Selective hydrodeoxygenation of levulinic acid to γ-valerolactone over Ru supported on functionalized carbon nanofibers

1 Selective hydrodeoxygenation of levulinic acid to γ‐valerolactone over Ru supported on functionalized carbon nanofibers Charf Eddine Bounouktaa,b, Cristina Megías-Sayagoc,*, Nuria Rendónc, Fatima Ammarib, Anna Penkovaa, Svetlana Ivanovaa , Miguel Ángel Centenoa, Jose Antonio Odriozolaa aDepartamento de Química Inorgánica e Instituto de Ciencia de Materiales de Sevilla, Centro Mixto CSIC-Universidad de Sevilla, 41092 Sevilla, Spain. bLaboratoire de Génie des Procédés ChimiquesLGPC, Département de Génie des Procéés, Faculté de Technologie, Université Ferhat Abbas Setif-1, 19000 Setif, Argelia. cDepartamento de Química Inorgánica e Instituto de Investigaciones Químicas, Centro mixto CSIC-Universidad de Sevilla, 41092 Sevilla, Spain * Corresponding author, email: [email protected]s, [email protected] Abstract In this work carbon nanofibers (CNF) have been successfully functionalized by using different approaches and finally used for the preparation of Ru based catalysts. The organometallic approach has demonstrated to be suitable for CNF functionalization, leading to well-defined Ru NPs (by adding organosilane, amino or mercapto functionalities, among others) in comparison with the mineral acid treatments conventionally used to activate and/or functionalize carbonaceous solids. All catalysts have been tested in the levulinic acid hydrodeoxygenation to γ‐valerolactone under mild conditions, being the impact of CNF functionalization on the catalysts’ performance fully discussed in comparison with the unmodified commercial CNF. 2 Keywords: GVL, levulinic acid, Ru based catalysts, carbon nanofibers, organometallic functionalization 1. Introduction Biomass is the only sustainable resource containing carbon for chemicals and fuels production. It is the ideal equivalent of petroleum for organic compounds production and in other words, the only viable answer to fulfil our future energy and materials requirements 1–3. In this sense, the main challenge for scientists and industries is the optimization of the processes and their sustainable and environmentally friendly implementation in large capacity biorefineries with economic costs. Levulinic acid (LA) appears as one of the top 12 promising building blocks derived from biomass, due to its multiple applications and availability 4,5. This platform chemical can be converted in important intermediates by different catalytic processes 6, which makes it a versatile material for biomass valorization processes. Concretely, during levulinic acid hydrodeoxygenation reaction (Scheme 1) the unsaturated structure is reduced, and the substrate undergoes oxygen removal via dehydration till gamma valerolactone (GVL) intermediate formation 7. The latter can be further upgraded to pentenoic and pentanoic acids and/or 1,4 pentanediol and methyl tetrahydrofuran (2MTHF), appropriate for monomers and branched hydrocarbons for diesel, gasoline and jet biofuels manufacturing 8–13. GVL have thus emerged as a key intermediate within biorefinery, standing out as an excellent aprotic polar solvent, gasoline additive, food ingredient, flavoring agent and as intermediate for the synthesis of polymers 12. The formation of GVL can proceed through LA reduction to 4-hydroxypentanoic acid (4-HPA) and further dehydration of the latter or via angelica lactone (AGL) (see Scheme 1). In any case, both reduction and dehydration reactions take place thus needing the presence of different active 3 sites to occur. The usually reported active sites for reduction are metallic nanoparticles, while dehydration is guaranteed by the presence of acidic