Furanic biofuels production from biomass using Cu-based heterogeneous catalysts

1 Furanic biofuels production from biomass using Cu-based heterogeneous catalysts Nerea Viar, Jesús M. Requies, Ion Agirre, Aitziber Iriondo, Pedro L. Arias Chemical and Environmental Engineering Department. Engineering Faculty of Bilbao, University of Basque Country (UPV/EHU), Plaza Ingeniero Torres Quevedo 1 – 48013 Bilbao (Spain) e-mail: [email protected], [email protected], ion.agir[email protected], aitziber.ir[email protected], [email protected] Abstract Biorefineries can be a good alternative to cover the current needs of energy, such as transportation fuels and also fine chemicals. 5-hydroxymethylfurfural (HMF) can be obtained directly from biomass, and it is considered one of the most promising intermediate for the synthesis of biofuels. In this work the production of 2,5-dimethylfuran (DMF) and 2,5dimethyltetrahydrofuran (DMTHF), two particularly attractive compounds to be used as biofuels due to its similar characteristics to gasoline, has been studied. For this purpose, a continuous fixed bed catalytic reactor was employed using HMF as raw material. Zr-supported Cu, Ce-Cu, Ru-Cu and Ni-Cu catalysts were tested. Moreover, the catalysts textural properties were also measured - before and after the activity tesst - in order to get a deeper understanding of the relationships among the measured DMF yields and the catalyst characteristics. Final DMF, DMTHF and butanol separation was also evaluated. Keywords 2,5-Dimethylfuran; 2,5-Dimethyltetrahydrofuran; 5-Hidroxymethylfurfural; Hydrogenation; Bimetallic Highlights x DMF & DMTHF production, two suitable biofuels, was studied x Ni avoids sinterization of Cu and increases the presence of more reduced Cu species x 15Ni15CuZr catalyst shows the highest DMF+DMTHF yield (53.2%) x 15Ni15CuZr catalyst is stable after 24 hours on stream for DMF (25% yield) x Ce and Ru incorporation did not have a positive effect in the DMF or DMTHF yield. *Corresponding Author: Phone: +34-94 601 7297; e-mail: [email protected]; University of the Basque Country (UPV/EHU). Engineering Faculty of Bilbao. Chemical and Environmental Engineering Department, Plaza Ingeniero Torres Quevedo 1, 48013 Bilbao, Spain This is the accepted manuscript of the article that appeared in final form in Energy 172 : 531-544 (2019), which has been published in final form at https://doi.org/10.1016/j.energy.2019.01.109. © 2019 Elsevier under CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) 2 1. INTRODUCTION One of the most serious world environmental concerns is related to the climate change due to the greenhouse effect generated by the anthropological CO2 emissions. Most of these emissions come from fossil fuels combustion [1,2]. Actually, according to the International Energy Agency (IEA) [3–5], oil, coal and natural gas represent the main energy resources (more than 80% of primary energy) and their usage is still increasing [6,7]. The transportation sector is responsible for around 70 % and 20 % of the CO and CO2 emissions, respectively [3]. Moreover, the geopolitical instability in the oil producer countries and the future depletion of the available reserves [1,4,8,9] demand sustainable alternatives. However, oil is not only the source of fuels but also of many chemicals for the production of multiple items like plastics and cloths [10]. Thus, the R&D searching for sustainable alternatives to oil and natural gas has become one of the main technological challenges in the near future. Biomass seems to be a good candidate as a direct source of energy but also as the raw material of platform products for the production of biofuels and fine chemicals [3,11]. It is clear that the biofuels present a high potential as renewable energy for transportation. In fact, in 2014, around 90 billion liters were produced worldwide [12]. Although the first generation of biofuels (biodiesel and bioethanol) generate controversy due to their competition for food crops, the second generation can be produced using building blocks from lignocellulosic biomass [13–15]. Special interest as these building blocks merit the ones derived from sugars produced by hydrolysis of the cellulosic and hemicellulosic fractions of this biomass. Among those building blocks, 5-hydroxymethylfurfural (HMF) was selected as a top chemical in 2010 [16] and it is considered as an intermediate between biomass and final biofuels (2,5dimethylfuran (DMF) and 2,5-dimethyltetrahydrofuran (DMTHF)) and biochemicals (ethyl levulinate (EL), 2,5-furandicarboxilic acid (FDCA) or levulinic acid (LA)) [3,8,11,17–19]. HMF can be produced by dehydration of C6 sugars like glucose, fructose and sucrose [16]. Actually, AVA Biomchem is already preparing its commercialization [20]. DMF and DMTHF furanic biofuels could be valuable substitutes of transportation fossil fuels [21]. A large number of models have published about the combustion of furanic compounds, mainly before 2000 [21]. This can be due to furanic biofuels being excellent octane boosters [22]. Regarding the DMF, some authors [2,8,23–28] conclude that it could be a promising additive to oil-based fuels but taking into consideration the prevailing pressure and temperature conditions. DMF, which shows a very low solubility in water [15,29], is reported as a high-quality biofuel or bioadditive, for conventional gasoline [4,19]. It presents some interesting chemical properties as its high octane number, higher than the one of conventional gasoline [3] and high energy density, similar to that of gasoline and higher than that of ethanol [8,15] (see Table 1). One advantage of its usage is that the large particles emission is decreased considerably, although it is not able to reduce significantly the number of particles smaller than 10 nm [30]. Some authors report its good performance as a fuel on direct injection spark ignition (DISI) engines [2,8], without any important modifications of the engine, but also in compression engines [31]. Chen et al. [24,25] tested diesel/DMF blends following different multi-injection combustion strategies in diesel engines and characterized the corresponding emissions. They observed that ignition-delay was prolonged and smoke decreased when the EGR (exhaust gas 3 recirculation) rate increased. Liu et al. [26] compared the emissions of four different diesel mixtures with a 20% blending ratio by volume (cetane + iso-cetane, n-heptane, DMF and DMF+2ethylhexyl nitrate) varying the EGR from 0% to 62%. They demonstrated that there is little difference in combustion and emissions among the first two blends and pure diesel. The lower cetane number of the DMF20 blend is the uppermost factor to control the combustion process, as having a more premixed combustion reduces soot formation and emission. However, it seems that diesel/DMF blends show higher NOx emissions when compared to pure diesel usage [27,28]. Table 1 Comparison of specifications among conventional and renewable fuels [21,24–28,32] Parameters Diesel Gasoline Biodiesel DMF DMTHF Ethanol nbutanol Molecular formula ϭϮоϮϱ ϰоϭϮ ϭϮоϮϰ C 6H8O C6H12O C2H5OH C4H9OH Octane number ϵϬоϵϵ 119 82 108 96 Cetane number ϰϬоϱϱ ϭϬоϭϱ ϰϳоϱϮ 9 17.23a 8 25 Oxygen content (wt%) 10 16.7 16 34.8 21.6 Stoichiometric air/fuel ratio 14.3 14.7 12.5 10.79 11.73 9.02 11.21 Lower heating value (MJ/kg) 42.5 42.7 38.81 33.7 38.3 26.8 33.1 Density at 20 °C (g/cm3) 0.82 0.745 0.87 0.89 0.83 0.79 0.81 Viscosity (mm2/s) at 40 °C 1.9–4.0 0.4–0.8b 4 0.65 0.47a 1.08 2.63 Flash point (°C) 65–88 оϰϱоϯϴ 166 оϭ 27 8 35 Latent heating (kJ/kg) at 25 °C ϮϳϬоϯϬϭ 380–500 200 333 348a 904 582 Boiling point (°C) ϭϴϬоϯϳϬ 25–215 ϮϲϮоϯϱϵ ϵϮоϵϰ 90-92 78.4 117.7 Saturation pressure (kPa) at 38 °C 1.86 31.01 12.8a 15.1a 13.8 2.27 Auto-ignition temperature (°C) 246 420 363 285.85 434 385 a Specification estimated by Aspen Plus, b at 20 ºC According to the last published findings in the open literature, DMF can be produced from selective hydrogenation of HMF [15] using non-noble and noble metal catalysts [3,4]. This conversion involves different pathways [3,15,33] in which dehydration and hydrogenation reactions are predominant. Figure 1 shows a simplified reaction scheme of DMF production from biomass. The hydrogenolysis of HMF to 5-methylfurfuryl alcohol (MFA) via 5-metylfurfural (MF) or via 2,5-bishydroxymethylfuran (BHMF) are the main routes for DMF production. Afterwards, the MFA hydrogenation produces DMF. 4 Figure 1 Production route of DMF and DMTHF via HMF from biomass (adapted from [34,35]). On the one hand, metal hydrogenation catalysts are necessary for the hydrogenation of the HMF. There are two types of metal-based catalysts, those containing noble metals such as Pd, Pt, Ru, Rh and Au [3,15,29,36,37], and those containing transition metals, such as Cu, Co, Ni and Fe [36,38,39]. Some authors [17] have also studied the combination of both type of metals obtaining a high selectivity (over 70 %) in the production of DMF using Cu-Ru supported on carbon at 220 ºC and 6.8 bar of H2 and using 1-butanol as solvent[17]. Chidambaram et al. [40] obtained a selectivity of 32 % to DMF with a 47 % HMF conversion using a Pd/C catalyst and ionic liquids at 120 ºC and 62 bar of H2. Zu et al. [4] reported a very high yield to DMF (93.4 %) using tetrahydrofuran (THF) as solvent at 130 ºC, and 7 bar of H2 partial pressure using a Ru/ Co3O4 catalyst. Ru participates in the hydrogenation while Co3O4 was the responsible for the HMF adsorption and the breakage of the C-O bonds. Goyal et al. [38] achieved total HMF conversion with a DMF selectivity over 98.7 % employing a Ni catalyst supported on mesoporous nitrogen rich carbon material and using liquid water as solvent at 200 ºC and 30 bar of H2 pressure. It seems that using Ni nanoparticles (below 5 nm) leaded to a high Ni dispersion associated to a high interaction with the support. On the contrary, with higher nickel particle sizes lower DMF selectivities were achieved [38]. However, all these data reported were developed using batch reactors and the corresponding catalyst stability was not tested. On the other hand, apart from the metals, the support also plays a key role in DMF production. The presence of acid supports could cause the loss of one of the radicals present in the HMF molecule or degradation of the DMF since acid sites, Brønsted or Lewis, favor the C-C bond cleavage. Thus, neutral-basic and weak acid supports are suggested for HMF hydrogenolysis to DMF. This type of neutral-basic supports seem to avoid C-C bond cleavage of the HMF tertiary carbon [13]. Carbon materials [17,41,42], zirconia [13] and metal oxides such as Co3O4 [4] seem to be good candidates for this purpose. Another interesting by-product that appears is DMF production compound is DMTHF. This compound, together with DMF, is considered as a suitable liquid biofuel due to their similar HMF BHMF MF MFA DMF BHMTF DMTHF MTHFA Ring OpeningProducts (ketones) Ring OpeningProducts (polyols) 5 properties (such as energy density, boiling point, and water solubility) to commercial gasoline and high octane number [35,43,44] (see Table 1). DMTHF can be obtained by over hydrogenation of MFA (hydrogenation of DMF) but also via hydrogenation+hydrogenolysis of MFA being 5-methyltetrahydrofurfuryl alcohol (MTHFA) the intermediate compound. However, this latter reaction pathway seems to be only catalyzed by noble metals such as Pd [35]. Additionally, some other byproducts, such as 1,6-hexanediol and 1,2,6-hexanetriol, can also be obtained due to the furanic ring opening. Moreover, some authors report the possibility of getting DMTHF by etherification of 2,5-hexanedione [45]. Etherification reactions are typically catalyzed by acid sites, suggesting that HD cyclization to DMTHF might be facilitated by Lewis acidic sites on Ru in the form of RuOx [30–32]. Although the production of DMF is currently in a very early stage, some authors like Kazi et al. [46] have already carried out a techno-economic assessment of DMF and HMF production using fructose as the raw material. The minimum selling prices for HMF and DMF were estimated as 1.3 $/L and 2 $/L, respectively. However, this analysis was able not only to estimate minimum selling prices but also to identify different areas of improvement. Thus, one of the most important conclusions is that the catalyst performance at the present level of development does not seem to be enough to produce commercial DMF to be used as a fuel. Moreover, in the sensitivity analysis they performed, the DMF yield is the parameter that affects the most the DMF minimum selling price. However, there are also some other important questions to deal with such as the required synergies in a biorefinery in order to valorize/utilize by-products, or the necessity to investigate in renewable sources for H2. Based on this background, the present paper focuses on the catalytic transformation of HMF into DMF and DMTHF, but using a continuous fixed bed reactor instead of batch reactors. High biofuel production rates require the use of continuous reactors. Besides that, real catalyst stability studies are also easier under continuous operating conditions. For this purpose, the use of inexpensive bifunctional catalysts was studied, analyzing the influence of the support and different metals on the DMF yields. Catalysts stability of the most promising catalysts was also been studied. 