Ru-(Mn-M)OX Solid Base Catalysts for the Upgrading of Xylitol to Glycols in Water

A series of Ru-(Mn-M)OX catalysts (M: Al, Ti, Zr, Zn) prepared by co-precipitation were investigated in the hydrogenolysis of xylitol in water to ethylene glycol, propylene glycol and glycerol at 200 °C and 60 bar of H2. The catalyst promoted with Al, Ru-(Mn-Al)OX, showed superior activity (57 h−1) and a high global selectivity to glycols and glycerol of 58% at 80% xylitol conversion. In comparison, the catalyst prepared by loading Ru on (Mn-Al)OX, Ru/(Mn-Al)OX was more active (111 h−1) but less selective (37%) than Ru-(Mn-Al)OX. Characterization of these catalysts by XRD, BET, CO2-TPD, NH3-TPD and TEM showed that Ru/(Mn-Al)OX contained highly dispersed and uniformly distributed Ru particles and fewer basic sites, which favored decarbonylation, epimerization and cascade decarbonylation reactions instead of retro-aldol reactions producing glycols. The hydrothermal stability of Ru-(Mn-Al)OX was improved by decreasing the xylitol/catalyst ratio, which decreased the formation of carboxylic acids and enabled recycling of the catalyst, with a very low deactivation.


Introduction
Ethylene glycol (EG) and propylene glycol (PG) are employed as raw or starting materials in various useful industrial applications, such as heat-transfer fluids and cosmetics, in the pharmaceutics and packaging sectors [1,2]. The global production of these glycols is currently based on the hydration of petroleum-based ethylene and propylene via epoxides intermediates [3], with 25 million metric tons being produced in 2017. Given recent strategies to diversify carbon sources, the use of bio-alditols, including xylitol and sorbitol, which can be obtained from inedible cellulose and hemicelluloses, appears to be a promising complementary process. Indeed, there are several advantages including the high C/O molar ratio of carbohydrates, the smaller environmental footprint, and their abundance [4][5][6].
In the last few decades, it has been reported that the upgrading of alditols to glycols can be performed in the aqueous phase under harsh conditions with temperatures of 160-230 • C and H 2 pressure (40-120 bar). The selective hydrogenolysis is challenging since parallel and consecutive reactions can occur and lead to a complex mixture of products [7][8][9][10]. The initial dehydrogenation of xylitol leads to an aldose (preferential route) or a ketose (minor route), according to the different rates of dehydrogenation of the primary or secondary hydroxyl groups. Then, C-C cleavage forms C 3 C2 intermediates by the retro-aldol reaction of the aldose or ketose catalyzed by base promotors, such as Ca(OH)2. They are then hydrogenated to glycols or glycerol. Therefore, the (de)hydrogenation reaction is catalyzed by a heterogeneous metal-based catalyst and the retro-aldol step by an alkaline base. The decarbonylation reaction may also be catalyzed by the metallic sites and forms C4 alditols (i.e., threitol) which can lead to the formation of butanediols (BDO) and CO via dehydroxylation and chain decarbonylation reactions [11]. In parallel, the epimerization reaction of xylitol produces other C5 alditols (arabitol and adonitol) on metallic sites [12]. Moreover, lactate (LA) and 2hydroxybutyrate (2-HBA) can be produced in the presence of a base by Cannizzaro-type reaction [9]. The main products observed in the hydrogenolysis of xylitol are reported in Figure 1. Thus, different strategies have been exploited in order to enhance the catalytic activity as well as glycols yield. Many metal-based catalysts, such as Cu, Ni, Ru, and Pt, have been studied for the hydrogenolysis of sugar alcohols to glycols, promoted or not by an alkali.
In the presence of Ca(OH)2 base, 90%Cu-SiO2 catalysts achieved glycols and glycerol yields of 50-60% in the aqueous phase at 200 °C under 60 bar of H2. However, the catalytic activity was relatively low [13]. Ni-based catalysts, such as 6%Ni-NaY [14] and bimetallic 10%Ni-80%Cu-SiO2 [15] catalysts displayed high selectivity to glycols and glycerol (up to 85%) at high conversion of the alditols in the presence of Ca(OH)2. The Ru-based catalysts seem to be the most interesting candidates due to their high activity and their ageing resistance. Sun and Liu [9] reported a selectivity to glycols of 67% in the presence of 4%Ru/C catalyst in the hydrogenolysis of xylitol (Ca(OH)2, 10 wt % xylitol in the aqueous phase, 200 °C and 40 bar H2). Zhao et al. [16] observed a selectivity of 61% to glycols and glycerol over 3% Ru/carbon nanofibers. More recently, N-containing supports (aminefunctionalized carbon nanotubes or covalent triazine framework) improved the performance of Rubased catalysts in the presence of a base, in terms of activity [17] or yield of glycols and glycerol up to 85% [18].
