Carbon-Based Catalysts from H₃PO₄ Activation of Olive Stones for Sustainable Solketal and γ-Valerolactone Production

Abstract

The use of activated carbon-based catalysts for the production of solketal and γ-valerolactone (GVL), two products of interest for biorefinery processes, was investigated. Activated carbons (ACs) were prepared by chemical activation of olive stones, an agricultural byproduct, using H₃PO₄ to olive stone mass impregnation ratios (IRs) of 1:1 and 3:1, and under nitrogen or air atmosphere. The ACs showed S_BET values of 1130–1515 m²/g, owing to the presence of micropores (0.45–0.60 cm³/g). The use of an IR of 3:1 delivered a wider pore size distribution, with mesopore volume increasing up to 1.36 cm³/g. XPS confirmed the presence of phosphorus groups with surface concentrations of 2.2–3.2 wt% strongly bonded the AC surface through C-O-P bonds. The ACs were tested as acid catalysts for the acetalization of glycerol in a stirred batch reactor at temperatures of 30–50 °C, glycerol concentrations of 1.5 to 3.4 mol/L, and 1–3 wt% catalytic loading. The catalytic activity was clearly correlated with the quantity of C-O-P acid groups determined by TPD, which increased when ACs were prepared under air atmosphere. The AC prepared with IR 3:1 under air achieved full selectivity to solketal, with activation energy of 49 kJ/mol and conversion of up to 70%, matching the equilibrium conversion value under the optimum reaction conditions. A bifunctional catalyst was prepared over this AC by deposition of 5 wt% zirconium and tested in stirred batch reactor for the hydrogenation of levulinic acid (LA) using isopropyl alcohol (IPA) as solvent and H₂ donor, with LA:IPA ratios from 1:1 to 1:7 and temperatures between 160–200 °C. The catalyst reached full LA conversion and a GVL yield higher than 80% after only 12 h at 200 °C. A test conducted in the presence of water revealed that it was an inhibitor of the reaction. The identification of isopropyl levulinate as an intermediate suggests that the most likely reaction pathway was dehydration, followed by hydrogenation and cyclization, to obtain GVL. Kinetic modelling of the results showed a value of 42 kJ/mol for the hydrogenation step. The reusability of the catalyst was tested for five consecutive reaction cycles, maintaining most of the activity and selectivity towards GVL.

1. Introduction

New policies around the world encourage the sustainable production of chemicals and biofuels from renewable sources to mitigate the global warming effects caused by the overuse of fossil feedstocks. The intensification and sustainability of biorefineries, where biomass is transformed into value-added products and materials [1], requires the development and implementation of zero-waste heterogenous catalytic systems, setting them in line with circular economy concepts [2]. Heterogenous catalysts are generally prepared using inorganic oxides such as silica, alumina, or zeolites, among others. However, activated carbons (ACs) can serve as support and even as catalysts [3], as they are low-cost materials with large specific surface area, high porosity and chemical and thermal stability, and they possess a surface chemistry that can be easily modified [4]. ACs can be cost-efficiently produced by chemical activation of lignocellulosic residues with H₃PO₄ under certain conditions [5,6], such as those obtained on an industrial scale in biorefineries. This process presents several advantages in comparison with other activation methods of lignocellulosic materials, namely, (i) high preparation yields, (ii) controllable pore size distribution, and (iii) generation of thermally and chemically stable phosphorus surface groups with moderate acidic strength [7,8,9,10,11]. Those phosphorus groups have been detected on ACs obtained by H₃PO₄ activation of lignin, olive stone, hemp residues, and other lignocellulosic wastes and are responsible for the excellent activity of those type of materials in alcohol dehydration, hydrodeoxygenation of bio-oils, and methoxylation of terpenes [12,13,14,15].

Biodiesel, a typical biorefinery product also known as fatty acid methyl esters (FAMEs), is produced through transesterification of fatty acids with an alcohol, typically methanol, in the presence of basic catalysts. The most common feedstocks for second-generation biodiesel production are waste cooking oils, animal fats, and lignocellulosic biomass [16]. During the transesterification process, crude glycerol is generated as a byproduct at a mass ratio of approximately 1:10 with respect to biodiesel. Currently, the sustainable and profitable utilization of this byproduct involves significant challenges, which have led the scientific community to explore numerous processes and applications for this byproduct [17]. The production of hydrogen by reforming glycerol [18], the carboxylation of glycerol for the synthesis of polyurethane [19], or the hydrogenation of glycerol to obtain chemical products and additives [20] are some of the proposed valorization methods that could be directly integrated into a biorefinery. Among these processes, the production of solketal by glycerol acetalization stands out for its special relevance in the framework of a lignocellulosic biorefinery [21]. Solketal is a versatile liquid that can be used as a fuel additive, reducing gelling, emissions, and consumption, and increasing octane ratings [22]. Selectivity to solketal in this process is usually high, as the byproduct, 2,2-dimethyl-1,3-dioxan-5-ol, has low thermal stability due to the axial position of its methyl group, which significantly favors the formation of the five-membered ring species (solketal). Acetalization of glycerol with acetone for solketal production requires acid catalytic activity, which traditionally was provided by liquid mineral acids such as HF, HCl, H₃PO₄, H₂SO₄, or p-toluenesulfonic acid [23]. However, current methods often involve the use of heterogeneous catalysis, using co-solvents such as ethanol [24] or a solvent-free approach with excess acetone. The most studied heterogeneous catalysts have been solid acids such as functionalized silica [25], zeolites [26], heteropolyacids [27], or ion exchange resins [28].

