MCM-41 Supported Co-Based Bimetallic Catalysts for Aqueous Phase Transformation of Glucose to Biochemicals

The transformation of glucose into valuable biochemicals was carried out on different MCM-41-supported metallic and bimetallic (Co, Co-Fe, Co-Mn, Co-Mo) catalysts and under different reaction conditions (150 ◦C, 3 h; 200 ◦C, 0.5 h; 250 ◦C, 0.5 h). All catalysts were characterized using N2 physisorption, Temperature Programmed Reduction (TPR), Raman, X-ray Diffraction (XRD) and Temperature Programmed Desorption (TPD) techniques. According to the N2-physisorption results, a high surface area and mesoporous structure of the support were appropriate for metal dispersion, reactant diffusion and the formation of bioproducts. Reaction conditions, bimetals synergetic effects and the amount and strength of catalyst acid sites were the key factors affecting the catalytic activity and biochemical selectivity. Sever reaction conditions including high temperature and high catalyst acidity led to the formation mainly of solid humins. The NH3-TPD results demonstrated the alteration of acidity in different bimetallic catalysts. The 10Fe10CoSiO2 catalyst (MCM-41 supported 10 wt.%Fe, 10 wt.%Co) possessing weak acid sites displayed the best catalytic activity with the highest carbon balance and desired product selectivity in mild reaction condition. Valuable biochemicals such as fructose, levulinic acid, ethanol and hydroxyacetone were formed over this catalyst.


Introduction
The depletion of fossil fuels along with the environmental problems associated with their utilization promote new processes for the generation of fuels based on renewable sources [1]. The use of biomass, typically lignocellulosic biomass, in the production of fuels, fuel additives or added-value chemicals has attracted considerable interest, becoming a potential research area [2]. Biomass can be converted into biofuels and valuable chemicals via chemical or thermochemical processes; among them, aqueous phase reaction is an effective method to convert lignocellulose into biochemicals [3]. Lignocellulosic biomass is principally constituted of cellulose, hemicellulose and lignin [4,5]. Due to the complex nature of biomasses and their chemical compositions, they have different reactivities, and conversion processes can occur via various reaction pathways. Thus, researchers often prefer to use typical

Catalysts Synthesis
A silica mesoporous support (MCM-41) was synthesized following the procedure described by Ghedini et al. [24]. Hexadecyltrimethylammonium bromide (CTAB, Aldrich) was first dissolved in a NaOH aqueous solution at room temperature (r.t) under stirring; then, the required amount of tetraethyl orthosilicate (TEOS) was added. The resulting mixture was aged in an autoclave at 150 • C for 22 h, and thereafter filtrated, thoroughly washed and dried at room temperature. The surfactant was removed by calcination at 500 • C for 6 h in air flow (50 mL/min).
For the monometallic catalyst, the active phase was introduced on a silica support by incipient wetness impregnation using an aqueous solution of Co(NO 3 ) 2 ·6H 2 O in order to obtain 20 wt.% of metal loading.
For bimetallic samples, the precursors were introduced by co-impregnation of the previous Co solution and the corresponding precursors including (NH 4 ) 6 3 .9H 2 O) in order to obtain a nominal value of 10 wt% for each metal. Finally, the samples were dried and calcined at 500 • C in air flow (50 mL/min) for 6 h.

Nitrogen Physisorption
Nitrogen physisorption measurements were performed at −196 • C using a Micromeritics Tristar II Plus sorptometer (MICROMERITICS, Norcross, GA, USA). The sample (~400 mg) was outgassed at 200 • C for 2 h in vacuum prior to the sorption experiment. The surface area was calculated using the BET equation [25], and the total pore volume, V tot , was measured as the adsorbed amount of N 2 at P/P 0 values near 0.98. Pore size distribution was determined by the BJH method [26] applied to the N 2 adsorption isotherm branch [27,28].

Temperature Programmed Reduction (TPR)
TPR measurements were carried out with a lab-made instrument at CATMAT laboratory, Ca' Foscari University of Venice. The analysis was performed under 5% H 2 /Ar (40 mL/min) from 25 • C to 800 • C with a heating rate of 10 • C/min. The H 2 consumption was analyzed by a Micrometrics TPD-TPR 2900 analyzer equipped with a TCD detector (Gow-Mac 24-550 TCD instrument CO, Bethlehem, PA, USA).

