Energy crises, depletion of fossil fuel reserves, and global warming have increased interest in the production of biobased materials [1]. In confronting these challenges, biorefined materials have emerged as an alternative to industrial carbon. In biorefining, lignocellulosic biomass is transformed into value-added chemicals and fuels [2]. The hydrogenation/hydrogenolysis of furans such as furfural (FAL) and 5-hydroxymethylfurfural (HMF) has been widely studied for the production of biobased fuels such as 2-methylfuran (2-MF), 2-methyltetrahydrofuran, pentane, and 2, 5-dimethylfuran (DMF) [3-6]. HMF is a valuable chemical because it can be converted to a variety of chemicals such as 2, 5-bis(hydroxymethyl)furan (BHMF), 2-hydroxymethyl-5-methylfuran (5-MFOL), DMF, and 2, 5-dimethyltetrahydrofuran (DMTHF) (Scheme 1) [7-9]. DMF is an important compound because it has potential use in gasoline blends and can be converted to p-xylene through a Diels-Alder reaction with ethylene [10-12].

Precious metals, i.e. Pt, Pd, Rh, Ru, and Au, are the main metals with potential uses in the hydrogenation of HMF to DMF; Pd and Ru have been reported to give the best selectivities for DMF [9, 13-17]. The first reported catalyst for hydrogenation of HMF to DMF was CuRu/C, which gave 71% yields of DMF from two different starting materials [9, 18]. It was reported that the activity of this catalyst gradually decreased because of deactivation of active centres and/or the presence of chloride ions [9]. Thananatthanachon and Rauchfuss reported the conversion of HMF to DMF using Pd/C as a catalyst and formic acid as an additive [8]. An excellent yield, i.e. 93%, with 99% HMF conversion was achieved, but only when formic acid and H₂SO₄ were used. However, because of their highly corrosive nature and environmental concerns, the use of both formic acid and H₂SO₄ is restricted. Luijkx et al. [19] used a combination of Pd/C and HCl to reduce HMF to DMF; The results were similar to those reported by Rauchfuss et al. [8]. Zu et al. [15] used Ru/Co₃O₄ as a catalyst for the selective conversion of HMF to DMF at 130 ℃ and 0.7 MPa; they clearly identified the roles of Ru and CoO_x, which are responsible for reduction and hydrogenolysis of carbonyl and hydroxyl groups, respectively.

Because of the high cost of noble metals, research on precious-metal-free catalysts which can effectively convert HMF to DMF is needed. Yang et al. [20] obtained a 76% yield of DMF over Ni/Co₃O₄ catalysts; the roles of Ni and CoO_x were similar to those reported for Ru and CoO_x by Zu et al. [15]. Cu/ZnO, Cu-Zn alloy, Cu/MgO/Al₂O₃, Ru-modified Cu/MgO/Al₂O₃, and Cu-PMO catalysts have been examined in the hydrogenation of HMF to DHMF and DMF [21-23]. However, the development of low-cost and environmentally benign robust catalytic systems for this reaction is still in progress.

Recently, supported bimetallic Cu-Co nanoclusters have emerged as versatile catalytic systems for hydrogenation/hydrogenolysis reactions because of the presence of Cu(0), CoO_x species, and mixed oxide phases [24-34]. It has been reported that early-transition-metal oxides such as CoO_x, ReO_x, and MoO_x can effectively split C-O and O-H bonds [15, 19, 35, 36]. A combination of Cu metal and CoO_x could therefore be effective in the production of DMF by hydrogenation/hydrogenolysis of HMF. In our previous work, we hydrogenated FAL to 2-MF over bimetallic Cu-Co catalysts supported on SiO₂, H-ZSM-5, and Al₂O₃ [29]. Cu-Co/Al₂O₃ was the best hydrogenolysis catalyst because of the presence of small Cu particles and Cu-Cox mixed phases, and synergies among Cu, Co, and Al₂O₃.

The present work was undertaken based on the background described above. In this study, instead of using SiO₂ and H-ZSM-5 as supports, materials which stabilize metal dispersion and improve oxygen storage (e.g. CeO₂ and ZrO₂) were examined and compared with Al₂O₃ as supports for Cu-Co bimetallic catalysts. The structures of these synthesized catalysts were studied using X-ray diffraction (XRD), the Brunauer-Emmett-Teller (BET) method, transmission electron microscopy (TEM), temperature-programmed desorption of NH₃ (NH₃-TPD), temperature-programmed reduction by H₂ (H₂-TPR), and CHNS analysis. The activities and selectivities of these catalysts in the hydrogenolysis of HMF to DMF were investigated. Cu-Co/Al₂O₃ catalysts were best for the hydrogenolysis of biomass-derived HMF to DMF; DMF was selectively obtained in 78% yield under mild conditions. The HMF conversion and DMF selectivity were optimized based on the reaction pathways and effects of the reaction parameters, namely temperature, H₂ pressure, and time.