2. EXPERIMENTAL PROCEDURE The experimental procedure used for this study is shown in Figure 2. First, catalysts and operating conditions were selected based on the literature. After this, a basic characterization of the catalysts was carried out followed by the catalytic tests. The results obtained from the tests and the basic characterization are then used to establish what catalysts required a deep characterization and assess the possibility of improving catalysts activity and stability by modifying them. 6 Figure 2 Outline followed for the catalytic study of the DMF and DMTHF production from HMF. 2.1. DMF and DMTHF production Synthetic HMF (Sigma-Aldrich, 99%) diluted in 1-butanol was used as the feed to a continuous fixed-bed reactor in order to produce DMF and DMTHF. 1-Butanol presents suitable properties to be used as a green solvent in DMF production and it is one of the most used solvent agents for this purpose [17,29,36,47]. 1.5 wt% of HMF was continuously fed to the reactor by means of a HPLC pump. 0.5 g of catalyst diluted with inert CSi (weight catalysts/CSI ratio = 1:9) were placed in the reactor and a WHSV (wt HMF/(wt cat·h)) of 0.15 h-1 was used. In order to know the effect of the catalyst reduction step in catalytic activity, in some experiments the catalyst was in-situ reduced before the activity test using H2 at 1 bar and 200 – 275 °C. After this, the activity tests were carried out with fresh pre-reduced and no pre-reduced fresh catalysts under different temperature and pressure conditions of 200 - 275 °C and 15 - 20 bar of H2, respectively. The indicated operation conditions were selected based on the results obtained in a previous work [13]. In order to report the activity tests results, the HMF conversion (%) and DMF or DMTHF yields (%) were used defined as follows: in out HMF HMF in HMF N (mol/min)-N (mol/min) Conversion (%) = ·100 N(mol/min) (1) out DMF/DMTHF in DMF/DMTHF N (mol/min) Yield (%) = ·100 N (mol/min) (2) 2.2. Analysis Reactor feed and output liquid streams were sampled and analyzed by off-line highperformance liquid chromatography (HPLC) and gas chromatography (GC). HMF content of both inlet and outlet streams was determined by HPLC 1260 Infinity equipped with HI-Plex H column and an infrared detector. An Agilent 6804 GC equipped with a flame ionization detector (FID) and using a Suprawax 280 capillary column was used to analyze DMF, DMTHF and other byproducts present in the outlet stream. 2.3. Catalyst preparation Different monometallic and bimetallic catalysts were prepared by the incipient wetness impregnation method. For this purpose, ZrO(NO3)2·xH2O (Sigma-Aldrich, 99%), Cu(NO3)2·2.5H2O Catalysts’ Preparation Impregnation Catalytic Performance Fresh Catalysts’ Basic Characterization Fresh/Used Catalysts’ Deep Characterization N 2 physisorption; ICP-AES; H 2 -TPR TPD; XRD; SEM; XPS Incorporation of a second metal and support modifier 7 (Alfa Aesar, 98%), RuCl3·xH2O (Merck, metallic base Ru 35-40%), CeN3O9·6H2O (Alfa Aesar, 99,5%) and Ni(NO3)2·6H2O (Sigma-Aldrich, 99.999%) reagents were used. Cu supported on ZrO2 catalysts were prepared by a single step impregnation and were designed as 15CuZr, 30CuZr, 45CuZr, in which the numerical factor indicates the nominal copper content as wt%. Subsequently, the Ru-Cu bimetallic catalysts, designed as 1RuXCuZr (being X the nominal weight of Cu), were prepared by impregnation of the previously calcinated 15CuZr and 30CuZr catalysts with a ruthenium precursor solution. Cu supported on CeZr was afterwards prepared in two sequential steps, first impregnating Ce on Zr and then Cu on CeZr, having dried and calcinated the CeZr before being impregnated the Cu. These catalysts were designed as 30CuNCeZr and 15CuNCeZr (were N refers to nominal weight of Ce). Finally, bimetallic 15Ni15CuZr was prepared in a single step co-impregnation. To understand the behavior of bimetallic catalyst, the monometallic 15NiZr catalyst was also prepared. After impregnation and drying, the samples were calcined in air at 250 °C for 2 h. 2.4. Catalyst characterization The catalysts were characterized by several physicochemical techniques such as N2 physisorption (BET method), Plasma Atomic Emission Spectroscopy (ICP-AES), Temperature Programmed Reduction (H2-TPR), ammonia Temperature Programmed Desorption (TPD), X-ray Diffraction (XRD), Scanning Electron Microscope (SEM) and X-ray Photoelectron Spectroscopy (XPS). 2.4.1. Textural properties BET surface area and pore characteristics of the calcined fresh catalysts were evaluated from the N2 adsorption–desorption isotherms obtained at -198 °C over the whole range of relative pressures. In the case of the Zr-supported Cu monometallic and bimetallic catalysts, this sample was characterized using Autosorb®-1-C/TCD (Quantachrome, USA) after outgassing solid samples at 150 °C for 24 h. 2.4.2. Plasma atomic emission spectroscopy (IPC-AES) Metal contents were determined by ICP-AES technique, using a Perkin-Elmer Optima 3300DV apparatus, previous dissolution of the ground samples in acid solutions (37.5% HNO3, 37.5% HF and 25% HCl, in volume). 2.4.3. Temperature Programmed Reduction with hydrogen (H2-TPR) An AutoChem II Instrument (Micromeritics, USA) equipped with a TCD detector was used to study the reducibility of the calcined catalysts. The samples were warmed up to 900 °C at a rate of 10 °C/min passing 50 mL/min of reducing gas (5% v/v H2 diluted in Ar) through the sample. 