Synergy between Sulfonic Functions and Ru Nanoparticles Supported on Activated Carbon for the Valorization of Cellulose into Sorbitol

: The production of sorbitol from biomass, and especially from its cellulosic component, has been studied as a sustainable method for producing platform molecules. Because it requires two steps, namely, hydrolysis and hydrogenation, bifunctional materials are required as catalysts for this transformation. This study reports a bifunctional catalyst composed of sulfonic functions grafted onto a carbon support for the hydrolysis step and RuO 2 nanoparticles for the hydrogenation step. As sulfur can easily poison Ru, synthetic optimization is necessary to obtain an efﬁcient bifunctional catalyst that surpasses a mere Ru/C catalyst. Kinetic studies highlight the better activity of the bifunctional catalysts compared to the reference monofunctional catalysts. Besides being active in hydrolysis reactions, sulfonic functions also have a role in avoiding the degradation of the sorbitol produced. The recyclability of the bifunctional catalyst is also superior to that of the monofunctional one.


Introduction
Lignocellulose is the dominant non-edible biomass resource in our environment. Its valorization into fine chemicals is one of the most important challenges to finding renewable and sustainable alternatives to fossil fuels [1][2][3][4]. It is composed of lignin (10-30%), hemicellulose (15-40%) and cellulose (30-65%) [5]. Lignin is a phenylpropanoid polymer linked randomly. Hemi-cellulose is a non-linear polymer of pentose, hexoses and sugar acids. Cellulose is a linear polymer of glucose units linked by β-1,4-glycosidic bonds. The treatment of lignocellulosic biomass can produce many useful molecules, such as glucose, 5-hydroxymethylfurfural, sorbitol, aromatic alcohols or aldehydes, furfural, xylitol, etc. [4,[6][7][8][9]. The transformation of cellulose into sorbitol has been extensively studied because sorbitol is a platform molecule of interest. It can be used as a sweetener or an excipient in the food and cosmetics industries [10,11]. However, this transformation is very challenging. Cellulose hydrolysis is the trickiest part because cellulose has a microcrystalline structure. Intra-and intermolecular as well as inter-sheet hydrogen bonds make it insoluble in almost all solvents, especially water [12]. Assistance in reducing the crystallinity of cellulose and breaking it into oligomers is often used in cellulose transformation processes. Ionic liquids can dissolve cellulose by preventing intermolecular hydrogen bonds, but these liquids are very expensive and toxic [13]. Supercritical water is also used to partially dissolve cellulose [14]. Ball milling can reduce mechanical cellulose crystallinity [15,16]. Microwave irradiation is also a common method to depolymerize cellulose via the in situ generation of heat [4,17]. Nowadays, mineral acids such as HCl In complement to the use of pre-treatment methods, acidic heterogeneous catalysts have been extensively studied in biomass valorization. Even though they will show less activity, they will display stability, recyclability and robustness. Various acidic heterogeneous catalysts have been investigated, such as Amberlyst-15 resin [22], hydrotalcites [23], zeolites [24], functionalized carbon materials [25,26] and so on. Glucose arising from cellulose depolymerization can then be hydrogenated into sorbitol. Industrially, this step is catalyzed by using Ni-Raney. However, many studies have tried to improve this transformation by using noble metals, such as Ru [27] or even the more expensive Pt [28]. Nowadays, the challenge is to directly transform cellulose into sorbitol using a bifunctional catalyst that will display acidic functions and Ru active sites. Different types of supports have been investigated in such contexts: MCM-48 [29], ZSM-5 zeolites [30], modified SiO 2 [31] and metal-organic framework (MOF) [32]. Carbon materials have been functionalized with acidic functions, such as COOH or SO 3 H, and have supported Ru nanoparticles [33,34].
The reaction pathways from cellulose to sorbitol are shown in Scheme 1. It has to be pointed out that by using bifunctional catalysts, it is possible to first hydrogenate cellobiose into 3-β-D-glucopyranosyl-D-glucitol (cellobitol) and then hydrolyze it into glucose and sorbitol or, as discussed above, to first hydrolyze and then hydrogenate. It has been shown that this alternative pathway (with hydrogenation first) is faster than hydrolyzing first [35]. Moreover, the hydrogenation of glucose just after its release from cellulose decreases the amount of degradation products [36]. Therefore, the one-pot production of sorbitol from cellulose through the action of an efficient bifunctional catalyst would be a milestone in biomass processing.
The purpose of the present contribution is to use sulfonic functions jointly with Ru nanoparticles to obtain a bifunctional carbon-supported catalyst that can transform cellobiose and, more importantly, cellulose directly into sorbitol. Sulfonic functions will be grafted covalently by using an organic reaction on the carbon surface to produce a highly acidic support. An optimal balance between the amount of sulfur and the amount of Ru will be targeted to avoid poisoning Ru and still display the best possible catalyst activity. These bifunctional catalysts will be first tested on cellobiose, as a water-soluble model for cellulose, allowing the study of its hydrogenolysis into sorbitol without the need for assistance systems to increase solubility. Eventually, the transformation of cellulose into sorbitol will be assessed with the best bifunctional catalysts to highlight the synergetic effects between the two types of active sites and benchmark our new materials.