Another process of industrial relevance in the context of a lignocellulosic biorefinery aimed at producing biofuels is the hydrogenation of levulinic acid to γ-valerolactone (GVL). This chemical is a promising fuel additive that may serve as an alternative to ethanol due to its low vapor pressure and the reduction of CO emissions and smoke. In addition, its lack of azeotropes reduces the energy consumption associated with water removal [29]. The production of γ-valerolactone from levulinic acid can proceed by two different routes: (i) hydrogenation of levulinic acid over a metal catalyst to form 4-hydroxypentanoic acid, followed by its dehydration over an acid catalyst, or (ii) dehydration of levulinic acid followed by hydrogenation of the α-angelic lactone to γ-valerolactone [30]. In both cases, hydrogen is required, but some authors have proposed the use of hydrogen promoters such as alcohols [31] or formic acid [32] as alternatives to gaseous hydrogen. Regarding catalysts, although in the recent decades Ru-based homogeneous catalysts have been widely studied in this reaction [33], current research has shifted towards the development of bifunctional noble metal heterogeneous catalysts, with Ru, Rh, Pd, Pt, or Au as metal phase on an acidic support [34]. However, due to their high cost, some authors have proposed alternatives to noble metals catalysts for the production of γ-valerolactone, like the Meerwein–Ponndorf–Verley reaction, in which alcohols are used as hydrogen promoters and metal oxides such as γ-Al₂O₃ or ZrO₂ as catalysts [31].

The objective of this work is to highlight the potential use of activated carbons prepared from lignocellulosic waste in reactions of interest within the field of biorefinery, to increase plant sustainability. The concept is based on the idea that, instead of using different acid catalysts based on inorganic oxides, such as alumina, silica or zeolites, for each biorefinery process, easy-to-prepare, acid ACs obtained by H₃PO₄ activation of biomass waste can serve as the starting point for preparing catalysts with different functionalities. In such a method, both the feedstocks and the catalysts would be obtained from renewable sources.

Specifically, acidic characteristics of carbon-derived catalysts from olive stone were studied for glycerol acetalization to solketal. For levulinic acid hydrogenation, zirconia was added to the catalyst showing the highest acid character, to provide the redox function needed for the production of GVL.

2. Results and Discussion

2.1. Characterization of Catalysts

Table 1. Preparation yields and main textural parameters measured by adsorption–desorption of N₂ and adsorption of CO₂ at −196 and 0 °C, respectively, for the carbon-derived catalysts.

Sample	Preparation	N2 Adsorption–Desorption			CO2 Adsorption
	Yield (%)	ABET (m2/g)	Vs (cm3/g)	Vmeso (cm3/g)	ADR (m2/g)	VDR (cm3/g)
OS1	42.0	1205	0.55	0.16	535	0.21
OS1A	43.4	1405	0.60	0.40	560	0.22
OS3	42.9	1245	0.47	1.12	435	0.17
OS3A	42.8	1515	0.55	1.37	530	0.21
OS3AZr	42.9	1130	0.46	0.50	460	0.19

presents the preparation yields of activated carbons prepared under inert and air atmospheres. Preparation yields over 40% were achieved owing to the catalytic effect of H₃PO₄ in the dehydration of lignin structure with cleavage and formation of phosphate and polyphosphate crosslinks, reducing the loss of volatiles from the carbon precursor and thus increasing the yield of activated carbon [35,36]. The yields remained almost equal between the activated carbon prepared under inert atmosphere and in the presence of oxygen, probably due to the inhibition of carbon oxidation exerted by the phosphorus groups on the surface of the AC [37,38].

Figure 1A shows the N₂ adsorption–desorption isotherms at −196 °C for the different carbon-derived catalysts. The activated carbons prepared under inert atmosphere at different impregnation ratios, OS1 and OS3, presented quite different N₂ adsorption–desorption isotherms. The OS1 isotherm was Type I(b) with a small hysteresis loop, type H4, characteristic of activated carbons with a broad micropore size distribution, also including narrow mesopores. On the other hand, the Type I + IV adsorption–desorption N₂ isotherm for OS3 revealed not only a highly developed mesoporosity but also the presence of microporosity. For these samples, the mean pore size increased with the impregnation ratio, owing to the production of phosphate ester bonds that widened the porosity. Thus, mesopore volume increased seven-fold with only a slight modification of the microporosity volume, as shown in Table 1. This may be useful to minimize mass transfer limitations usually present in liquid phase reactions. The A_BET as compiled in Table 1 showed only a small increase for OS3 compared with OS1, due to the lower contribution of wide mesopores to the surface area compared with narrow mesopores and micropores. The lower value of micropore volume measured by CO₂ adsorption at 0 °C (${\mathrm{V}}_{\mathrm{S}}^{{\mathrm{N}}_2}>{\mathrm{V}}_{\mathrm{D}\mathrm{R}}^{\mathrm{C}{\mathrm{O}}_2}$) indicates that all the samples had a wide microporosity (mean micropore size > 0.7 nm) [39].

