Theoretical Modeling of Pathways of Transformation of Fructose and Xylose to Levulinic and Formic Acids over Single Na Site in BEA Zeolite

Abstract

The aim of our work is to theoretically model the conversion of C6 and C5 carbohydrates derived from lignocellulosic biomass waste into C1–C5 carboxylic acids such as levulinic, oxalic, lactic, and formic acids. Understanding the mechanism of these processes will provide the necessary knowledge to better plan the structure of zeolite. In this article, we focus on the theoretical modeling of two carbohydrates, representing C5 and C6, namely xylose and fructose, into levulinic acid (LE) and formic acid (FA). The modeling was carried out with the participation of Na-BEA zeolite in a hierarchical form, due to the large size of the carbohydrates. The density functional theory (DFT) method (StoBe program) was used, employing non-local generalized gradient-corrected functions according to Perdew, Burke, and Ernzerhof (RPBE) to account for electron exchange and correlation and using the nudged elastic band (NEB) method to determine the structure and energy of the transition state. The modeling was performed using cluster representations of hierarchical Na-Al₂Si₁₂O₃₉H₂₃ and ideal Al₂Si₂₂O₆₄H₃₄ beta zeolite. However, to accommodate the size of the carbohydrate molecules in reaction paths, only hierarchical Na-Al₂Si₁₂O₃₉H₂₃ was used. Sodium ions were positioned above the aluminum centers within the zeolite framework.

1. Introduction

The valorization of lignocellulosic biomass into platform chemicals is a key objective in the development of sustainable chemical processes. Among the biomass components, the hemicellulosic fraction—rich in C5 sugars such as xylose and fructose—plays a particularly vital role due to its abundance and industrial relevance. Efficient conversion of these carbohydrates into C1–C5 carboxylic acids, including levulinic acid (LA), formic acid (FA), lactic acid (LA), and oxalic acid (OA), offers a promising route to renewable chemical production [1,2,3,4,5,6,7,8].

Recent studies, including our own, have demonstrated the potential of zeolite-based catalysts, particularly Na-BEA, in facilitating the one-pot transformation of both C6 and C5 sugars into valuable acids [2,4]. However, the complexity of carbohydrate structures and the confined environment of zeolite pores pose significant challenges for catalyst design. In this context, theoretical modeling emerges as a powerful tool, enabling detailed insights into adsorption mechanisms, reaction pathways, and energy profiles. By simulating molecular interactions within zeolite frameworks, computational approaches accelerate the rational design of catalysts and optimization of reaction conditions, reducing the need for extensive experimental screening.

Levulinic acid (LE) has emerged as one of the most promising bio-based platform chemicals, recognized by the U.S. Department of Energy for its versatility and potential for replacing fossil-derived intermediates. It serves as a precursor to a wide array of high-value products, including biofuels such as γ-valerolactone (GVL), green solvents, plasticizers, and pharmaceutical intermediates. What makes levulinic acid particularly attractive is its feasible production from both C6 sugars (like glucose and fructose) and C5 sugars (such as xylose and arabinose), which are abundant in the hemicellulosic fraction of lignocellulosic biomass [1,2,5]. Through acid-catalyzed hydrolysis and dehydration processes, these sugars can be efficiently converted into LE using either homogeneous or heterogeneous catalytic systems [9,10]. Recent advances have demonstrated that even raw and waste biomass can serve as effective feedstock [11,12], aligning the production of LE with the principles of green chemistry and the circular economy [13]. Moreover, the valorization of hemicellulose-derived sugars into LE not only enhances the economic viability of biorefineries [14] but also contributes to the sustainable transformation of agricultural residues and forestry waste into valuable chemicals [15]. Levulinic acid can be efficiently produced from both fructose and xylose [16], making it accessible from a wide range of C6 and C5 sugars present in hemicellulosic biomass. In aqueous acidic media, fructose undergoes dehydration to form 5-hydroxymethylfurfural (HMF), which is subsequently rehydrated to yield levulinic acid and formic acid. This transformation is typically catalyzed by mineral acids such as H₂SO₄, HCl, or organic sulfonic acids, which act as homogeneous proton donors facilitating both dehydration and rearrangement steps [17,18]. The process is highly sensitive to reaction parameters such as temperature, acid concentration, and reaction time, with optimal levulinic acid yields generally achieved at 160–200 °C. Despite the simplicity and high activity of homogeneous acid systems, challenges remain in terms of catalyst recovery, corrosion, and waste generation, which have spurred interest in integrating homogeneous steps with downstream heterogeneous or separation technologies [18]. Nevertheless, homogeneous catalysis remains a benchmark for fructose valorization due to its efficiency and mechanistic clarity, especially in laboratory-scale and kinetic studies. Among the most prominent products is levulinic acid, which is formed via acid-catalyzed dehydration of fructose to 5-hydroxymethylfurfural (HMF), followed by rehydration and rearrangement. Solid acid catalysts such as sulfated zirconia, niobium phosphate, and heteropolyacids supported on silica and zeolites have demonstrated high activity and selectivity in this transformation, offering advantages in catalyst recovery and process sustainability [19]. The reaction typically proceeds under aqueous or biphasic conditions at elevated temperatures (140–200 °C), where the acidity and porosity of the catalyst play a crucial role in minimizing side reactions such as humin formation. Recent studies have also explored carbon-based catalysts, including sulfonated carbons and graphene-supported systems, which provide tunable surface acidity and hydrophobicity, enhancing the yield of levulinic acid while maintaining catalyst stability [19]. These heterogeneous systems not only align with green chemistry principles but also enable continuous processing, making them attractive for industrial-scale biomass valorization.

