration a number of factors such as known processes, industrial viability, economical feasibility, the market size and the capacity of the compound to serve as a platform chemical in bio-refinery, the DOE report has listed 12 chemical opportunities. The only furanic compound included in this list was 2,5-furan dicarboxylic acid. In 2010, Bozell and Petersen revised that list, adding and excluding a few sugar derived compounds.²⁰ Most importantly, ethanol, furfural and 5-hydroxymethyl furfural (HMF) were included in the new list. The conversion of lignocellulose to ethanol is the most established technology for carbohydrates derived biofuel production. In this two-step methodology, cellulose and hemicellulose are hydrolyzed with diluted acid or an enzyme into sugar monomers followed by fermentation of the mono-sugars to ethanol using yeast or bacteria.²¹ However, bio-ethanol production from lignocellulose is expensive and requires a unique pretreatment step in order to achieve high yields. Furthermore, the process is time consuming and has low atom economy since only two third of the reactant's carbon atoms are retained in the product.^7,22 More appealing compounds included in Bozell and Petersen report, which retain all of the reactant's carbon atoms, are 5-hydroxymethyl furfural (HMF) and furfural. HMF and furfural can be produced from hexoses and pentoses found in cellulosic and hemicellulosic biomass, respectively.²³ Biofuel preparation from furfural and HMF was vastly investigated. Yan et al. and Nakagawa et al. have reviewed the conversion of furfural and HMF to various valuable compounds and fuel candidates via hydrogenation and hydrogenolysis.^24,25 Catalysts investigated and reported for hydrogenation or hydrogenolysis of furfural or HMF include Pd, Ni, Ru, Rh, Cu–Cr, Cu–Fe etc.^26–32 In hydrogenation/hydrogenolysis of furfural and HMF, the full or partial hydrogenation of the aldehyde and aromatic groups can take place. Typical compounds formed by furfural hydrogenation/hydrogenolysis include furfuryl alcohol, 2-methylfurn, tetrahydrofurfuryl alcohol and 2-methyltetrahydrofuran. Possible hydrogenation/hydrogenolysis products that can be formed from HMF include 5-methyl furfural (MF), 5-methyl furfuryl alcohol, 2,5-dihydroxymethylfuran, 2,5-dimethylfuran or any tetrahydro-form of these compounds. These furanic compounds are potential biofuels or fuel additives. In addition, diesel range fuel can be achieved by esterification, etherification or condensation of furfural or HMF (cross-condensation or self aldol-condensation of hydrogenated products). These reactions were reported in reviews published by Nakagawa et al. and James et al.^15,25 The catalytic production of diesel range fuels by C–C bond formation has been reported by several authors. Liu and Chen demonstrated coupling of HMF and other furanic derivatives via Umpolung coupling catalyzed by heterocyclic carbene.³³ Huang et al. have reported C–C bond formation of furfural and other aromatic aldehydes via reductive coupling.³⁴ Furfural is attained from hemicellulose, while cellulose is the major ingredient of lignocellulose and conversion methods for cellulosic material are highly significant. HMF can be produced from cellulosic material, but HMF is limited by its instability in storage and in processing, as reflected by its decomposition to levulinic acid and polymerization to humic materials.^35,36 The formation of halogenated furfurals, 5-chloromethyl furfural (CMF) and 5-bromomethyl furfural (BMF), from pure cellulose or from cellulosic waste was proven to be more efficient than HMF formation.^37,38 In these processes more than 80% yield of isolated CMF and BMF could be obtained. Nevertheless, halogenated furfurals cannot be use as biofuels directly, and therefore they serve only as platform compounds for biofuel production in a subsequent process. However, studies regarding subsequent conversions of CMF and BMF into fuel candidates are rather scarce. Mascal and Nikitin have reported that CMF can react with ethanol, replacing chloride ions with ethoxide and produce 5-ethoxymethyl furfural in high yields. Alternatively, CMF can undergo hydrogenation with palladium chloride catalyst and transform into 5-methyl furfural (MF) in high yields. However, the utilization of homogeneous catalyst might limit the applicability of this process.³⁷ Kumari et al. have also demonstrated the reaction of BMF with ethanol and the formation of 5-ethoxymethyl furfural;³⁸ however hydrogenation of BMF was not demonstrated in their report or in any other publication to the best of our knowledge.

In this study we explored the catalytic hydrogenation of CMF and BMF over heterogeneous palladium catalysts. We found that high yields of MF, that can be considered as fuel candidate or fuel additive, are obtained in this protocol. In addition, we observed that in the presence of some types of catalysts, coupling of furfural molecules proceeds rather than hydrogenation. The products obtained from coupling reactions can be used as platform for diesel range fuel, while MF is more suitable as a gasoline range fuel.

2. Experimental

2.1 Materials

CMF and BMF were prepared in our laboratory as published in a previous work.³⁹ 1,2-Dichloroethane (DCE) was purchased from Sigma-Aldrich and used as a solvent without any further purification. Several activated carbon supported palladium were examined as catalysts. Commercial 5% palladium on carbon (abbreviated as Pd/C^S) was purchased form Sigma-Aldrich. Activated carbon Darco and activated charcoal Norit® (abbreviated as Pd/C^D and Pd/C^N, respectively) were purchased from Sigma-Aldrich and examined as supports for palladium. Palladium loadings of 5%, 3%, 1% and 0.5% over activated carbon were prepared in our laboratory according to the following protocol: 1 g of activated carbon (Darco or Norit) was pretreated in an oven at 200 °C for 1 h. Afterwards, the treated activated carbon was placed in round bottom flask containing 10 ml aqueous solution of palladium nitrate (0.4686, 0.2816, 0.0937 and 0.0468 mmol for loading of 5%, 3%, 1% and 0.5%, respectively) and the slurry was thoroughly stirred for 1 h. Then, potassium formate, purchased from Sigma-Aldrich, was added drop-wise to the slurry (with molar ratio of 20 : 1 of potassium formate to palladium nitrate), and the slurry was stirred for 24 h. At the last step the slurry was filtered and the cake has been washed 3 times with deionized water and dried under air flow.