Among strategies investigated to avoid both the addition of alkaline promoters and the formation of carboxylates, the use of bifunctional catalysts combining both metallic and basic sites has been studied. For instance, Ni-MgO catalysts [19] displayed a total selectivity to glycols and glycerol of 80% at 68% sorbitol conversion (in water, 200 °C, 40 bar H2, 4 h). Some mixed oxide catalysts were also developed. Ni supported on fly ash (a waste mainly composed of SiO2, Al2O3 and Fe2O3) showed a high global selectivity to glycols and glycerol (58% at 41% conversion) as well as high hydrothermal stability for four consecutive runs [20]. Catalysts supported on Ca-containing supports or directly on Ca(OH)2 reached up to 60-70% selectivity to glycols at high conversion under base-free conditions. These include Cu-CaO-Al2O3 [21], Ni-Ru/Ca(OH)2 (not stable because of the solubility of the support) [22], and more recently Ni-Ca(Sr)/hydroxyapatite catalysts [23]. Ni-Mg-Al-O hydrotalcite-like catalysts with different Mg/Al ratios were reported to be relatively stable during five runs with reasonable production of glycols [24]. In a previous study, we investigated the Mn promotion of a Ru/C catalyst for hydrogenolysis of xylitol in water [10]. We reported high activity (221 h −1 ) and reasonable selectivity to glycols and glycerol (32%) over a 3%Ru/MnO(4.5%)/C catalyst. Moreover, we observed that the activity increased with the Mn-loading up to 384 h −1 and the selectivity to glycols and glycerol to 50% over 3.1%Ru/MnO(19.3%)/C catalyst. The selectivity over the same 3%Ru/MnO(4.5%)/C catalyst was then enhanced up to 70% in a water/alcohol mixture Many metal-based catalysts, such as Cu, Ni, Ru, and Pt, have been studied for the hydrogenolysis of sugar alcohols to glycols, promoted or not by an alkali.
In the presence of Ca(OH) 2 base, 90%Cu-SiO 2 catalysts achieved glycols and glycerol yields of 50-60% in the aqueous phase at 200 • C under 60 bar of H 2 . However, the catalytic activity was relatively low [13]. Ni-based catalysts, such as 6%Ni-NaY [14] and bimetallic 10%Ni-80%Cu-SiO 2 [15] catalysts displayed high selectivity to glycols and glycerol (up to 85%) at high conversion of the alditols in the presence of Ca(OH) 2 . The Ru-based catalysts seem to be the most interesting candidates due to their high activity and their ageing resistance. Sun and Liu [9] reported a selectivity to glycols of 67% in the presence of 4%Ru/C catalyst in the hydrogenolysis of xylitol (Ca(OH) 2 , 10 wt % xylitol in the aqueous phase, 200 • C and 40 bar H 2 ). Zhao et al. [16] observed a selectivity of 61% to glycols and glycerol over 3% Ru/carbon nanofibers. More recently, N-containing supports (amine-functionalized carbon nanotubes or covalent triazine framework) improved the performance of Ru-based catalysts in the presence of a base, in terms of activity [17] or yield of glycols and glycerol up to 85% [18].
Among strategies investigated to avoid both the addition of alkaline promoters and the formation of carboxylates, the use of bifunctional catalysts combining both metallic and basic sites has been studied. For instance, Ni-MgO catalysts [19] displayed a total selectivity to glycols and glycerol of 80% at 68% sorbitol conversion (in water, 200 • C, 40 bar H 2 , 4 h). Some mixed oxide catalysts were also developed. Ni supported on fly ash (a waste mainly composed of SiO 2 , Al 2 O 3 and Fe 2 O 3 ) showed a high global selectivity to glycols and glycerol (58% at 41% conversion) as well as high hydrothermal stability for four consecutive runs [20]. Catalysts supported on Ca-containing supports or directly on Ca(OH) 2 reached up to 60-70% selectivity to glycols at high conversion under base-free conditions. These include Cu-CaO-Al 2 O 3 [21], Ni-Ru/Ca(OH) 2 (not stable because of the solubility of the support) [22], and more recently Ni-Ca(Sr)/hydroxyapatite catalysts [23]. Ni-Mg-Al-O hydrotalcite-like catalysts with different Mg/Al ratios were reported to be relatively stable during five runs with reasonable production of glycols [24]. In a previous study, we investigated the Mn promotion of a Ru/C catalyst for hydrogenolysis of xylitol in water [10]. We reported high activity (221 h −1 ) and reasonable selectivity to glycols and glycerol (32%) over a 3%Ru/MnO(4.5%)/C catalyst. Moreover, we observed that the activity increased with the Mn-loading up to 384 h −1 and the selectivity to glycols and glycerol to 50% over 3.1%Ru/MnO(19.3%)/C catalyst. The selectivity over the same 3%Ru/MnO(4.5%)/C catalyst was then enhanced up to 70% in a water/alcohol mixture solvent, which avoided the competitive decarbonylation reaction on the Ru sites [25]. Nevertheless, the catalyst had a poor hydrothermal stability.
Hereby, we developed promising mixed-oxide catalysts Ru-(Mn-M)O X catalysts (M: Al, Ti, Zr, Zn) synthetized by co-precipitation, calcination and reduction. We compared different methods of Ru incorporation, which influenced the activity and the product distribution. Finally, we demonstrated that the Ru-(Mn-Al)O X catalyst stability could be improved by varying the operating conditions.