2.4.4. Temperature programmed desorption with ammonia (TPD) The acidity of the calcined catalysts was measured by temperature programmed desorption of NH3. For this purpose, as well as in the H2-TPR technique, an AutoChem II instrument was used. Firstly, samples were flushed by He at 250 °C for 30 min and then reduced with a 5% H2/He mixture. Afterwards the catalyst was cooled down to 40 °C and NH3 loading was maintained for 8 30 min. Before starting the desorption process, physically absorbed NH3 was removed using He at 85 °C until no further desorption was recorded. Finally, the release of chemically adsorbed NH3 was recorded increasing the temperature from 85 °C to 800 °C at a rate of 10 °C /min. 2.4.5. X-ray Diffraction (XRD) X-ray diffraction patterns were obtained using a Seifert XRD 3000P diffractometer, equipped with a PW 2200 Bragg-Brentano u/2u goniometer, bent graphite monochromator and automatic slit, using Cu K radiation (0.15418 nm) and 0.028° steps for scanning. The Scherrer equation was used to calculate the average particle size of the crystalline species registered for reduced and used catalysts. 2.4.6. Scanning Electron Microscope (SEM) The morphological appearance of the fresh reduced and the used catalysts was studied by SEM analysis in a JEOL JSM-6400 with W filament and a resolution of 3.5 nm. The fresh reduced and the used samples were prepared via dispersion into isopropanol as solvent and placed on a carbon-coated copper grid (300 Mesh) followed by drying under vacuum. 2.4.7. X-ray photoelectron spectroscopy (XPS) This technique was used to evaluate the surface characteristics (oxidation state of the species formed, interactions, atomic ratios, etc.) of the samples. The measurements were carried out with a VG Escalab 200R spectrometer equipped with a hemispherical electron analyzer and an Al K1 (h = 1486.6 eV) 120W X-ray source. The powdered samples were deposited on a stainless steel sample holder, placed in the pre-treatment chamber and degassed at 300 °C. The spectrometer base pressure was typically 10о9 torr (0.0133 bar). The spectra were collected at a pass energy of 20 eV, which is typical of high resolution conditions. Both fresh reduced and spent catalysts were analyzed with this technique. 3. RESULTS AND DISCUSSION 3.1. Catalysts characterization results 3.1.1. Chemical and textural properties According with the N2 adsorption-desorption results shown in Table 2, the calcined Zr presents high BET area and pore volume.The presence of Ce on the Zr support reduces a little bit the surface area and the pore volume due to the Ce deposition on the pores. Nevertheless, after the increment of Ce on the catalyst from 5 to 20 %, the BET area was practically the same (due to the equipment experimental error is r 3 m2/g) and also pore volume. However, the presence of a higher amount of ceria decreased the pore diameter. Probably the incorporation of more Ce favours the formation of larger ceria particles. The addition of Cu onto the Zr and CeZr mesoporous supports causes a decrease in the specific surface area of the catalysts possibly due to the presence of CuO crystallites in the pores, limiting the adsorption of N2 during the adsorption-desorption experiment [48]. It is remarkable that this reduction tendency of the surface area is more important for low Cu contents than for high Cu contents, except for the 15Cu20CeZr catalyst. This behavior was also observed in the catalysts with 15 wt% of Cu like 1Ru15CuZr, 15Ni15CuZr and 15Cu5CeZr. It seems that the presence of 9 high ceria loadings modified the way in which Cu was incorporated on the catalytic surface. In the case of bimetallic catalysts, the presence of a second metal reduces the surface area and the pore volume. Regarding the 15NiZr catalyst, the nickel incorporation reduced drastically the surface area. As well as Cu, the metallic nickel clusters blocked the surface, resulting in a low dispersion of the metal [49]. The 15RuZr catalyst presents similar behavior than the monometallic catalyst as the Ru incorporation decreased both the BET surface and the pore volume. Table 2 Nomenclature and textural properties for calcined supports and Cu, Ru, RuCu and NiCu catalysts. Group Catalysts BET (m2/g) Pore Volume (cm3/g) Average pore diameter (nm) Supports Zr 137.6 0.267 7.2 5CeZr 112.3 0.228 7.9 20CeZr 114.1 0.211 3.7 Cu, Ru and RuCu catalysts 15CuZr 35.1 0.094 10.7 30CuZr 81.8 0.267 13.1 45CuZr 52.0 0.043 4.2 1RuZr 89.3 0.401 18.0 1Ru15CuZr 22.0 0.068 12.1 1Ru30CuZr 49.0 0.062 4.9 CeZr-supported Cu catalysts 15Cu5CeZr 40.8 0.073 7.2 15Cu20CeZr 110.3 0.204 7.2 30Cu5CeZr 60.1 0.099 6.4 Ni and NiCu catalysts 15NiZr 5.0 0.033 25.3 15Ni15CuZr 18.1 0.034 7.7 The real metal contents of the catalysts were determined by ICP-AES and the results are presented in Table 3. As it can be observed, in general, the real values were lower than the nominal ones for both monometallic and bimetallic catalysts. This difference was specially high when ceria was deposited on the zirconia support. These big differences in the ceria-doped supports also were measured, as expected, in the corresponding catalysts (CuCeZr). The difference of the real amount of ceria incorporated on the CeZr supports is so low that this can explain the small difference between the BET surface areas of ceria doped and non-doped zirconia supports. Probably the surface of the ZrO2 is not very reactive with the cerium, i.e., the ZrO2-Ce interaction is very weak, and therefore the Ce does not incorporate on ZrO2 surface efficiently. Table 3 Real metal load of the catalysts determined by ICP-AES. Group Catalysts Cu (%) Ru (%) Ni (%) Ce (%) Supports 5CeZr 20CeZr 0.4 2.4 Cu, Ru and RuCu catalysts 15CuZr 13.0 30CuZr 17.0 45CuZr 28.4 1Ru15CuZr 10.9 0.26 1Ru30CuZr 20.6 0.50 16 3.1.5. Morphological characteristics The SEM measurements were carried out for fresh reduced and used catalyst. As it can be observed in Figure 8, Cu is well dispersed on Zr, in both fresh and used catalysts, and it appears in particles smaller than 1000 nm. Zr particles show a smother surface and appear as bigger particles, with around 0.1-0.5 mm. In case of the 30Cu20CeZr and 15Cu20CeZr catalysts, EDX measurements certify the Ce presence only in the 30Cu20CeZr catalyst but not in all the studied points (see Figure S2). Moreover, it is not possible to distinguish the Ce from Cu in the SEM pictures. These results are in good agreement with ICP-AES and XRD, where the cerium content on the catalytic surface was very low. Concerning the used catalysts, EDX analysis certified the presence of coke particles on the catalyst surface. Cu still remained well dispersed with a similar particle size than the one in the fresh catalyst. Zr particles also showed a similar particle size except in the case of the 30CuZr catalyst, where the deposited coke covers Zr and Cu particles and it is not possible to determine their particle sizes. The presence of coke was also detected by XPS (see Table 7). As well as with the fresh catalysts, it is not possible to distinguish Ce from Cu in the case of the 30Cu20CeZr and 15Cu20CeZr catalysts. Nevertheless, contrary to the fresh catalysts, no Ce could be detected by EDX in any of the studied areas. 