2.2 Catalysts characterization

Morphological observations in low and high magnifications of catalysts were performed with high resolution scanning electron microscope, HR-SEM, Sirion (EFI Company, USA). The HR-SEM unit is equipped with high stable Schottky field emission source filament, SE and BSE detectors, and acceleration voltage of 200 V to 30 kV.

Specific surface area and porous texture characterizations of supported palladium catalysts were carried out by N₂ adsorption at 77.3 K using NOVA-1200e instrument (Quantachrome Instruments, USA). Surface areas were determined using Brunauer–Emmett–Teller (BET) equation, while pore radiuses and pore volumes were determined using N₂ t-plot method and Barrett–Joyner–Halenda (BJH) method, respectively. All samples were degassed for 6 hours prior to each experiment under nitrogen flow at 300 °C.

Acidity measurements of the Darco and Norit supports were carried out according to a protocol reported by Sugumaran et al.⁴⁰ According to this protocol 1% (w/w) suspension of the activated carbon in deionized water was heated to 90 °C while stirring. Then the suspension was cooled to room temperature and the pH was measured. The pH values were measured with a bench top pH meter equipped with glass electrode from MRC Laboratories.

2.3 CMF and BMF hydrogenation

Hydrogenation of CMF and BMF was carried out according to the following protocol: 2.075 mmol of CMF (0.300 g) or BMF (0.392 g) were dissolved in 40 ml 1,2-dichloroethane (DCE) and placed in a 300 ml autoclave reactor (Parr Instrument Company, USA). The reactor has an inner coating of Hastelloy, in order to achieve high corrosion durability, and it is operating in batch mode. Then 0.030–0.100 g of Pd/C catalyst was added to the CMF or BMF solution. The reactor was sealed and purged 3 times with argon and then 3 times with hydrogen. After purging, the reactor was loaded with the desired hydrogen pressure (1–20 bar). The desired temperature (60–100 °C) was set and heating was executed by a thermocouple connected to controller. At the end of reaction, after the reactor was cooled down, the excess gas was removed and the reactor purged 3 times with argon, to remove hydrogen for safety reason. Finally, the solution was filtered in order to separate and recover the catalyst, and the solvent was evaporated in order to obtain the crude product.

The identification of products was carried with GC-MS (Thermo Scientific, USA) constructed from Trace 1300 GC, equipped with 30 m Rxi®-1ms non polar column, connected to a single quadruple mass spectrometer ISQ QD. Assay of reaction products was performed with gas chromatography (Focus GC, Thermo Scientific, USA) equipped with a low polarity 30 m ZB-5 column. Complementary analyses for products identification was carried out using NMR and FT-IR spectroscopy. ¹H-NMR spectra were recorded with Bruker DRX-400 instrument (Bruker Corporation, USA). Samples for NMR analyses were prepared by dissolving the crude products in DMSO-d6 and placing the solutions in NMR tubes. FT-IR spectra were recorder with FT-IR spectrometer model Alpha (Bruker Optics, Bruker Corporation, USA). Leaching of palladium from support to reaction medium was determined using ICP-AES (Perkin-Elmer Corp, Optima 3000). ICP-AES calibration was carried out with 10 ml solutions of 0.01–10 ppm palladium nitrate dissolved in 0.1 M nitric acid. ICP-AES samples were prepared by mixing the reaction crude products in 5 ml 0.1 M nitric acid solutions followed by filtration of solutions with 0.2 μm filter.

Hydrochloric acid and hydrobromic acid rejuvenation was carried out by connecting the reactor to a tube containing 21 ml of deionized water and bubbling of the gaseous phase remained in reactor at the end of reaction, after the reactor was cooled down to room temperature, into the deionized water. The tube was immersed in an ice bath in order to cool the water. After bubbling of reaction gaseous phase, the reactor gaseous content was flushed 3 more times into the collection tube with 10 bar N₂. The collected acid content was determined by titration with 0.1 M sodium hydroxide solution.

3. Results and discussion

3.1 Catalysts screening and reaction products

Various catalysts were investigated in CMF and BMF hydrogenation. With most of the examined catalysts the hydrogenation reaction proceeded via hydrodehalogenation, namely the replacement of halogen atom with hydrogen atom. Then, the hydrogenation product is 5-methyl furfural (MF), as identified by GC-MS analysis (see ESI†), and hydrochloric acid or hydrobromic acid are formed as a by-products from CMF or BMF, respectively (see in Fig. 1).

Fig. 1 A schematic representation of catalytic hydrogenation of (1) CMF and (2) BMF to MF (with the formation of HCl or HBr as by-products).

In order to find the most active and selective catalyst for CMF and BMF hydrogenation a catalysts screening has been performed and the results are summarized in Table 1. The abbreviations Pd/C^S, Pd/C^N and Pd/C^D are assigned for commercial activated carbon supported Pd (purchased from Sigma), Norit type activated carbon supported Pd and Darco type activated carbon supported Pd, respectively (the chemical and textural differences between these catalysts will be elaborated in the following sections).