sites. Scheme 1. LA hydrodeoxygenation transformation reactions. Several heterogeneous catalysts have been reported to be active in the levulinic acid HDO 14–16, being the Ru-based ones the most preferred due to their intrinsic ability to hydrogenate levulinic acid C=O functionality 17. The observed catalytic performance depends on the support type, being catalyst’ reusing possibilities strongly linked to the support’ stability in aqueous environment, 4 especially in acidic conditions 18,19. In contrast to metal oxides, carbonaceous supports have demonstrated to be superior, maintaining Ru/C catalytic performance after different operation cycles 16,20,21. The unique properties of carbon nanofibers (CNF) have turned these materials in very promising in numerous applications. Nevertheless, in most cases, their huge potential is limited by their hydrophobic and inert surfaces, reason why different modification techniques have been developed. Approaches such as thin film deposition or chemical/biochemical functionalization have been successfully applied to activate and tailor CNF’ surface 22. The latter sometimes results crucial for their use as catalytic supports since CNF surface can further affect the nanoparticles deposition process 23. Adding functionalities of different nature and concentration usually results in the creation of new anchoring sites for metal nanoparticles, affecting metal/support interaction and nanoparticles stabilization, size and dispersion 24. In addition to it, CNF’s surface modification can change the role of CNF in the catalytic process, from just being a mechanical support to be an active participant in the reaction if properly functionalized 25,26. Exploring the possibilities of CNF multifunctionalization is thus convenient not only to improve the catalyst preparation but also to enhance their catalytic performance in the selected process. In this work, commercially available carbon nanofibers have been functionalized using several approaches, including different acid treatments (nitric and sulfuric acids, respectively) and chemical grafting (adding organosilane, amino, mercapto or imidazolium derivatives functionalities) being the principal aim of the functionalization, the introduction of sulfur or nitrogen heteroatoms via groups with different acid/base character. After the different modifications, Ru supported CNF catalysts have been prepared and tested as catalysts for the selective hydrodeoxygenation of levulinic acid to γ‐valerolactone in aqueous media. The impact 5 of CNF functionalization on the catalysts’ properties has been discussed in detail and compared with unmodified commercial CNF. 2. Experimental 2.1 Carbon nanofibers functionalization Preparation of CNF-O and CNF-S. Commercially available carbon nanofibers CNF (GANF13, Grupo Antolin Ingeniería) were treated with nitric acid HNO3 (37 %) and sulfuric acid H2SO4 (96 %) at 60 ºC and 150 °C respectively in reflux equipped round bottom flask. The resulting solids were abundantly washed till neutral pH, dried overnight at 60ºC and grinded in a mortar to get finally CNF-O and CNF-S samples (O and S standing for nitric and sulfuric acid treatments, respectively). Preparation of CNF-APTMS and CNF-MPTMS. For the modification of CNF with organosilane, amino or mercapto functionalities 2 g of CNF-O sample were suspended in 50 mL of toluene and 1.5 mL of amino/mercapto -propyltrimethoxysilane (APTMS or MPTMS) and stirred under nitrogen and reflux overnight. The obtained solid is filtered, washed three times with toluene and dried under nitrogen. Finally, the functionalized CNF were labelled as CNF-APTMS and CNFMPTMS respectively and dried at 100 ºC in an oven overnight. Preparation