Catalyst Characterization
The series of Ru-(Mn-M)O X catalysts were obtained by thermal decomposition of co-precipitated precursors and reduction. For comparison, a Ru/(Mn-Al)O X was prepared by impregnation of a Ru salt on (Mn-Al)O X mixed oxide. The catalyst synthesis methods are reported in the Experimental Section. The M metal loading, the specific surface areas, the Ru crystallite sizes, and the number of acidic or basic sites are shown in Table 1. The loading of Ru could not be measured experimentally by ICP-OES, since Ru could not be totally solubilized, either in the solution of H 2 SO 4 and HNO 3 , or in concentrated HF solution. However, in the filtrate recovered after impregnation of Ru precursor, no traces of Ru were detected. The actual M loadings (i.e., Al, Ti, Zn, Zr) were very different from the nominal ones; in particular up to 97% of Al was lost ( Table 1, entry 1). This might originate from the too high pH applied during the synthesis of the material, which is outside the limits of precipitation of aluminum hydrate and therefore alkaline dissolution of amorphous Al(OH) 3 to soluble sodium aluminate occurs. ICP-OES analysis of the solids after each synthesis step demonstrated that M loss took place during the filtration step; the M precursor had not been entirely precipitated despite the basic pH (12), the temperature (80 • C) and the ageing time (20 h). Meanwhile, whatever the method, the experimental Mn loadings were close to the nominal loadings. The specific surface area of the mixed-oxide catalysts varied in a large range from 13 to 61 m 2 g −1 . The catalyst where the Ru was deposited on the mixed-oxide (Ru/(Mn-Al)O X , Table 1, entry 1) possessed a lower surface area than the catalyst where the Ru precursor was co-precipitated (Ru-(Mn-Al)O X , Table 1 [26]. The lowest specific area is associated with the lowest M content in the mixed oxide catalysts. This is not surprising, since MnO X exhibits low surface area of 24 m 2 g −1 [10].
The XRD patterns of the Ru-(Mn-Al)O X catalyst after each step of the synthesis highlighted the modification of the oxide species during the different thermal treatments, namely drying, calcination at 460 • C under air, and reduction at 450 • C under H 2 flow (Figure 2a). After drying, the only detectable crystalline phase was Mn 3 O 4 , whereas no peak corresponding to Al 2 O 3 could be observed. The calcination step led to a mixture of Mn 3 O 4 and Mn 5 O 8 , as reported in the literature under the same atmosphere and similar temperature [27]. The Mn 3 O 4 species consists of tetrahedral Mn(II) and octahedral Mn(III) units, which stem from partial oxidation of Mn(II) under hydrothermal conditions, as shown on the Pourbaix diagram [28,29]. The Mn 5 O 8 species is a metastable oxide of equal amounts of Mn(II) and Mn(IV) [30]. Meanwhile, no peak corresponding to RuO 2 was observed, suggesting an amorphous phase or a small crystallite size (<2 nm). After reduction treatment, only peaks attributed to MnO were detected. Ru-(Mn-Zr)O X , Ru-(Mn-Zn)O X and Ru-(Mn-Ti)O X samples presented similar patterns (Figure 2b). Furthermore, despite the high M amount in some of the catalysts, no peak attributed to the M oxide was observed, except ZnO in the Ru-(Mn-Zn)O X material. This observation suggests that TiO 2 and ZrO 2 were amorphous, unlike in a Cu-ZrO 2 -MgO material obtained after similar thermal treatment [31]. It is worth noting that the pattern of Ru-(Mn-Al)O X recorded with a higher time per step ( Figure A1) exhibited peaks associated with Mn 3 O 4 in the 2θ region of 28-38 • . This phase might also be present in the other Ru-(Mn-M)O X . The Ru 0 crystallites were detected in all XRD patterns with different sizes depending on the synthesis method ( Table 1). The Ru-(Mn-M)O X catalysts prepared by co-precipitation exhibited higher Ru mean crystallite sizes (13-29 nm) than the catalyst prepared by deposition of Ru over the (Mn-Al)O X support (Ru/(Mn-Al)O X , 6 nm). This is probably due to the calcination step. As reported in the literature for 4%Ru/Al 2 O 3 catalyst [32], small RuO 2 crystallite sizes (<2 nm) are formed upon calcination, however, sintering of Ru particles occurs during reduction. On the other hand, the Ru/(Mn-Al)O X catalyst underwent the reduction step without the initial calcination step leading to a small mean crystallite size of Ru. same atmosphere and similar temperature [27]. The Mn3O4 species consists of tetrahedral Mn(II) and octahedral Mn(III) units, which stem from partial oxidation of Mn(II) under hydrothermal conditions, as shown on the Pourbaix diagram [28,29]. The Mn5O8 species is a metastable oxide of equal amounts of Mn(II) and Mn(IV) [30]. Meanwhile, no peak corresponding to RuO2 was observed, suggesting an amorphous phase or a small crystallite size (<2 nm). After reduction treatment, only peaks attributed to MnO were detected. Ru-(Mn-Zr)OX, Ru-(Mn-Zn)OX and Ru-(Mn-Ti)OX samples presented similar patterns ( Figure 2b). Furthermore, despite the high M amount in some of the catalysts, no peak attributed to the M oxide was observed, except ZnO in the Ru-(Mn-Zn)OX material. This observation suggests that TiO2 and ZrO2 were amorphous, unlike in a Cu-ZrO2-MgO material obtained after similar thermal treatment [31]. It is worth noting that the pattern of Ru-(Mn-Al)OX recorded with a higher time per step ( Figure A1) exhibited peaks associated with Mn3O4 in the 2θ region of 28-38° This phase might also be present in the other Ru-(Mn-M)OX. The Ru 0 crystallites were detected in all XRD patterns with different sizes depending on the synthesis method ( Table 1). The Ru-(Mn-M)OX catalysts prepared by co-precipitation exhibited higher Ru mean crystallite sizes (13-29 nm) than the catalyst prepared by deposition of Ru over the (Mn-Al)OX support (Ru/(Mn-Al)OX, 6 nm). This is probably due to the calcination step. As reported in the literature for 4%Ru/Al2O3 catalyst [32], small RuO2 crystallite sizes (<2 nm) are formed upon calcination, however, sintering of Ru particles occurs during reduction. On the other hand, the Ru/(Mn-Al)OX catalyst underwent the reduction step without the initial calcination step leading to a small mean crystallite size of Ru.   Figure 3a) showed a homogeneous size of small Ru particles (3-7 nm) gathered together as wormlike nanoparticles. These TEM observations are in accordance with the Ru crystallite size measured by XRD. Significant differences in the size and distribution of Ru particles may influence the activity and the product distribution in hydrogenolysis of xylitol.