Meanwhile, activation in the presence of air indicates an additional positive feature of porosity development. Samples OS1A and OS3A showed larger N₂ uptakes at low relative pressures and a steeper line at medium and high relative pressures (Figure 1A). Consequently, micropore and mesopore volumes, as well as A_BET, increased between 15 and 20% (Table 1). The increases in porosity of H₃PO₄-activated carbons prepared under air atmosphere was associated with the inhibition of aromatization reactions by oxygen, rendering lower bulk densities of the resulting carbons [36]. Figure 1B compares the pore size distribution (PSD) of the carbon samples. The increase in the impregnation ratio from 1 to 3 promoted the formation of wide mesopores in the 15–25 nm region. It also broadened micropores, as depicted by the decrease in the 0.4–1 nm peak and the increase in narrow mesopores at 2–5 nm observed for OS3. The higher values of the PSD profiles of air-activated samples (dotted lines) indicate the generation of new pores across the whole size range.

Figure 1A also shows the changes in the N₂ adsorption isotherms of OS3A after loading the Zr, OS3AZr. The main differences between the OS3A catalyst and the bifunctional catalyst (OS3AZr) were observed at medium and high relative pressures, corresponding to the mesopores. As indicated by the NLDFT pore size distributions in Figure 1B, the mesoporosity was reduced over the entire range (2 to 30 nm) for OS3AZr. Despite these changes, the micropore volume remained practically intact, as the metal phase was mainly deposited in the mesopores. However, the pore size distribution in the micropore region became narrower, probably due to metal deposition in the walls of the narrower mesopores and the wider micropores, reducing their sizes to the range of 0.7–1.2 nm, as can be seen in Figure 1B.

Figure 2 shows the temperature-programmed desorption profiles of CO and CO₂ for the activated carbons. At temperatures around 300 °C, an increase in the CO₂ desorption profile observed, associated with carboxylic acid decomposition, followed by the decomposition of anhydrides at approximately 470 °C, visible in both the CO and CO₂ profiles. Between 600 and 700 °C, the desorption of phenols (releasing CO) and lactones (evolving as CO₂) was observed in the corresponding TPD profiles [40]. The sharp and significant evolution of CO observed at 800 °C was attributed to the presence of phosphate groups, with a small peak in the CO₂ profile at the same temperature related to secondary CO reactions. These groups are generated during the chemical activation of biomass waste and remain strongly attached to the carbon surface through C-O-P bonds. Previous studies have proven that they confer both acid and redox functions to the carbon surface C-O-P bonds are thermally stable up to 800 °C, where the decomposition of C-O-P bonds to C’-P bonds takes place with the release of CO [10].

Similar trends were observed in all profiles, although the desorption rates (i.e., the amount of functional groups) varied between the air-activated and nitrogen-activated samples.

The amount of evolved CO and CO₂ during TPD was calculated and is reported in

Table 2. Summary of amounts of evolved CO and CO₂ from TPD, and mass surface concentration derived from XPS analyses of the carbon-derived catalysts.

Sample	TPD (mmol/g)			XPS (wt%)
	CO	CO2	O	C	O	P	Zr
OS1	1.51	0.19	1.88	87.7	9.1	3.2	-
OS1A	4.02	0.81	5.64	77.6	20.2	2.2	-
OS3	1.82	0.24	2.30	87.9	9.6	2.5	-
OS3A	4.85	0.95	6.76	80.5	16.8	2.7	-
OS3AZr	1.76	0.42	2.61	76.7	13.5	2.7	6.7

. The total oxygen amount, O, was estimated from the amounts of CO and CO₂. These quantities were much higher (about three times greater) for air-activated ACs. The excess oxygen on the surface of OS1A and OS3A samples seems to have been mostly in the form of phosphate functional groups and also as CO₂-evolving acid groups, to a lower extent. The presence of the former groups is clearly demonstrated in Figure 2A; the CO desorption rate at 800 °C associated with the thermal decomposition of C-O-P increased from ca. 1–1.3 (OS1 and OS3) up to 2.6–3.2 µmol/g·s (OS1A and OS3A). The current finding of a marked shoulder at 625 °C in OS1A and OS3A has been also observed for P-containing carbons treated in air at 200–350 °C [10] and electrooxidized at harsh conditions in acid electrolyte [8]; it is therefore attributable to the oxidation of P functional groups during the activation step in air, which were transferred to the carbon surface to form phenolic groups. In addition, the use of a larger impregnation ratio also increased the O content in approximately 20% for both nitrogen- and air-activated carbons (see differences between OS1–OS3 and OS1A–OS3A, Table 2). This increase was again mostly associated with higher CO release at 800 °C. The increase in CO evolution at this temperature could indicate higher phosphorus fixation during the activation process.

Table 2 also shows the surface concentrations of the chemical species found on the carbon-based catalysts measured by XPS. It was observed that the carbon content in all samples was close to 80%, as expected for activated carbons. The XPS results confirmed the presence of P, which was present in relevant amounts in all samples. This phosphorus came from the formation of phosphate and polyphosphate ester bonds during the chemical activation of lignocellulosic materials with H₃PO₄ [35]. The activated carbons whose activation process was carried out in air atmosphere, i.e., OS1A and OS3A, showed higher oxygen content on the surface, in agreement with the results obtained in the thermal-programmed desorption analyses. The bifunctional catalyst OS3AZr presented a carbon and phosphorus surface concentration similar to OS3A, with an increase in oxygen content, attributable to the oxidation process during the decomposition of the metal precursor carried out in air at 250 °C. It is also noteworthy that the surface Zr content was ca. 35% higher than the nominal concentration (5 wt%). This observation is in agreement with the findings presented in the results for porous texture, where the significant reduction in N₂ adsorption at high relative pressures was associated with the preferential deposition of zirconium in the mesopores. Figure 3A shows the P_2p spectra of the OS3A catalyst and the OS3AZr bifunctional catalyst. These spectra showed a shift towards higher binding energies (at about 134.6 eV), which has been previously related to the presence of zirconium phosphate species [15]. Figure 3B, which represents the Zr_3d spectrum of OS3AZr, also confirms the formation of this functional group, showing a prominent peak at 183.2 eV, typical of the zirconium phosphate species [15].