Tetrahydrofuran (THF), HMF, and all the products shown in Scheme 1 (all GC grade) were purchased from Sigma-Aldrich, Mumbai, India. The supports, i.e. CeO₂, ZrO₂, and Al₂O₃ (purity 99%), were purchased from Sterling Chemical Pvt., Ltd., Surat, India. Bimetallic Cu-Co catalysts with Cu:Co molar ratios of 1, 2, and 4 were synthesized on CeO₂, ZrO₂, and Al₂O₃ at fixed Cu loadings of 10 wt% by impregnation with aqueous solutions of Co(NO₃)₂·6H₂O and Cu(NO₃)₂·6H₂O. Details of the synthesis and characterization of the supported bimetallic Cu-Co catalysts are available elsewhere [29].

Hydrogenation of HMF was performed in a 100-mL autoclave reactor (Amar Equipment Pvt., Ltd., Mumbai, India). Briefly, the catalyst (0.5 g) was reduced at 280 ± 2 ℃ in a H₂ flow at 1 MPa for 3 h. The reactor was cooled to room temperature and flushed with N₂. The reactor was charged with HMF (2.5 mmol) in THF (20 mL). The catalytic activities of the prepared catalysts were investigated at 160-230 ℃ under a H₂ pressure of 3.0 MPa with stirring at 1000 r/min. The reproducibility of each reaction was checked by repeating it twice. Product samples were analysed using a Sigma GC system with a flame ionization detector (FID). An AB-5 capillary column of dimensions 30 m × 0.25 mm × 0.5 μm was used. In a typical procedure, the initial oven temperature was held at 50 ℃ for 2 min and increased to 300 ℃ at a ramping rate of 10 ℃/min and held for 2 min. The product suspension (1 μL) was injected into the capillary column using a 20:1 split ratio, with N₂ as the carrier gas. The carbon balance was calculated in each run by measuring the carbon in the liquid phase using GC-FID; it was found to be >92% in each run because of unidentified products. The unidentified products were quantified by subtracting the overall carbon balance at each step from 100%. The following equation was used to determine the amount of carbon in the reaction mixture: carbon balance (%) = moles of carbon after reaction (reactant and products)/moles of carbon before reaction (reactant) x 100.

The effects of the support on the crystalline phases of the bimetallic Cu-Co (Cu/Co = 1) catalysts were investigated using XRD. The XRD patterns of the calcined and reduced catalysts are shown in Figs. 1-3. The XRD patterns of the ZrO₂ and CeO₂ catalysts show the presence of face-centred cubic (fcc) ZrO₂ and (fcc) CeO₂ crystalline structures (Fig. 1), based on the JCPDS files (JCPDS 83-0936 and JCPDS 81-0792) [37]. However, no (fcc) structure was observed for Al₂O₃; only a weak diffraction peak at 2θ = 67° indicated the presence of Al₂O₃. To gain further insights into the effects of the supports on active site formation in the bimetallic catalysts, we examined the XRD patterns of monometallic Cu (10 wt%) and Co (10 wt%) on all three supports and compared them with the patterns of the bimetallic counterparts (Fig. 1). The intensities of the diffraction peaks from all three supports decreased after doping with 10 wt% Cu and 10 wt% Co in the monometallic catalysts. The intensities decreased further after doping simultaneously with Cu and Co. This implies that the crystallinities of the supports decreased as a result of high metal loadings. The diffraction patterns of the monometallic Cu and Co catalysts over all three supports showed diffraction peaks from CuO at 2θ = 35.5° and 38.7°, which are assigned to the (111) and (111) planes of monoclinic copper oxide (JCPDS 80-1917 and 45-0937), and from Co₃O₄at 2θ = 37°, 44°, and 56° (JCPDS 43-1003) [28, 29].