of CNF-ILs. For the immobilization of ionic liquid based on imidazolium over CNF surface, the oxidized CNF-O sample (2g) was reacted with thionyl chloride under nitrogen at 70 °C during 24 h. The excess of SOCl2 was removed by washing with anhydrous THF and the sample was finally dried under nitrogen. The next step of immobilization consisted on reacting the solid with 3-chlor-1-propanol at 120 °C for 24 h under reflux in nitrogen. The excess of 3-chlor-1- 6 propanol was removed by washing with dichloromethane. Finally, the as prepared solid was reacted with 1-methyl imidazole under reflux at 95 °C during 24 h to allow the formation of the final immobilized ionic liquid on CNF surface. The catalyst was labeled CNF-ILs. Summarizing, the following samples will be used in the next step: CNF, CNF-O, CNF-S, CNFMPTMS, CNF-APTMS and CNF-ILs. 2.2 Metal impregnation Ruthenium (1 wt. % nominal loading) was deposited on the different functionalized carbon nanofibers by means of incipient wetness impregnation method using ethanol/water mixture as a solvent and ruthenium (III) nitrosil nitrate (Johnson Matthey; purity = 14.34%) as metal precursor. In this way, the required amount of precursor was dissolved in 50 mL solvent mixture and mixed with 2 g of functionalized CNF. Then, the solvent was evaporated in rotary evaporator and the resulting solid dried overnight at 100 °C. The samples were reduced at 400 °C during 2 h (10 °C/min heating ramp) under 100 mL/min N2/H2 (1/1 composition) prior to reaction. 2.3 Characterization techniques. The textural properties of the samples were evaluated by means of N2 physisorption in a Micromeritics TRISTAR II equipment. The samples were degassed for 12 hours prior to the analysis. Raman spectra of the different samples were recorded on a dispersive Horiba Jovin Yvon LabRam HR800 Confocal Raman Microscope equipped with a green laser (λ = 532.14 nm) and working at 5 mV power using a 600 grooves/mm grating. The microscope used a 50x objective with a confocal pinhole of 1000 μm. 7 The samples’ acidity was estimated over 50 milligrams of solid dispersed in 50 ml of aqueous solution by using a pH-electrode (Metrohm). The pH values were collected with time until the data values remain constant. TEM images were obtained with a FEI Talos electron microscope operating at 200 kV acceleration voltage and equipped with a Field Emissions filament and Ceta 16M camera. The samples were previously supported on a holey carbon-coated copper grid without using any liquid. The mean particle size was estimated by measuring more than 200 particles according to the equation (1): 𝐷𝑃=∑𝑛𝑖𝑑𝑖3 ∑𝑛𝑖𝑑𝑖2 (1) where ni is the number of particles with di diameter. The metal dispersion was estimated from the average particle size obtained from TEM and using a mathematical model for cuboctahedral or hemispherical particles 27. The metal contents for all carbon supported catalysts were determined by inductively coupled plasma atomic emission spectroscopy (ICP). To this aim, a Horin Jobin Yvon ICP spectrometer was used, being the samples previously digested in HF. SEM/EDX analysis were performed on Hitachi S4800 SEM-FEG high resolution (5 nm) scanning electron microscope provided with SE and BSE detectors and a Bruker X Flash Detector 4010 EDX analyser with a resolution of 133 eV in the Mn Kα line. Operating conditions were varied for every experiment. XPS measurements were carried out on a SPECS spectrometer equipped with PHOIBOS 150 MCD analyzer working with fixed pass energy of 40 eV and 0.1 eV resolution for the studied zones. Al Kα radiation (1486.6 eV) was used working on 250W and 12.5 kV voltage. The analytical chamber operates at ultra-high vacuum at around 10−10 mbar pressure. Prior the analysis, 8 the samples were pressed into a thin disk. All XPS spectra were recorded at room temperature with binding energy calibration on C1s at 284.6 eV with an uncertainty ±0.2. The obtained data were processed using the CasaXPS software. The decomposition into components was performed by using Gaussian-Lorentzian functions with Shirley background subtraction. 