Table 1 Summary of CMF and BMF hydrogenation experiments with different catalysts. All experiments were carried out with 0.300 g CMF or 0.393 g BMF as substrate, 40 ml 1,2-dichloroethane (DCE) as reaction solvent, 11 bar H₂ and reaction time of 5 h

Substrate	Temperature	Catalyst	Conversion	MF yield	Others yield
a With 0.5% Pd/C^N catalyst, 70 °C and 0.1 g catalyst were needed in order to achieve full conversion of CMF to MF.
CMF	50 °C	5% Ru/C	0%	—	—
CMF	50 °C	5% Rh/C	9.1%	9.1%	—
CMF	80 °C	65% Ni/SiAl	18.5%	18%	0.5%
CMF	50 °C	5% Pt/C	24.7%	14.5%	10.2%
CMF	50 °C	5% Pd/C^S	100%	6.2%	93.8%
CMF^a	70 °C	0.5% Pd/C^N	100%	99%	1%
BMF	50 °C	5% Ru/C	5.1%	5.1%	—
BMF	50 °C	5% Rh/C	53%	53%
BMF	50 °C	65% Ni/SiAl	0%	—	—
BMF	50 °C	5% Pt/C	86.7%	1.2%	85.5%
BMF	50 °C	5% Pd/C^S	100%	98%	2%
BMF	50 °C	0.5% Pd/C^N	31.9%	31.9%	—

---

Catalysts screening results reveal that Ru/C, Rh/C and Ni/SiAl are poor catalysts for CMF and BMF hydrogenation, and they result in low conversions under the reaction conditions specified in Table 1 (although Rh/C shows a medium activity in hydrogenation of BMF with 53% conversion to MF). Pt/C catalyst brings to CMF and BMF conversions of 24.7% and 86.7%, respectively. However, the selectivity of this catalyst towards MF is poor since only 14.5% and 1.2% yields of MF from CMF and BMF, respectively, have been obtained. The other products formed were identified as 5-methylfurfuryl alcohol and condensation products (53.3% yield of 5-methylfurfuryl alcohol and 32.2% yield of condensation products from BMF). Pd/C catalysts were found to be the most active catalysts for the conversion of halogenated furfurals into MF. 100% conversion of BMF and 98% yield of MF were obtained with commercial 5% palladium on active carbon (Pd/C^S), and 100% conversion of CMF and 98% yield of MF were obtained with 0.5% palladium on active carbon (Pd/C^N), which was prepared in our lab. Surprisingly, the commercial palladium catalyst was also very active for CMF hydrogenation, but it yielded mainly a complex products mixture comprised from condensation products rather than MF. The reaction products formed with commercial Pd/C catalyst, Pd/C^S, are depicted in Fig. 2.

Fig. 2 Identified reaction products obtained in CMF hydrogenation catalyzed by Pd/C^S.

When CMF was hydrotreated in the presence of Pd/C^S catalyst, only 6.2% yield of MF has been formed. In addition to condensation products, a small amount of non-condensation products could be identified. These include 5-methyltetrahydrofurfuryl alcohol (yield of 5.9%) and traces of saturated and partially saturated chlorinated compounds (total yield of 1.5%). The main products formed with Pd/C^S catalyst were condensation reaction products, with varying degrees of saturation. The identified compounds are presented in Fig. 2 and their total yield was 69.5%. Analysis of products structures was carried with GC-MS and supported by complementary information obtained by H-NMR spectroscopy and FT-IR spectroscopy (see ESI†). The proton NMR spectrum indicates the absence of aldehyde hydrogen and aromatic hydrogen in the products mixture, while the FT-IR spectrum indicates the existence of carbonylic and hydroxyl groups.

3.2 Catalysts characterization

3.2.1 The acidity of carbon supports. The acidities of the commercial palladium catalyst, 5% Pd/C^S, and the different carbon supports, Norit and Darco, were determined by pH measurements according to the protocol detailed in Section 2.2. Norit activated carbon was found to be slightly basic with pH of 8.88, while the Darco activated carbon was determined as acidic with pH value of 3.15. A pH value of 8.39 has been recorded for the commercial 5% Pd/C^S catalyst. The acidic or basic characteristics of the supports influence, as we shall see in the following sections, on the dispersion of palladium particles, and as a consequence on the yields of MF obtained from CMF and BMF.

3.2.2 SEM characterization of catalysts. The different catalysts tested in this work were examined with scanning electron microscopy for the characterization of morphology and dispersion of palladium particles over supports. SEM images are presented in Fig. 3.

Fig. 3 Scanning electron microscopy images of (a) 0.5% Pd/C^D, (b) 5% Pd/C^D, (c) 0.5% Pd/C^N, (d) 1% Pd/C^N, (e) 3% Pd/C^N, (f) 5% Pd/C^N and (g) 5% Pd/C^S.