15CuZr (fresh) 15CuZr (used) 30CuZr (fresh) 30CuZr (used) 45CuZr (fresh) 45CuZr (used) 17 15Cu20CeZr (fresh) 15Cu20CeZr (used) 30Cu20CeZr (fresh) 30Cu20CeZr (used) Figure 8. SEM measurements of fresh and used catalysts. 3.1.6. Surface properties Table 7 summarizes the XPS results where surface properties, such as surface atomic ratios and Cu species ratios, are presented. The ratio of metallic Cu to Cu ions (Cu0/Cu++Cu2+) was calculated to identify if the catalysts were reduced or oxidized during their use in the studied reaction. For this ratio calculation, different energy bands were employed. Cu 2p binding energys hows a value of 934 eV for Cu2+ [75]. However, in this region it is not possible to distinguish between Cu0 and Cu+, because both of them exhibit the same energy binding of 932 eV [75]. Therefore Cu LMM region was then analyzed, where Cu0and Cu+species present a peak at 335 eV and 337.5 eV [75–78], respectively. As it can be observed, the Cu/Zr surface atomic ratio of fresh reduced monometallic catalysts increased when higher Cu loadings were incorporated. In general, the presence of other cometals like Ru and Ce increased the Cu/Zr ratio improving the Cu presence on catalyst’s surface. The addition of a low amount of a noble metal, for instance Ru, lead also to a higher metal dispersion and metallic surface area, and high degree of reduction, as the TPR profiles suggest, favoring the formation of metallic Cu. 18 The Ni incorporation improved the fresh reduced Cu/Zr ratio when comared to the monometallic one. The global effect is lower than the one associated to Ru and Ce incorporation. The incorporation of a noble metal to a non-noble monometallic catalyst generates an intimate contact and a synergy between both metals, as the XRD data corroborates, resulting in a better surface exposure, reduction capacity, metal dispersion and metal surface area [79,80]. In the case of the 1RuCuZr catalysts, this was not fulfilled. Probably, the coke deposition on big metal particles provoked the reduction of diffraction patterns intensity of used catalysts, resulting in lower average particle size when compared with the data determined for the reduced ones. Nevertheless, after the activity tests, all the catalysts, the monometallic ones like the bimetallic ones, showed a lower Cu/Zr ratio, which is an indication of some sintering taking place. In addition, the 15Ni15CuZr catalyst also showed some Ni sintering. These results are in good agreement with the XRD results. Moreover, the C/Zr ratio of the used 15Ni15CuZr catalyst was very high if it is compared with the rest of the 15Cu monometallic and bimetallic catalysts. Probably the presence of the Ni favored the C-H bond cleavage, increasing the coke formation. Some catalysts showed a lower Cu0/(Cu++Cu2+) ratio after reaction, meaning that they have suffered some active phase oxidation during the reaction, but in general the variation of this ratio is small. Only the 30Cu5CeZr and 30Cu20CuZr suffered some significant oxidation phenomena.. Table 7 XPS results of Zr-supported Cu and RuCu, and XCuNCeZr catalysts. Group Catalyst Cu/Zr Ƶ϶ (Cu+нƵϸЀͿ C/Zr Ce/Zr Ru/Zr Ni/Zr Cu catalysts 15CuZr Fr. red. 0.24 0.73 1.04 Used 0.12 0.70 10.61 30CuZr Fr. red. 0.49 0.43 1.74 Used 0.32 0.60 17.78 45CuZr Fr. red. 1.36 0.31 5.38 Used 0.22 0.54 144.77 CeZrsupporte d Cu catalysts 15Cu5CeZr Fr. red. 0.51 0.48 3.95 0.0042 Used 0.16 0.53 16.47 0.0022 15Cu20CeZr Fr. red. 0.30 0.58 1.71 0.0027 Used 0.28 0.54 4.51 0.0009 30Cu5CeZr Fr. red. 0.42 0.43 2.76 0.0041 Used 0.33 0.37 12.83 0.0019 30Cu20CeZr Fr. red. 0.70 0.61 3.61 0.0025 Used 0.23 0.45 22.77 0.0015 Ru and RuCu catalysts 1RuZr Fr. red. 1.21 0.08 Used 14.60 0.85 1Ru15CuZr Fr. red. 1.31 0.55 7.10 0.44 Used 0.24 0.63 22.61 1.51 1Ru30CuZr Fr. red. 11.23 0.30 13.55 1.03 Used 1.10 0.33 7.72 0.49 NiCu catalysts 15Ni15CuZr Fr. red. 0.54 3.95 0.43 Used 0.25 73.65 0.10 19 In general, the use of catalysts in the HMF hydrogenolysis reactiona provoked the formation of great quantities of coke, except in the case of the 1Ru30CuZr and 15Cu20CeZr catalysts. In this sense, the usage of Ce clearly limited coke production [81]. Thus although the amount of Ce on the catalytic surface was very small, at the highest Ce/Zr ratios the best coke resistance ability results were obtained. The measured Ce/Zr ratios on the XCuNCeZr catalyts’ surface were smaller than the expected ones, but they are in a good agreement with the ICP-AES, XRD and SEM results [81]. In the case of the 1Ru30CuZr the high dispersion reached on the catalyst surface disfavored the coke production, as small metal particles do not favor coke production. 3.2. Activity results An initial catalytic screening was carried out in order to set the most optimal operating conditions, varying the temperature, H2 partial pressure and the need or not of catalyst pretreatment (see Supplementary Material). Thus, the activity experiments were carried out at the following operating conditions: T = 275 °C, P = 15 bar of H2, no pre-treatment (reduction) of the catalysts and a WHSV of 0.15 h-1 3.2.1. Metal and support influence on catalytic activity Different catalysts were tested in order to know the influence of metal and support on the catalytic activity and selectivity. After this study, the most promising catalyst was selected for a long test. The activity results of the tested catalysts are summarized in Table 8. This table presents the maximum yield of DMF, DMTHF and their summation, emphasizing in brackets the time on stream where these maximum yields were obtained. Moreover, the yield at 6 h on stream is also detailed. In all the cases, complete conversion of HMF was reached. Additionally, a graph of the reaction evolution of each catalyst with the yields of both products appears in Figure S1. Table 8 Activity results of different catalysts at 275 ºC, 15 bar of H2 and space velocity of 0.15 h-1 . Catalyst DMF yield (%) DMTHF yield (%) DMF + DMTHF yield (%) Max. 6h Max. 6h Max. 6h 15CuZr 