Cellobiose Transformation into Sorbitol
The synthesis of bifunctional catalysts with both sulfonic functions and Ru is a challenge because it is known that sulfur, especially under high-pressure and high-temperature catalytic testing conditions, can strongly poison noble metals [37], especially Ru [38]. Therefore, it was decided to first functionalize the carbon support with sulfonic functions before adding the Ru nanoparticles to limit the poisoning of the active metal phase. The method employed for carbon functionalization has already been discussed in another study [26] and is described in the experimental section. To keep the integrity of the sulfonic functions, a neutral method to add the RuO 2 nanoparticles was used, namely, the adsorption of preformed colloidal RuO 2 suspension. The Ru is reduced in situ by using H 2 during the reaction. Catalytic tests were carried out with monofunctional, either with sulfonic functions alone or with only Ru nanoparticles, and bifunctional catalysts (Table 1). The hydrogenation pathway is favored with the RuO 2 /AC catalyst, which essentially provided cellobitol and sorbitol. The SO 3 H/AC acid monofunctional catalyst yielded a high amount of glucose, as expected in the absence of Ru to catalyze the hydrogenation. Unfortunately, with the corresponding bifunctional catalyst, prepared in exactly the same way, with both SO 3 H and Ru on the support, sorbitol was not obtained. However, the results were very similar to those obtained with the acidic catalyst without Ru. This means that all Ru is poisoned by the sulfur of the sulfonic functions, even if the metal component was introduced in the second position. A test with the two monofunctional catalysts mechanically mixed was carried out, and the result was still similar to SO 3 H/AC. Even if the sulfonic functions were not on the same support, they managed to poison the Ru. Sulfonic functions are likely leaching during the reaction and can poison Ru on other catalyst grains.
In order to enhance bifunctional catalyst activity, several parameters were studied to optimize the ratio between sulfur and ruthenium and avoid this poisoning effect. To control the amount of sulfur for the preparation of bifunctional catalysts, the modulation of the sulfonic functions grafted onto the carbon supports by heat treatment (under an N 2 atmosphere) was investigated ( Table 2). By increasing the heat treatment temperature, a decrease in the O and S surface amounts was observed. While the acidity (determined by Boehm titration) did not decrease at 200 • C, it was lower after treatment at 300 • C and 400 • C. This indicates that the sulfonic functions present in the carbon materials are logically more abstracted when the temperature of the treatment increases. It has been shown by TPD-MS that these grafted sulfonic acid groups do not start to decompose at all before 200 • C, with the main decomposition temperature above 300 • C [26]. 1 Total S at. % (addition of sulfonic peak and thiol peak).
These new catalysts with varying acidities were assessed in the hydrolysis of cellobiose into glucose under a nitrogen atmosphere ( Table 3). As expected, the activity of the catalysts decreased with the applied heat treatment temperature. The low selectivity into glucose obtained for the catalysts heated at 300 • C and 400 • C is the same selectivity obtained when the test was conducted without any catalyst (blank test). The remaining sulfonic functions were no longer useful after these heat treatments. This confirms that sulfonic functions are indeed present in lower amounts when the temperature of the heat treatment is increased. Moreover, this is consistent with our previous work, which showed that a minimum amount of SO 3 H is needed to trigger the hydrolysis of cellobiose [26]. It was also shown that the heat treatments at 200 • C and 300 • C did not modify the nature of the sulfonic functions. Indeed, XPS analyses showed that the sulfur peak did not shift (S1 in the Supplementary Materials). However, the heat treatment at 400 • C formed thiols (at 164 eV) on the support surface in addition to the remaining sulfonic functions (at 168.5 eV) (S1 in the Supplementary Materials). These supports, possessing different amounts of sulfonic functions, were then used to prepare a new set of bifunctional catalysts by depositing RuO 2 nanoparticles. The hydrogenolysis of cellobiose was tested for 2 h with these materials to investigate the impact of the Ru/S ratio on the performance of the catalysts (Table 4). The results show that the pathway of the reaction is different depending on the ratio between Ru and the sulfonic functions. When the functionalized support is treated at 200 • C, sulfonic groups are still present in high amounts, poisoning the Ru. Therefore, the hydrolysis of cellobiose into glucose is the main reaction, and sorbitol is not produced. After treatment at 300 • C, sulfonic functions are degraded partially (as confirmed by the XPS results displayed in S3 section in the Supplementary Materials), and both pathways are followed in parallel. Glucose (the hydrolysis product) and cellobitol (the hydrogenation product) are produced simultaneously, while sorbitol production remains really low due to the very short testing duration. The material pre-heated at 400 • C produced more sorbitol, while glucose selectivity was rather low compared to cellobitol selectivity. This means that, in this case, the hydrogenation pathway is the major one. All these observations were confirmed with kinetic studies with each of these catalysts (S2 section in the Supplementary Materials). After 24 h, it was observed that the bifunctional catalysts pre-heated at 300 • C and 400 • C produced sorbitol, while the lower-temperature-treated bifunctional catalyst barely managed to form some of it. The diminution in sulfonic function amounts allows the Ru to be less poisoned, and the resulting bifunctional catalysts can produce sorbitol. However, the sorbitol production was not better than a monofunctional catalyst with only Ru deposited on the unmodified carbon support. Therefore, other ways of enhancing sorbitol productivity were investigated.
In order to study a different method that could prevent the poisoning of Ru by sulfur, consecutive reactions with both monofunctional catalysts were carried out (Table 5). Cellobiose conversion and sorbitol yield are always given with respect to the first starting compound, even in the case of consecutive reactions. First, hydrogenation using the RuO 2 /AC catalyst was carried out, and then, the solid was filtered out. Consecutively, the SO 3 H/AC catalyst was added to catalyze the hydrolysis. By doing so, a sorbitol yield of 46% in 2 h was obtained, which was, by far, better than with any other catalyst tested so far. It is important to highlight the fact that when cellobitol is hydrolyzed, it provides one equivalent of glucose and one equivalent of sorbitol. When staged in this order, the two reactions cannot provide 100% of sorbitol because the glucose produced needs to be further hydrogenated. If the two consecutive reactions are performed in the opposite order (first hydrolysis and then hydrogenation), it was noticed that the sorbitol yield was dramatically lower (11%). This result can be explained by the fact that during hydrolysis some sulfonic functions will leach and go into solution. Indeed this was confirmed as the pH went down from 5 to 3 after the hydrolysis step. Then, when the RuO 2 /AC was added for the second reaction, it was poisoned by the sulfur compounds present within the solution, decreasing the activity of Ru. As sulfonic functions can leach and poison Ru nanoparticles, a pre-treatment under the same conditions as the catalytic tests was conducted to remove all labile functions and keep solely the sulfonic groups well grafted onto the carbon support. This type of pre-treatment has been shown in previous work to provide 116 mmol/100 g of acidity, with good results for the hydrolysis of cellobiose into glucose (98% selectivity toward glucose) [26]. XPS data of the pre-treated SO 3 H/AC can be found in the Supplementary Materials (S3 section). Two consecutive reactions were carried out with first the pre-treated SO 3 H/AC catalyst and then with the RuO 2 /AC catalyst.