The SEM images in Fig. 3(c–f) reveal that for the Norit support (the basic activated carbon), as palladium loading is higher, larger agglomerated particles were formed. In all samples, polydispersed particles with size range of 3–475 nm were found. A higher magnification reveals that all aggregates are built from small particles with size range of 1.75–4.75 nm (see ESI†). It also appears that in higher palladium loadings over support the proximity between particles is higher, i.e. the particles are dispersed in a denser manner. In lower palladium loadings, smaller and less frequent agglomerates were formed, and the distances between particles are large. From the SEM images in Fig. 3(a) and (b), for 0.5% and 5% of Pd/C^D, respectively, we can conclude that also for the Darco support as palladium loadings are higher larger agglomerates were formed. However, in the case of the acidic Darco support, palladium particles are dispersed less densely than palladium particles supported over the basic Norit activated carbon. In addition, as shown in Fig. 3(a) and (b), the Darco supported palladium particles are distant regardless of the metal loading. The differences between Norit and Darco supports in palladium particles agglomerations and dispersions might be attributed to the acidity of the supports.^41,42 The acidic Darco support has a very low pH, and densely assembles acidic surface groups that interact strongly with palladium precursors prevent the formation of highly dispersed palladium particles. On the other hand, the formation of highly dispersed palladium particles is possible with the slightly basic Norit activated carbon since the interactions between precursor and surface groups of the support are not too strong. In the case of Pd/C^S, in Fig. 3(g), the occurrence of large agglomerated particles seems to be more frequent than the other catalysts. Higher magnification SEM image of Pd/C^S reveals also the presence of dense highly dispersed small sized palladium particles, even more densely dispersed than the other examined catalysts (see ESI†). It is not likely that the differences between Pd/C^S and Pd/C^N are caused by different surface acidity since they both possess a week basic surface (see Section 3.2.1). Therefore, differences might derive from several factors, which are unknown for the commercial catalyst, such as the treatment of the active carbon, reduction temperature, differences in metal precursors etc.

3.2.3 Surface area and pores size measurements. The surface area (m² g⁻¹), pore radius (Å) and pore volume (cm³ g⁻¹) were determined for all activated carbon supported palladium catalysts. The results are summarized in Table 2.

Table 2 Summary of surface area, pore radius and pore volume of palladium on active carbon catalysts with different palladium loadings and different activated carbon type

Catalyst	Surface area (m^2 g^−1)	Pore radius (Å)	Pore volume (cm^3 g^−1)
0.5% Pd/C^N	860.409	15.536	0.170
1% Pd/C^N	659.473	24.742	0.177
3% Pd/C^N	645.811	24.709	0.168
5% Pd/C^N	614.820	24.764	0.149
0.5% Pd/C^D	1241.495	15.396	0.320
5% Pd/C^D	1149.588	17.291	0.290
5% Pd/C^S	847.629	24.768	0.257

---

In general, we can conclude from Table 2 that the acidic Darco supported catalysts have larger surface areas than the basic Norit supported catalysts. The higher tendency of palladium particles to agglomerate over the acidic Darco support, in comparison to Norit support, might contribute to the higher surface areas of Darco supported palladium catalysts as a result of lower occupancy of surface and pores by palladium particles. The lower surface and pores occupancy is also consistent with the larger pore volumes of Darco supported catalysts in comparison with the pore volumes of Norit supported catalysts. The surface area and pore volume values determined for 5% Pd/C^S are similar to the values determined for 0.5% Pd/C^N and 5% Pd/C^D, respectively. In addition, as the support is loaded with larger amount of palladium the surface area become smaller. The same phenomenon was observed for the pore volume, as palladium loading is higher the pore volume is smaller (although 1% Pd/C^N behaves exceptionally and it has larger pore volume than 0.5% Pd/C^N). These phenomena are reasonable since higher palladium loading will bring to higher surface coverage and higher occupancy of pore volume. The pore radius values also seem to be related to the palladium loading; in general as the palladium loading is higher the pore radius is larger (with the exception of 3% Pd/C^N, which shows a smaller pore radius than 1% Pd/C^N). The increase of pore radiuses in catalysts with higher palladium loadings might caused by increased number of particles occupying pores forcing the carbon pores to be broadened. Generally, Darco supported palladium catalysts show smaller pore radiuses than Norit supported palladium catalysts. 0.5% Pd/C^N and 0.5% Pd/C^D have the smallest pore radiuses, 15.536 Å and 15.396 Å, respectively. The commercial catalyst, 5% Pd/C^S, exhibits the largest pore radius, 24.768 Å.

3.3 Catalysts and reaction parameters

3.3.1 Effect of palladium loading. Palladium loadings of 0.5–5% over Norit support were tested in CMF and BMF hydrogenation. The results are presented in Fig. 4.

Fig. 4 Effect of palladium loading (%wt) on hydrogenation products yields from (a) BMF and from (b) CMF. All reactions were carried out with 2.075 mmol of reactant (0.300 g CMF or 0.392 g BMF), 0.030 g catalyst, 40 ml DCE, 11 bar H₂, reaction time of 5 h and temperature of 50 °C or 70 °C for BMF and CMF, respectively.