X-ray Powder Diffraction (XRD)
A XRD (PW1769, Philips Analytical, Eindhoven, The Netherlands) using Cu-Kα (Ni-filtered) radiation was used for crystalline phase determination. The measured 2θ angle range was 10.0 • -70.0 • with a step size of 0.02 • and a counting time of 1.25 s per step. The size of the metal particle phase was obtained using the Scherrer equation [26]. The correction for instrument broadening was applied after background subtraction and curve-fitting procedures on the assumption of Lorentzian peak profiles.

Raman Spectroscopy
Raman spectra were collected on powder samples at room temperature in back-scattering geometry using an inVia Renishaw 1000 spectrometer equipped with an air-cooled, charge-coupled device (CCD) detector and edge filters. A 488.0 nm emission line from an Ar + laser was focused on the sample using a Leica DLML microscope with 5 × or 20 × objectives and an incident beam power of about 5 mW. A solid-state laser emitting at 785 nm with low power to avoid sample damage (about 2 mW) was used to analyze the 10Fe10CoSiO 2 catalyst. The spectra were calibrated using the 520.5 cm −1 line of a silicon wafer. The spectral resolution was 3 cm −1 . Data analyses included baseline removal and curve fitting  (NH 3 -TPD) NH 3 -TPD analyses of samples were carried out using lab-made equipment at CATMAT laboratory, Ca' Foscari University of Venice in order to study the acidity of the catalysts. First, 100 mg of the catalyst was charged in a quartz reactor and degassed in He with a flow rate of 40 mL/min at 500 • C for 90 min. The catalyst was then cooled to room temperature (25 • C) prior to adsorption of ammonia. Then, the adsorption of 5% NH 3 /He with a flow rate of 40 mL/min at 25 • C for 30 min was performed. The physisorbed ammonia was removed from the catalyst surface by passing He (40 mL/min) at room temperature for 10 min. The desorption profile of NH 3 -TPD was recorded using a Micrometrics TPD-TPR 2900 analyzer equipped with a thermal conductivity detector TCD (Gow-Mac 24-550 TCD instrument CO, Bethlehem, PA, USA) from 25 to 1000 • C at a heating rate of 10 • C/min under the flow of He (40 mL/min).

Aqueous Phase Transformation Catalytic Tests
Aqueous phase reforming (APR) tests were carried out in a 300 mL stainless-steel Parr autoclave loaded with a 0.3-3.0 wt.% solution of glucose in water and 0.45 g of catalyst. All the catalysts were pelletized and reduced at 500 • C for 3 h under a 10 % (v/v) H 2 /N 2 flow before each test.
The experiments were performed by placing the catalyst and glucose water solution (50 mL) into the autoclave. Thereafter, the sealed autoclave was first purged under N 2 flow for a few minutes to remove oxygen in the gas phase, and then heated to the desired temperature at 4.2 • C/min. When the desired temperature was reached, the reaction was started. The heating period was not considered in the reaction time. All the reactions were performed in a temperature range of 150-250 • C at autogenous pressure for different durations. At 150 • C, the tests were conducted for 3 h to allow glucose conversion and product formation, while at higher temperatures, the reaction was carried out for 0.5 h to avoid complete glucose transformation into humins. Considering the heating step, the following reaction times were employed: 3.5 h at 150 • C (0.5 h heating); 1.25 h at 200 • C (0.75 h heating); and 1.5 h at 250 • C (1,0 h heating). At the end of the reaction time, the autoclave was quenched in ice and allowed to cool to room temperature over 30-40 min. The reaction mixture was analyzed using Agilent HPLC over a Rezex ROA Organic Acid column (0.0025 M H 2 SO 4 eluent, oven temperature 60 • C and 0.6 mL/min flux) with a RID detector. Gas analyses were performed in an off-line Thermo Focus GC with a carbon molecular sieve column (Carbosphere 80/100 6 * 1/8) and TCD detector.
The glucose conversion, carbon balance and product yields were calculated using the following equations: Carbon balance (%) = i (mmol out)·(C atoms) (mmol subin ·C atoms glc)

·100
(2) where i represents the general product of the reaction. A couple of randomly chosen center-points were duplicated according to a statistical approach to estimate the variability of the results. The maximum standard deviation exceeded 0.03 and 3 for the conversion and the carbon balance, respectively.