The sorbitol yield at the end of the two consecutive reactions was significantly higher when the SO 3 H/AC was pre-treated in this way ( Table 5, bottom). This shows, indeed, that the pre-treatment can lower the quantity of labile sulfonic functions that can poison the Ru nanoparticles. However, there is still a portion of sulfonic functions that inhibit the Ru. Indeed, the obtained yield is not as good as when the two consecutive reactions were carried out in the opposite order. Nevertheless, this pre-treatment on the sulfonic functions was used to prepare new bifunctional catalysts that would suffer less from the poisoning of Ru. This time the Ru/S ratio was studied by varying the amount of Ru. New catalysts with 1 wt.%, 3 wt.% and 5 wt.% of Ru were synthesized using pre-treated SO 3 H/AC as the support. They were assessed as catalysts in comparison with corresponding mono-and bifunctional catalysts without pre-treatment (Table 6) in the one-pot direct hydrogenolysis of cellobiose into sorbitol.
The results obtained with the bifunctional catalyst with 1 wt.% of Ru are similar to the bifunctional catalyst prepared without any pre-treatment of the sulfonic functions. Indeed, even if the amount of sulfonic functions decreases with the pre-treatment, it is still enough to totally inhibit the Ru active phase. For the monofunctional catalyst with 3 wt.% of Ru, it can be observed that the sorbitol yield was the same that the monofunctional catalyst with 1 wt.%. Nevertheless, if the quantity of cellobitol is lower, the sorbitol yield should be higher. This means that sorbitol seems to react further or suffer from degradation. The bifunctional catalyst with 3 wt.% of Ru is more active than the one with 1 wt.%. Interestingly, the apparition of cellobitol and sorbitol in the products can be seen. This indicates that both pathways (as seen in Scheme 1) are followed and both Ru nanoparticles and sulfonic functions are active at the same time. However, the sorbitol yield was still not higher than with the monofunctional catalyst. It can also be observed that the bifunctional catalyst with 3 wt.% of Ru without the pre-treatment of sulfonic functions displays a lower yield of sorbitol. The monofunctional catalyst with 5 wt.% of Ru provides a higher yield of sorbitol, which was expected as the Ru quantity increased. Finally, the bifunctional catalyst with 5 wt.% provides the best results. Indeed, the sorbitol yield reached 53% in 2 h, indicating that a synergy between the two active sites occurred with this ratio between Ru and SO 3 H functions (10 Ru atoms for 1 S atom based on XPS analysis). The sorbitol yield decreased drastically to 13% when the SO 3 H functions were not pre-treated. This highlights the importance of this pre-treatment in order to have an efficient bifunctional catalyst. Kinetic studies were carried out on the monofunctional and bifunctional catalysts with 5 wt.% of Ru to highlight the differences between them ( Figure 1). It can be observed that the bifunctional catalyst provided a higher sorbitol yield than the monofunctional catalyst during all investigated timeframes. The conversion of cellobiose and sorbitol selectivity/yield are presented in the Supplementary Materials (S4 section). The sorbitol yield was the same after 6 h and after 24 for the monofunctional catalyst, whereas cellobitol decreased. This confirms our assumption that sorbitol reacts further during the reaction. This did not happen with the bifunctional catalyst. The combination of SO 3 H functions and Ru nanoparticles in this optimized ratio stabilizes the sorbitol, thereby preventing it from reacting further or degrading. This effect has already been reported in the literature [31]. This explains the higher yield of sorbitol at the end of the reaction. The bifunctional catalyst is not more active than the monofunctional catalyst; indeed, the conversion is lower at the beginning of the reaction with the RuO 2 -pre-treated SO 3 H/AC catalyst, but it stabilizes the sorbitol, which provides all in all a much higher yield. This effect is striking after 24 h of reaction, with a yield of 11% for the monofunctional catalyst compared to 79% for the bifunctional catalyst.
The degradation of sorbitol with the RuO 2 /AC and bifunctional catalyst was studied. Sorbitol (the product of the reaction), rather than cellobiose, was placed in the presence of the two catalysts under the same conditions as in the previous tests. The results show that RuO 2 /AC transformed 50% of the sorbitol under these conditions, while the bifunctional catalyst transformed only 7% of the sorbitol within 2 h. The presence of SO 3 H functions highly stabilizes the sorbitol molecule and prevents it from degradation. Unfortunately, the products formed by sorbitol degradation could not be identified. A list of molecules that were injected in HPLC and that can be ruled out from possible side products can be found in the Supplementary Materials (S5 section). To further optimize the Ru/S ratio, modification of the sulfur amount with thermal pretreatment was investigated with 5 wt.% of Ru loading ( Table 7). The bifunctional catalyst with sulfonic functions treated at 200 • C provided the same amount of glucose as sorbitol. This means that there were still too many sulfonic groups and that the Ru nanoparticles were poisoned. When the sulfonic functions were treated at 300 • C, the results were the same as when the sulfonic functions were pre-treated in water. The bifunctional catalyst is better than the reference RuO 2 /AC, showing synergy between the two types of active sites. After heat treatment at 400 • C, the bifunctional material displayed a lower activity. Indeed, the sulfonic functions were more degraded, leading to more active Ru nanoparticles and sorbitol degradation. The synergy between the two active sites was no longer observable.