Xylose, a five-carbon aldopentose sugar commonly found in hemicellulose, serves as a versatile platform molecule for the synthesis of various carboxylic acids through different homogeneous catalytic pathways. Depending on the reaction conditions, xylose can undergo full oxidation to yield xylaric acid—a dicarboxylic acid formed by oxidizing both the aldehyde and hydroxyl groups—or partial oxidation to produce xylonic acid, where only the aldehyde group is converted to a carboxyl group [20,21,22]. Through selective oxidation, xylose can first isomerize to xylulose, a ketopentose, which can then be further oxidized to generate hydroxycarboxylic acid derivatives. In biological systems, xylose can be fermented by engineered or native microorganisms into a range of valuable acids, including lactic, succinic, acetic, propionic, and even adipic acid under industrial conditions [23]. Of particular interest is the acid-catalyzed conversion of xylose to levulinic acid, a key platform chemical in green chemistry. This transformation involves the dehydration of xylose (or its isomer xylulose) to furfural, followed by hydrolysis and rearrangement in the presence of strong mineral acids such as H₂SO₄ or HCl at elevated temperatures (150–200 °C), ultimately yielding levulinic acid and formic acid as a by-product [24]. These diverse pathways highlight the potential of xylose as a renewable feedstock for sustainable chemical production. Xylose can also be efficiently converted into a variety of carboxylic acids through heterogeneous catalytic processes, offering a promising route for biomass valorization. In contrast to homogeneous systems, heterogeneous catalysts—such as supported metal oxides, zeolites, and carbon-based materials—enable easier separation and reuse, making them attractive for industrial applications [18]. For instance, the hydrogenation of xylose over carbon-supported copper (Cu@C) catalysts [25] has demonstrated high selectivity toward xylitol, a valuable sugar alcohol, under relatively mild hydrogen pressures, benefiting from the stability and inertness of the carbon support. Similarly, mixed oxide-supported ruthenium catalysts (e.g., Ru/NiO-TiO₂) have been shown to facilitate the conversion of xylose to xylitol with high efficiency, following pseudo-first-order kinetics with low activation energy, and optimized via response surface methodology [26]. Beyond xylitol, the dehydration of xylose to furfural and subsequent transformation to levulinic acid can also be catalyzed by solid acid catalysts such as sulfonated carbons or zeolites, which provide the necessary acidic sites for dehydration and rearrangement reactions. These heterogeneous systems not only enhance catalyst longevity and process sustainability but also align with green chemistry principles by minimizing waste and enabling continuous processing. As such, the development of robust, selective, and recyclable heterogeneous catalysts remains a central focus in the advancement of xylose valorization technologies.