2.4 Catalytic reaction and analysis The reaction was performed in a 50 mL Parr 4597 autoclave equipped with 4848 P.I.D. controller and stirrer. Typically, 20 mg of catalyst and 10 mL of levulinic acid aqueous solution (0.5 M) were added to a Teflon liner adapted to a high-pressure autoclave reactor. The sealed autoclave was then pressurized/depressurized three times with 5 bars of nitrogen and finally pressurized with 10 bars of H2 at room temperature. Then, the reactor was rapidly heated to the desired temperature and the conditions maintained the desired time. Once the reaction finished, the reactor was cooled down to room temperature using an ice/water bath. Once depressurized, the reaction mixture was separated from the catalyst with a syringe filter (0.45 μL syringe Nylon filter) and directly analyzed with a HPLC Infinity 1260 instrument, equipped with both Diode array (DAD) and Refractive index (RID) detectors and a Hi-Plex H 300 × 7.7 mm column. For the recycling experiments, the catalysts were separated from the liquid, then washed with water, dried at room temperature and activated at 400 °C in nitrogen/hydrogen flow before the next run. The levulinic acid conversion, yield and selectivity to the GVL have been calculated as follows: 𝐋𝐞𝐯𝐮𝐥𝐢𝐧𝐢𝐜 𝐚𝐜𝐢𝐝 𝐜𝐨𝐧𝐯𝐞𝐫𝐬𝐢𝐨𝐧 (%)= 𝐦𝐨𝐥𝐞𝐬 𝐨𝐟 𝐫𝐞𝐚𝐜𝐭𝐞𝐝 𝐋𝐀 𝐦𝐨𝐥𝐞𝐬 𝐨𝐟 𝐢𝐧𝐢𝐭𝐢𝐚𝐥 𝐋𝐀 ∗𝟏𝟎𝟎% (2) 𝐏𝐫𝐨𝐝𝐮𝐜𝐭 𝐲𝐢𝐞𝐥𝐝(%)=𝐦𝐨𝐥𝐞𝐬 𝐨𝐟 𝐟𝐨𝐫𝐦𝐞𝐝 𝐩𝐫𝐨𝐝𝐮𝐜𝐭 𝐦𝐨𝐥𝐞𝐬 𝐨𝐟 𝐢𝐧𝐢𝐭𝐢𝐚𝐥 𝐋𝐀 ∗𝟏𝟎𝟎% (3) 9 𝐏𝐫𝐨𝐝𝐮𝐜𝐭 𝐬𝐞𝐥𝐞𝐜𝐭𝐢𝐯𝐢𝐭𝐲 (%)=𝐦𝐨𝐥𝐞𝐬 𝐨𝐟 𝐟𝐨𝐫𝐦𝐞𝐝 𝐩𝐫𝐨𝐝𝐮𝐜𝐭 𝐦𝐨𝐥𝐞𝐬 𝐋𝐀 𝐜𝐨𝐧𝐯𝐞𝐫𝐭𝐞𝐝 ∗𝟏𝟎𝟎% (4) 𝐓𝐎𝐅 = 𝐦𝐨𝐥𝐞𝐬 𝐋𝐀 𝐜𝐨𝐧𝐯𝐞𝐫𝐭𝐞𝐝 𝐦𝐨𝐥𝐞𝐬 𝐑𝐮∗𝐡∗𝐝𝐢𝐬𝐩𝐞𝐫𝐬𝐢𝐨𝐧 (5) 3. Results and discussion The textural properties of all prepared solids were analyzed by N2 physisorption and summarized in Table 1. After Ru impregnation, CNF’s textural parameters did not change significantly. BET surface variations of less than 10% are observed with a slight increase observed in the most cases, exception made by CNF and CNF-S. This surface increase can be assigned either to the metal surface contribution or to some surface reorganization and functionalization groups loss during the degassing procedure. The average pores volume and diameter size decreases after impregnation suggesting Ru nanoparticles introduction within the pores with a positive contribution to the total surface area but a negative contribution to the porous structure. Table 1. Textural parameters of the used catalysts and initial supports (indicated in parenthesis). Samples BET area catalyst (support) (m2/g) Pore volume catalyst (support) (cm3/g) Average pore size catalyst (support) (nm) Ru/CNF 146 (160) 0.596 (0.691) 13.8 (15.5) Ru/CNF-O 183 (178) 0.542 (1.188) 10.31 (15.4) Ru/CNF-S 160 (178) 0.63 (0.780) 13.6 (13.0) 16 in type and intensity corresponding to carbon-nitrogen bond in aliphatic amine (Ru/CNF-APTMS), or N-heterocyclic carbene (Ru/CNF-ILs). As for ruthenium state, Figure 2c shows the XPS spectra of Ru3p region. The Ru3p signals are considered instead of 3d, due to overlapping between the Ru3d with C1s and in order to evite the ambiguity in determining the exact oxidation state of ruthenium31. The peak position indicates metallic Ru state for both samples. Nevertheless, the peak intensity is different suggesting differences in dispersion. Indeed, XPS analysis confirms the better dispersion for the Ru/CNF-APTMS sample also reflected by the surface composition presented in Table 5. Both samples present the same metal loading (Table 3) but it is Ru/CNF-APTMS that showed 96% of the targeted values on the surface, suggesting that higher surface acidity results in higher metal uptake and dispersion. Table 5. Superficial XPS composition of two selected catalysts. Samples C1s, at.