Catalysts Characterization
A TEM image of MCM-41 as a catalyst support is shown in Figure 1. It demonstrates the presence of a highly ordered array and layered structure.

Catalysts Characterization
A TEM image of MCM-41 as a catalyst support is shown in Figure 1. It demonstrates the presence of a highly ordered array and layered structure. The 10Mn10Co SiO2 sample exhibits a broad peak at 2θ of 36.6° and barely detectable features at 2θ of 59° and 65° which were assigned to the most intense Co3O4 (311), (511), (440) planes; no peaks of manganese-containing phases were detected. The average crystallite size of Co3O4, determined by Scherrer equation (Lorentzian peak profile; 2θ, 36.6°), was ~30 nm for the 20CoSiO2 sample, ~ 10 nm for the 10Fe10CoSiO2 sample and ~ 6 nm for the 10Mn10CoSiO2 sample. An average crystallite size of about 10 nm was determined for Fe3O4 (2θ, 35.4°) in the 10Fe10CoSiO2 sample. These results indicate that in the bimetallic catalysts, the spreading of surface species was favored, with an ensuing reduction in material crystallinity [29].
As for the 10Mo10CoSiO2 system, the XRD pattern shows reflections due to MoO3 [JCPDS card 5-0508] and to CoMoO4, identified by the peaks at 2θ values of 26.4 and 31.9° and 35.5° [JCPDS card . The average crystallite size results of ~100 nm for MoO3 and ~50 nm for CoMoO4 are indicative of highly crystalline materials [29] Raman Spectroscopy was applied to obtain information concerning the chemical structure and molecular interactions among the various components in both monometallic and bimetallic samples ( Figure 2A). The 10Mn10Co SiO 2 sample exhibits a broad peak at 2θ of 36.6 • and barely detectable features at 2θ of 59 • and 65 • which were assigned to the most intense Co 3 O 4 (311), (511), (440) planes; no peaks of manganese-containing phases were detected. The average crystallite size of Co 3 O 4 , determined by Scherrer equation (Lorentzian peak profile; 2θ, 36.6 • ), was~30 nm for the 20CoSiO 2 sample,~10 nm for the 10Fe10CoSiO 2 sample and~6 nm for the 10Mn10CoSiO 2 sample. An average crystallite size of about 10 nm was determined for Fe 3 O 4 (2θ, 35.4 • ) in the 10Fe10CoSiO 2 sample. These results indicate that in the bimetallic catalysts, the spreading of surface species was favored, with an ensuing reduction in material crystallinity [29].
As for the 10Mo10CoSiO 2 system, the XRD pattern shows reflections due to MoO 3 [JCPDS card 5-0508] and to CoMoO 4 , identified by the peaks at 2θ values of 26.4 and 31.9 • and 35.5 • [JCPDS card . The average crystallite size results of~100 nm for MoO 3 and~50 nm for CoMoO 4 are indicative of highly crystalline materials [29] Raman Spectroscopy was applied to obtain information concerning the chemical structure and molecular interactions among the various components in both monometallic and bimetallic samples ( Figure 2A).  The 20CoSiO2 sample (Figure 2A), curve a) exhibits the sharp Raman-active modes (F2g, Eg, F2g, F2g and A1g, respectively) predicted for the Co3O4 spinel structure [30], confirming the XRD results. Co3O4 has a normal spinel structure with Co 2+ positioned at the tetrahedral site and Co 3+ at octahedral site. The most intense mode (A1g) is attributed to the octahedral site symmetry, whereas the weakest modes (F2g and Eg) are related to the combined vibrations of tetrahedral sites and octahedral oxygen motions [31]. In the 10Fe10CoSiO2 spectrum ( Figure 2A, curve b), the most intense bands (190, 472 and 676 cm −1 ) are assigned to Co3O4 modes [24], while the broad and low intensity one at about 305 cm −1 clearly identifies Fe3O4 nanoparticles [25], in line with XRD analysis. The presence of both Fe3O4 and Co3O4 is further confirmed by the broad features at 508 and 610 cm −1 and the asymmetric shape of the main band at 676 cm −1 , resulting from the superimposition of some of their bands. The 20CoSiO 2 sample (Figure 2A), curve a) exhibits the sharp Raman-active modes (F 2g , E g , F 2g , F 2g and A 1g , respectively) predicted for the Co 3 O 4 spinel structure [30], confirming the XRD results. Co 3 O 4 has a normal spinel structure with Co 2+ positioned at the tetrahedral site and Co 3+ at octahedral site. The most intense mode (A 1g ) is attributed to the octahedral site symmetry, whereas the weakest modes (F 2g and E g ) are related to the combined vibrations of tetrahedral sites and octahedral oxygen motions [31]. In the 10Fe10CoSiO 2 spectrum (Figure 2A, curve b), the most intense bands (190, 472 and 676 cm −1 ) are assigned to Co 3 O 4 modes [24], while the broad and low intensity one at about 305 cm −1 clearly identifies Fe 3 O 4 nanoparticles [25], in line with XRD analysis. The presence of both Fe 3 O 4 and Co 3 O 4 is further confirmed by the broad features at 508 and 610 cm −1 and the asymmetric shape of the main band at 676 cm −1 , resulting from the superimposition of some of their bands.