Specific XRD patterns were observed for doped Cu-Co over all three supports (Fig. 1). For the calcined catalysts, either segregated CuO or a spinel oxide consisting of Co₃O₄/CuCo₂O₄ was clearly observed. Because of their similar cubic frameworks and unit cell parameters (Co₃O₄, a = 8.177 Å and CuCo₂O₄, a = 8.122 Å), Co₃O₄ and CuCo₂O₄ are not easily distinguishable based on their XRD patterns [29]. The Cu-Co/CeO₂ and Cu-Co/ZrO₂ catalysts showed weak but clear diffraction peaks at 2θ = 35.5° and 38.7°, which are assigned to the (111) and (111) planes of monoclinic copper oxide (JCPDS 80-1917 and 45-0937), and diffraction peaks attributed to Co₃O₄ at 2θ = 37°, 44°, and 56° (JCPDS 43-1003). For Cu-Co/Al₂O₃, the weak diffraction peaks at 2θ = 35.5° and 38.7° are attributed to highly dispersed Cu on porous Al₂O₃ [28, 29]. The full-widths at half-maximum (FWHM) of the diffraction peak at 2θ = 37°, which is attributed to CuCoO₄, increased in the order CeO₂ < ZrO₂ < Al₂O₃; this is because Cu-Co/Al₂O₃ has the smallest crystals.

In the case of the reduced catalysts (Fig. 2), Cu-Co/ZrO₂ showed weak diffraction peaks from (fcc) Cu metal and Cu₂O. However, diffraction peaks from (fcc) Cu₂O were not observed for the other two catalysts, i.e., Cu-Co/CeO₂ and Cu-Co/Al₂O₃; they only displayed diffraction peaks from (fcc) Cu metal. The FWHMs of the strongest peak, at 2θ = 44.3°, decreased in the order of CeO₂ < ZrO₂ < Al₂O₃. This confirms that the metal particles in Cu-Co/Al₂O₃ are smaller than those in the other two catalysts. The sizes of the Cu metal crystallites in the reduced catalysts were calculated using the Scherrer formula, based on the Cu (111) peak at 2θ = 44.3°. The CeO₂-supported catalyst had the largest Cu particles, of average size 25 nm; the average Cu particle size for the Al₂O₃-supported catalyst was 15 nm. The differences among the phases and crystallite sizes obtained with different supports can be attributed to the nature of the supports and/or metal-support interactions. The oxide supports used in this study interact weakly with Cu and Co, and the order of the interaction strengths is CeO₂ > ZrO₂ > Al₂O₃ [38]. The stronger interactions between both the metals and CeO₂ produced large segregated CuO and Co₃O₄ phases, whereas the strong interactions between copper and cobalt oxides and their weaker interactions with Al₂O₃ promoted formation of mixed copper-cobalt oxides. To understand the synergy between Cu and Co over Al₂O₃, the reduced bimetallic Cu-Co (Cu/Co = 1) catalyst was compared with its monometallic counterparts, i.e. Cu (10 wt%)/Al₂O₃ and Co (10 wt%)/Al₂O₃ (reduced at 280 ℃), as shown in Fig. 3. The monometallic Cu catalyst contained large Cu crystals but the monometallic Co catalyst contained both oxide and metallic forms because of the reducing conditions used in this study. Monometallic Co₃O₄ is reduced at or above 450 ℃. In contrast to the monometallic catalysts, bimetallic Cu-Co/Al₂O₃ (Cu/Co = 1) gave diffraction peaks from Cu and/or Co metal and partially reduced CoO_x. The low intensities of the diffraction peaks from Cu metal suggest that co-doping of Co and Cu on porous Al₂O₃ significantly decreased the size of the metal crystallites. The particle size of monometallic Cu, based on the peak at 44.3°, was 25-30 nm. However, as reported earlier, the corresponding size was 15 nm for bimetallic Cu-Co/Al₂O₃ (Cu/Co = 1). These observations are in good agreement with previously reported results for Cu-Co catalysts [24-34].