% O1s, at.% N1s, at.% Ru3p3/2, at.% Si2p, at.% Ru/CNF-ILs 95.6 3.16 1.09 0.14 - Ru/CNF-APTMS 86.3 9.8 0.97 0.58 2.2 17 Figure 2. XPS a) O 1s region, b) N 1s region and c) Ru 3p region of the selected Ru CNF samples. 18 4. Catalyst screening The levulinic acid hydrodeoxygenation reaction has been carried out over all prepared samples and the obtained results are presented in Figure 3. The performances of the Ru/CNF catalysts are estimated by measuring the LA conversion and GVL yield in aqueous solutions at 100 ºC under 10 bar H2 using 20 mg of catalyst charge and 10 mL of 0.5 M LA. The catalytic activity is clearly influenced by the support nature. The non-treated CNF sample presents the lowest LA activity with only 29% of conversion and 14% GVL yield. The supports functionalized to contain nitrogen heteroatoms (with aminosilane or ionic liquids based on imidazolium) are found to be more active than those containing sulfur. The observed differences in textural properties and elemental analysis within the series cannot explain completely the increase of the activity for the N-containing samples and some electronic effects must be considered due to the highest electronegativity of nitrogen. Those effects allow in principle higher electron density and stronger metal/support interaction. Several studies have reported the benefits of introducing nitrogen in carbon structures 32,33 with the major benefit provided by the nitrogen sites being act as metal anchorage centers for metal dispersion. Therefore, better-dispersed particles provide higher available catalytic surface resulting in higher LA conversion and GVL yield, as confirmed also by Li et. al 34. Moreover, the nitrogen groups could also participate actively in the HDO reaction. LA hydrodeoxygenation reaction is initiated by adsorption and dissociation of molecular hydrogen on the active metal surface and with the formation of Metal-H bond 35. Then, the LA molecule is adsorbed in vicinity through its carbonyl group integrating the hydrogen atom and generating an intermediate converted by several mechanisms to GVL. 19 Figure 3. LA HDO to GVL over Ru CNF catalysts. Reaction conditions: 120 min, 100°C, 10 ml 0.5 M [LA], 20 mg of catalyst, 10bars PH2. Within the catalyst series, Ru/CNF-APTMS sample presents the highest LA HDO activity with a total selectivity to GVL followed by the Ru/CNF-ILs sample with 96% conversion and 90% yield. It appears that amine and ionic liquid moieties exercise an important positive effect on LA conversion and GVL yield, being the common between them the presence of N centers. Although present in both samples, the N centers acts differently. While the Ncenters in APTMS act as strong donors to the metal centers facilitating on first place, the metal dispersion and helping the electron transfer from LA C=O bond to the metal 36 the IL containing sample seems to remain active even at low metal dispersion. The later indicates that the prevailing mechanisms of activity promotion must be different for both samples. If we consider that the IL exercise the same effect on the activity as when used as a solvent, we can only speculate that the grafting results in a very 0 20 40 60 80 100 % LA conversion GVL yield GVL selectivity 20 closely positioned Ru *** N-IL sites and that the N acts as an ion pairing center extracting one proton from the H-H adsorbed on the Ru surface facilitating