The acid-base properties were investigated by CO 2 and NH 3 TPD. In the case of Al-containing materials, the result was dependent on the method of introduction of Ru. In CO 2 -TPD profiles (Figure 4a), three main peaks at 175, 250 and 450 • C were detected and attributed to weak, moderate and strong basic sites, respectively. Ru/(Mn-Al)O X catalyst contained mostly weak basic sites. The signal at 450 • C would be associated with CO 2 adsorbed on strong basic sites or on Ru particles, according to the literature [10,33]. The Ru-(Mn-M)O X catalysts showed different desorption patterns that are a function of the nature of M. Catalysts Ru-(Mn-M)O X with M: Zr, Zn, or Ti contained a majority of weak and moderate basic sites due to the presence of possible ZrO 2 , ZnO and TiO 2 phases [34,35]. Stronger basic sites were observed on Ru-(Mn-Al)O X catalysts, which is likely due to the higher content of Mn in this material. The total number of basic sites (in µmol g −1 ) was as follows: Table 1). Since the M oxides have an amphoteric character, acid sites were also analyzed ( Figure 4b). They are present in lower amounts than basic sites (18 to 83 µmol g −1 , i.e., acid density of 1.3-2.2 µmol m −2 ). weak and moderate basic sites due to the presence of possible ZrO2, ZnO and TiO2 phases [34,35]. Stronger basic sites were observed on Ru-(Mn-Al)OX catalysts, which is likely due to the higher content of Mn in this material. The total number of basic sites (in µmol g −1 ) was as follows: Ru/(Mn-Al)OX (28) (Table 1). Since the M oxides have an amphoteric character, acid sites were also analyzed ( Figure 4b). They are present in lower amounts than basic sites (18 to 83 µmol g −1 , i.e., acid density of 1.3-2.2 µmol m −2 ).
(a) (b)   weak and moderate basic sites due to the presence of possible ZrO2, ZnO and TiO2 phases [34,35]. Stronger basic sites were observed on Ru-(Mn-Al)OX catalysts, which is likely due to the higher content of Mn in this material. The total number of basic sites (in µmol g −1 ) was as follows: Ru/(Mn-Al)OX (28)     To summarize, the introduction of a M promoting element modified the specific surface area and total number of acid and basic sites, whereas the MnO X phases (Mn 3 O 4 and MnO) and the Ru crystallite size remained relatively the same. Meanwhile, the two methods of addition of Ru (via co-precipitation and deposition methods) showed that Ru/(Mn-Al)O X contained small Ru particles, however, they were gathered while Ru-(Mn-Al)O X contained small and large Ru particles with a heterogeneous distribution. These differences may influence considerably the catalytic performances in hydrogenolysis of xylitol.

Hydrogenolysis of Xylitol
The reactions were conducted in water at 200 • C and 60 bar H 2 , without any alkali addition. The results of activity and selectivity to the main products are shown in Table 2. The selectivities are given at 80% conversion when the maximum concentration of the desired products was reached, as described previously [10,25]. The total carbon balance (comparison between TOC measured and TOC calculated from HPLC analysis) was around 90%. The catalytic performance depended on the mode of introduction of Ru in the Al-promoted materials. Indeed, the activity was halved from 111 h −1 for Ru/(Mn-Al)O X to 57 h −1 for Ru-(Mn-Al)O X . In parallel, the selectivities towards glycols were enhanced in the presence of Ru-(Mn-Al)O X (21% to EG and 28% to PG) vs. Ru/(Mn-Al)O X (12% to EG and 17% to PG). The selectivity to lactic acid (LA) and to glycerol (GLY) remained the same. On the contrary, the cumulated selectivity to C 4 and C 5 alditols decreased from 17% to 7%. Moreover, the products transferred into the gas phase (determined as the difference of initial TOC and TOC measured in the liquid phase) were produced in lower amounts over Ru-(Mn-Al)O X . This observation suggests that the side reactions of cascade decarbonylation leading to the degradation of glycols, particularly EG, were minimized [25]. This was confirmed by the calculation of the molar ratio of C 3 to C 2 products, which was 1.3 and 1.9 for the reactions in the presence of Ru-(Mn-Al)O X and Ru/(Mn-Al)O X , respectively, while the theoretical molar ratio is 1.0 for a totally selective hydrogenolysis reaction of xylitol to EG plus PG. Therefore, the Ru-(Mn-Al)O X catalyst favored the retro-aldol reaction instead of the decarbonylation reactions. In contrast, Ru/(Mn-Al)O X contributed to the decarbonylation reactions. Since the latter reactions take place on Ru particles, the larger particle sizes of Ru-(Mn-Al)O X (as shown by TEM and XRD) may explain why side-reactions observed at 80% xylitol conversion are not significant, in addition to showing lower activity. Otherwise, the presence of a higher number of basic sites on Ru-(Mn-Al)O X than on Ru/(Mn-Al)O X favors the retro-aldol reaction to the detriment of epimerization or cascade decarbonylation reactions, which explains the decrease in catalytic activity.