25.8 (6h) 25.8 7.9 (2h) 1.1 26.9 (6h) 26.9 30CuZr 24.1 (6h) 24.1 7.1 (2h) 2.1 27.2 (4h) 26.2 45CuZr 14.5 (3h) 3.5 2.6 (3h) 0.3 17.1 (3h) 3.8 1Ru15CuZr 23.5 (3h) 23.1 8.1 (2h) 1.1 29.7 (3h) 24.2 1Ru30CuZr 22.2 (4h) 21.6 5.4 (3h) 3.7 27.0 (4h) 25.3 1RuZr 0.8 (2h) 0.4 0.1 (1h) 0.0 0.8 (2h) 0.4 15Cu5CeZr 24.0 (6h) 24.0 7.1 (2h) 2.5 27.3 (5h) 26.5 15Cu20CeZr 14.7 (5h) 14.2 7.0 (3h) 5.8 21.6 (5h) 20.0 30Cu5CeZr 21.9 (6h) 21.9 5.6 (3h) 4.8 26.7 (6h) 26.7 30Cu20CeZr 23.5 (6h) 23.5 10.3 (2h) 3.8 31.7 (4h) 27.3 15NiZr 21.9 (4h) 15.2 0.0 0.0 21.9 (4h) 15.2 15Ni15CuZr 15.1 (6h) 15.1 40.2 (4h) 32.9 53.5 (4h) 48.0 Regarding DMF+DMTHF yield, the obtained results for the 15CuZr and 30CuZr catalysts are very similar. In both cases, a total yield of more than 25 % was achieved and it was stable after 6 h 20 on stream (see Figure S1). In contrast, the 45CuZr catalyst reaches a maximum yield of 17 % after 3 h of reaction and then the catalyst suffers from deactivation. The similar selectivity results obtained for the 15CuZr and 30CuZr catalysts suggest that the total amount of Cu in the catalytic system or on the catalytic surface is not a key factor (see XPS results in Table 7 and ICP-AES results in Table 3). It seems that the Cu dispersion is the one playing a key role in the DMF production. The catalyst with the lowest average particle size (measure by XRD, see Table 5) also presented the best activity and stability results. The acidity is also an important factor, but all Cu monometallic catalysts showed a similar acidity. But the acidity of the catalysts played an important role on the decrease of DMF yield through the formation of degradation products derived from an excess of C-C bond cleavage favoring also carbon/coke particles formation. For instance, as it was expected, due to the similar acidity, strength of acidic sites, TPR reduction profiles and dispersion of the Cu on catalytic surface (XPS), the activity behavior of the 15CuZr and 30CuZr catalysts did not differ too much and comparable DMF and DMTHF yields were achieved under the studied operating conditions. At 6 hours on stream, the 15CuZr and 30CuZr catalysts presented a stable behavior although the Cu/Zr ratio in the used catalysts decreased along the test, while the C/Zr ratio increased. Probably, for longer times on stream some deactivation would have been detected, as it can be observed for the 45CuZr catalyst. This 45CuZr catalyst presented the lowest acidity and a higher reducibility than 15CuZr and 30CuZr catalysts. This lowest acidity is an indication of a low Cu-Zr interaction increasing the Cu reducibility, but this weak metal-support interaction favored the sintering processes (Cu/Zr ratio decreases from 1.32 to 0.22), as it can be observed in the Table 7. The increment of the particle size favors the coke production (the C/Zr ratio of the 45CuZr is the highest one, 141.3). This production seems to take place on the catalytic surface of the 45CuZr [82]. In order to improve the catalytic activity of the most stable catalysts, 15CuZr and 30CuZr, Ru was added using a second impregnation. The Ru is a widely used metal for heterogeneous catalytic hydrogenation [83]. To examine the effect of Ru in the reaction, the 1RuZr catalyst was tested. This catalyst presented a total conversion of HMF but very low or negligible DMF or DMTHF yields. This means that the presence of Ru implies another reaction path (see Figure 1) that the HMF hydrogenolysis. The monometallic Ru catalyst breaks preferentially the C-C bond instead of the C-O bond, which results in degradation products [84,85]. The 1RuZr catalyst was the catalyst with the highest acidity and with the strongest acid sites as TPD-NH3 data corroborated. This characteristic favors the C-C bond cleavage of the HMF molecule and therefore, the presence of acid supports could cause the loss of one of the radicals present in the HMF molecule or degradation of the DMF [12]. This strong acidity could also explain the total HMF conversion and the negligible DMF and DMTHF yield. In Figure S1 it can be observed that the impregnation of Ru in CuZr catalysts implied a higher production of DMTHF, but the total yield of DMF and DMTHF was not improved. Moreover, a decrease of the total yield after 3 – 4 h time on stream could be detected. Nevertheless, the presence of Ru improves the reducibility and therefore, the Cu/Zr ratio on the catalytic surface was quite higher than on the monometallic one. These catalysts suffered from high problems of deactivation as it can be observed in the lower Cu/Zr ratios for the used catalysts than the ones measured for the fresh reduced catalysts. Probably, the presence of Ru weakens the Cu-Zr interaction (see TPR results), favouring de Cu sintering (see XPS results). Due to the characteristic 21 of the RuCuZr catalyst, it can be expected higher DMF and DMTHF yields, but a similar yields to the ones corresoponding to the 15CuZr and 30CuZr catalysts were reached. This can be ascribed to the fact that the ruthenium could not catalysed the hydrogenolysis reaction as well as it was tested for the 1RuZr, where although a full conversion was reached, no DMF or DMTHF was detected. This means that there are two different main routes catalysed by Cu and Ru, and they compete for the same molecule (HMF). According to literature [86,87], the mixture of ZrO2 and CeO2 has been found to improve the thermal stability, catalytic activity and oxygen storage capacity. The latter refers to the ability to deliver oxygen from the lattice to the gas phase or to solid (adsorbed) carbon. Based on the general results, this effect can be related to a slight improvement of stability of the catalysts supported on Ce-Zr comparing with Zr supported catalysts. To analyse if this affects to the reaction, Ce was added to the Zr support. As it can be observed in Table 8, the presence of Ce on some catalysts hardly improved the selectivity to DMF, but in the case of the 30Cu5CeZr catalyst increased the stability of the 30CuZr catalyst, reducing the coke production (C/Zr used catalysts in Table 7). However, in the rest of XCuNCeZr catalysts this effect was negligible due to the low amount of cerium that it was incorporated to the Zr structure (see ICP-AES, XRD, SEM-EDX and XPS results). This low amount of impregnated Ce results on higher DMTHF yields (see Figure S1). The good behaviour registered for the 15CuZr catalyst, when compared