For the 10Mo10CoSiO 2 sample (Figure 2A), curve d), bands characteristic of the α-MoO 3 crystal phase [33] (sharp peaks at about 240, 285, 337, 376, 660, 815 and 994 cm −1 ) and of the CoMoO 4 structure [34] (broad feature at about 942 cm −1 ) were identified. As for the α-MoO 3 phase, the narrow peak at 994 cm −1 could be assigned to the terminal oxygen (Mo = O) stretching mode, and the peaks at 811 cm −1 and at 660 cm −1 were attributed to the doubly (Mo 2 -O) and triply coordinated (Mo 3 -O) oxygen stretching mode, respectively, whereas the low intensity of peaks in the 200-400 cm −1 region were due to the Mo-O bending modes.
In Figure 2B, the broad feature representing the 10Mn10CoSiO 2 sample (curve a) is inspected with reference to the spectral features of Co 3 O 4 (curve b) and Mn 3 O 4 (curve c) oxides. Both the oxides showed sharp peaks indicative of a crystalline structure, with the Mn 3 O 4 spectrum being characterized by an intense peak at 654 cm −1 assigned to Mn-O vibrations of manganese (II) ions in tetrahedral coordination [35]. With reference to the Co 3 O 4 spectrum, the addition of manganese species caused bands shift to a lower frequency, broadening and coalescence of some of the vibration modes in the 500-700 cm −1 region. These changes could have arisen from the formation of nanostructured species, which caused changes in the coordination and local symmetry of the pure oxide components [32,36]. Bands at 566 and 615 cm −1 identified by Curve fitting results ( Figure 2B), curve a inset) suggested that surface Mn 3 O 4 and Co-Mn mixed oxide species had formed [37,38], as also supported by bands at 487 cm −1 and at 657 cm −1 , arising from coalescence (bands at 471 and 512 cm −1 ) and down shift of Co 3 O 4 modes.
In order to investigate the specific surface areas and pore size distribution of the catalysts, nitrogen physisorption was performed. The adsorption-desorption profile of pristine silica support exhibited a type IV isotherm, which was typical of a high surface area mesoporous material ( Figure S2 black line), in accordance with IUPAC classification [28]. The BET surface area of the obtained material was 1000 m 2 /g and the pore volume of 0.4 cm 3 /g. A high surface area and the mesoporous structure of a catalyst have direct and indirect effects on the reaction results, i.e., increasing the active metal dispersion on the catalyst surface, which could make the active phase more accessible for the reactant and improve the activity of the catalyst. Moreover, a mesoporous catalyst structure made the reactant diffusion and product formation more efficient.