2.2. Catalytic Glycerol Acetalization

The acetalization of glycerol was investigated using acidic ACs prepared under both inert and oxidizing atmospheres. In addition, various reaction parameters such as temperature, catalyst loading, and acetone-to-glycerol ratio were varied to determine the effects on yield of solketal. The presence of any product different than solketal was not detected in the experiments performed in this study, achieving full selectivity towards this product.

Figure 4A shows the evolution of the yield of solketal (coinciding with conversion of glycerol, due to its total selectivity) as a function of the reaction time for the different carbon-based catalysts operating at 50 °C with a molar ratio of acetone to glycerol of 6:1 and catalyst loading of 1 wt% with respect to the limiting reagent. The yield of solketal with the ACs activated in an oxidizing atmosphere was higher than their inert-atmosphere counterparts, following the order OS3A > OS1A > OS3 > OS1. As can be seen, there was a direct correlation between solketal yield and the amount of oxygen evolved during TPD. In this sense, glycerol acetalization activity is directly related to the quantity and strength of the catalysts’ acidic sites, as in the case of sulfonated grafted silica [25] or zeolites [41]. In the case of ACs prepared by chemical activation with H₃PO₄, the acidity is directly related to the presence of phosphorus C-O-P surface groups, which decompose as CO around 860 °C [11]. As indicated in Figure 2, the yield of solketal followed the same tendency as the CO evolution at high temperatures.

Figure 4A reveals that the formation of solketal was faster for OS3A, probably due to the higher accessibility to the acidic sites caused by the presence of mesopores and larger micropores.

OS3A was selected as the most active catalyst for solketal production; so, the optimization of the experimental conditions was carried out with this catalyst. The effect of reaction temperature was studied at 30, 40, and 50 °C using 1% catalyst loading (catalyst/glycerol) and an acetone–glycerol molar ratio of 6:1. The use of higher temperatures has been reported to produce deactivation by fouling [42]; therefore, the studied temperatures were no higher than 50 °C. The solketal yield versus time reflected an important increase in the catalytic activity when the temperature was raised from 30 to 40 °C (Figure 4B), and a slight rise in the activity of the catalyst was observed at 50 °C, probably related to the exothermic nature of the glycerol acetalization process [28,43]. In this sense, under the reaction conditions used in this study, conversion is similar to the equilibrium value (ca. 70% at 40–50 °C [28,44]).

Additionally, the reaction behavior was studied by increasing the relative amount of catalyst while maintaining a constant temperature of 40 °C and an acetone-to-glycerol ratio of 6:1, as shown in Figure 4C. A rise in solketal yield was observed when the catalyst amount increased from 1% to 3 wt%, attributed to the greater number of acidic sites available in the reaction medium. Specifically, the increase in catalyst loading doubled the yield of solketal at 30 min and allowed it to reach a conversion rate close to thermodynamic equilibrium after 75 min.

Figure 4D shows the yields of solketal at three different glycerol concentrations with 3 wt% catalyst loading and 40 °C temperature. The conversion remained similar for concentrations up to 2.3 mol/L; however, a significant decrease was observed for the highest concentration tested. As the glycerol concentration increased, it moved into the concentration range where saturation of the active site with glycerol was achieved in accordance with the Langmuir–Hinshelwood mechanism. This situation rendered a reaction order for glycerol concentration between 0 and 1, explaining the lower conversion observed. In addition, the adsorption constant of glycerol is higher than that of acetone, and increasing the concentration of glycerol in the solution resulted in a decrease in solketal yield as less acetone was adsorbed at acidic sites near the adsorbed glycerol. On the other hand, the use of a lower glycerol concentration could decrease selectivity towards solketal, favoring the production of 5-hydroxy-2-dimethyl-1,3-dioxane ketal [45]. However, this byproduct was not detected in the product stream for the concentration range studied in this work.

In this context, where reactants interact through a Langmuir–Hinshelwood mechanism and assuming that the solketal desorption constant is high, it can be concluded that the reaction has an irreversible character. Moreover, the concentration of acetone is much higher than glycerol, so the reaction rate could be represented as an approximation by a pseudo-first order reaction under the conditions presented. In this study, the main kinetic constants were calculated using the Arrhenius equation, using the method of initial rates. It was found that the activation energy of OS3A was lower than that observed for the other ACs, at 48 kJ/mol, and significantly lower than other values reported in the literature achieved using ionic exchange resins [46]. Other studies have reported similar activation energy values, but using materials of significant higher cost [28,47] or zeolites [48].