The Ru-(Mn-Zn)O X material was not active at all and no conversion was measured after a reaction time of 30 h. A comparable result was observed in our previous study using Ru/ZnO/C catalyst; the catalytic activity decreased from 220 to 13 h −1 in comparison with Ru/MnO/C catalyst [25]. Yet, previous studies have reported that the combination of Cu and ZnO was active and selective in the hydrogenolysis of alditols to glycols [21,[36][37][38]. The activity of Ru-(Mn-Zr)O X and Ru-(Mn-Ti)O X catalysts was 33 h −1 , i.e., lower than the activity of Ru-(Mn-Al)O X (57 h −1 ). The product distribution was quite similar whatever the promoter. At 80% xylitol conversion, the cumulated selectivity to glycols and GLY was as high as 59% and the values of selectivity to C 4 and C 5 products were maintained in the range 7-10%. The selectivity to gaseous products increased slightly in the presence of Ru-(Mn-Zr)O X and Ru-(Mn-Ti)O X to 23% and 26%, respectively, vs. 20% over Ru-(Mn-Al)O X catalyst. Thus, in general, despite the relatively high amount of M introduced (10 wt % Zr and 7.0 wt % Ti) and the various total number of acid and basic sites, the use of M promotor had no significant effect on the product distribution. This suggests that a low number of basic sites was sufficient to efficiently perform the retro-aldol reaction, whereas the acid sites favor the cascade decarbonylation to produce gaseous products. Moreover, the activity of Ru-(Mn-Al)O X is higher due to the higher amount of Mn in the bulk. It has been previously shown that the amount of Mn introduced on the support before deposition of Ru accelerated the conversion of xylitol in base-free medium [10].
Compared to Ru-based catalysts described in the literature for hydrogenolysis of alditols in water under base-free conditions, the Ru-(Mn-Al)O X catalyst is the most selective to glycols and GLY (58%). This selectivity was 41% from sorbitol over Ru/Ca(OH) 2 [22] and 32% from xylitol over Ru/MnO/C [10]. The selectivity of Ru-(Mn-Al)O X is equal to that of a few Ni-based catalysts, such as Ni-(Mg-Al)O X [24] or Ni/fly ash [20]. It is close to the one of the best catalytic systems, such as Ni-CaO/C [39] and Cu/CaO-Al 2 O 3 [21], which displayed a selectivity to GLY and glycols of 69% and 73%, respectively; however, the latter Ni catalysts were less active in the hydrogenolysis under similar reaction conditions of alditols with values of 20 and 0.4 h −1 respectively, compared to Ru-(Mn-Al)O X (57 h −1 ).
The stability of the catalysts was investigated by analysis of the leaching of the catalyst in the final solution and of the used catalyst by XRD. No Ru leaching was observed (detection limit ICP-OES; <0.2 mg L −1 , <0.15 wt %), Zr and Ti leaching was less than 1 wt %, while 13-16 wt % of Al were leached. These observations are in line with the literature on the higher stability of TiO 2 and ZrO 2 than Al 2 O 3 as support for Ru under hydrothermal conditions under H 2 atmosphere [40]. On the contrary, a Mn leaching as high as 42-48 wt % was observed for all catalysts. Although this value was still too high, the Mn leaching from Ru-(Mn-M)O X catalysts was below the one from Ru/MnO/C of 63-70 wt % under the same reaction conditions [10]. The Mn leaching is associated with the production of carboxylic acids, such as lactic or 2-hydrobutyric acids, which were analyzed in the final reaction medium [10,33]. While Mn 3 O 4 and MnO were observed in the fresh catalyst, only MnCO 3 was detected in the used catalyst. Similarly, calcium carbonate was detected over Ru/Ca(OH) 2 catalyst after hydrogenolysis of sorbitol under similar conditions [22].

Influence of Operating Conditions
The operating conditions (H 2 pressure, temperature) may have a significant influence on the product distribution, as shown over a Ru/C catalyst in hydrogenolysis of xylitol [9]. We investigated the H 2 pressure in order to decrease the carboxylic acids production and thus avoid the Mn leaching. Figure 5 shows the selectivity to the main products as a function of pressure over Ru-(Mn-Al)O X catalyst.