with the 15Cu5CeZr and 15Cu20CeZr catalysts, could be due to two reasons: i) a better metal dispersion (see Table 5), favoured by stronger metal-support interaction observed by TPR, and ii) its low acidity (see Table 4), which reduced coke formation [88] (see Table 7). The expected effect of the cerium as eliminator of solid adsorbed carbon did not take place in the 15CuNCeZr catalysts. In the case of Zrand CeZr-supported 30Cu catalysts, the lower DMF and DMTHF yield obtained could be related to a lower metal dispersion and a higher acidity, which favoured higher coke formation. Based on the literature, Ni enables the cleavage of C–O bond [89,90], so NiCu bimetallic catalyst was also prepared by the co-impregnation method. It can be observed (See Table 8) that DMTHF yield achieves a maximum of 40.2 % after 4 h of reaction, with a total yield of DMF and DMTHF of 53.5 %, providing the highest products yields. According to TPR and XRD characterization data, this catalyst presents a synergic effect of Ni and Cu through the formation of a NiCu alloy species (high resolution XRD). These facts can be the explanation of this highest yield, since the literature reports that NiCu alloy phases are extremely active for hydrogenolysis reactions [91]. 3.2.2. Stability test of the 15Ni15CuZr catalyst Since the 15Ni15CuZr catalyst showed the most promising results in terms of DMF+DMTHF yields, the stability of this catalyst was tested measuring its activity for longer times on stream. Thus, as it can be observed in Figure 9, this catalyst is stable for the production of DMF, obtaining a yield of 25% after 24 h on stream. However, in the case of DMTHF, its yield starts to decrease after achieving a maximum yield of 40% after 4 h on stream. The stability of this catalyst could be ascribed to the synergic effect of Ni and Cu, which avoids any significant sintering effect as it can be observed analysing this catalyst data determined by high resolution XRD. In addition to this, the DMTHF yield decreased after 4 hours on stream, while the DMF yield increased after 22 22 hours on stream, remaining almost constant. It seems that the 15Ni15CuZr catalyst lost part of its initial hydrogenation capacity, and therefore its activity to farther hydrogenate DMF to DMTHF decreased along the time on stream. This loss of hydrogenation activity could be related to the of nickel dispersion decrease, as XPS data corroborated. It is well known that Ni presents a high hydrogenation capacity, higher than the one of Cu. That is why the loss of nickel surface area could decrease the catalyst hydrogenation capacity. Moreover, the recrystallization of NiCu observed in high resolution XRD results could also influence this DMTHF and DMF production variation. Figure 9. Stability results of the 15Ni15CuZr catalyst ( DMF yield and DMTHF yield). 3.3. Preliminary purification results The biofuels production not only include the development of efficient catalytic processes to convert biomass or biomass-derived building blocks into these biofuels. Farther processing is required to reach the purity required for the final liquid fuels formulations. In our case, as all the activity experiments were carried out using n-butanol as solvent and just 1.5 wt% of HMF as feed, the produced DMF and DMTHF are very diluted in the solvent and a purification train is required. For that purpose, a distillation process after the reactor outlet has been studied by means of process simulations using Aspen Plus. This simulation has been simplified using DMF, DMTHF and n-butanol as feed in the proportions obtained when operating with the bet developed catalyst (15Ni15CuZr). It can be concluded that a quite simple distillation process (20 equilibrium stages, 3.8 of reflux ratio and using 6.3 kW as reboiler duty) is enough in order to get a liquid fuel composed by 21.4 mol% of DMF, 63.2 mol% of DMTHF and 15.0 mol% of nbutanol as the distillate. Additionally, n-butanol liquid stream of 99.99 mol% would be recovered from the bottoms to be recycled. It must be indicated that butanol is also a good biofuel if produced by fermentation. However, if lower n-butanol concentration is required in the fuel, the height of the tower as well as the reboiler duty starts increasing considerably. Table 9 shows the fuel properties of the purified fuel with 15 mol% of n-butanol. Moreover, it can be observed 2 4 6 8 10 22 24 0 10 20 30 40 50 60 DMF and DMTHF yield (%) Time on stream (h) 23 that different DMF and DMTHF product distribution in the fuel does not affect considerably to the fuel properties. Table 9 Specifications of three different DMF, DMTHF and n-butanol compositions DMF mol% 0.21 0.64 0.43 DMTHF mol% 0.64 0.21 0.43 BUT mol% 0.15 0.15 0.15 Cetane number 16.90 15.64 16.24 Oxygen content (% weight) 16.8% 17.1% 16.9% Stoichiometric air/fuel ratio 11.43 10.82 11.13 Lower heating value (MJ/kg) 36.70 34.68 35.70 Density at 20 °C(g/cm3) 0.79 0.75 0.77 Viscosity (mm2/s) at 40 °C 0.66 0.69 0.687 Boiling point (°C) 95.26 96.48 95.86 4. CONCLUSIONS DMF and DMTHF present good characteristics for their aplicattion as high-quality biofuels or bio-additives for conventional liquid transportation fuels. These compounds can be continuously produced from HMF using a catalytic fixed-bed reactor. Through an extensive experimental work trying to maximize stable DMF and DMTHF production, the developed 15Ni15CuZr catalyst showed a total yield of DMF and DMTHF of 53.5 % after 4 hours on stream, maintaining only a stable DMF yield of 25 % after 24 hours operating at 15 bar, 275 ȗĂŶĚwith no pretreatment (reduction) of the catalyst. In order to reach commercial production some catalyst improvements are required. From this work some future orientations can be selected: i) zirconia is an adequate support because of its high surface area and limited acidity, ii) the Cu dispersion is a key parameter, therefore its necessary to improve the Cu dipersion to increase the DMF yield and reduce coke formation, iii) Ni addition generates the very active and stable NiCu solid alloy but excess Ni favors the production of over-hydrogenation products such as DMTHFand iv) a noble metal incorporation (Ru) or the zirconia modification using ceria did not generate positive effects. The buthanol is an excellent solvent for this hydrogenolysis process (from HMF to DMF) and its separation through distillation is a comparatively cheap process provided that part of the buthanol was also accepted as a component of the biofuel mixture to be commercialized. Acknowledgements This work was supported by University of the Basque Country (UPV/EHU), Spanish Ministry of Economy and Innovation and European Union through the European Regional Development Fund (FEDER) (Projects: CTQ2015-64226-C3-2-R), and Basque Country Government (Project: IT993-16). The authors also express their gratitude to the Analysis General Services (SGIker) attached to the UPV/EHU University for its assistance with the SEM analysis, and to the Pedro J. Maireles (University of Malaga) for its assistance with the XPS results. 