The results in Fig. 4(a) indicate that for BMF conversion to MF, as palladium loading over the carbon support is higher the yields of MF are higher. The maximal yield of MF, 93.3%, is obtained with 5% Pd/C^N catalyst. Then, BMF conversion is also about 93.3% and only traces of humic materials were formed. The correlation between catalyst loading and BMF conversion is reasonable since higher loadings bring to larger amounts of catalytic sites. In the case of CMF conversion, in Fig. 4(b), higher loadings also bring to higher CMF conversions (here too, the total yields of products are identical to conversions). However, as Norit support is loaded with larger amounts of palladium, the yields of the other products (mostly condensation products) are increased in place of MF yields. The reason for this phenomenon is not well understood, but several explanations derived from the morphological and textural characters of the supported catalysts are possible. First, selectivity and activity of catalysts can be influenced by particles sizes.^43,44 As we saw in Section 3.2.2, increasing palladium loading over Norit support result in larger agglomerated palladium particles. In larger palladium agglomerates, the active sites might be blocked for CMF hydrodehalogenation and the competitive slower condensation reactions can take place more rapidly. Another possible explanation might relate to the fact that larger palladium particles have higher activity towards hydrogenation of the aromatic ring.⁴⁵ Ring hydrogenation is required in order to enable condensation reactions. As observed by SEM images, in Fig. 3, higher palladium loadings lead to the formation of larger Pd particles. As a result, condensations reactions can proceed over catalysts with higher loadings over Norit support. A third hypothesis is related to the proximity of the dispersed Pd particles. The condensation of two halogenated furfurals molecules can take place only if they are in close proximity, as required in bimolecular reactions.⁴⁶ A more densely packaging of particles over support has been observed for higher palladium loadings. The close proximity of palladium particles enables the contact of reactants and the formation of condensation products. Oppositely, when particles are distant from each other in lower palladium loadings, as observed for 0.5% Pd/C^N, the contact of two halogenated furfural molecules is less likely. Finally, the texture of support might also influence on the occurrence of condensation reactions. BET results in Table 2 reveal that pore radiuses are larger for higher palladium loadings (although 3% Pd/C^N behaves exceptionally). As pore radius is larger, the accessibility of two reactants to the active catalytic site is more feasible. Therefore, condensation products could not be detected when 0.5% Pd/C^N was used since it has the smallest pore radius, 15.536 Å. It should be pointed that the formation of condensation products is reasonably influenced by two, or more, of the above factors simultaneously. It is also clear from Fig. 4 that condensation products were formed from CMF and not from BMF over Pd/C^N. Bromine atom is a better leaving group than chlorine atom and thus BMF reacts faster than CMF with hydrogen. Therefore, hydrogenation of BMF over the catalytic sites and MF desorption from the catalytic sites are too fast for the competitive condensation reactions to occur. On the other hand, in the case of CMF hydrogenation, the chlorine atom is substituted with hydrogen slowly enough for the competitive condensation reactions to occur.

3.3.2 Effect of catalyst type. Three types of palladium supported catalysts were examined in the hydrogenation of CMF and BMF: Darco supported Pd catalyst (Pd/C^D), Norit supported Pd catalyst (Pd/C^N) and commercial 5% Pd/C catalyst (Pd/C^S). The activities of Darco and Norit activated carbon supports were tested with palladium loadings of 0.5% and 5%. The results are summarized in Fig. 5.

Fig. 5 Effect of catalyst type on hydrogenation products yields from (a) BMF and from (b) CMF. All reactions were carried out with 2.075 mmol of reactant (0.300 g CMF or 0.392 g BMF), 0.030 g catalyst, 40 ml DCE, 11 bar H₂, reaction time of 5 h and temperature of 50 °C or 70 °C for BMF and CMF, respectively.

The results in Fig. 5(a) indicate that for palladium loading of 0.5%, the Darco support has a little higher activity than the basic Norit support towards BMF hydrogenation to MF. However, with higher palladium loading of 5%, Norit support is more active than Darco support towards the formation of MF from BMF, with MF yields of 93.3% and 77.6%, respectively. In the case of CMF hydrogenation, in Fig. 5(b), it is clear that Norit support has higher activity than Darco support with both lower and higher palladium loadings. A 99% conversion of CMF was recorded with 5% Pd/C^N catalyst while the CMF conversion obtained with 5% Pd/C^D catalyst was 92.8%. However, the acidic Darco support has higher selectivity towards MF than the basic Norit support. MF yield of 78.9% and other products yield (mostly condensation products) of 13.9% have been recorded with 5% Pd/C^D catalyst, while with 5% Pd/C^N catalyst only 42% yield of MF has been obtained and the other products yield was 56.5%. The difference in activity towards CMF hydrogenation was more markedly with palladium loading of 0.5%. Utilization of 0.5% Pd/C^N brought to full selectivity of MF under the specified reaction conditions with 72.8% yield of MF, while utilization of 0.5% Pd/C^D brought to 15.8% yield of MF and 1.2% yield of other compounds In general, Norit support result in more active catalysts than Darco support, probably as a result of the highly dispersed Pd particles over Norit support. In addition, it is clear from Fig. 5 that CMF is more prone for condensation over Norit supported catalysts than over Darco supported catalyst. The lower yields of condensation products formed over Darco supported catalysts might result from differences in supports acidities, differences in Pd dispersion over supports and different metal–support interactions. In addition, geometrical limitation imposed by the low pore radius of the Darco carbon support might also contribute to the lower yields of condensation products formed with Darco supported catalyst. The commercial catalyst, 5% Pd/C^S, shows high activity toward BMF and CMF conversion. It is the most active catalyst for BMF hydrogenation into MF, with 98.6% yield of MF. This catalyst is also highly active for CMF conversion. However, only to 6.2% yield of MF has been obtained with 5% Pd/C^S catalyst, while the rest of the products are mainly condensation products. The enhanced activity of this catalyst is probably derived from a combination of large pore radius and pore volume and close proximity of highly dispersed particles over support. It should be pointed that the differences in activities of the acidic Darco supported catalysts and the basic Norit supported catalysts, as well as the commercial catalyst, might also derive from the hydrophilic or hydrophobic character of the supports surfaces.⁴⁷ Catalysts with acidic surface are more hydrophilic than the catalysts with the basic surface and therefore wetting of the catalyst acidic surface with the organic reactant is not as efficient as wetting of the basic catalysts.

In conclusion 0.03 g of 5% Pd/C^S shows the best performance in selective BMF hydrogenation to MF. On the other hand, 0.5% Pd/C^N seems to be the most selective catalyst towards MF formation from CMF, but a larger quantity of 0.1 g catalyst is needed to achieve full conversion of CMF.