Eventually, the recyclability of both the monofunctional and bifunctional catalysts was assessed. Five catalytic tests were carried out for each catalyst (Figure 2). Both catalysts suffered from deactivation after the first run, but the impact was less important on the bifunctional catalyst than on the monofunctional catalyst. The activity was then stable over the next runs. TOC analysis of the solution after catalytic tests showed that the carbon balance was respected. XPS characterizations were performed on the monofunctional and bifunctional catalysts before and after the catalytic tests. The results are presented in Table 8, while the XPS narrow scans can be found in the Supplementary Materials (S6 section). As the Ru amount in the catalysts is important, Ru 3d has to be subtracted from the carbon C 1s signal; otherwise, the carbon will be overestimated, leading to erroneous percentages for other elements. Ru 3p was also analyzed to confirm the Ru 3d data, but only the Ru 3d results will be displayed in this article in order to quantify the Ru surface at. % and Ru oxidation states. The Ru 3d (IV) amount was higher for the bifunctional catalyst than the monofunctional one before the catalytic test. The same statement can be made for the catalysts with 3 wt.% of Ru (S7 section in the Supplementary Materials). The in situ activation of Ru in the reactor was confirmed by these analyses. Indeed, as expected, the Ru(0) amount was drastically increased after the catalytic tests. It is known that Ru can be reduced during XPS analyses, explaining the small amount of Ru(0) in the catalysts before the tests. It should be noted that not all Ru(IV) was reduced during the catalytic tests. The main assumption based on the XPS results is that the core of the nanoparticles was not reduced, and only the surface became Ru(0), giving core-shell structures. When superposing the Ru 3p regions of the catalysts before and after multiple runs (corresponding to Figure 2), it is obvious that Ru was reduced during the catalytic tests (see Figure 9 in the Supplementary Materials). The amount of S, the position of the S 2p peak and the increase in O after sulfonic functionalization corroborate the success of the reaction, as presented above. These results were confirmed by the bulk analysis of the bifunctional catalyst (before catalytic reaction). This analysis revealed 3.7 wt.% of Ru and almost 3 wt.% of S. The nominal value of 5 wt.% of Ru was not reached due to the fact that Ru forms refractory oxides during sample preparation for ICP analysis, so this analysis underestimates the metal content. The S amount is quite important, explaining the significant Ru poisoning.
XRD analyses were also conducted on these catalysts ( Figure 3). The wide peak between 12 • and 28 • comes from the carbon support (AC) [39]. The sharp peak at~27 • is probably due to graphitic C or S but cannot be ascribed to Ru species. The Ru peak is around 44 • [40]. This is very low for the catalysts before the reaction because it is essentially RuO 2 . After the reaction, Ru was reduced and the Ru peak increased because Ru(0) is crystalline in the case of the monofunctional catalyst, but surprisingly not for the bifunctional catalyst. It seems that the presence of sulfonic functions interferes with the crystallinity of the Ru(0) domains and decreases it. HR-TEM images were obtained for the four catalysts ( Figure 4). The RuO 2 nanoparticle preparation method can be adapted to control their sizes, and a nanoparticle size of 1-2 nm was targeted here [41]. The images show that the RuO 2 nanoparticles have a mean size of around 1 nm, as expected. The distribution and size of the nanoparticles are the same without or with sulfonic functions (see Figure 4a,b for comparison). EDX mapping analyses were performed and confirmed the good dispersion of Ru on the carbon support (S8 section in the Supplementary Materials). When the support was functionalized by the sulfonic functions, sulfur was found everywhere on the carbon support. There was no specific area that could be identified where Ru and S were more present together. The catalysts recovered after the catalytic reaction (five runs) were also analyzed by using HR-TEM (Figure 4c,d) and EDX (S8 section in the Supplementary Materials). The TEM images show that the Ru nanoparticles were still well dispersed on the support. However, due to the high temperature during the reaction, a small increase in nanoparticle size could be observed after catalyst recovery, but the nanoparticle size was still between 1 and 2 nm. The presence of sulfur all over the support for the bifunctional catalyst after five runs was also confirmed by EDX analyses.

Cellulose Transformation into Sorbitol
Due to its insolubility in water, cellulose has difficulties reacting in this medium. This is why a variety of pre-treatments can be conducted to make cellulose more reactive, as explained in the Introduction [42][43][44][45]. In our case, a ball-milling step for 24 h prior to catalysis was chosen. The conditions are detailed in the Experimental Section. First, the reactivity of our bifunctional catalyst with and without prior ball milling of the cellulose was assessed (Table 9). Table 9. Hydrogenolysis of cellulose with bifunctional catalyst (RuO 2 (5%)-pre-treated SO 3 H/AC). cellulose conversion; xylitol, mannitol and sorbitol selectivities; sorbitol yield (150 • C, 30 bar of H 2 , 2 h, 150 mg of catalyst, 1 g of cellulose). The estimated confidence interval for cellulose conversion and selectivities is 5%. Cellulose conversion is calculated, as described in the Experimental Section. As this is calculated by weighing the cellulose after the catalytic reaction, the conversion of cellulose presented in the article is the lowest possible conversion (underestimation) because all the recovered mass is attributed to cellulose, but it could potentially be side products that are also solid, such as humins. Cellulose conversion under these soft conditions (150 • C) is low. The ball milling of cellulose enhanced its reactivity, and increases in both sorbitol conversion and selectivity was obtained compared to without milling. Two side products were identified, namely, xylitol and mannitol. Mannitol comes from the isomerization of sorbitol, while xylitol is produced by the decarbonylation of sorbitol [46]. These side products were not present in the catalytic tests on cellobiose. The mechanism of hydrolysis and hydrogenation must be slightly different when starting from a soluble glucose dimer (cellobiose) than from an insoluble polymer (cellulose), which explains the difference in side-product formation. Perhaps small oligomers with more than two units are formed from cellulose, which can go through other reaction routes. Traces (<1%) of cellobiose and glucose can also be seen on the chromatograms, but this was not indicated in the results table for the sake of clarity and readability.