Based on the results obtained from the characterization of the carbon-based catalysts and the glycerol acetalization reaction tests, a proposed reaction pathway is outlined in Scheme 1. This reaction mechanism details the interaction of the main reactants, i.e., acetone and glycerol, with a model active site present in the carbon-based catalysts analyzed in this study. As illustrated, the active site is a C-O-P group, whose oxygen remains bonded to a proton forming an OH group. Such functional groups, according to TPD results, could predominate in activated carbons prepared in an oxidizing atmosphere, such as OS1A and OS3A. As Scheme 1 shows, the catalyzed reaction begins with the protonation of the oxygen of the carbonyl group of acetone by hydrogen from the P-OH group of the active site, with the subsequent nucleophilic attack involving a hydroxyl group of glycerol and the carbon bonded to the protonated oxygen (step I) [49]. Then, a proton is transferred from the protonated adduct to the active site of the catalyst (step II). Subsequently, another hydroxyl group of glycerol transfers a proton to the active site, followed by the nucleophilic attack of that oxygen on the quaternary carbon of the adduct to form the five-membered ring (step III). The reaction concludes with the release of a water molecule, the restoration of the active site, and the formation of the final reaction product (solketal) (step IV). Another possible reaction pathway includes a carbocation as an intermediate, through which the protonated adduct directly releases water, forming a closed form of carbocation that transfers a proton to the negatively charged active site to finally obtain solketal and the initial active site.

2.3. Catalytic Levulinic Acid Hydrogenation

Acidic character is an important feature in catalytic levulinic acid hydrogenation. To produce γ-valerolactone from LA, both acidic and redox functions are necessary in order to promote the involved dehydration, cyclization, and hydrogenation steps [50]. P-free carbon supports usually have weak basic surface character, and additional functionalization treatments would be needed to generate acid sites. For this reason and based on the performance of activated carbon for glycerol acetalization, OS3A impregnated with zirconium to obtain OS3AZr was chosen as the most promising catalyst to transform levulinic acid (LA) into products.

Figure 5A illustrates the yields of the different reaction products in the presence of OS3A or OS3AZr after 24 h under reaction conditions at 200 °C. Three main products were detected: isopropyl levulinate (iPrL), γ-valerolactone (GVL), and methyl 4-isopropoxybutanoate (mPrB). Both OS3A and OS3AZr showed high activity in the reaction, enabling 78% and 93% of LA conversion, respectively. Nevertheless, their yields of GVL were quite different. While OS3A had a GVL yield around 3% (the rest was converted into iPrL), OS3AZr showed a GVL yield around 90%, with small amounts of iPrL and mPrB. In order to deepen in the reaction pathway for the OS3AZr catalyst, an experiment was performed in which the same temperature was maintained but the reaction time ranged from 0 h to 48 h (Figure 5B). The figure shows full LA conversion after only 6 h, but the product distribution differed. After 6 h of the experiment, similar quantities of GVL and iPrL were obtained, while only traces of mPrB were found. Increasing the time to 12 h produced a reduction in iPrL amount, while GVL yield increased significantly. Accordingly, mPrB slightly increased. After 24 h under reaction conditions, iPrL was detected only at trace concentration but the GVL yield reached a maximum of 95%. The amount of mPrB again slightly increased. At the longest reaction time (48 h), iPrL remained at a trace level, GVL yield decreased to some degree, and mPrB yield increased to 6%.

To obtain more insights about the reaction pathway, two series of experiments were performed in which temperature and ratio of formic acid (FA) against levulinic acid were changed, with the reaction time set at 24 h (Figure 6). In all of these experiments, the LA conversion was almost complete but different product distributions could be found. The change in temperature was screened (Figure 6A) and similar conclusions to those obtained for different reaction times could be reached. At the lower temperature (160 °C), the major product was GVL, while there was still a considerable yield of iPrL and only small quantities of mPrB. At higher temperatures, the yield of GVL increased to a maximum of around 96%, which was maintained at 175 and 185 °C but was slightly reduced to 93% at 200 °C. The yield of iPrL decreased as the temperature increased, being only 2% at 175 °C and 0.3% at 200 °C. In contrast, the yield of mPrB gradually rose as the temperature increased, to a value of 3.5% at 200 °C.

Under the reaction conditions studied, IPA was used as hydrogen donor. Nevertheless, some other authors [31] have proposed FA as hydrogen donor, as this compound is a typical byproduct when LA is obtained from biomass. For this reason, a series of experiments in which the FA/LA ratio was changed was performed to assess the presence of FA in hydrogenation (Figure 6B). As this ratio increased, conversion slightly decreased, moving from total levulinic acid conversion at an FA/LA ratio of 1 to 93% conversion at a ratio of 7. A similar trend was observed for GVL yield, as it decreased from 96% to 89% when the ratio increased from 1 to 7. This may have been associated to the possible interaction of FA with the active sites, reducing its activity. Yields of iPrL and mPrB were low compared with GVL, following opposite trends. iPrL increased slightly with the ratio, probably due to the homogenous catalysis of this esterification by the acidic pH, while mPrB slightly decreased, which was probably associated with the small decrease in GVL concentration (as it is supposed to be a product of GVL decomposition). GVL decomposition to produce mPrB can be caused by decyclization of the GVL molecule [51] followed by the addition of an isopropoxy radical, forming 5-isopropoxy pentanoic acid or 4-methyl-4-isopropoxybutanoic acid, which undergoes a methyl transposition to generate mPrB. Another possibility could be the formation of 5-isopropoxy pentanoic acid or 4-methyl-4-isopropoxybutanoic acid, as they have similar mass spectra to mPrB and can be derived from decyclization reactions of GVL.