The adsorption-desorption profiles of Co-supported catalysts ( Figure S2) were similar to the isotherm profile of SiO 2 , proving that the pristine support kept its structure after the deposition of the metal phase. However, the isotherm and surface area values of the 10Mo10CoSiO 2 sample substantially changed compared to the support ones. The textural properties, including surface areas and pore volumes, decreased somewhat (Table 1) with the deposition of metal active phase. As shown in Table 1, the 10Mo10CoSiO 2 catalyst presented the lowest surface area; this was ascribed to the deposition of a large number of cobalt species inside the pores of the mesoporous silica structure. Moreover, according to the XRD analysis, the sample consisted of large crystallites of MoO 3 and CoMoO 4 , leading to a decrease in specific surface area due to pore blocking. A H 2 -TPR analysis was carried out to determine the reducibility of metal species and the possible interaction between the metals present on each catalyst. Figure 3 shows the profiles related to hydrogen consumption as a function of temperature for mono and bimetallic Co-based samples. A H2-TPR analysis was carried out to determine the reducibility of metal species and the possible interaction between the metals present on each catalyst. Figure 3 shows the profiles related to hydrogen consumption as a function of temperature for mono and bimetallic Co-based samples. The TPR profiles of the catalysts showed several peaks indicating the formation of different Cobalt/Promoters species and their various interactions with the support. The pattern of the monometallic 20CoSiO2 sample presented two major reduction peaks at 330 °C and 365 °C. The peak at a lower temperature (330 °C) was assigned to the reduction of Co3O4 species to CoO, while the second one (365 °C) was ascribed to the reduction of CoO to Co 0 [39][40][41]. The presence of a broad feature situated between 400 °C and 850 °C could instead be related to the reduction of cobalt species strongly interacting with the support [40].
The TPR profile of the 10Fe10CoSiO2 catalyst was similar to that of the 20CoSiO2 sample, even though the position of the second peak was slightly shifted toward a higher temperature (from 365 °C to 400 °C) and the broad feature centered at 490 °C disappeared, indicating stronger interaction with the support. In addition, the second peak at 400 °C could be ascribed to the reduction of both CoO to Co 0 and Fe3O4 to Fe 0 [42].
The TPR profile of the 10Mo10CoSiO2 catalyst showed two well-resolved reduction peaks centered at 515° and 640 °C , and a low intensity broad peak at 355 °C. With regards to the peak at 515 °C, it was not possible to assign it to specific metal species, since it may have been due to the reduction step of CoO to Co° shifting toward higher temperatures, or to the reduction of MoO3 to MoO2 which occurs over a temperature range of 450-650 °C [43]. On the other hand, the peak at 650°, which was not present in TPR profile of the monometallic catalysts (20CoSiO2), could be ascribed to the reduction of MoO2 to Mo° [44], and to the reduction of CoMnO4 [45].
Finally, compared to the 10CoSiO2 catalyst, the presence of Mn in the 10Mn10CoSiO2 catalyst led to a shift to higher temperature of the first reduction peak, which was related to the harder reduction of Co3O4→ Co 0 ; this was proven by using less hydrogen. However, the reduction profile of the 10Mn10CoSiO2 sample showed broad peaks at 560° and 790 °C , that could be related to the reduction of Mn 4+ and Mn 3+ to Mn 2+ 37 , and to the reduction of Co species, which have strong metalsupport interactions, and Co-Mn mixed oxide species, as also identified by Raman analysis [46,47].
NH3-TPD was performed in order to study the acidic features of the catalysts. The NH3-TPD spectra of all the catalysts are presented in Figure 4. Two peaks in two temperature ranges were found in the spectra of the catalysts, indicating the presence of two types of acidity. The peak in the temperature range of around 100-150 °C was associated with the week acidic sites, while that in the temperature range of approximately 700-850 °C was ascribed to the strong acidic sites [48]. 20CoSiO2 showed a small peak in a higher temperature range (750-850 °C), indicating the presence of slight amounts of strong acidic sites, probably due to the acidity of SiO2 and the formation of CoOx over the support. Compared to the Co monometallic catalyst, mixed Fe-Co and Mo-Co phases exhibited weak The TPR profiles of the catalysts showed several peaks indicating the formation of different Cobalt/Promoters species and their various interactions with the support. The pattern of the monometallic 20CoSiO 2 sample presented two major reduction peaks at 330 • C and 365 • C. The peak at a lower temperature (330 • C) was assigned to the reduction of Co 3 O 4 species to CoO, while the second one (365 • C) was ascribed to the reduction of CoO to Co 0 [39][40][41]. The presence of a broad feature situated between 400 • C and 850 • C could instead be related to the reduction of cobalt species strongly interacting with the support [40].
The TPR profile of the 10Fe10CoSiO 2 catalyst was similar to that of the 20CoSiO 2 sample, even though the position of the second peak was slightly shifted toward a higher temperature (from 365 • C to 400 • C) and the broad feature centered at 490 • C disappeared, indicating stronger interaction with the support. In addition, the second peak at 400 • C could be ascribed to the reduction of both CoO to Co 0 and Fe 3 O 4 to Fe 0 [42].