3.3.3 Effect of reaction time. Reduction of reaction time has a great significance for industrial application. The optimal time of reaction was determined for the formation of MF from CMF and BMF. The results are presented in Fig. 6.

Fig. 6 MF yields as a function of reaction time (h) from CMF and from BMF. All reactions were carried out with 2.075 mmol of reactant (0.300 g CMF or 0.392 g BMF), 40 ml DCE and 11 bar H₂. For BMF hydrogenation temperature of 50 °C and 0.030 g 5% Pd/C^S catalyst were used. For CMF hydrogenation temperature of 70° and 0.100 g 0.5% Pd/C^N catalyst were used.

The optimal reaction time for the conversion of BMF into MF is 4 h, and then MF yield is 98.6%. In the case of CMF conversion into MF the optimal reaction is 5 h, and then MF yield is 99.2%. In both cases 100% conversion was recorded and traces of humic materials were formed. The reaction time of BMF is shorter than the reaction time of CMF since bromine atom is a better leaving group than chlorine atom and it will be replaced with hydrogen atom more easily (and thus more rapidly). It should be pointed that shorter and longer reaction time lead to lower MF yields. In longer reaction times MF yields are lower as a result of increased condensation products formation, this time from MF. The condensation products were identified to be the same as detailed in previous sections as confirmed by GC analysis.

It should also be pointed that the effects of the hydrogen pressure, 1–20 bar, and temperature, 60–100 °C, were also examined. However, these parameters have no impact or a negative impact on MF yields. Investigating the effect of hydrogen pressure revealed that higher hydrogen pressures did not bring to subsequent hydrogenation of MF, since the aromatic bonds or the aldehyde group were not hydrogenated under the reaction conditions. However, lower MF yields have been obtained with lower hydrogen pressures. Temperature examination showed that lower MF yields have been obtained when higher and lower temperatures were applied. Higher temperatures bring to larger amount of condensation products, and even increased formation of humic materials, and lower temperatures bring to lower MF yields as a result of insufficient energy to promote reaction.

Therefore, the optimal hydrogen pressure for BMF and CMF conversion to MF is 11 bar and the optimal temperatures of reaction are 50 °C and 70 °C for BMF and CMF conversion to MF, respectively.

3.4 Catalyst and acid recycling

The most active catalysts in CMF and BMF hydrogenation, 0.5% Pd/C^N and 5% Pd/C^S, were tested in sequential reaction cycles. The catalysts were separated after each cycle and returned into the reactor for another cycle without any further treatment. The catalytic performance of the recovered catalysts in CMF and BMF hydrogenation is presented in Fig. 7.

Fig. 7 MF yields from CMF and BMF hydrogenation in four sequential cycles. All reactions were carried out with 2.075 mmol of reactant (0.300 g CMF or 0.392 g BMF), 40 ml DCE, 11 bar H₂. For BMF hydrogenation a reaction time of 4 h, temperature of 50 °C and 0.030 g 5% Pd/C^S were used. For CMF hydrogenation a reaction time of 5 h, temperature of 70 °C and 0.100 g 0.5% Pd/C^N were used.

It can be observed from Fig. 7 that in the second and third cycles of both catalysts MF yields are slightly decreasing. In the case of CMF hydrogenation, MF yields of 99.2%, 97.3% and 94% have been obtained in the first, second and the third cycles, respectively. However, in the fourth repetition of reaction with 0.5% Pd/C^N catalyst a substantial lost of activity was observed and the yield of MF was dropped to 61.6%. The same phenomenon has been observed in BMF hydrogenation with 5% Pd/C^S catalyst. MF yields of 98.6%, 98.5% and 91.52% were obtained in the first, second and third catalytic cycles, respectively. However, in the fourth cycle of the recovered catalyst, a significant lost of activity has been observed and a yield of 63.6% of MF was recorded. The trend of activity decrease must be a consequence of a slow deactivation of catalytic sites followed by a rapid deactivation that proceed when a critical point had been reached. Several factors might contribute to activity loss. Sintering of palladium particles and loss of surface area is a known factor to promote deactivation. This phenomenon occurs more markedly with activated carbon supported palladium catalysts than metal oxide supported palladium catalysts as a result of weak metal–support interactions.⁴⁸ However, deactivation of Pd/C as a result of particles sintering is driven from the presence of hydrogen and elevated temperatures. In our study, at low temperatures, palladium particles are stable to sintering, as supported by SEM analyses performed on catalysts after reaction (see ESI†). The stability of Pd particles on carbon supports in low temperature hydrodechlorination was also reported by Janiak and Okal.⁴⁹ Leaching of palladium from supports might explain deactivation and thus leaching tests for samples collected from the three first cycles of both 5% Pd/C^S in BMF hydrogenation and 0.5% Pd/C^N in CMF hydrogenation have been performed. For 5% Pd/C^S catalyst, a palladium leaching of 0.216 ppm, which corresponds to 0.144% of the palladium content, was determined in the first reaction cycle. In the second and third cycles no palladium content could be detected in samples and therefore we can conclude that or no leaching of palladium did occur or the palladium content in samples was under the detection limit of the ICP instrument (and thus leaching was taking place only in very small amounts). The results for 0.5% Pd/C^N in CMF hydrogenation showed similar behavior. At the first cycle palladium leaching of 0.04 ppm could be detected (inaccurate result because of the value's closeness to the detection limit) but in second and third cycles no palladium could be detected in samples.