In light of all the aforementioned data and based on the reaction pathway proposed by Chen et al. [52], Scheme 2 shows two possible reaction pathways for GVL production. In the upper proposal, IPA is dehydrated to propylene as a first step, due to the already reported catalytic activity of similar Zr-loaded P-containing activated carbon in dehydration reactions [15]. After that, the compound reacts with levulinic acid to form iPrL that subsequently interacts with IPA to generate isopropyl 4-hydroxypentanoate. This compound is finally submitted to a cyclization reaction to obtain GVL and IPA. According to the product distribution with time, the production of iPrL is the first step and is carried out quite quickly, being totally completed after only 6 h at 200 °C. After that, the hydrogenation of iPrL seems to have been the limiting step, as no amount of isopropyl 4-hydroxypentanoate was detected, which implies that its cyclization should be very fast.

In this context, an experiment in which water was mixed with IPA (50:50) was carried out and the reaction was completely avoided. However, another experiment was performed using only 1% of water in IPA, and a reduction in GVL yield of around 85% was detected. This result suggests that water acts as an inhibitor of the reaction, in concordance with the reaction pathway proposed in which catalytic dehydration of IPA is the first step.

The other possible pathway proposed by Chen et al. [52] begins with the hydrogenation of the ketone group in the levulinic acid, followed by the dehydration of the 4-hydroxypentanoic acid to obtain the GVL. This reaction pathway cannot be disregarded, but as 4-hydroxypentanoic acid was not detected in the reaction medium, it is not considered to apply to this catalytic system.

Scheme 3 and Scheme 4 illustrate the reaction mechanism at the heterogeneous active site for iPrL and GVL production. In the case of iPrL production in Scheme 3, the heterogeneous active site takes the form of Zr-O-P. After the adsorption of an IPA molecule, the active site breaks and while the isopropoxy radical is bonded to the phosphorus, a hydroxyl group remains bonded to the Zr. Then, LA is adsorbed into the Zr-OH group and the oxygen of that hydroxyl group carries out a nucleophilic attack against the central carbon of the isopropoxy group. At the same time, oxygen bonded to phosphorus interacts with the hydrogen of the hydroxyl group. After that, iPrL is released and hydroxyl groups remain attached to both the P and Zr atoms. These groups interact with each other to release a water molecule, regenerating the initial active site in a was similar to that proposed by Palomo et al. [53].

Conversely, the obtention of GVL from iPrL and IPA begins with the same active Zr-O-P site (Scheme 4). Nevertheless, in this mechanism, after the adsorption of an IPA molecule, the isopropoxy group remains bonded to the Zr while a hydroxyl group is bonded to the phosphorus. Then, the Zr atom interacts with the carbonyl group of an LA molecule, forming a six-membered ring [54,55] in which the isopropoxy group adsorbed on the Zr is also involved. After the hydrogenation of LA, the isopropyl 4-hydroxypentanoate remains bonded to the zirconium atom and is repositioned, allowing interaction with the hydroxyl group bonded to the phosphorus. Finally, a nucleophilic attack by the oxygen bonded to the zirconium is carried out against the carbon of the carboxyl group, causing the cyclization of the structure to form GVL at the same time that IPA is released, and the active site recovers its initial structure.

Finally, an experiment was performed to assess the reutilization of the catalyst (Figure 7). In this experiment, the same catalyst was submitted to several cycles of reaction with fresh reactants. Between cycles, the spent catalyst was washed with water and dried before it was introduced into the reaction medium. With a focusing on the conversion, it remained unchanged after five cycles. Moreover, GVL yield decreased from 97% to around 88% after the fourth cycle (remaining similar in the fifth cycle). This decrease in GVL yield coincided with the increase in iPrL yield, as mPrB yield remained similar in every cycle. As stated previously, iPrL was the first product formed, and it was produced even when the catalyst without zirconium (OS3A) was used. So, after five runs, there were still enough active sites to perform the conversion of LA to iPrL. The GVL yield decreased accordingly, due to coke deposition on the zirconium phosphate active site, as some authors have suggested [56]. Finally, the mPrB yield remained similar. This could have been associated with mPrB formation being associated only with the amount of GVL in the reaction medium, with this amount achieving comparable values after every reaction cycle.

2.4. GVL Production Kinetic Study

Based on Scheme 2, a kinetic model of GVL production from LA (between 160–200 °C and initial LA concentration of 0.26 M) is proposed. The kinetic model supposes negligible heat and mass transfer limitations and ideal batch reactor behavior. The mass balance equation, which represents the discontinuous autoclave reactor, is represented in Equation (1):

$$\left(-r_{LA}\right)={\displaystyle \frac{d{C}_{LA}}{dt}}$$

(1)

where C_LA represents the concentration of levulinic acid (mmol/L), t is the time (h), and −r_LA is the reaction rate (mmol/(g·h)).