The textural properties of the supports and the Cu-Co catalysts supported by CeO₂, ZrO₂, and Al₂O₃ are summarized in Table 1. Because of its ordered porous structure, Al₂O₃ had the highest BET surface area, i.e. 166 m²/g, and a pore volume of 0.54 cm³/g. CeO₂ had the lowest BET surface area, (112 m²/g) and pore volume, (0.28 cm³/g). The decreases in the surface area, pore volume, and pore size of CeO₂ after doping with Cu and Co were smaller than those for the other two supports; Al₂O₃ showed the largest reductions in surface area, pore volume, and pore size. The differences among the textural properties of the synthesized catalysts can be attributed to the nature of the supports and their interactions with both metals. The small decline in the pore volume and pore size of CeO₂ after doping could be the result of deposition of metal particles on the external surface of the support rather than in the pores. However, the Al₂O₃-supported catalyst showed large reductions in pore volume and pore size; this can be attributed to deposition of metal particles on the surface and in the pores. It can be assumed that CeO₂ can stabilize metal particles on the surface because it has strong interactions with both the metals used in this study. However, compared with CeO₂, Al₂O₃ has weak interaction with Cu and Co, and this may result in highly dispersed metal particles. These results are in agreement with those reported by Wang et al. [39], who suggested that strong interactions between Cu and Co and weaker interactions with Al₂O₃ lead to highly dispersed metal particles on porous Al₂O₃. This may result in a high surface area compared with those of the other two catalysts.

TEM images of freshly reduced Cu-Co (Cu/Co = 1) on different supports are shown in Fig. 4. The dispersion and morphology/ordering of the Cu-Co nanoparticles on different supports vary depending on the nature of the support and the discrete interactions between the support and both Cu and Co. Agglomerated large metal crystallites (25-30 nm) were observed in the CeO₂-supported catalyst because of the strong interactions between CeO₂ and the metal particles. Similar-sized but segregated, non-uniformly distributed Cu-Co nanoparticles were observed on ZrO₂. Cu-Co/Al₂O₃ had well dispersed, ordered, uniformly spherical small nanoparticles. The metal particles in Cu-Co/Al₂O₃ were mainly of uniform size about 20 nm. This morphology/ordering of Cu-Co nanoparticles on CeO₂, ZrO₂, and Al₂O₃ can be explained based on a combination of a porous structure and metal-support interactions.

The more uniform distributions of Cu-Co particles on the Al₂O₃ support can be ascribed to the interactions between uniformly porous Al₂O₃ and both Cu and Co species being weaker than the metal-support interactions in the other cases [29].

The reductive properties of Cu-Co/CeO₂, Cu-Co/ZrO₂, and Cu-Co/Al₂O₃ were investigated using H₂-TPR. Metal-support interactions play a vital role in stabilizing and reducing metal oxides to metallic sites. In particular, for bimetallic catalytic systems, the interactions between each metal component and the support significantly affects the chemical environment of the metals and/or the reduction temperature. Weakly interacting metal oxide species are easily reduced to the metallic state [38]. Conversely, if the metal oxides are strongly adsorbed on the support, they are difficult to reduce, and this leads to incomplete reduction and formation of low-valence metals [37, 38]. The reductive properties of the Cu-Co (Cu/Co = 1) catalysts supported on CeO₂ and ZrO₂ were determined using H₂-TPR and the profiles were compared with that of Cu-Co/Al₂O₃ [29] (Fig. 5). The nature of the support significantly affected the reducibility of the bimetallic catalysts. The differences among the reduction temperatures showed that the CeO₂ (broad hump at 220-370 ℃) support had stronger interaction with both metals. Al₂O₃ and ZrO₂ showed broad humps which were deconvoluted to several peaks. The TPR profile of the ZrO₂-supported catalyst showed three major reduction peaks. The first peak, at 200 ℃, is attributed to the reduction of small CuO particles, which interact weakly with ZrO₂. The second peak, at 250-370 ℃, is ascribed to the reduction of larger CuO particles, which are strongly attached to ZrO₂. The third peak, at around 400 ℃, is attributed to the reduction of Co₃O₄ species on the ZrO₂ surface.

The TPR profile for Cu-Co/Al₂O₃ was different from those for the other two catalysts, and reduction occurred at a lower temperature; this is ascribed to the reduction of mixed copper-cobalt oxides and small Cu particles. These observations are supported by the XRD and TEM results. The presence of multiple reduction peaks indicates two-step reduction of Cu and partial reduction of CuCo₂O₄/Co₃O₄ [16, 35]. Among the three supports, Al₂O₃ has the weakest interactions with the copper-cobalt oxide phases and this enhances the reducibility of the copper oxides and mixed CuCo₂O₄, which increases the number of Cu-CoO_x sites in Cu-Co/Al₂O₃. To clarify the effect of synergy between Cu and Co on the reductive properties, the reduction profiles of monometallic Cu (10 wt%)/Al₂O₃ and Co (10 wt%)/Al₂O₃ were also obtained (Fig. 5), and compared with those of their bimetallic counterparts. The differences among these reductive behaviours were discussed at length in our previous work [29]. Table 2 shows the experimental H₂ consumptions during TPR. It can be deduced that Co and Cu are both essentially completely reduced by 400 ℃ on all three supports, with Cu-Co/CeO₂ being slightly less reduced (82%) compared with the other two (88% and 96%).