the hydrogen coordination and transfer to the molecule to be hydrogenated 37. This effect suggests the presence of Ru-N coupled sites which could assist subsequently the π-back-bonding of C=O(LA) facilitating the formation of active C=O(LA)-H intermediates that in total, result in higher GVL yield. Of course, these mechanisms of promotion are conditioned by the position of the N centers (obeying generally on - stacking), its amount, proximity to the Ru sites and nature. All this would result in a different mechanism of hydrogen activation, but the reaction will proceed always through the formation of active structures of stabilized N-Ru-C=O(LA)-H complexes where the imidazolium cation and the amino group (-NH2) act as one and two isolated electron donor ligands for the metal sites. Thus, by adding the right N-center the HDO activity can be promoted indifferently to the dispersion of the active site. As for the other samples, the Ru/CNF-O sample is significantly less active (81% conversion and 68% yield) than the samples mentioned above but still better than sulfur containing catalysts. The sulfur containing catalysts seems to follow the same trend, the simultaneous presence of sulfur and silicium for the Ru/CNF-MPTS vs. Ru/CNF-S catalysts improves the LA conversion and GVL yields but still stays far away from the nitrogen containing catalysts. 4.1. Reaction parameters effect The LA HDO reaction is optimized over Ru/CNF-ILs and Ru/CNF-APTMS catalysts varying different reaction parameters. Changing the reaction time from 15 to 240 min (Figure 4) at 100 °C shows a continuous increase of the conversion and GVL yield. Even in the first 15 min LA conversion attains 40 % with 58% GVL yield for Ru/CNF-ILs and 63 % and 82% for Ru/CNFAPTMS respectively. Both catalysts reach full LA conversion in 180 and 120 min respectively. 21 Further increase of the reaction time causes changes in yield and selectivity. The maximum yield of GVL ∼99 % is reached at 240 min for Ru/CNF-ILs while the maximum yield of GVL for Ru/CNF-APTMS is achieved after 120 min after which it drops significantly due to a further hydrogenation of GVL to 1,4 pentandiol. Figure 4. Time effect of LA HDO to GVL over a) Ru/CNF-ILs and b) Ru/CNF-APTMS. Reaction conditions: 100°C, 10ml of H2O, [LA] 0.5M, 20mg of catalyst, 10 bars PH2. 0 20 40 60 80 100 15 30 60 90 120 180 240 % Time (min) a) LA conversion GVL yield GVL selectivity 0 20 40 60 80 100 15 30 60 90 120 180 240 % Time (min) b) LA conversion GVL yield GVL selectivity 22 The reactions in batch reactor are difficult to normalize and the use of TON number is well accepted (Table 3), as it considers the surface availability of the active sites. It is clear that the presence of Nspecies, no matter the Ru dispersion, increases the TON of the reaction presenting the Ru/CNF-ILs, Ru/CNF-APTMS and Ru/CNF-O samples the highest turnover number. On the other hand, and despite being not very significant for reactions without reactive renew, the turnover frequency (TOF) can be also used to compare the reaction results at different time and durations and hydrogen pressure (Figure 5) with the most of the literature data. Our best catalysts, at the time of their lowest possible conversion, showed TOF of 17222.2 h-1 and 9887.8 h1 (Ru/CNF-ILs and Ru/CNF-APTMS, respectively), values more than comparable with several studies at optimum reaction conditions of temperature, pressure, and Ru content. Indeed, Ru/CNF-ILs catalyst exhibits the highest observed TON and TOF values, which showcases the catalyst’s superiority despite of the lower obtained dispersion. Figure 5. TOF versus pressure. Literature data 21,38–43. Legend notes: reduced graphene oxide (RGO), activated biochar