The catalytic valorization of xylose into carboxylic acids using heterogeneous catalysts has gained significant attention as a sustainable and efficient route for biomass conversion. According to Sobuś et al. [4], a one-pot transformation strategy enables the conversion of hemicellulose-derived sugars, including xylose, into valuable carboxylic acids such as formic, acetic, and levulinic acids. This process employs solid acid catalysts—such as zeolites—which provide both Brønsted and Lewis acid sites essential for dehydration, isomerization, and subsequent rearrangement reactions. The study highlights that the catalytic performance is strongly influenced by the acidity and porosity of the support, as well as the reaction conditions such as temperature and solvent system. Notably, the use of water or biphasic systems enhances the selectivity toward levulinic acid while minimizing humin formation. These findings underscore the potential of heterogeneous catalysis not only for improving process efficiency and catalyst recyclability but also for aligning with green chemistry principles in the production of bio-based chemicals from renewable feedstocks like xylose.

A thorough understanding of the reaction mechanisms of hemicellulose-derived carbohydrates—such as xylose and fructose—over zeolite catalysts is essential for optimizing process parameters toward the selective formation of desired products. By elucidating these pathways, researchers can more effectively tailor catalytic systems and reaction conditions to enhance yield, selectivity, and overall process efficiency in biomass valorization. Therefore, considering our experience in modeling reactions within zeolite pores, this study was extended to investigate the reaction mechanisms of fructose and xylose conversion into levulinic and formic acids using density functional theory (DFT) methods.

2. Results and Discussion

The heterogeneous process using Na-BEA zeolite revealed several plausible pathways for acid formation at active sites, consistent with experimental observations of mixed acid products. The proposed mechanisms highlight the role of Na⁺ centers in facilitating fructose and xylose adsorption and decomposition within the BEA zeolite framework. These findings provide valuable insights into the catalytic behavior of both systems and offer guidance for the rational design of zeolite-based catalysts for biomass valorization.

Figure 1 shows the several paths of fructose and xylose to carboxylic acid, which involves the steps of protonation/deprotonation, dehydration, and oxidation reactions.

By analyzing the experimental transformation of C5 and C6 sugars using a Na-BEA zeolite catalyst and the resulting product mixture, we proposed several reaction pathways. In the case of xylose, the formation of levulinic acid proceeds via a double dehydration process, where water molecules are removed. The conversion of fructose also yields formic acid through a dehydration mechanism. For xylose, two oxidation-based pathways are also possible: in the first, lactic acid and two molecules of formic acid are formed; in the second, oxidation is coupled with deprotonation, resulting in lactic acid and oxalic acid. In the case of fructose, oxidation may be coupled with protonation and dehydration, leading to the formation of lactic acid, oxalic acid, and formic acid. Additionally, deprotonation of fructose leads to the formation of lactic acid and pyruvic acid. At this stage, we have successfully reconstructed the levulinic acid formation pathway for both sugars, xylose and fructose, by dehydration reactions. However, the remaining pathways, particularly those involving oxidation and multiple product formation, are expected to proceed through more complex mechanisms that likely involve multiple active centers within the zeolite framework. Accurate modeling of these transformations will require an expanded cluster model to capture the cooperative effects and spatial constraints of the catalytic environment. These aspects are currently under investigation as part of ongoing work.

Therefore, we now turn to a detailed discussion of the transformation pathways of fructose and xylose into individual products—levulinic acid in the case of xylose and both levulinic and formic acids in the case of fructose. The next two figures (Figure 2 and Figure 3) show the mechanism and energy diagram of fructose and xylose adsorption and decomposition into levulinic and formic acids at Na-BEA.

Fructose adsorbs onto the sodium cation located at the pore of the BEA zeolite via the oxygen atom of the ketone group at the C2 position and the hydroxyl group at C1 (Figure 2b), forming a bidentate coordination with sodium. This adsorption is energetically favorable, with an adsorption energy of −0.34 eV. In the next step, a water molecule is formed through the rearrangement of hydroxyl groups at the C5 and C6 carbon atoms (Figure 2c), although this step requires a significant energy input (0.63 eV). Subsequently, formic acid is generated from the C1 carbon through the detachment via the OH group from sodium and a rearrangement involving the hydroxyl group at C3 (Figure 2d), which is energetically favorable (−0.13 eV). The simultaneous desorption of water and formic acid (Figure 2e) is also energetically favorable (−0.36 eV). Levulinic acid remains coordinated to the sodium site, and its desorption is associated with a moderate energy cost of +0.26 eV.