An increase of H 2 pressure from 60 to 100 bar at 200 • C doubles the molar fraction of H 2 solubilized in aqueous solution [41]. However, catalytic activity increased only slightly from 52 to 59 h −1 , suggesting that the solubility of H 2 in water was not the limiting step. As for the product distribution, the cumulated selectivity to glycols decreased slightly from 48% to 43%. The selectivity to GLY increased gradually from 7% to 10%, and the selectivity to carboxylic acids (LA + 2-HBA) was reduced from 8.5% (7.5% LA; 1.0% 2-HBA) to 3.4% (2.7% LA; 0.7% 2-HBA). According to the scheme of the proposed reaction, GLY and LA come from the same intermediate, namely glyceraldehyde, via hydrogenation of the terminal carbonyl or via dehydration followed by Cannizzaro-like reaction; thereby, the pressure of H 2 favored GLY production at the expense of LA [9]. Nevertheless, despite the decrease in carboxylic acid production, the crystalline phase of the catalyst at the end of the reaction was MnCO 3 , and Mn leaching still occurred. In addition, the substrate to catalyst ratio was investigated at 200 °C and 60 bar H2 by modifying the amount of catalyst. The mass ratio was varied from 30 to 3, i.e., a molar xylitol/Ru ratio from 764 to 77. The activity, the product distribution, and the Mn leaching at the end of reaction are shown in Figure 6. Normalized to the Ru moles, the catalytic activity increased from 57 h −1 to 101 h −1 as the catalyst mass increased (molar ratio from 764 to 255, Reaction #1-3), confirming that the Mn amount in the reaction system enhanced the activity, as observed in Table 2 and as reported previously [10]. When normalized to the total moles of Ru and Mn, the catalytic activity also varied by a factor of 2 between 1.1 and 2.0 molxyl molRu+Mn −1 h −1 . Assuming that Mn is responsible for the retro-aldol reaction and Ru for the (de)hydrogenation reactions, the difference in the catalyst activity suggests a pseudo-order reaction higher than 1 for the retro-aldol reaction. The product distribution was notably similar whatever the mass of catalyst: the cumulated selectivity to GLY and glycols remained in the range In addition, the substrate to catalyst ratio was investigated at 200 °C and 60 bar H2 by modifying the amount of catalyst. The mass ratio was varied from 30 to 3, i.e., a molar xylitol/Ru ratio from 764 to 77. The activity, the product distribution, and the Mn leaching at the end of reaction are shown in Figure 6. Normalized to the Ru moles, the catalytic activity increased from 57 h −1 to 101 h −1 as the catalyst mass increased (molar ratio from 764 to 255, Reaction #1-3), confirming that the Mn amount in the reaction system enhanced the activity, as observed in Table 2 and as reported previously [10]. When normalized to the total moles of Ru and Mn, the catalytic activity also varied by a factor of 2 between 1.1 and 2.0 molxyl molRu+Mn −1 h −1 . Assuming that Mn is responsible for the retro-aldol reaction and Ru for the (de)hydrogenation reactions, the difference in the catalyst activity suggests a pseudo-order reaction higher than 1 for the retro-aldol reaction. The product distribution was notably similar whatever the mass of catalyst: the cumulated selectivity to GLY and glycols remained in the range PG, In addition, the substrate to catalyst ratio was investigated at 200 °C and 60 bar H2 by modifying the amount of catalyst. The mass ratio was varied from 30 to 3, i.e., a molar xylitol/Ru ratio from 764 to 77. The activity, the product distribution, and the Mn leaching at the end of reaction are shown in Figure 6. Normalized to the Ru moles, the catalytic activity increased from 57 h −1 to 101 h −1 as the catalyst mass increased (molar ratio from 764 to 255, Reaction #1-3), confirming that the Mn amount in the reaction system enhanced the activity, as observed in Table 2 and as reported previously [10]. When normalized to the total moles of Ru and Mn, the catalytic activity also varied by a factor of 2 between 1.1 and 2.0 molxyl molRu+Mn −1 h −1 . Assuming that Mn is responsible for the retro-aldol reaction and Ru for the (de)hydrogenation reactions, the difference in the catalyst activity suggests a pseudo-order reaction higher than 1 for the retro-aldol reaction. The product distribution was notably similar whatever the mass of catalyst: the cumulated selectivity to GLY and glycols remained in the range EG, In addition, the substrate to catalyst ratio was investigated at 200 °C and 60 bar H2 by modifying the amount of catalyst. The mass ratio was varied from 30 to 3, i.e., a molar xylitol/Ru ratio from 764 to 77. The activity, the product distribution, and the Mn leaching at the end of reaction are shown in Figure 6. Normalized to the Ru moles, the catalytic activity increased from 57 h −1 to 101 h −1 as the catalyst mass increased (molar ratio from 764 to 255, Reaction #1-3), confirming that the Mn amount in the reaction system enhanced the activity, as observed in Table 2 and as reported previously [10]. When normalized to the total moles of Ru and Mn, the catalytic activity also varied by a factor of 2 between 1.1 and 2.0 molxyl molRu+Mn −1 h −1 . Assuming that Mn is responsible for the retro-aldol reaction and Ru for the (de)hydrogenation reactions, the difference in the catalyst activity suggests a pseudo-order reaction higher than 1 for the retro-aldol reaction. The product distribution was notably similar In addition, the substrate to catalyst ratio was investigated at 200 °C and 60 bar H2 by modifying the amount of catalyst. The mass ratio was varied from 30 to 3, i.e., a molar xylitol/Ru ratio from 764 to 77. The activity, the product distribution, and the Mn leaching at the end of reaction are shown in Figure 6. Normalized to the Ru moles, the catalytic activity increased from 57 h −1 to 101 h −1 as the catalyst mass increased (molar ratio from 764 to 255, Reaction #1-3), confirming that the Mn amount in the reaction system enhanced the activity, as observed in Table 2 and as reported previously [10]. When normalized to the total moles of Ru and Mn, the catalytic activity also varied by a factor of 2 between 1.1 and 2.0 molxyl molRu+Mn −1 h −1 . Assuming that Mn is responsible for the retro-aldol reaction and Ru for the (de)hydrogenation reactions, the difference in the catalyst activity suggests a pseudo-order reaction higher than 1 for the retro-aldol reaction. The product distribution was notably similar whatever the mass of catalyst: the cumulated selectivity to GLY and glycols remained in the range In addition, the substrate to catalyst ratio was investigated at 200 • C and 60 bar H 2 by modifying the amount of catalyst. The mass ratio was varied from 30 to 3, i.e., a molar