Another plausible reason for the catalysts activity lost in CMF and BMF hydrogenation is the formation of coke on the catalytic sites from traces of humic materials formed during reaction. Humic materials can adsorb and accumulate on the catalytic sites after each cycle until a critical amount of coke form over the catalytic sites and MF yields drop sharply. Additionally, acid by-products can also adsorb and accumulate on the catalytic sites and inhibit halogenated furfurals hydrogenation. In order to recover the catalysts, they were washed 3 times with 1 M NaOH solutions followed by washing twice with deionized water (in addition to washing 3 times with DCE). The basic solution can remove humic materials (especially if humic acids were formed) and HCl or HBr and keep the catalytic sites free and available for the following cycles. Catalytic performance results of 5% Pd/C^S and 0.5% Pd/C^N catalysts in four sequential cycles of CMF and BMF hydrogenation to MF when the catalysts were treated with basic solution after each cycle are presented in Fig. 8.

Fig. 8 MF yields from CMF and BMF hydrogenation in four sequential cycles when catalysts treated with 1 M NaOH solution after each cycle. All reactions were carried out with 2.075 mmol of reactant (0.300 g CMF or 0.392 g BMF), 40 ml DCE, 11 bar H₂. For BMF hydrogenation a reaction time of 4 h, temperature of 50 °C and 0.030 g 5% Pd/C^S were used. For CMF hydrogenation a reaction time of 5 h, temperature of 70 °C and 0.100 g 0.5% Pd/C^N were used.

We can conclude from Fig. 8 that after the catalysts were washed after each cycle with 1 M NaOH solution their activities in the fourth reaction cycles were improved compared to the untreated catalysts. Now in the fourth cycles of the catalysts, MF yields of 83.8% (instead of 63.4% without treatment) and 81.6% (instead of 61.6% without treatment) were recorder with 5% Pd/C^S and 0.5% Pd/C^N catalysts, respectively.

CMF and BMF are formed by treatment of carbohydrates with concentrated hydrochloric or hydrobromic acid, respectively. In CMF and BMF hydrogenation HCl and HBr are regenerated as side products, and their recycling (and reusing in halogenated furfurals preparation) is economically significant for the overall process. Therefore, experiments of HCl and HBr recovery have been carried out as described in Section 2.3. For HBr, 5.1 ml of 0.1 M NaOH solution was needed for titration of the acid collected in the water tube, and calculation shows that it equals to 24.57% of the theoretical hydrobromic acid content. In the case of HCl, 10 ml of 0.1 M NaOH was needed for titration, that imply on recovery of 48.07% of the hydrochloric acid content.

4. Conclusions

In this work the catalytic hydrogenation of biomass derived halogenated furfurals, CMF and BMF, into 5-methyl furfural (MF) over active carbon supported palladium catalysts is demonstrated. MF itself can be used as fuel additive or fuel candidate, or it can be further convert into other valuable products. It should be pointed that in the current technological status, the utilization of MF as a fuel additive is more likely. Active carbon supported palladium catalysts are the most active catalysts towards MF formation via CMF or BMF hydrogenation, and MF yields of 99.2% and 98.6% can be achieved from CMF and BMF, respectively. Several types of Pd/C catalysts were examined in this research: acidic type Darco supported catalyst, basic type Norit supported catalyst and commercial Pd/C catalyst. It turns out that the catalysts characters, including their palladium loading and the support type can determine not only MF yield but also the selectivity towards MF production compared to other products, identified mostly as condensation products, which can form. The most active and selective catalysts towards MF formation from BMF and CMF are 5% Pd/C^S and 0.5% Pd/C^N, respectively. In addition to the catalytic parameters, the optimal reaction parameters were also determined. Finally, the catalyst and the acid by-product (HCl or HBr) recovery were examined. The catalysts performance in four reaction cycles was demonstrated, and they showed a high deactivation in the fourth cycle. The catalysts could be partially rejuvenated by treatment with 1 M sodium hydroxide solution.

Acknowledgements

We thank the Israel Strategic Alternative Energy Fund (I-SAEF) for financial support of this research. Special thanks for supporting in part the Ph.D. scholarship of E. M. We also like to thanks Ms Evgenia Blayvas from the unit for nanoscopic characterization at The Hebrew University Center For Nanoscience And Nanotechnology for her contribution in the catalyst characterization using SEM.