Xylose adsorbs onto the Na-BEA zeolite surface via the carbonyl oxygen of its aldehyde group at the C1 position, with an adsorption energy of −0.41 eV (Figure 3b). The subsequent formation of two water molecules involves an energy barrier of 1.26 eV (Figure 3c). This is followed by a dehydration step, which is energetically favorable, exhibiting a reaction energy of −0.05 eV (Figure 3d). Finally, the desorption of levulinic acid from the zeolite surface requires an energy input of 0.26 eV (Figure 3e).

The adsorption energies of both fructose and xylose are quite similar, as illustrated in Figure 2b and Figure 3b, with values around −0.40 eV. However, the subsequent transformation of fructose proceeds with an energy barrier that is 0.60 eV lower than that of xylose, suggesting that this process occurs more readily in the case of fructose.

Therefore, under elevated temperature conditions and in a mixture containing both carbohydrates, the two processes are expected to proceed at comparable rates. It is important to note, however, that a key distinction lies in the mode of adsorption of the two carbohydrates: fructose adsorbs in a bidentate manner—through two oxygen atoms—whereas xylose binds monodentately, forming a single bond via one oxygen atom.

Figure 4 presents the bond orders and lengths, along with Mulliken charge distributions, for Na-BEA zeolite systems involved in the valorization of fructose and xylose. The panels include (a1) isolated fructose, (b1) fructose adsorbed on Na-BEA, (c1) fructose adsorbed in closer proximity to Na-BEA, (a2) isolated xylose, and (b2) xylose adsorbed on Na-BEA.

Analyzing the bond order data for fructose adsorption, we observe the initial weakening of the bond between carbon atoms C1 and C2. This weakening suggests the onset of bond cleavage, which ultimately supports the formation of formic acid in the subsequent steps of fructose conversion to levulinic acid (as shown in Figure 2d). In contrast, no such changes in C–C bond strength are observed for xylose, further corroborating the proposed reaction mechanism illustrated in Figure 3. Moreover, the strength of interaction between fructose or xylose and the sodium site is comparable, with an average bond strength of approximately 0.14 and distances around 2.5 Å. This indicates that the nature of their adsorption is similarly stable, and differences in reactivity are more likely attributed to structural and electronic factors rather than binding affinity alone.

To provide a broader perspective on the presented analyses, several key insights emerge from this study. The computational modeling of fructose and xylose adsorption and transformation on Na-BEA zeolite catalysts has revealed a set of energetically feasible reaction pathways leading to the formation of carboxylic acids such as levulinic acid (LA), formic acid (FA), pyruvic acid (PI), and oxalic acid (OA). These pathways are in line with experimental findings and highlight the catalytic relevance of sodium centers in facilitating the conversion of biomass-derived carbohydrates.

Importantly, the mechanistic focus of this work was intentionally placed on the simplest reaction routes, which can proceed via a single sodium active site within the zeolite framework. This strategic choice allowed for a clear and controlled analysis of the fundamental steps involved in carbohydrate valorization. However, we recognize that more complex pathways, particularly those involving multiple intermediates or oxidation steps, are likely to require cooperative interactions between several active centers and a more sophisticated cluster model to capture the full catalytic environment. These aspects are currently being explored in ongoing research. While the current study centers on sodium-based Lewis acid sites, the potential role of Brønsted acid sites (Al–OH) and alternative adsorption geometries remains an open and promising area for future investigation. These factors may significantly influence both the adsorption behavior and the reaction energetics, and their inclusion in future models will provide a more comprehensive understanding of zeolite-catalyzed carbohydrate transformations.