xylitol/Ru ratio from 764 to 77. The activity, the product distribution, and the Mn leaching at the end of reaction are shown in Figure 6. In addition, the substrate to catalyst ratio was investigated at 200 °C and 60 bar H2 by modifying the amount of catalyst. The mass ratio was varied from 30 to 3, i.e., a molar xylitol/Ru ratio from 764 to 77. The activity, the product distribution, and the Mn leaching at the end of reaction are shown in Figure 6. Normalized to the Ru moles, the catalytic activity increased from 57 h −1 to 101 h −1 as the catalyst mass increased (molar ratio from 764 to 255, Reaction #1-3), confirming that the Mn amount in the reaction system enhanced the activity, as observed in Table 2 and as reported previously [10]. When normalized to the total moles of Ru and Mn, the catalytic activity also varied by a factor of 2 between 1.1 and 2.0 molxyl molRu+Mn −1 h −1 . Assuming that Mn is responsible for the retro-aldol reaction and Ru for the (de)hydrogenation reactions, the difference in the catalyst activity suggests a pseudo-order reaction higher than 1 for the retro-aldol reaction. The product distribution was notably similar Normalized to the Ru moles, the catalytic activity increased from 57 h −1 to 101 h −1 as the catalyst mass increased (molar ratio from 764 to 255, Reaction #1-3), confirming that the Mn amount in the reaction system enhanced the activity, as observed in Table 2 and as reported previously [10]. When normalized to the total moles of Ru and Mn, the catalytic activity also varied by a factor of 2 between 1.1 and 2.0 mol xyl mol Ru+Mn −1 h −1 . Assuming that Mn is responsible for the retro-aldol reaction and Ru for the (de)hydrogenation reactions, the difference in the catalyst activity suggests a pseudo-order reaction higher than 1 for the retro-aldol reaction. The product distribution was notably similar whatever the mass of catalyst: the cumulated selectivity to GLY and glycols remained in the range 57-60%, while the selectivity to C 5 and C 4 products was 10%; the selectivity to carboxylic acids decreased from 5% to less than 1% as the xylitol/catalyst ratio decreased. Surprisingly, Mn leaching at the end of reaction dropped from 48% to 10% as the xylitol/Ru molar ratio decreased from 764 to 255. Meanwhile, Al leaching decreased from 15% to less than 1%, and the Ru leaching remained below the detection limit. Therefore, in order to emphasize this effect, the xylitol/catalyst molar ratio was further decreased to 77, using a 1.0 wt % xylitol solution ( Figure 6, Reaction #4). Activity decreased to 25 h −1 suggesting mass transfer limitations. Interestingly, Mn leaching was below 1%, i.e., 45 mg L −1 Mn in the final reaction medium. Furthermore, XRD analysis of the used Ru-(Mn-Al)O X catalyst ( Figure 7) showed that the main crystalline phase was MnCO 3 after reactions using xylitol/Ru molar ratio between 764 and 255 (reactions #1-3). However, no peak attributed to MnCO 3 was detected in the used catalyst after reaction at molar ratio of 77 (reaction #4). Only peaks of Mn 3 O 4 and large peaks of the hydrated MnO phase, Mn(OH) 2 were observed [42]. Therefore, the stability of Ru-(Mn-Al)O X was enhanced by diminishing the xylitol/catalyst mass ratio (molar ratio 77), which made the recycling of the catalyst interesting. 57-60%, while the selectivity to C5 and C4 products was 10%; the selectivity to carboxylic acids decreased from 5% to less than 1% as the xylitol/catalyst ratio decreased. Surprisingly, Mn leaching at the end of reaction dropped from 48% to 10% as the xylitol/Ru molar ratio decreased from 764 to 255. Meanwhile, Al leaching decreased from 15% to less than 1%, and the Ru leaching remained below the detection limit. Therefore, in order to emphasize this effect, the xylitol/catalyst molar ratio was further decreased to 77, using a 1.0 wt % xylitol solution ( Figure 6, Reaction #4). Activity decreased to 25 h −1 suggesting mass transfer limitations. Interestingly, Mn leaching was below 1%, i.e., 45 mg L −1 Mn in the final reaction medium. Furthermore, XRD analysis of the used Ru-(Mn-Al)OX catalyst ( Figure 7) showed that the main crystalline phase was MnCO3 after reactions using xylitol/Ru molar ratio between 764 and 255 (reactions #1-3). However, no peak attributed to MnCO3 was detected in the used catalyst after reaction at molar ratio of 77 (reaction #4). Only peaks of Mn3O4 and large peaks of the hydrated MnO phase, Mn(OH)2 were observed [42]. Therefore, the stability of Ru-(Mn-Al)OX was enhanced by diminishing the xylitol/catalyst mass ratio (molar ratio 77), which made the recycling of the catalyst interesting. After the first run, the used Ru-(Mn-Al)OX catalyst was washed with distillated water and reduced at 450 °C for 3 h in order to eliminate the surface hydroxyls and to reactivate Ru. The XRD analysis of the reactivated catalyst in Figure 7 shows that MnO was again the main crystalline phase. Small peaks associated with Mn3O4 can also be observed. It has been reported that formation of amorphous MnOX and Mn3O4 phases leads to an increase of specific surface area [43]; this was After the first run, the used Ru-(Mn-Al)O X catalyst was washed with distillated water and reduced at 450 • C for 3 h in order to eliminate the surface hydroxyls and to reactivate Ru. The XRD analysis of the reactivated catalyst in Figure 7 shows that MnO was again the main crystalline phase. Small peaks associated with Mn 3 O 4 can also be observed. It has been reported that formation of amorphous MnO X and Mn 3 O 4 phases leads to an increase of specific surface area [43]; this was observed with the reactivated catalyst whose specific area was enhanced from 28 to 60 m 2 g −1 . These changes did not significantly modify the performance in hydrogenolysis of 1.0 wt % of xylitol in water with a molar ratio xylitol/Ru of 77, as shown in Table 3. Indeed, the activity only decreased from 26 to 17 h −1 , while the cumulated selectivity to GLY and glycols was 50% instead of 57%. The selectivity to C 4 and C 5 alditols increased slightly from 11% to 17% and the selectivity to gaseous products increased from 10% to 15%. Therefore, the recycling of catalyst favored the epimerization and cascade decarbonylation reactions and the Mn leaching after the second run was less than 2.0 wt % (55 mg L −1 in the final mixture). The XRD analysis of the reactivated catalyst (after 2nd run, Figure 7) revealed that the proportion of Mn 3 O 4 in the crystalline bulk and amorphous phase had increased. These results showed that the structure of Ru-(Mn-Al)O X catalyst was slightly altered under these hydrothermal conditions. After a reactivation under H 2 , the selectivity to GLY and glycols remained at a good level.