Pre-Treatment
The temperature was varied to see its impact on the reaction. In order to improve cellulose reactivity, the temperature was increased to 170 • C and 190 • C ( Table 10). The impact of temperature on cellulose conversion is striking: a slight increase at 170 • C and a huge improvement at 190 • C were observed. However, the sorbitol selectivity stayed around 30%, with a maximum of 34% at 170 • C. Indeed, it is known that increasing the temperature will favor sorbitol degradation [42]. A test at 190 • C for 24 h was conducted and confirmed that sorbitol selectivity drops when high temperatures and long reaction times are applied. However, the cellulose conversion reached almost 100%, and the global yield of sorbitol (20%) was the same for both conditions. The optimal conditions to maximize the sorbitol yield should be somewhere between 2 h and 24 h. Unfortunately, kinetic studies are difficult to build on this reaction. Indeed, as cellulose is not soluble in water, the solid has to be weighed after filtration to calculate the cellulose conversion by subtracting the catalyst mass from the final combined mass. Aliquots cannot be taken during the reaction to have enough data points in order to build a full kinetic curve. Table 10. Hydrogenolysis of cellulose with bifunctional catalyst (RuO 2 (5%)-pre-treated SO 3 H/AC). Cellulose conversion; xylitol, mannitol and sorbitol selectivities; sorbitol yield (30 bar of H 2 , 150 mg of catalyst, 1 g of ball-milled cellulose (300 rpm)). The estimated confidence interval for cellulose conversion and selectivities is 5%. A comparison between the monofunctional and the bifunctional catalysts was conducted with two different reaction conditions: at 150 • C for 24 h and at 190 • C for 2 h (Table 11). The results indicate that the same cellulose conversion was obtained with both catalysts. However, the bifunctional catalyst produced almost twice the amount of sorbitol. Side products, such as xylitol and mannitol, formed less with the bifunctional catalyst. The same effect of sorbitol stabilization, as was shown in the previous section, can be observed here. This highlights the synergy between the Ru and sulfonic functions and confirms that the bifunctional catalyst is definitely better than the monofunctional catalyst for sorbitol production. In order to maximize the sorbitol yield, a new pre-treatment was set up. As cellulose is not soluble in water, increasing its proximity to the catalyst should enhance its reactivity. This is why combined ball-milling pre-treatment of the cellulose with the catalyst was implemented just before starting the reaction. This ball milling is shorter than the pretreatment of cellulose alone, only 2 h, and is detailed in the Experimental Section. Cellulose is still ball-milled for 24 h before the second pre-treatment. The results obtained by doing so are presented in Table 12. If these results are compared with the results without the ball milling of both cellulose and catalyst together (Table 10), it can be seen that combined ball milling has a positive impact on the results. Cellulose conversion increased for all temperatures. The sorbitol selectivity remained the same for the test at 150 • C, but it increased drastically for the tests at 170 • C and 190 • C. With this new method, almost 40% of sorbitol yield can be reached at 190 • C in only 2 h directly from cellulose. A catalytic test with the monofunctional catalyst was conducted similarly, and the results were also better when implementing combined ball milling. However, it was confirmed that the bifunctional catalyst is far better than the monofunctional one even under the new conditions. Table 12. Hydrogenolysis of cellulose with monofunctional and bifunctional catalysts. Cellulose conversion; xylitol, mannitol and sorbitol selectivities; sorbitol yield (30 bar of H 2, 2 h, 150 mg of catalyst, 1 g of ball-milled cellulose (300 rpm) with combined ball milling). The estimated confidence interval for cellulose conversion and selectivities is 5%. As the combined ball milling of the cellulose and the catalyst extended the total ball-milling time of cellulose by 2 h, it is reasonable to question if this increase in activity is due to the extra ball-milling duration or primarily due to the proximity between the cellulose and the catalyst. A catalytic test with cellulose pre-treated for 26 h was carried out to address this issue (Table 13). The results show that the ball milling of cellulose for 26 h slightly increased the activity compared to 24 h ball milling. However, the combined ball milling of both cellulose and the catalyst still provided the best results. It can be concluded that combined ball milling has a real effect on the reactivity between cellulose and the catalyst. Table 13. Hydrogenolysis of cellulose with bifunctional catalyst (RuO 2 (5%)-pre-treated SO 3 H/AC). Cellulose conversion; xylitol, mannitol and sorbitol selectivities; sorbitol yield (170 • C, 30 bar of H 2 , 2 h, 150 mg of catalyst, 1 g of ball-milled cellulose (300 rpm) with and without combined ball milling). The estimated confidence interval for cellulose conversion and selectivities is 5%.