The surface acidities of the bimetallic Cu-Co (Cu/Co = 1) catalysts supported on CeO₂ and ZrO₂ were determined using NH₃-TPD and the results were compared with those for Cu-Co/Al₂O₃ (Fig. 6 and Table 3). Generally, the strengths of the acidic sites, based on the NH₃ desorption temperature, can be classified as weak (< 250 ℃), medium (250-400 ℃), and strong (> 400 ℃) acidic sites. The overall acidity of a catalyst can be calculated from the relative peak areas in the NH₃-desorption curves [29]. Cu-Co/CeO₂ (Cu/Co = 1) had few weak and strong acidic sites, and had only medium acidity. Cu-Co/Al₂O₃ (Cu/Co = 1) showed three peaks (at 145, 335, and 555 ℃), which can be ascribed to NH₃ desorption from weak, medium, and strong acidic sites. Cu-Co/ZrO₂ (Cu/Co = 1) showed two peaks (minor 100-250 ℃, major 350-550 ℃), which can be ascribed to NH₃ desorption by weak and strong acidic sites (Fig. 6 and Table 3). A comparison of the results for the three catalysts shows that Cu-Co/Al₂O₃ (Cu/Co = 1) had the maximum acidity, i.e. 0.49 mmol/g cat. The order of the total acidities was Al₂O₃ > ZrO₂ > CeO₂. It is worth noting that five different types of OH groups are present on the surface of γ-Al₂O₃, giving a net electric charge (σ), depending on the number of Al neighbours and Al coordination [40, 41]. The removal of OH groups during high-temperature treatment (calcination) creates unsaturated surface cations, with mainly tetrahedrally[Al(IV)] and octahedrally[Al(VI)] coordinated Al [42, 43]. Both Lewis and Brnsted acidic sites may therefore be present on Al₂O_3, depending on the degree of dehydration.

We previously reported a liquid-phase hydrogenation of FAL to 2-MF using a highly efficient non-noble metal, environmentally benign, bimetallic Cu-Co catalyst [29]. The advantage of this catalytic system, which was prepared using an impregnation method, over other Cu-based catalysts has been thoroughly discussed. In our previous work, we found that SiO₂ and H-ZSM-5 supports gave low selectivities for C-O hydrogenolysis. In this study, we therefore used materials which can stabilize metal dispersion and improve oxygen storage, i.e. CeO₂ and ZrO_2, as supports for Cu-Co (Cu/Co = 1) and compared the performance of these catalysts with that of Cu-Co/Al₂O₃ (Cu/Co = 1) in the hydrogenation/hydrogenolysis of HMF to DMF at 200 ℃ and 3.0 MPa (Fig. 7). The support plays a significant role in the selective hydrogenation/hydrogenolysis of HMF to DMF. Cu-Co/CeO₂ showed the lowest catalytic activity (78% conversion) and Cu-Co/Al₂O₃ showed the highest catalytic activity (98% conversion) in HMF hydrogenation. The order of the catalytic activities was CeO₂ < ZrO₂ < Al₂O₃. HMF contains three functional groups, namely C=O, O-H, and the furan ring (Scheme 1). Over Cu catalysts, HMF hydrogenation proceeds by reduction of the C=O bond to give the diol intermediate BHMF. Further hydrogenolysis of BHMF leads to the formation of DMF via another intermediate, i.e. 5-MFOL [20-23]. Nagaraja et al. [44] proposed that the presence of a larger amount of metallic Cu than of CuO promoted reduction of the C=O bond in FAL to give furfuryl alcohol. In our previous work, we also observed that the conversion of FAL to furfuryl alcohol via C=O reduction increased with increasing amount of surface metallic Cu [27]. The activities of all three catalysts used in this study can therefore be explained based on (1) the degree of reduction, i.e. the amount of surface metallic Cu species, indicated by H₂ consumption during TPR, and (2) the number of copper-cobalt mixed oxide sites available. Cu-Co/CeO₂ gave the lowest H₂ consumption, and Cu-Co/Al₂O₃ consumed the highest volume of H₂.