24 REFERENCES [1] Zhang M, Tong X, Ma R, Li Y. Catalytic transformation of carbohydrates into 5hydroxymethyl furfural over tin phosphate in a water-containing system. Catal Today 2016;264:131–5. doi:10.1016/j.cattod.2015.06.031. [2] Qian Y, Zhu L, Wang Y, Lu X. Recent progress in the development of biofuel 2,5dimethylfuran. Renew Sustain Energy Rev 2015;41:633–46. doi:10.1016/j.rser.2014.08.085. [3] Nishimura S, Ikeda N, Ebitani K. Selective hydrogenation of biomass-derived 5hydroxymethylfurfural (HMF) to 2,5-dimethylfuran (DMF) under atmospheric hydrogen pressure over carbon supported PdAu bimetallic catalyst. Catal Today 2014;232:89–98. doi:10.1016/j.cattod.2013.10.012. [4] Zu Y, Yang P, Wang J, Liu X, Ren J, Lu G, et al. Efficient production of the liquid fuel 2,5dimethylfuran from 5-hydroxymethylfurfural over Ru/Co3O4 catalyst. Appl Catal B Environ 2014;146:244–8. doi:10.1016/j.apcatb.2013.04.026. [5] International Energy Agency (IEA). Key world energy statistics. Paris: 2016. [6] International Energy Agency. World Energy Outlook 2015. Paris: 2015. [7] International Energy Agency. Key Oil Trends. 2016. [8] Yang P, Cui Q, Zu Y, Liu X, Lu G, Wang Y. Catalytic production of 2,5-dimethylfuran from 5-hydroxymethylfurfural over Ni/Co3O4 catalyst. Catal Commun 2015;66:55–9. doi:10.1016/j.catcom.2015.02.014. [9] De S, Saha B, Luque R. Hydrodeoxygenation processes: Advances on catalytic transformations of biomass-derived platform chemicals into hydrocarbon fuels. Bioresour Technol 2015;178:108–18. doi:10.1016/j.biortech.2014.09.065. [10] Green Chemistry. Fraunhofer magazine 2.12. 2012. [11] Rosatella AA, Simeonov SP, Frade RFM, Afonso CAM. 5-Hydroxymethylfurfural (HMF) as a building block platform: Biological properties, synthesis and synthetic applications. Green Chem 2011;13:754–93. doi:10.1039/C0GC00401D. [12] Urbanchuk JM. Contribution of the Ethanol Industry To the Economy of the United States in 2014. 2015. [13] Iriondo A, Mendiguren A, Güemez MB, Requies J, Cambra JF. 2,5-DMF production through hydrogenation of real and synthetic 5-HMF over transition metal catalysts supported on carriers with different nature. Catal Today 2017;279:286–95. doi:10.1016/j.cattod.2016.02.019. [14] Menon V, Rao M. Trends in bioconversion of lignocellulose: Biofuels, platform chemicals & biorefinery concept. Prog Energy Combust Sci 2012;38:522–50. doi:10.1016/j.pecs.2012.02.002. [15] Climent MJ, Corma A, Iborra S. Conversion of biomass platform molecules into fuel additives and liquid hydrocarbon fuels. Green Chem 2014;16:516–47. doi:10.1039/C3GC41492B. [16] Choi S, Song CW, Shin JH, Lee SY. Biorefineries for the production of top building block chemicals and their derivatives. Metab Eng 2015;28:223–39. doi:10.1016/j.ymben.2014.12.007. 25 [17] Roman-Leshkov Y, Barrett CJ, Liu ZY, Dumesic JA. Production of dimethylfuran for liquid fuels from biomass-derived carbohydrates. Nature 2007;447:982–5. [18] Zhao J, Zhou C, He C, Dai Y, Jia X, Yang Y. Efficient dehydration of fructose to 5hydroxymethylfurfural over sulfonated carbon sphere solid acid catalysts. Catal Today 2016;264:123–30. doi:10.1016/j.cattod.2015.07.005. [19] Zhang Y, Zhang J, Su D. 5-Hydroxymethylfurfural: Akey intermediate for efficient biomass conversion. J Energy Chem 2015;24:548–51. doi:10.1016/j.jechem.2015.09.005. [20] Kläusli T. AVA Biochem: commercialising renewable platform chemical 5-HMF. Green Process Synth 2014;3:235–6. doi:10.1515/gps-2014-0029. [21] Tran LS, Sirjean B, Glaude PA, Fournet R, Battin-Leclerc F. Progress in detailed kinetic modeling of the combustion of oxygenated components of biofuels. Energy 2012;43:4– 18. doi:10.1016/j.energy.2011.11.013. [22] Brocard JC, Baronnet F, O’Neal HE. Chemical kinetics of the oxidation of methyl tertbutyl ether (MTBE). Combust Flame 1983;52:25–35. doi:10.1016/0010-2180(83)901189. [23] :ħǏĂŬ^njŝĚĂDŽƌħďƐŬŝD,ŝŐŚƉƌĞƐƐƵƌĞƉŚLJƐŝĐŽĐŚĞŵŝĐĂůƉƌŽƉĞƌƚŝĞƐŽĨϮmethylfuran and 2,5-dimethylfuran – second generation biofuels. Fuel 2016;184:334– 43. doi:10.1016/j.fuel.2016.07.025. [24] Chen G, Di L, Zhang Q, Zheng Z, Zhang W. Effects of 2,5-dimethylfuran fuel properties coupling with EGR (exhaust gas recirculation) on combustion and emission characteristics in common-rail diesel engines. Energy 2015;93:284–93. doi:10.1016/j.energy.2015.09.066. [25] Chen G, Shen Y, Zhang Q, Yao M, Zheng Z, Liu H. Experimental study on combustion and emission characteristics of a diesel engine fueled with 2,5-dimethylfuran-diesel, nbutanol-diesel and gasoline-diesel blends. Energy 2013;54:333–42. doi:10.1016/j.energy.2013.02.069. [26] Liu H, Xu J, Zheng Z, Li S, Yao M. Effects of fuel properties on combustion and emissions under both conventional and low temperature combustion mode fueling 2,5dimethylfuran/diesel blends. Energy 2013;62:215–23. doi:10.1016/j.energy.2013.09.057. [27] Zheng Z, Wang XF, Zhong X, Hu B, Liu H, Yao M. Experimental study on the combustion and emissions fueling biodiesel/n-butanol, biodiesel/ethanol and biodiesel/2,5dimethylfuran on a diesel engine. Energy 2016;115:539–49. doi:10.1016/j.energy.2016.09.054. [28] Zheng Z, Xia M, Liu H, Wang X, Yao M. Experimental study on combustion and emissions of dual fuel RCCI mode fueled with biodiesel/n-butanol, biodiesel/2,5-dimethylfuran and biodiesel/ethanol. Energy 2018;148:824–38. doi:10.1016/j.energy.2018.02.015. [29] Shi J, Wang Y, Yu X, Du W, Hou Z. Production of 2,5-dimethylfuran from 5hydroxymethylfurfural over reduced graphene oxides supported Pt catalyst under mild conditions. Fuel 2016;163:74–9. doi:10.1016/j.fuel.2015.09.047. [30] Conturso M, Sirignano M, Anna AD. Effect of 2,5-dimethylfuran doping on particle size distributions measured in premixed ethylene / air flames. Proc Combust Inst 2016;In press:1–8. doi:10.1016/j.proci.2016.06.048.