Benchmarking
As the superiority of bifunctional catalysts over the monofunctional catalyst has now been demonstrated, it is time to compare our results with the literature. In order to do that, two catalytic tests were carried out using another cellulose grade: Avicel PH-101 (Table 14). Indeed, this grade was selected as it is the most common type that can be found in the literature in similar studies [45]. Surprisingly, the results here are almost the same without and with the combined ball milling of both catalyst and cellulose before the reaction. Only a slight increase in cellulose conversion can be observed when implementing combined ball milling. If these results are compared with the ones obtained with the other cellulose type, as described above, it can be noticed that the Avicel PH101 cellulose produces a little less sorbitol under the same conditions. As these results are almost similar to the ones obtained with the first cellulose grade, all the conclusions that are established above can obviously stand for this second cellulose type. Table 14. Hydrogenolysis of cellulose with bifunctional catalyst (RuO 2 (5%)-pre-treated SO 3 H/AC). Cellulose conversion; xylitol, mannitol and sorbitol selectivities; sorbitol yield (190 • C, 30 bar of H 2 , 2 h, 150 mg of catalyst, 1 g of ball-milled cellulose (300 rpm) with and without combined ball milling). The estimated confidence interval for cellulose conversion and selectivities is 5%. The results obtained under the best possible conditions, using the Avicel PH-101 cellulose, allow us to compare our results with the literature and to benchmark the performance obtained in comparison to all the other catalysts envisaged for sorbitol production. It should be noted that comparisons must be performed based on sorbitol yields, which are quantified in the catalytic test filtrates, rather than relying on solid residues, which may also contain humins. Table 15 lists the sorbitol production obtained directly from cellulose with various Ru-based catalysts. A comparison between the literature results and our work is not an easy task because of the differences in the reaction conditions. By looking at entry 1, it can be observed that sorbitol was produced from crystalline cellulose with a monofunctional Ru catalyst in a short reaction time but at a very high reaction temperature (245 • C versus 190 • C in the present study). The best catalyst presented in this work shows a similar result at lower temperatures with joint ball milling and longer reaction times (entry 13). Entry 2 showed similar sorbitol productivity and used the same reaction conditions as our work, except for the reaction time and Ru loading. This highlights the improvement in catalyst activity by the sulfonic functions and the ball-milling procedure. Entries 3 and 4 used bifunctional catalysts with Ru nanoparticles on AC doped with heteropolyacids. Despite the harsh reaction conditions (lower cellulose/catalyst ratios, longer reaction times and higher temperatures), the sorbitol yields obtained by these authors were lower than the ones obtained with the catalysts that were made in this work. The fact that the joint ball-milling step slightly decreased the sorbitol yield, in their case, is unexpected. Similar bifunctional catalysts using Ru and SO 3 H-AC (sulfonation was achieved by using concentrated H 2 SO 4 ) were used (entries 5 and 6) with 10 wt.% of Ru loading. The difference without and with joint ball milling is striking, and this shows that joint ball milling can lead to excellent sorbitol production. Higher Ru loading, lower substrate/catalyst ratios and longer reaction times can explain the better activity of this catalyst compared to ours despite the lower reaction temperature. Entries 7 and 8 used the same type of catalyst as us with the same Ru loading. However, in their case, 5 wt.% of Ru led to lower selectivity in hexitols (the authors did not separate the sorbitol and mannitol selectivities). Despite using lower substrate/catalyst ratios and longer reaction times, the results they obtained are quite similar to ours, which indicates that our catalyst again seems to be more active. Nevertheless, such comparisons are very delicate because the impact of each variation in experimental conditions is difficult to predict. Sulfonated silica with Ru can be used to produce sorbitol at low temperatures with low cellulose/catalyst ratios and longer reaction times (entry 9). Entry 10 provided the results for a bifunctional catalyst with Ru on an acidic zeolite. A low yield of sorbitol was obtained, but this can be attributed to the use of microcrystalline cellulose as a substrate. The last two entries (11 and 12) used a mix of two catalysts-Ru on mesoporous carbon and an acidic catalyst (ZrP). Once again, the joint ball milling of cellulose with the catalysts drastically improved the sorbitol yield. Globally, our best catalyst is very efficient, and the sorbitol yield we obtained is competitive with the literature results. Moreover, our reaction conditions with high cellulose/catalyst ratios and short reaction times are beneficial for possible industrial applications.

Acidic Functionalization
The carbon functionalization was carried out by a diazonium coupling method (Scheme 2) [54]. Typically, 1 g of carbon powder was dispersed in 60 mL of distilled water. A total of 1.5 g of sulfanilic acid was added, and the suspension was stirred at 70 • C for 10 min. The suspension was cooled down to 30 • C before adding 1.2 mL of isopentyl nitrite. The mixture was stirred for 16 h. Then, it was filtered out and washed with distilled water and ethanol. The resulting material was dried overnight at 100 • C. Scheme 2. Diazonium coupling on carbon materials for sulfonic acid grafting.

Heat Treatment
Catalysts were thermally treated at different temperatures to modulate the amount of surface sulfonic functions. In order to do this, solids were placed in a tubular oven and heated under N 2 to 200 • C, 300 • C or 400 • C with an increase in temperature of 100 • C/h. When the targeted temperature was reached, it was maintained for 2 h before being decreased to room temperature naturally.

Ru Nanoparticles Deposition
A colloidal suspension of RuO 2 was synthesized via slow hydrolysis/condensation of RuCl 3 by using H 2 O 2 as oxidizing agent. It was adapted from a literature procedure [41]. In order to have 1 wt.% of Ru on the carbon support, 24.8 mg of RuCl 3 was added to 40 mL of demineralized water (solution A), and 1 mL of H 2 O 2 (35% in volume) was added to 20 mL of demineralized water (solution B). Solution B was added drop-by-drop to solution A. The final mixture was agitated for 5 min. After this, a Berlin glass flask capped with aluminum foil containing the mixture was placed in an oven at 95 • C for 2 h. The solution was cooled down to room temperature and added to 1 g of AC. The dispersion was agitated for 30 min. The solid was recovered by evaporation of the solvent in a rotavapor at 60 • C under reduced pressure, and it was dried overnight at 100 • C.