The irreversible elemental steps represented in Scheme 2 can be described as pseudo-first-order equations (Equations (2)–(5)), as IPA is in such an excess that the concentration can be considered constant, and adsorption and desorption constants in the liquid phase are high compared with the reaction constant. Kinetic constants are considered to follow the Arrhenius law.

$$\left(-r_1\right)=k_1·C_{LA}·C_{IPA}={k^{\prime }}_1·C_{LA}$$

(2)

$$\left(-r_2\right)=k_2·C_{iPrL}$$

(3)

$$\left(-r_3\right)=k_3·C_{iPrH}·C_{IPA}=k_3^{\prime }·C_{iPrH}$$

(4)

$$\left(-r_4\right)=k_4·C_{GVL}$$

(5)

where C_LA, C_IPA, C_iPrH, and C_GVL are the concentrations of levulinic acid, isopropyl alcohol, isopropyl 4-hidroxypentanoate, and γ-valerolactone (mmol/L), respectively. In this study, r₃ was observed to be fast compared with the rest, as isopropyl 4-hidroxypentanoate (iPrH) was not detected in the reaction medium and, as proposed in Scheme 4, this compound was not released from the active site.

To obtain the kinetic parameters, a set of ordinary differential equations were solved simultaneously, in accordance with Equations (2), (4) and (5) and based on a modified second-order Rosenbrock formula. The objective function (OF) to be minimized through a non-linear fitting method is represented in Equation (6), comparing the measured experimental concentrations and the concentrations calculated by the previously mentioned set of differential equations:

$$OF=\sum {\left({C_i}_{exp}-{C_i}_{calc}\right)}^2$$

(6)

where C_i are the experimental and calculated concentrations (mmol/L) of the different species presented on the reaction medium.

Table 3. Kinetic parameters.

	Ea (kJ·mol−1)	k0 (h−1)
−r1	39	7.0 × 103
−r2	42	1.1 × 104
−r4	63	4.9 × 104

lists the kinetic parameters of the aforementioned reactions and Figure 8 represents the data calculated by the kinetic model (lines) and the experimental data (symbols).

Figure 8 shows the good prediction of GVL concentration achieved by the kinetic model. This model also predicted low values of activation energy for reactions 1 and 2, along with a higher preexponential factor for reactions 1 and 2, explaining the high selectivity towards GVL achieved after 12 h. The activation energy of reaction 4 was higher, in concordance with the low amount of mPrB detected at low temperatures, achieving significant values only at 200 °C.

3. Materials and Methods

3.1. Carbon-Based Catalysts Preparation

Olive stone was used as raw material for the preparation of the activated carbons. The carbon precursor was impregnated with H₃PO₄ (85% w/w, Labkem) at two different phosphoric acid/olive stone mass ratios, i.e., 1/1 and 3/1. After impregnation, the mixture was dried 24 h at 60 °C and then activated in a tubular furnace with nitrogen (99.999%, Linde, Dublin, Ireland) or air flow of 150 cm³/min STP for 2 h at 500 °C, using a heating rate of 10 °C/min. The activated samples were washed with deionized water at 60 °C until constant pH was reached in the eluate. The carbon materials were dried, ground, and sieved between 100–300 µm. The samples were denoted by “OS” followed by a number corresponding to the impregnation ratio. Those samples prepared under air atmosphere were identified by adding an A at the end of the name. The activated carbon OS3A was impregnated via the incipient wetness impregnation method with zirconium (IV) oxynitrate hydrate (ZrO(NO₃)₂·xH₂O, purity 99%, Sigma Aldrich, St. Louis, MO, USA). The activated carbon was loaded with zirconium, dried overnight at 120 °C, and then calcined at 250 °C for 2 h in a tubular furnace under continuous air flow (150 cm³/min STP) to obtain a final Zr concentration of 5 wt%. The Zr-loaded carbon-based catalyst was denoted OS3AZr.

3.2. Carbon-Based Catalysts’ Characterization

The porosity of the samples was characterized by N₂ adsorption–desorption at −196 °C and CO₂ adsorption at 0 °C, which were carried out in an ASAP 2020 model from the Michromeritics Instruments Corporation (Norcross, GA, USA). Samples were outgassed at 150 °C. From the N₂ adsorption isotherm, the apparent surface area (A_BET) was calculated by the application of the BET equation and considering micropore volume (V_s) using the α_s method, using a nonporous carbon black sample (Elftex-120, Cabot Corporation, Boston, MA, USA) as standard. The mesopore volume (V_mes) was determined as the difference between the adsorbed volume at a relative pressure of 0.98 and the micropore volume (V_s). Based on the results obtained for CO₂ adsorption, the narrow micropore volume (V_DR) and surface area (A_DR) were obtained by applying the Dubinin–Radushkevich method. The pore size distribution was estimated by applying a 2D-NLDFT heterogeneous surface model [57] to the N₂ and CO₂ adsorption isotherms of all the samples using SAIEUS 3.0 (Solution of Adsorption Integral Equation Using Splines) software (http://www.nldft.com/, accessed on 26 November 2024). X-ray photoelectron spectroscopy (XPS) was performed to analyze the oxidation state of surface species in the catalysts. The analyses were conducted using a PHI 5700 (Ulvac Phi Inc., Chigasaki, Japan) instrument with a monochromatic Al Kα radiation source (1486.6 eV). The maximum of the C_1s peak was positioned in the XPS photoemission energy of C–C and C=C bonds (284.5 eV) as a reference point. Carbon materials contain several oxygenated functional groups that decompose into CO and/or CO₂ at specific temperature ranges, allowing their identification and quantification through temperature-programmed desorption (TPD). For the TPD experiments, a 100 mg sample of dried carbon-based catalyst was placed in a quartz reactor, which was heated from room temperature to 900 °C at a rate of 10 °C/min under a nitrogen flow of 150 cm³/min. The release of CO and CO₂ during the heating process was continuously measured using a non-dispersive infrared (NDIR) ULTRAMAT 23 (Siemens, Berlin, Germany) gas analyzer. The inorganic composition of the Zr-loaded carbon-based catalyst was assessed by X-ray fluorescence (XRF) using THERMO ARL PERFORM’X (Thermo Fischer Scientific, Waltham, MA, USA) equipment.