The results obtained for the adsorption and valorization of fructose and xylose into levulinic and formic acids on sodium centers of a zeolite catalyst are both intriguing and significant. Modeling the adsorption of carbohydrates—relatively large molecules—within the confined pores of zeolites requires a deep understanding of zeolite topology. This challenge is further amplified when considering hierarchical zeolites, where the use of periodic models becomes impractical due to the necessity of extremely large unit cells, often comprising several thousand atoms.

In this context, the application of a cluster model proves to be not only appropriate but also essential. However, this approach demands precise knowledge of the zeolite’s structural framework and careful construction of representative clusters—an aspect we have successfully addressed in our study. Therefore, the modeling of fructose and xylose adsorption in the Na-BEA zeolite presented here marks an important step forward, opening new avenues for virtually exploring the valorization potential of lignocellulosic biomass.

In the long term, such modeling efforts may enable the prediction of optimal process conditions and the design of active site architectures in zeolites tailored for the selective transformation of hemicellulosic carbohydrates. Nevertheless, the current study, focused on sodium centers and adsorption via oxygen atoms at C1 and C2 positions, does not exhaust all possible scenarios. For instance, the role of Brønsted acid sites (Al–OH) or alternative adsorption configurations remains to be explored. These aspects are the subject of ongoing investigations within our research group.

Overall, this work establishes a computational foundation for the rational design of zeolite-based catalysts and the predictive modeling of biomass valorization processes. The insights gained here offer valuable guidance for optimizing reaction conditions and tailoring catalyst architecture. Moreover, this study demonstrates the indispensable role of theoretical modeling in elucidating adsorption geometries, reaction energetics, and mechanistic pathways at the atomic level—ultimately contributing to the development of efficient catalytic systems for sustainable chemical production.

3. Materials and Methods

3.1. Computational Details

The density functional theory (DFT) method was used to calculate the electron structure of the presented clusters using the StoBe program (version 2014, Department of Inorganic Chemistry, Fritz-Haber-Institut der MPG, Berlin, Germany) [27]. The non-local generalized gradient corrected functionals according to Perdew, Burke, and Ernzerhof (RPBE) [28,29] was used to account for the electron exchange and correlation. Kohn–Sham orbitals were represented by linear combinations of atomic orbitals (LCAOs) using contracted Gaussian basis sets for atoms [30]. Mulliken populations [31] and Mayer bond order factors [32,33] were used to precisely analyze the electron structure of the clusters.

Double valence zeta polarization (DZVP) functional bases were used for Si and Al (6321/521/1), Na (6321/411/1), O and C (621/41/1), and H (41) orbital basis sets. Auxiliary functional bases were also used to adjust the electron density and the exchange potential of the correlation of individual atoms: Si and Al (5,4;5,4), Na (5,4;5,4), O and C (4,3;4,3), and H (4,0;4,0) [27].

The energy difference between consecutive steps in the reaction mechanism was calculated according to the following expression:

$$E_{diff}=E_{step(n+1)}-E_{step\left(n\right)}\pm E_{substrates/products}\left[eV\right]$$

where E_step(n) and E_step(n+1) represent the total electronic energies of the system at two consecutive stages of the reaction pathway (e.g., reactant, intermediate, transition state, or product); E_{substrates/products} accounts for the energy of any species that are formed or removed between the two steps (e.g., water, formic acid). All energies are reported in electron volts (eV) and correspond to single-point total energies obtained from optimized geometries.

All calculations were performed with a total charge of 0. Several spin multiplicities were tested, and in all cases, the lowest-energy structures corresponded to the singlet state, which was therefore used throughout this study. To identify the structure and energy of the transition state, we employed the nudged elastic band (NEB) method as implemented in the StoBe software package. This approach enables reaction path optimization through both the standard NEB and the climbing image NEB (CI-NEB) techniques [34,35,36,37], which are designed to locate saddle points along the minimum energy path (MEP).

The optimization of individual images along the reaction coordinate was carried out using the quasi-Newton BFGS algorithm, while the overall path refinement was supported by a trajectory extrapolation scheme. Elastic coupling between adjacent images was controlled using spring constants of 0.5 for the main coupling and 0.1 for the variation applied to the lowest-energy images, ensuring a balanced and stable optimization of the reaction pathway.