In terms of product selectivity, the main compounds identified in the liquid reaction products were BHMF, 5-MF, 5-MFOL, and DMF, over all three catalysts. However, the selectivity for the desired product (DMF) is affected by the support, the second metal, or both. For example, Cu-Co/CeO₂ gave the highest selectivity for BHMF because of preferential reduction of the C=O bond in HMF. Small amounts of hydrogenolysis products such as 5-MF, 5-MFOL, and DMF, and small amounts of the products FOL and MF, which were produced by decarbonylation followed by hydrogenolysis, were also detected. Cu-Co/Al₂O₃ was the most active in polarizing O-H and C-O bonds, resulting in DMF production via sequential hydrogenation/hydrogenolysis of HMF. These results show that Cu effectively promotes hydrogenation and moderately promotes hydrogenolysis. In addition, Cu can break the C=O bond during hydrogenation of FAL and HMF [21, 28, 44]. It has also been reported that Cu metal gives selective production of propylene glycol via glycerol hydrogenolysis [43]. Sithisa et al. [44] proposed that aldehydes were preferentially adsorbed in the η¹(O) configuration on Cu metal, which allows interactions between surface Cu and the oxygen of the FAL carbonyl group, to yield furfuryl alcohol. It has also been reported that surface Cu can induce O-H bond scission to form stable alkoxide (CH₃O) intermediates during methanol decomposition [45].

The above findings suggest that the increased selectivity of Cu-Co/Al₂O₃ for DMF production via C-O and O-H hydrogenolysis can be attributed to synergy between surface metallic Cu sites and partially reduced CoO_x sites. The XRD and TPR results show that Cu-Co/Al₂O₃ has the maximum amounts of surface metallic Cu and mixed oxide phases, resulting in an increased rate of hydrogenolysis [29]. It has been reported that the reduced forms of spinel CuCoO₄ catalysts involve metallic Cu dispersed on CoO_x [34, 35]. It has also been reported that early-transition-metal oxides such as ReO_x, MoO_x, and CoO_x are effective catalysts in the hydrogenolysis of biomass-derived components [35, 36]. Previous studies [47, 48] have shown that dissociative adsorption of H₂ on Cu(0) has a low activation energy [49]. Similar effects have been observed for promotion of Cu catalysts with noble metals [50]. Recently, our group reported that an increase in the Co loading increased Cu dispersion because of the strong interactions between Cu and Co. A combination of Cu-CoO_x was found to be effective in the one-step hydrogenation of FAL to 2-MF via reduction of the C=O bond followed by C-O hydrogenolysis [29]. These previous reports and the present experimental findings suggest that the improved dispersion of Cu in the presence of Co and the synergy between Cu and CoO_x species as a result of strong interactions between Cu and Co facilitate the hydrogenolysis of HMF to DMF. This is supported by the XRD, BET, and H₂-TPR results. Metallic Cu and mixed Cu-Co particles can therefore activate H₂ to enable partial reduction of Co₃O_4, which is difficult to reduce at low temperatures. This reducibility of the bimetallic catalysts is responsible for the enhanced activity of the Cu-Co/Al₂O₃ catalyst.

Cu-Co/ZrO₂ had low selectivity for DMF (about 40%) and produced a combination of ring-saturation and over-hydrogenolysis products such as DMTHF and other by-products. The properties of ZrO₂ and the presence of strong acidic sites may be responsible for this behaviour; the strong acidic sites may cause side reactions such as ring opening or ring saturation. Bui et al. [51] reported that ring opening by hydrolytic cleavage of the furanic C-O bond was usually promoted by Brnsted acidic sites in acidic materials. In addition, traces of 5, 5-oxybis(methylene)bis(2-methylfuran), formed by dehydration of 5-MOL, were also observed with Cu-Co/ZrO₂. This is in agreement with a previous report [15].

We studied the hydrogenolysis of 5-methylfurfural (5-MF) over the same catalysts and under the same reaction conditions (Fig. 8) to explore the reaction pathways and clarify the reasons for the superior hydrogenolysis activities of some of the catalysts. The hydrogenolysis activity of Cu-Co/CeO₂ was inferior to those of the other two catalysts. About 86% 5-MF was converted over Cu-Co/CeO₂, with 32% selectivity for DMF. The major product selectivity was for 5-MFOL (48%), formed via reduction of the C=O bond in 5-MF. Following the trend observed for HMF hydrogenation, Cu-Co/Al₂O₃ showed excellent hydrogenolysis activity and produced about 82% DMF with 99% conversion of 5-MF. However, although 99% conversion of 5-MF was achieved over Cu-Co/ZrO₂, the selectivity for DMF (56%) was low and large amounts of ring-saturation and ring-opening products were obtained. These results are in agreement with those obtained for one-step hydrogenolysis of HMF to DMF.