The TPR profile of the 10Mo10CoSiO 2 catalyst showed two well-resolved reduction peaks centered at 515 • and 640 • C, and a low intensity broad peak at 355 • C. With regards to the peak at 515 • C, it was not possible to assign it to specific metal species, since it may have been due to the reduction step of CoO to Co 0 shifting toward higher temperatures, or to the reduction of MoO 3 to MoO 2 which occurs over a temperature range of 450-650 • C [43]. On the other hand, the peak at 650 • , which was not present in TPR profile of the monometallic catalysts (20CoSiO 2 ), could be ascribed to the reduction of MoO 2 to Mo 0 [44], and to the reduction of CoMnO 4 [45].
Finally, compared to the 10CoSiO 2 catalyst, the presence of Mn in the 10Mn10CoSiO 2 catalyst led to a shift to higher temperature of the first reduction peak, which was related to the harder reduction of Co 3 O 4 →Co 0 ; this was proven by using less hydrogen. However, the reduction profile of the 10Mn10CoSiO 2 sample showed broad peaks at 560 • and 790 • C, that could be related to the reduction of Mn 4+ and Mn 3+ to Mn 2+ 37 , and to the reduction of Co species, which have strong metal-support interactions, and Co-Mn mixed oxide species, as also identified by Raman analysis [46,47]. NH 3 -TPD was performed in order to study the acidic features of the catalysts. The NH 3 -TPD spectra of all the catalysts are presented in Figure 4. Two peaks in two temperature ranges were found in the spectra of the catalysts, indicating the presence of two types of acidity. The peak in the temperature range of around 100-150 • C was associated with the week acidic sites, while that in the temperature range of approximately 700-850 • C was ascribed to the strong acidic sites [48]. 20CoSiO 2 showed a small peak in a higher temperature range (750-850 • C), indicating the presence of slight amounts of strong acidic sites, probably due to the acidity of SiO 2 and the formation of CoO x over the support. Compared to the Co monometallic catalyst, mixed Fe-Co and Mo-Co phases exhibited weak acid sites, while the mixed Mn-Co phase increased the intensity of the strong acidity peak. The various bimetallic phases showed different acidic properties [48][49][50].
Processes 2020, 8, x FOR PEER REVIEW 9 of 16 acid sites, while the mixed Mn-Co phase increased the intensity of the strong acidity peak. The various bimetallic phases showed different acidic properties [48][49][50].

Catalytic Test
In order to understand the reactivity of glucose under different reaction conditions, and to determine the effect of the catalysts, blank test were performed. The results obtained at different reaction conditions (150 °C for 3.0 h, 200 °C and 250 °C for 0.5 h) are shown in Figures 5 and 6 and Table 2.
The conversion of glucose without any catalyst was only 22% at 150 °C ( Figure 5). Fructose, C4-C5 sugars, glycerol and lactic acid were also detected. This suggests that the isomerization, retroaldol condensation and dehydration reactions occurred under reaction conditions without any catalyst. Further increases in temperature up to 200 and 250 °C resulted in the almost complete conversion of glucose (98 and 99%, respectively). However, the yields of the identified products did not increase appreciably. The obtained data evidenced high conversion and carbon loss at temperatures above 200 °C due to changes in the reaction pathways; at high temperatures, glucose was converted into both insoluble humins and soluble polymeric byproducts [51]. This could be attributed to the oligomerization reaction of both glucoses, HMF and other reaction products (fructose, other C4-C5 sugars). This underlines one of the main drawbacks of working with sugars [12]. For this reason, different catalysts under the same reaction conditions were screened to identify a reliable method to reduce these byproducts.
In the presence of 20Co/SiO2, even at a low reaction temperature (150 °C), the conversion of glucose was 58%, and fructose was identified as the main product according to NH3-TPD tests that showed the presence of slight number of strong acidic sites active in the isomerization of glucose to fructose. However, the temperature was too low to allow further conversion to occur of fructose to other products. Hence, at 200 °C, the presence of Co nearly completely converted the glucose, and the yield of lactic acid increased to 19%, along with larger amounts of HMF (14%) from the dehydration of fructose. As shown in the experimental data, it seems that the Co-based catalyst with few strong acid sites according to NH3-TPD analyses favored dehydration and the breakage of C-C bonds, in comparison to the blank experiment [52,53].