3.2. Geometrical Models

The structure of the BEA zeolite was taken from the Database of Zeolite Structure [38]. The tetragonal phase of the BEA framework type is described by the space group P 41 2 2 (No. 91) with lattice constants a = b = 12.6320 Å, and c = 26.1860 Å. The crystal unit cell contains 192 atoms. Figure 5 presents the structure of the BEA zeolite viewed along three crystallographic directions.

The BEA framework (purely siliceous silicalite-1 and aluminum-containing BEA) has a three dimensional channel system consisting of straight channels along the a and b axes and tortuous channel along the c axis [39]. The pore-limiting diameter of the largest pore is 6.9 Å (Figure 5) [40]. The maximum sphere diameter that can diffuse along BEA pores is 5.95 Å [38]. Considering the size of molecules such as fructose relative to the pore dimensions of zeolite, it is essential to incorporate hierarchical structuring during model development to avoid steric hindrance.

To create the cluster required for the calculations, a fragment of BEA crystal structure was cut out. Figure 6 shows the structures that were used to perform the fructose and xylose adsorption processes. The structures were formed by adsorption of dimers near the active site, which is represented by two aluminum atoms. Cluster models of beta zeolite (ideal pore Al₂Si₂₂O₆₄H₃₄, Figure 6a, and simulating hierarchical zeolite Na-Al₂Si₁₂O₃₉H₂₃, Figure 6b) have been used with Na⁺ ion above aluminum centers in the zeolite frame.

The broken bonds were saturated with a point charge, which was represented by a hydrogen atom. The hydrogen was placed at a distance of 0.97 Å in a direction consistent with that presented by the broken Si-O bond. The fragment contained important active sites for the catalyst including two aluminum atoms. This structure was successfully used in earlier studies [41]. During geometry optimization, only the atoms in the central part of the cluster and the adsorbates were relaxed, while the peripheral atoms, including terminal hydrogens, were kept fixed to maintain the structural integrity of the zeolite framework.

To simulate the mechanism of xylose and fructose valorization into carboxylic acids, hierarchical cluster structures were used (Figure 6b).

4. Conclusions

To summarize the presented analyses, several essential findings can be confirmed. The computational investigation into the adsorption and valorization of fructose and xylose on Na-BEA zeolite catalysts has revealed a set of energetically plausible reaction pathways leading to the formation of carboxylic acids. These pathways are consistent with experimental observations and underscore the catalytic relevance of sodium centers in facilitating carbohydrate transformation. However, they required more complex model of zeolite. Therefore, as a starting point, we deliberately focused on the simplest mechanistic routes, which can proceed via a single Na active site within the zeolite framework. The more complex pathways, including those involving other intermediates, are expected to require multiple active centers and a more elaborate modeling approach. These aspects are currently under investigation and will be addressed in future work.

A key insight is the comparable adsorption strength of fructose and xylose onto Na⁺ sites, with average bond energies around −0.40 eV and bond strengths of approximately 0.14. Despite this similarity, fructose undergoes decomposition with a slightly lower energy barrier (0.63 eV) compared to xylose (1.26 eV), making its overall transformation to levulinic acid more energetically favorable. This subtle difference suggests that under elevated temperatures and in mixed carbohydrate systems, both sugars may react competitively, though fructose may dominate certain pathways.

This study also highlights the importance of adsorption geometry: fructose binds in a bidentate fashion via two oxygen atoms, while xylose forms a monodentate interaction. This structural distinction influences the subsequent bond rearrangements and product formation. Notably, bond weakening between C1 and C2 in fructose supports the formation of formic acid, a step not observed in xylose, further validating the proposed mechanisms.

While the current focus is on sodium centers, the role of Brønsted acid sites (Al–OH) and alternative adsorption configurations remains unexplored and represents a promising direction for future research. Overall, this work lays a strong foundation for the virtual design of zeolite-based catalysts and the predictive modeling of biomass valorization processes, offering valuable guidance for optimizing reaction conditions and zeolite catalyst architecture. Theoretical modeling has proven to be an indispensable tool for elucidating adsorption geometries, reaction energetics, and mechanistic pathways at the atomic level, thereby enabling the predictive design of zeolite catalysts and the acceleration of process optimization in biomass valorization systems for future industrial solutions.