For synthesizing the bifunctional catalysts with 3 wt.% and 5 wt.% of Ru, the amount of RuCl 3 engaged in the syntheses was adapted to fit the targeted percentages, with all other variables kept constant.

Cellulose Pre-Treatment
Ball milling was performed with a Fritsch Pulverisette 7 premium line apparatus. All experiments were carried out in stainless-steel milling bowls of 45 mL (Fristch™ 50.9750.00). A total of 10 g of cellulose and five stainless-steel balls of 10 mm (Fritsch™ 55.0100.09) were used for each pre-treatment. Millings were carried out at room temperature (20 • C), the local heating during milling was not monitored in situ, but milling pauses (breaks) were introduced to limit potential overheating. The complete cycle was composed of 15 min of ball milling at 300 rpm followed by 5 min of pause. A total of 96 cycles were performed for a total ball-milling time of 24 h.
Combined ball milling with both catalyst and cellulose was carried out with the same device but with different conditions: 1 g of cellulose and 150 mg of catalyst were placed together with three stainless-steel balls. This treatment was performed for 8 cycles at 300 rpm (total of 2 h).

Catalytic Tests
The tests were carried out in a 250 mL stainless-steel Parr autoclave. A total of 1 g of cellobiose or cellulose was added to 150 mg of catalyst in 120 mL of mQ water. This low substrate concentration ensures absence of diffusional limitations. Then, the autoclave was sealed, and the system was purged three times with nitrogen and heated up under autogenic pressure of N 2 for the hydrolysis tests and under 30 bar of H 2 for the hydrogenation tests (this pressure was set when the temperature was reached). For the controlled temperature, the agitation was started at 1700 rpm. After the fixed duration of catalytic test, the system was then cooled down to room temperature, and the solution was filtrated. The solid catalyst was washed with mQ water and dried. The filtrate was then diluted to 250 mL with mQ water in order to always have the same final volume so that all results could be strictly comparable and analyzed by using HPLC. For kinetic studies, at various time intervals, liquid aliquots (~1-2 mL) were taken and diluted two times with mQ water before analyzing it similarly. The pressure within the autoclave was brought back to 30 bar of H 2 after each sampling.
Consecutive reactions were performed as follows: First, hydrogenation using the RuO 2 /AC catalyst was carried out, as described above, under 30 bar of H 2 , and then, the solid was filtered out. Consecutively, the SO 3 H/AC catalyst was added to the filtrate, and the hydrolysis test was performed, as described above under N 2 .
The opposite consecutive reactions were performed with the same methodology by doing first the hydrolysis reaction using the SO 3 H/AC catalyst under N 2 . Then, the solid was filtered out and RuO 2 / AC catalyst was added to the filtrate. Finally, the hydrogenation was performed under 30 bar of H 2 .
Recyclability tests were performed by re-using catalysts obtained by filtration after a catalytic test. The recovered solid was first washed with mQ water and dried overnight at 100 • C in air. If the catalyst mass was below 150 mg, fresh catalyst that had not been used before was added to maintain the same catalyst quantity for each test. The added mass was never more than 10% of the total.
HPLC analyses were performed with a Waters system equipped with Waters 2414 refractive index (RI) detector (detector temperature, 30 • C). The column used for the analyses is an Aminex HPX 87C column, with mQ H 2 O (18 MΩ.cm at 25 • C) as eluent, a flux of 0.5 mL/min, a column temperature of 85 • C and 25 µL of injected volume. For cellobitol quantification, a calibration curve for cellobiose was used, with the assumption that they have similar refractive indexes.
The conversion of cellobiose is calculated as follows: Cellobiose conversion (%) = n cellobiose converted initial n cellobiose × 100 The glucose selectivity is calculated as follows: Glucose selectivity (%) = n glucose produced 2 × (n cellobiose converted) × 100 The sorbitol selectivity is calculated as follows: Sorbitol selectivity (%) = n sorbitol produced 2 × (n cellobiose converted) × 100 The cellobitol selectivity is calculated as follows: Cellobitol selectivity (%) = n cellobitol produced n cellobiose converted × 100 The glucose yield from cellobiose is calculated as follows: Glucose yield (%) = n glucose produced 2 × initial n cellobiose × 100 The sorbitol yield from cellobiose is calculated as follows: Sorbitol yield (%) = n sorbitol produced 2 × initial n cellobiose × 100 For the catalytic reaction on cellulose substrate, the formulae are slightly different. The conversion of cellulose is calculated as follows: Cellulose conversion (%) = (initial cellulose mass − recovered cellulose mass) initial cellulose mass × 100 Recovered cellulose mass = Total mass after filtration − initial catalyst mass The glucose selectivity is calculated as follows: Glucose selectivity (%) = n glucose produced (n glucose unit converted) × 100 The sorbitol selectivity is calculated as follows: Sorbitol selectivity (%) = n sorbitol produced (n glucose unit converted) × 100 The xylitol selectivity is calculated as follows: Xylitol selectivity (%) = n xylitol produced (n glucose unit converted) × 100 The mannitol selectivity is calculated as follows: Mannitol selectivity (%) = n mannitol produced (n glucose unit converted) × 100 The glucose yield from cellulose is calculated as follows: Glucose yield (%) = n glucose produced initial n glucose unit × 100 The sorbitol yield from cellulose is calculated as follows: Sorbitol yield (%) = n sorbitol produced initial n glucose unit × 100 For all the formulae above, the n glucose unit is calculated as follows: n anhydroglucose unit = cellulose mass molar mass of anhydroglucose unit

Characterization Methods
XPS analyses were carried out at room temperature with an SSI-Xprobe (SSX 100/206) photoelectron spectrometer from Surface Science Instruments (SSI, Moutain View, CA, USA), equipped with a monochromatized microfocus Al X-ray source. Samples were stuck onto small sample holders with double-sided adhesive tape and then placed on an insulating ceramic carousel (Macor ® , Moutier, Switzerland). Charge effects were avoided by placing a nickel grid above the samples and using a flood gun set at 8 eV. The binding energies were calculated with respect to the C-(C, H) component of the C 1s peak fixed at 284.4 eV. Data treatment was performed using the CasaXPS program (Casa Software Ltd., Teignmouth, Devon, UK). The peaks were decomposed into a sum of Gaussian/Lorentzian (85/15) after subtraction of a Shirley-type baseline. Only Ru 3d peaks were used to determine Ru at. % and its oxidation state in all catalysts.