However, by further increasing the temperature to 250 °C, the yield sum was lower, with hydroxyacetone being the main product, followed by lactic acid. This caused a decrease in the carbon

Catalytic Test
In order to understand the reactivity of glucose under different reaction conditions, and to determine the effect of the catalysts, blank test were performed. The results obtained at different reaction conditions (150 • C for 3.0 h, 200 • C and 250 • C for 0.5 h) are shown in Figures 5 and 6 and Table 2. balance values, together with an increase in conversion due to the formation of humins at higher temperature. Among the tested bimetallic catalysts, the best conversion was achieved over 10Mo10CoSiO2 (77%), followed by 10Mn10CoSiO2 (64%) at 150 °C. Compared to the Co monometallic catalyst, 10Mn10CoSiO2 with strong acidity enhanced the C-C bond cleavage, facilitating the generation of C3 products (lactic acid, hydroxyacetone, glycerol) [54]. In contrast, the 10Fe10CoSiO2 catalyst with weak acidity exhibited the lowest activity for glucose conversion (48%). Thus, it was confirmed that the presence, amount and strength of acid sites are important variables in determining the extent of glucose conversion into biochemicals. This notwithstanding, the use of 10Fe10CoSiO2 resulted in a complete carbon balance, thereby inhibiting the pathway to the formation of humins due to the slight acidity of the catalyst. It was demonstrated that strong acidity can increase product polymerization and the production of humins, and therefore, cause a decrease in product yields and carbon balance. This effect was also demonstrated by the formation of levulinic acid for both 10Fe10CoSiO2 and 10Mo10CoSiO2 catalysts with weak acid sites [54,55].
At higher temperatures (200 and 250 °C), 10Mo10CoSiO2 showed lower performance in terms of glucose conversion and sum yields of products compared to 20CoSiO2 only.
Even though the substitution by Mn or Fe did not significantly affect the conversion of glucose and the carbon balance at 200 °C compared to those obtained using only Co, different reaction product distributions occurred. For instance, the presence of Mn promoted the formation of a wide range C3 products, C4-C5 sugars and methanol, thus enhancing C-C cleavage and dehydration. In comparison, 10Fe10CoSiO2 promoted the production of ethanol and hydroxyacetone. This might stem from the synergistic effect between the two types of metal species and the difference of acidity in the two catalysts [54,56].
Completely different behavior was observed with the bimetallic catalysts at 250 °C. The data obtained from the experiment using 10Mn10CoSiO2 show the lowest carbon balance among the tested catalysts. Considering the reaction product distribution, it seems that the formation of humins could be due to the condensation reaction of C4-C5 sugars, one of the main products at 200 °C, under harsh reaction conditions, as demonstrated in NH3-TPD results.
The best balance (46%) was obtained at 250 °C, with complete conversion using the 10Fe10CoSiO2 catalyst, owing to its lower acidity, as shown by NH3-TPD tests. The main product was 5-HMF, due to the dehydration of fructose. Ethanol was produced with a yield of 11%, while the yield of lactic acid reached 9%.   The conversion of glucose without any catalyst was only 22% at 150 • C ( Figure 5). Fructose, C4-C5 sugars, glycerol and lactic acid were also detected. This suggests that the isomerization, retro-aldol condensation and dehydration reactions occurred under reaction conditions without any catalyst. Further increases in temperature up to 200 and 250 • C resulted in the almost complete conversion of glucose (98 and 99%, respectively). However, the yields of the identified products did not increase appreciably. The obtained data evidenced high conversion and carbon loss at temperatures above 200 • C due to changes in the reaction pathways; at high temperatures, glucose was converted into both insoluble humins and soluble polymeric byproducts [51]. This could be attributed to the oligomerization reaction of both glucoses, HMF and other reaction products (fructose, other C4-C5 sugars). This underlines one of the main drawbacks of working with sugars [12]. For this reason, different catalysts under the same reaction conditions were screened to identify a reliable method to reduce these byproducts.