The effect of the Cu/Co molar ratio on the catalytic activity of bimetallic Cu-Co/Al₂O₃ was investigated by studying catalysts with various molar ratio (Cu/Co = 1, 2, and 4) at 200 ℃ and 3.0 MPa. The results were compared with those for the monometallic Cu (10 wt%)/Al₂O₃ and Co (10 wt%)/Al₂O₃ counterparts (Table 4). The catalytic activity of Co (10 wt%)/Al₂O₃ was low, which confirms that Co₃O₄ species alone have little effect on HMF hydrogenolysis under the present conditions. This result is in agreement with that reported by Zu et al. [15], who found that Co₃O₄ had low catalytic activity in HMF hydrogenation. When 10 wt% Cu/Al₂O₃ was used, BHMF, formed via activation of the carbonyl group, was the major product (45.5%). However, the bimetallic catalysts gave lower selectivities for BHMF and increased selectivities for DMF; the DMF selectivity reached a maximum of about 68.4% over Cu-Co/Al₂O₃ (Cu/Co = 1). BHMF was the main intermediate formed over all the catalysts, and it was converted to DMF via 5-MF and MFOL by subsequent hydrogenation and hydrogenolysis under the present conditions [22-26]. It can be deduced that hydrogenation of the C=O bond followed by hydrogenolysis of C-O and O-H bonds in HMF is favoured at higher temperatures because of activation of the C-O and O-H bonds, which results in formation of DMF as the major product and BHMF, MFA, and MFOL as minor products over bimetallic Cu-Co catalysts [46]. These results can be explained based on previously reported XRD, H₂-TPR, and X-ray absorption near-edge structure studies, which showed that the amounts of surface metallic Cu and mixed Cu-Co phases increased with increasing Co loading; this is responsible for the enhanced hydrogenolysis activity of the bimetallic catalysts [29].

We then studied the effects of temperature, H₂ pressure, and time on reactions using the Cu-Co/Al₂O₃ (Cu/Co = 1) catalyst to improve the DMF yield.

The effect of temperature on the selective hydrogenolysis of HMF to DMF was studied over the temperature range 160-230 ℃ at a constant H₂ pressure of 3.0 MPa (Fig. 9). The HMF conversion was low (68%) at 160 ℃, but increased with increasing temperature. It reached the maximum, i.e. 100% at 220 ℃, and remained constant with further temperature increases to 230 ℃. These results can be attributed to an increase in the reaction rate because of facilitation of C=O bond breaking over metallic Cu, resulting in increased activation of molecular H₂ in the reaction mixture [22, 23]. Hydrogenation of the C=C bond (bond energy = 615 kJ/mol) is easier than C=O bond (bond energy = 715 kJ/mol) hydrogenation [45]. However, the presence of the conjugated furan ring makes hydrogenation of the C=O bond easier than C=C bond hydrogenation in the selective hydrogenation of HMF to DMF, which is generally promoted at higher temperatures [46]. At 160 ℃ the selectivity for DMF was low (40.6%), with formation of by-products such as BHMF, 5-MF, and 5-MFOL. However, with increasing temperature, the selectivity for DMF improved significantly and reached a maximum of 78% at 220 ℃. It is worth noting that hydrogenation of the C=O bond in HMF was favoured at low temperatures over Cu catalysts because of the higher activation energies of C-O and O-H, resulting in the formation of BHMF as a major product. However, higher temperatures promote C-O and O-H hydrogenolysis, which results in further conversion of BHMF to DMF. A high temperature is therefore important for C-O and O-H hydrogenolysis of the intermediates BHMF, 5-MF, and/or 5-MFOL. Further increases in temperature resulted in over-hydrogenolysis products and the yield of DMF dropped. This suggests that the conversion of HMF to DMF had reached a maximum at 220 ℃, and the effect of further increases in the reaction temperature was insignificant. A reaction temperature of 220 ℃ was therefore used in subsequent experiments.