Catalysts Preparation
The Ru-mixed oxide based catalysts were prepared by a standard co-precipitation method according to the literature [24,26]. In the synthesis of Ru-(Mn-Al)O X catalyst, an aqueous solution (330 mL) of Mn(NO 3 ) 2 ·4H 2 O (9.1 g), Al(NO 3 ) 3 ·9H 2 O (3.4 g) and Ru(NO)(NO 3 ) 3 ·xH 2 O (0.4 g) and an aqueous solution (264 mL) of NaOH (10.5 g) and Na 2 CO 3 (8.7 g) were added simultaneously dropwise under vigorous stirring at RT into a round-bottom 1 L flask and the pH was controlled at 12 ± 0.5. The obtained slurry was aged for 20 h at 80 • C, filtered, and washed a couple of times with hot deionized water (70 • C) until pH 7. Then, the cake was dried at 100 • C overnight, calcined under air flow (100 mL min −1 ) at 460 • C for 2 h (3 • C min −1 ) and reduced under H 2 flow (100 mL min −1 ) for 3 h, and finally passivated under 1% O 2 /N 2 gas mixture flow at RT for 30 min. The Ru-(Mn-Zr)O X and Ru-(Mn-Zn)O X catalysts were prepared analogously to this procedure. The Ru-(Mn-Ti)O X was prepared as previously, however, Mn(OAc) 2 was used as the precursor due to the use of Ti(OiPr) 4 precursor. A support (Mn-Al)O X was prepared by co-precipitation of Mn(NO 3 ) 2 ·4H 2 O and Al(NO 3 ) 3 ·9H 2 O, according to the literature [24,26]. The material was dried and calcined, as previously. Ru/(Mn-Al)O X catalyst was then synthetized by wet impregnation of the aqueous solution of Ru precursor over (Mn-Al)O X , followed by reduction and passivation, according to the previous procedure [25].

Catalytic Testing
Hydrogenolysis of xylitol was carried out in a Hastelloy Parr autoclave (300 mL) at a stirring speed of 1000 rpm. In a typical run, 135 mL of 10 wt % xylitol aqueous solution and 0.5 g catalyst were introduced into the autoclave. After purging with Ar, the reactor was heated to the reaction temperature. It was then pressurized with H 2 , which corresponded to the reaction time t = 0.
The products in the liquid samples were identified and quantified using two different HPLC instruments (Shimadzu, Marne la Vallée, France) connected to refractive index diffraction (RID10A) and UV detectors (SPD-M10A), as described previously [10]. The total organic carbon (TOC) in the liquid phase was measured using a TOC analyzer (Shimadzu TOC-VCHS equipped with ASI-automatic sampler) and compared to initial TOC to estimate the formation of products transferred to the gaseous phase. The carbon balance BC T was determined by comparison of TOC measured and TOC calculated from HPLC. Conversion, selectivity (on a carbon basis %), and initial reaction activity (mol xylitol mol Ru −1 h −1 ) were calculated as previously [10]. The metal leaching of the catalyst was detected by ICP-OES analysis of the final reaction medium.

Conclusions
To conclude, the Ru-(Mn-M)O X (M: Ti, Al, Zr) catalysts displayed high performances in the valorization of xylitol under base-free conditions. Interestingly, due to the high number of basic sites, the catalysts were able to selectively produce glycols and glycerol with a very low amount of carboxylic acids, such as lactic acid. The Ru-(Mn-Al)O X catalyst presented the highest activity (57 h −1 ) and high cumulated selectivity (58%) to glycols and glycerol. That makes Ru-(MnO-Al)O X one of the most selective Ru-based catalysts under these conditions. The comparison of Ru/(Mn-Al)O X and Ru-(Mn-Al)O X catalysts revealed that Ru/(Mn-Al)O X was more active (111 h −1 ), however, it produced more C 5 -C 4 alditols as well as gas phase products (i.e., CH 4 , C 2 H 6 ). The reason is the smaller Ru particles observed by XRD and TEM, which favor decarbonylation, epimerization as well as cascade decarbonylation reactions. However, Ru-(Mn-Al)O X catalyst was not stable during the reaction since 42 wt % of Mn-leaching and MnCO 3 phase were detected after reaction. The amount of catalyst in the system has a strong effect on the catalyst performance. The Ru-(Mn-Al)O X activity increased up to 101 h −1 as the molar ratio xylitol/Ru was decreased from 764 to 255. Moreover, the decrease in Mn-leaching allowed us to perform a recycling test. After catalyst reactivation, low deactivation was observed in the second run. Therefore, under the appropriate reaction conditions, Ru-(Mn-Al)O X is a promising catalyst for the synthesis of glycols and glycerol from biomass-derived alditols. Funding: This work was funded by Ecole Doctorale de Chimie de Lyon and IFP Energies Nouvelles.