Powder X-ray diffraction (PXRD) patterns were recorded at room temperature on a Siemens D5000 diffractometer equipped with a Ni filter using CuKα radiation (Bragg-Brentano geometry) operated at 40 kV and 40 mA. Diffractograms were taken between 5 • and 80 • (2θ) with a step size of 0.02 • (2θ).
Boehm titration method was used to evaluate catalyst acidity [55,56]. NaOH solutions were prepared via dilution of Titrisol ampoules (VWR) containing precise and known quantities of sodium hydroxide. HCl solutions were prepared via the dilution of concentrated hydrochloric acid. The HCl concentrations were determined by titration with the standard NaOH solutions. These solutions were prepared with mQ water that was previously decarbonated by nitrogen flushing. For titrating the acid groups, 60 mg of sample was dispersed in 30 mL of NaOH 0.01 mol/L, and the solution was decarbonized for 1 h under Ar flux. The mixture was then agitated for 23 h under Ar atmosphere. The suspension was then filtrated and 10 mL of the resulting filtrate was back-titrated (repeated twice) under Ar flux using a HCl 0.005 mol/L solution. The indicator used was phenolphthalein. The amount of acid functions on the catalyst was determined by calculating the difference between the initial amount of NaOH and the amount of NaOH titrated by the HCl.
HR-TEM analyses were carried out on a JEOL 2100 FEG S/TEM working at 200 kV voltage, equipped with a probe corrector for spherical aberrations. For these measurements, samples were dispersed in an ethanol solution and a drop of each suspension was deposited on a copper grid covered with a holey carbon membrane.
STEM analyses were carried out on a JEOL 2100 FEG S/TEM by using a spot size of 0.13 nm with a current density of 140 pA and a camera focal distance of 8 cm, corresponding to inside and outside annular detector diameters of 73 and 194 mrad. EDX mapping was performed using a JEOL Silicon Drift Detector (SDD) with sensor size of 60 mm 2 .
Total organic content (TOC) analyses were performed on solutions after catalytic tests by using a Shimadzu TOC-L analyzer with an ASI-L autosampler using the combustion catalytic oxidation method at 680 • C.

Conclusions
Bifunctional catalysts with sulfonic functions and RuO 2 nanoparticles were synthesized with the aim of directly transforming cellulose into sorbitol (two-step reaction of combined hydrolysis/hydrogenation). The sulfonic function amount was controlled by heat treatment at different temperatures and with pre-treatment under the same conditions as the catalytic test. These modifications allowed us to obtain different ratios between the sulfonic functions and the Ru amounts. It was shown that high amounts of sulfonic functions completely inhibit Ru active sites, and in this case, mainly glucose (a hydrolysis product) was obtained at the end of the reaction, starting from cellobiose as the model molecule. With a higher amount of Ru (5 wt.%), a very efficient bifunctional catalyst was obtained, and its activity outmatched the monofunctional catalyst. The kinetic studies on the monofunctional and bifunctional catalysts showed a striking difference between the two systems. Indeed, with the monofunctional catalyst, sorbitol degraded during the reaction, and the sorbitol yield after 24 h was very low, while the bifunctional catalyst played a stabilizing role for sorbitol and provided a very high yield of it. This was confirmed via a degradation test of sorbitol with both catalysts. The synergy between the sulfonic functions and the Ru nanoparticles was demonstrated by these results. Sulfonic functions stabilize the sorbitol and, therefore, increase the yield; however, it is a subtle balance, as too many sulfonic functions can poison Ru. The recyclability tests showed that the bifunctional catalyst was more stable than the monofunctional one, even if both catalysts suffered from slight deactivation during the first run. Catalyst stability was confirmed with TEM and EDX analyses before and after the reaction. Ru nanoparticle size only slightly increased after five consecutive reactions, and sulfur remained evenly distributed over the entirety of the support.
Finally, the best bifunctional catalyst was assessed for the direct transformation of cellulose into sorbitol. Sorbitol was successfully obtained from cellulose, and the bifunctional catalyst clearly provided better activity than the monofunctional one. The importance of cellulose pre-treatment was demonstrated with the better results obtained when ball milling was applied. Moreover, the combined ball milling of cellulose with the catalyst before the reaction showed outstanding results, as it doubled sorbitol yield (from 21% to 39% within 2 h). The transformation of another crystalline cellulose (Avicel PH101) was carried out with the bifunctional catalyst in order to more easily compare our results with the existing literature. Comparisons with other active metals for this transformation are important, and using a low amount of Ru to achieve a high sorbitol yield could be more interesting than using high quantities of Ni for the same price, especially given that Ru is more resistant to oxidation and leaching than Ni. Moreover, more active metals are more suitable for complex substrates, such as raw cellulose. This comparison with the literature shows that our bifunctional catalyst is efficient and can compete with other systems for sorbitol production.