In the presence of 20Co/SiO 2 , even at a low reaction temperature (150 • C), the conversion of glucose was 58%, and fructose was identified as the main product according to NH 3 -TPD tests that showed the presence of slight number of strong acidic sites active in the isomerization of glucose to fructose. However, the temperature was too low to allow further conversion to occur of fructose to other products. Hence, at 200 • C, the presence of Co nearly completely converted the glucose, and the yield of lactic acid increased to 19%, along with larger amounts of HMF (14%) from the dehydration of fructose. As shown in the experimental data, it seems that the Co-based catalyst with few strong acid sites according to NH 3 -TPD analyses favored dehydration and the breakage of C-C bonds, in comparison to the blank experiment [52,53].
However, by further increasing the temperature to 250 • C, the yield sum was lower, with hydroxyacetone being the main product, followed by lactic acid. This caused a decrease in the carbon balance values, together with an increase in conversion due to the formation of humins at higher temperature.
Co Among the tested bimetallic catalysts, the best conversion was achieved over 10Mo10CoSiO 2 (77%), followed by 10Mn10CoSiO 2 (64%) at 150 • C. Compared to the Co monometallic catalyst, 10Mn10CoSiO 2 with strong acidity enhanced the C-C bond cleavage, facilitating the generation of C3 products (lactic acid, hydroxyacetone, glycerol) [54]. In contrast, the 10Fe10CoSiO 2 catalyst with weak acidity exhibited the lowest activity for glucose conversion (48%). Thus, it was confirmed that the presence, amount and strength of acid sites are important variables in determining the extent of glucose conversion into biochemicals. This notwithstanding, the use of 10Fe10CoSiO 2 resulted in a complete carbon balance, thereby inhibiting the pathway to the formation of humins due to the slight acidity of the catalyst. It was demonstrated that strong acidity can increase product polymerization and the production of humins, and therefore, cause a decrease in product yields and carbon balance. This effect was also demonstrated by the formation of levulinic acid for both 10Fe10CoSiO 2 and 10Mo10CoSiO 2 catalysts with weak acid sites [54,55].
At higher temperatures (200 and 250 • C), 10Mo10CoSiO 2 showed lower performance in terms of glucose conversion and sum yields of products compared to 20CoSiO 2 only.
Even though the substitution by Mn or Fe did not significantly affect the conversion of glucose and the carbon balance at 200 • C compared to those obtained using only Co, different reaction product distributions occurred. For instance, the presence of Mn promoted the formation of a wide range C3 products, C4-C5 sugars and methanol, thus enhancing C-C cleavage and dehydration. In comparison, 10Fe10CoSiO 2 promoted the production of ethanol and hydroxyacetone. This might stem from the synergistic effect between the two types of metal species and the difference of acidity in the two catalysts [54,56].
Completely different behavior was observed with the bimetallic catalysts at 250 • C. The data obtained from the experiment using 10Mn10CoSiO 2 show the lowest carbon balance among the tested catalysts. Considering the reaction product distribution, it seems that the formation of humins could be due to the condensation reaction of C4-C5 sugars, one of the main products at 200 • C, under harsh reaction conditions, as demonstrated in NH 3 -TPD results.
The best balance (46%) was obtained at 250 • C, with complete conversion using the 10Fe10CoSiO 2 catalyst, owing to its lower acidity, as shown by NH 3 -TPD tests. The main product was 5-HMF, due to the dehydration of fructose. Ethanol was produced with a yield of 11%, while the yield of lactic acid reached 9%.

Conclusions
MCM-41-supported Co and bimetallic Co-Mn, Co-Mo and Co-Fe catalysts were investigated regarding the aqueous phase transformation of glucose into biochemicals, operating under different reaction conditions. With increases in temperature and the number and strength of catalyst acid sites, the conversion of glucose increased, but resulted in a low carbon balance due to the formation of humins. Isomerization was the predominant reaction at lower temperatures, while humins prevailed at higher temperatures, longer reaction times and higher catalyst acidity. The synergy of Fe, Mn and Mo with the Co increased the activity at 150 and 200 • C, while 250 • C hydrothermal conditions favored the retro aldol condensation reaction of glucose and its intermediates (fructose, C4-C5 sugars and 5-HMF) to form humins. The best catalytic reactivity was obtained under mild reaction conditions with the weak acidic sites of the 10Fe10CoSiO 2 catalyst, yielding valuable biochemicals such as fructose, levulinic acid, ethanol and hydroxyacetone.