3.3. Catalytic Acetalization of Glycerol

The catalytic activity of carbon-based catalysts for the acetalization of glycerol was evaluated in a batch thermostatic reactor with constant stirring (800 rpm). In a typical experiment, the catalyst was placed in a 100 cm³ round-bottom flask, also adding a measured amount of acetone (HPLC grade, 99.9%, Sigma Aldrich). The flask was closed with rubber stoppers and heated to a temperature in the range of 30 and 50 °C. After reaching the set point temperature, glycerol (99.0%, Panreac, Castellar del Vallès, Spain) was introduced into the reaction mixture, which marked the beginning of the experiment. The acetone-to-glycerol molar ratio was adjusted to 6:1, and the catalyst loading ranged from 1 to 3 wt% relative to the glycerol. The glycerol concentration also varied from 1.3 to 3.5 mol/L. During all experiments, aliquots (0.2 cm³) were taken at different time intervals to follow the progress of glycerol conversion. Solketal production was analyzed by gas chromatography using a PerkinElmer Clarus 500 GC system equipped with an Elite-WAX column (Perkin Elmer, Waltham, MA, USA, length: 30 m, I.D.: 0.32 mm, F.D.: 0.5 mm) and a flame ionization detector (FID). Chromatograms were obtained using the following conditions: an isothermal step at 50 °C for 2 min, followed by a temperature increase of 20 °C/min up to 180 °C, with a hold of 2 min. In this process, 1-pentanol (99.0%, Fluka, Buchs, Switzerland) was added to the reaction samples as an external standard.

3.4. Catalytic Hydrogenation of Levulinic Acid

γ-valerolactone synthesis was carried out in a high-pressure stirred autoclave reactor with 100 cm³ of capacity. The setup was equipped with a temperature and pressure control system, magnetic stirrer, and sample collection port. In a typical test, a 200 mg sample of catalyst was introduced in the reactor with a mixture of formic acid (purity 98%, Sigma Aldrich) and levulinic acid (purity 98%, Sigma Aldrich), using isopropyl alcohol (IPA, purity 99.97%, Sigma Aldrich) as solvent. The molar ratio of formic to levulinic acids varied between 1/1 and 7/1. The reactions took place at temperatures in the range of 160–200 °C. Reactants and products were analyzed with a 7000D Triple Quadrupole GC/MS (Agilent, Santa Clara, CA, USA), equipped with a HP5-UI column.

The conversion of glycerol and levulinic acid and the selectivity and yield of the products from the solketal and γ-valerolactone production reactions are defined as follows:

$$\mathrm{X}={\displaystyle \frac{{C_A}_0-C_A}{{C_A}_0}}$$

(7)

$$S_i={\displaystyle \frac{{\mathrm{C}}_{\mathrm{i}}}{\sum {\mathrm{C}}_{\mathrm{i}}}}$$

(8)

$$Y_i=\mathrm{X}·S_i$$

(9)

where C_A0 is the initial concentration of the limiting reactant, i.e., glycerol or levulinic acid, C_A is the time-dependent concentration of the limiting reactant, and C_i is the concentration of the product i.

3.5. Catalyst Regeneration

After the reaction, the reaction mixture was filtered to separate the spent catalyst from the liquid phase. Regeneration was attempted by washing in deionized water at room temperature for 24 h. Then, the washed catalyst was dried overnight and introduced into the reactor again for a new reaction cycle.

4. Conclusions

Activated carbons (ACs) with acid character were prepared by chemical activation of olive stones with phosphoric acid under nitrogen/air atmosphere; the olive stones were a lignocellulosic residue from agricultural industry. These ACs were and successfully tested as catalysts for the valorization of two biorefinery coproducts, glycerol and levulinic acid, by their conversion into solkeltal and γ-valerolactone, respectively.

The combination of the characterization of the surface chemistry and the results obtained from the glycerol acetalization indicated that chemical activation in air increased the acidity by promoting the formation of stable C-O-P groups, which were able to selectively catalyze the formation of solketal. Activation in air did not compromise the preparation yield or the porosity of the AC. These are relevant findings from the point of view of the sustainability of the catalyst preparation process, allowing the use of less demanding conditions.

The most acidic AC was modified to prepare a bifunctional Zr-loaded catalyst. The resulting catalyst successfully promoted the selective production of γ-valerolactone from levulinic acid using isopropyl alcohol (IPA) as solvent and H₂ donor. The catalytic experiments revealed that the acidic function of the carbon material promoted dehydration as a first step, producing isopropyl levulinate, while the metallic function was able to catalyze the hydrogenation of this intermediate to form γ-valerolactone. The presence of water acted as an inhibitor. This catalyst can be reused without observable decay of the catalytic activity.

These results underscore the viability of the use of carbon-based catalysts derived from lignocellulosic waste in the valorization reactions of the byproducts proposed in the work, and thereby validates the potential for integration of these processes in a lignocellulosic biorefinery.