In hydrogenation/hydrogenolysis reactions, the solubility of H₂, as well as other parameters (temperature and time), is important. A high H₂ pressure is needed to accelerate the hydrogenation of biobased molecules [22]. The optimum H₂ pressure for selective hydrogenation of HMF to DMF was identified by performing reactions at 220 ℃ and various pressures in the range 10-40 bar (Fig. 10). The conversion and selectivity both increased with increasing pressure up to 30 bar, and reached 100% and 78%, respectively. The HMF conversion did not change significantly with further increases in pressure. The conversion of HMF to DMF increased with increasing H₂ pressure up to 3.0 MPa because of the increasing solubility of H₂ in the reaction mixture. However, pressures above 3.0 MPa did not significantly change the HMF conversion because of the larger amount of chemisorbed hydrogen (H_ads) on the catalyst, which became saturated. However, above 3.0 MPa, a slight decrease in DMF selectivity was observed. This can be attributed to further opening and hydrogenation of the furan ring. A H₂ pressure of 3.0 MPa was therefore chosen for subsequent experiments.

The effects of time on the HMF conversion and the selectivity for DMF were investigated over the range 2-10 h at 220 ℃ and 3.0 MPa. Fig. 11 shows that only 21% conversion of HMF and 5.6% selectivity for DMF were achieved in 2 h. When the reaction time was prolonged from 2 to 8 h, the conversion reached 100% and the selectivity increased to 78%. The increased selectivity for DMF with time corresponds to conversion of increasing amounts of intermediates such as BHMF, 5-MF, and 5-MFOL to DMF. When the reaction time was further increased to 10 h, the HMF conversion remained constant at 100%; however, the selectivity for DMF decreased slightly. This can be attributed to the formation of increasing amounts of by-products when the reaction time increased beyond a certain value. A reaction time of 8 h was therefore used in subsequent experiments. The transformation of HMF to DMF proceeds via the intermediates BHMF or 5-MF, and then 5-MFOL. Conversion of both intermediates to DMF is considered to be slow and requires more time over Cu-based catalysts.

The Cu-Co/Al₂O₃ catalyst, which gave the best catalytic performance, was recycled twice to evaluate its stability. The HMF conversion and selectivity for DMF dropped from 100% to 62% and 78% to 42%, respectively, in the second run (Table 5). The decrease in the catalytic activity was investigated by examining the differences between the structures of the spent and freshly reduced catalysts using TEM and CHNS analysis. TEM and CHNS analysis showed the presence of carbon after use. The TEM results showed that carbon was deposited on the active sites and the particle size increased after the second cycle; these factors could cause the significant drop in the activity of the catalyst. These observations agree with our previous work on FAL hydrogenation, in which coke formation was observed [29]. To confirm the reasons for the loss of catalytic activity, the used catalyst (Cu-Co/Al₂O₃ after the second cycle) was calcined at 500 ℃ and activated in situ at 280 ℃ to remove deposits, and then tested under same reaction conditions. As expected, the conversion of HMF and selectivity for DMF both increased, to 94% and 72%, respectively. It can therefore be concluded that the catalyst lost activity because coke formed during HMF hydrogenation covered the active sites on the catalyst.

Three supported bimetallic catalysts, i.e. Cu-Co/CeO₂, Cu-Co/ZrO₂, and Cu-Co/Al₂O₃, with Cu/Co molar ratios of 1, 2, and 4 were used in the hydrogenolysis of HMF to DMF. The intrinsic properties of the carriers and the strong interactions between Cu and Co significantly affected the catalytic performances of the Cu-based bimetallic catalysts. The activity of Cu-Co/CeO₂ in C-O bond breakage was modest because of its large Cu particles and low acidity. Cu-Co/Al₂O₃ was much more active in hydrogenation/hydrogenolysis of HMF because of the presence of metallic Cu and mixed copper-cobalt oxides, and appropriate acidic sites. Cu-Co/ZrO₂ showed similar activity but different product selectivity because of morphological differences and the presence of strong acidic sites. The catalytic activity of Cu-Co/Al₂O₃ in the conversion of HMF to DMF increased with increasing Co loading and the activities of the bimetallic catalysts were higher than those of their monometallic counterparts. The catalyst with the highest loading (Co = Cu = 10 wt%) gave an excellent catalytic performance; this is attributed to the presence of mixed copper-cobalt oxide species on porous Al₂O₃. Time and temperature were important parameters in the selective conversion of HMF to DMF. Based on these promising results, further studies will focus on the reaction mechanism and kinetics of HMF conversion.