References

1. F. Cherubini, Energy Convers. Manage., 2010, 51, 1412–1421.
2. U.S. Energy Information Administration, use of energy in the united states explained, http://www.eia.gov/Energyexplained/?page=us_energy_transportation, accessed August 2016.
3. G. W. Huber and A. Corma, Angew. Chem., Int. Ed., 2007, 46, 7184–7201.
4. D. M. Alonso, J. Q. Bond and J. A. Dumesic, Green Chem., 2010, 12, 1493–1513.
5. K. Narasimharo, A. Lee and K. Wilson, J. Biobased Mater. Bioenergy, 2007, 1, 19–30.
6. F. Ma and M. A. Hanna, Bioresour. Technol., 1999, 70, 1–15.
7. M. Balat, H. Balat and C. Oz, Prog. Energy Combust. Sci., 2008, 34, 551–573.
8. C. N. Hamelinck, G. van Hooijdonk and A. P. C. Faaij, Biomass Bioenergy, 2005, 28, 384–410.
9. S. N. Naik, V. V. Goud, P. K. Rout and A. K. Dalai, Renewable Sustainable Energy Rev., 2010, 14, 578–597.
10. S. Lestari, I. Simakova, A. Tokarev, P. Maki-Arvela, K. Eranen and D. Y. Murzin, Catal. Lett., 2008, 122, 247–251.
11. R. A. Sheldon, I. Arends and U. Hanefeld, Green Chemistry and Catalysis, Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, 2007.
12. G. Avgouropoulos and T. Tabakova, Environmental Catalysis Over Gold-Based Materials, Royal Society of Chemistry, Cambridge, 2013.
13. H. Kobayashi, H. Ohta and A. Fukuoka, Catal. Sci. Technol., 2012, 2, 869–883.
14. D. M. Alonso, J. Q. Bond and J. A. Dumesic, Green Chem., 2010, 12, 1493–1513.
15. O. O. James, S. Maity, L. A. Usman, K. O. Ajanaku, O. O. Ajani, T. O. Siyanbola, S. Sahu and R. Chaubey, Energy Environ. Sci., 2010, 3, 1833–1850.
16. J. N. Chheda, G. W. Huber and J. A. Dumesic, Angew. Chem., Int. Ed., 2007, 46, 7164–7183.
17. J. P. Lange, E. van der Heide, J. van Buijtenen and R. Price, ChemSusChem, 2012, 5, 150–166.
18. R. J. van Putten, J. C. van der Waal, E. de Jong, C. B. Rasrendra, H. J. Heeres and J. G. de Vries, Chem. Rev., 2013, 113, 1499–1597.
19. T. A. Werpy and G. Petersen, Top value added chemicals from biomass, U.S. department of Energy, Pacific Northwest National Laboratory, 2004.
20. J. J. Bozell and G. R. Petersen, Green Chem., 2010, 12, 539–554.
21. Y. Sun and J. Cheng, Bioresour. Technol., 2002, 83, 1–11.
22. F. W. Bai, W. A. Anderson and M. Moo-Young, Biotechnol. Adv., 2008, 26, 89–105.
23. J. N. Chheda, Y. Roman-Leshkov and J. A. Dumesic, Green Chem., 2007, 9, 342–350.
24. K. Yan, G. Wu, T. Lafleur and C. Jarvis, Renewable Sustainable Energy Rev., 2014, 38, 663–676.
25. Y. Nakagawa, M. Tamura and K. Tomishige, ACS Catal., 2013, 3, 2655–2668.
26. K. Yan and A. Chen, Energy, 2013, 58, 357–363.
27. K. Yan and A. Chen, Fuel, 2014, 115, 101–108.
28. L. Baijun, L. Lianhai, W. Bingchun, C. Tianxi and K. Iwatani, Appl. Catal., A, 1998, 171, 117–122.
29. W. Yang and A. Sen, ChemSusChem, 2010, 3, 597–603.
30. Y. Zhao, Environ. Chem. Lett., 2014, 12, 185–190.
31. Y. Nakagawa, H. Nakazawa, H. Watanabe and K. Tomishige, ChemCatChem, 2012, 4, 1791–1797.
32. Y. Roman-Leshkov, C. J. Barrett, Z. Y. Liu and J. A. Dumesic, Nature, 2007, 441, 982–986.
33. D. Liu and E. Y. X. Chen, ChemSusChem, 2013, 6, 2236–2239.
34. Y. B. Huang, Z. Yang, J. J. Dai, Q. X. Guo and Y. Fu, RSC Adv., 2012, 2, 11211–11214.
35. M. J. Antal, W. S. L. Mok and G. N. Richards, Carbohydr. Res., 1990, 199, 91–109.
36. Y. J. Pagan-Torres, T. Wang, J. M. R. Gallo, B. H. Shanks and J. A. Dumesic, ACS Catal., 2012, 2, 930–934.
37. M. Mascal and E. B. Nikitin, Angew. Chem., 2008, 120, 8042–8044.
38. N. Kumari, J. K. Olesen, C. M. Pedersen and M. Bols, Eur. J. Org. Chem., 2011, 7, 1266–1270.
39. E. Meller, A. Aviv, Z. Aizenshtat and Y. Sasson, RSC Adv., 2016, 6, 36069–36076.
40. P. Sugumaran, V. P. Susan, P. Ravichandran and S. Seshadri, Journal of Sustainable Energy & Environment, 2012, 3, 125–132.
41. J. Li, L. Ma, X. Li, C. Lu and H. Liu, Ind. Eng. Chem. Res., 2005, 44, 5478–5482.
42. V. Z. Radkevich, T. L. Senko, K. Wilson, L. M. Grishenko, A. N. Zaderko and V. Y. Diyuk, Appl. Catal., A, 2008, 335, 241–251.
43. T. Harada, S. Ikeda, M. Miyazaki, T. Sakata, H. Mori and M. Matsumura, J. Mol. Catal. A: Chem., 2007, 268, 59–64.
44. C. Corolleur, F. G. Gault, D. Juttard, G. Maire and J. M. Muller, J. Catal., 1972, 27, 466–468.
45. D. Scholz, C. Aellig and I. Hermans, ChemSusChem, 2014, 7, 268–275.
46. J. Dwyer, K. Karim and A. F. Ojo, J. Chem. Soc., Faraday Trans., 1991, 87, 783–786.
47. D. J. Suh, T. J. Park and S. K. Ihm, Ind. Eng. Chem. Res., 1992, 31, 1849–1856.
48. P. H. Jen, Y. H. Hsu and S. D. Lin, Catal. Today, 2007, 123, 133–141.
49. T. Janiak and J. Okal, Appl. Catal., B, 2009, 92, 384–392.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra21472j

---