from loofah sponge lignocellulosic plant (LC), nitrogen carbon spheres 23 (NCS), manganese octahedral molecular sieves (OMS), Mn doped MCM-49 zeolite (Mn(0.7)MCM-49), activated carbon (AC) A further optimization was made by varying the temperature for Ru/CNF-ILs catalyst (Figure 6). The catalyst shows a very good conversion even at 60ºC. Nearly full yield is achieved at 140 °C in 2h of reaction time. Figure 6. Temperature effect of LA HDO to GVL over Ru/CNF-ILs catalyst. Reaction conditions: 120 min, 10mL of H2O, [LA] 0.5M, 20 mg of catalyst, 10 bars PH2. The hydrogen pressure effect is studied in the 5-20 pressure range at 1 h and temperature of 100°C over Ru-CNF-ILs catalyst (Figure 7). The LA conversion, GVL yield and selectivity increase with H2 pressure, attaining a constant value of GVL yield after 10 bars. Shifting to 15 and 20 bars increases the conversion but does not change the GVL yield, lowering the GVL selectivity with the formation of pentanoic acid. At the lowest pressure the water coverage on the reduced Ru surface is more important than the H-species, which might be the reason for the observed low selectivity and conversion. 0 20 40 60 80 100 60 80 100 120 140 % Temperature (ºC) LA conversion GVL yield GVL selectivity 24 Figure 7. Pressure effect of LA HDO to GVL over Ru/CNF-ILs catalyst. Reaction conditions: 100 °C, 120 min, 10ml of H2O, [LA] 0.5M, 20mg of catalyst. 4.2. Catalyst reuse The recyclability is studied over Ru/CNF-ILs and Ru/CNF-APTMS catalyst in five reaction cycles. The corresponding results of GVL yield are presented in Figure 8. Results show a stable GVL yield for Ru/CNF-APTMS sample but a decreasing yield for the Ru/CNF-ILs sample with significant changes after the first run (full LA conversion in all cycles for both samples). Post reaction characterization shows slight changes in particle size after the 4th cycle (1.9 nm and 6.5 nm for Ru/CNF-APTMS and Ru/CNF-ILs respectively) and in N content (0.98 wt.% N for Ru/CNF-APTMS and 3.6 wt% N for Ru/CNF-ILs). It appears that those changes does not influence the Ru/CNF-APTMS recycling behavior remaining with a dispersion of 80%. Nevertheless, the selectivity for GVL decreases for Ru/CNF-ILs in favor to 4-HPA, indicating active center modification. One could suspect that the imidazolium N-centers in some cases remain protonated thus decreasing the help that the IL could provide for hydrogen activation. For all this, 0 20 40 60 80 100 510 15 20 % Pressure (bar) LA conversion GVL yield GVL selectivity 25 we can conclude that Ru/CNF-APTMS sample in an excellent candidate to be used as heterogeneous catalyst for biomass valorization via HDO reaction in terms of activity and stability. Figure 8. GVL yield profile during reuse over Ru/CNF-APTMS and Ru/CNF-ILs catalysts. Reaction conditions: 100 °C, 120 min, 10ml of H2O, [LA] 0.5M, 20mg of catalyst, 10bars PH2. 5. Conclusions A series of functionalized carbon nanofibers and Ru catalysts are successfully synthetized and applied in the reaction of levulinic acid transformation to γ‐valerolactone under mild conditions. In this particular case, the organometallic approach proves to be highly versatile for the functionalization of CNF, leading to the generation of well-defined NPs in comparison with the mineral acid treatment, normally used for carbon materials functionalization. Both, organosilanes APTMS and MPTMS functionalized CNF are essential to control the size of the resulting Ru nanoparticles. Especially important appears the presence of Si-R-N centers that could act as anchoring sites to induce high Ru dispersion and as a consequence higher active surface exposure 0 20 40 60 80 100 1 2 3 4 5 GVL yield (%) Run Ru/CNF-APTMS Ru/CNF-ILs