Investigation for Enhancement of 5-hydroxymethylfurfural extraction into 2, 5-furandicarboxylic acid: an Ab initio study

Furandicarboxylic acid (FDCA) is recognized as a valuable product of hydroxymethylfurfural (HMF) derived from cellulosic materials that could nd future bioplastic application if a feasible separation process is developed. To nd a commercially available solvent that can selectivity separate FDCA and HMF as well as the downstream process was supported by Py-GC-MS experiments with density functional theory (DFT). Evaluation of the sigma potential and sigma surface analysis showed that benzene and ethyl acetate have better extraction and selectivity of HMF, whereas FDCA indicated ideal behavior in DMF and DMSO solvents, where the hydrophobicity is changed by improving the hydrogen-bonding interaction between them. The up-down selection of classes of solvents based on the experimental data found by GC-MS revealed that polar molecular solvents (ethanol-water) are more compatible with carboxylic acids and alcohol compounds, while n-hexane is a desirable solvent for phenolic compounds. It is discovered that levoglucosan retains a signicant fraction of water compared to other solvents that need to be considered for further economic and environmental analyses under the multifaceted framework of biomass-derived products. the thermodynamic cellulose derivatives obtained from biomass. Five samples, including HMF, DFF, HMFCA, FDCA, and FFCA, were studied. Flash points, boiling points, Henry’s constant, activity coecients, vapor-liquid equilibrium (VLE) and liquid-liquid equilibrium (LLE) were obtained for various solvents and different dielectric constants. The products were identied and predicted via thermodynamic properties that have not been reported before. This study aims to evaluate the eciency of different solvents for the desirable extraction of HMF and FDCA for future bioplastic industries. A multiscale COSMO-RS model on a desirable solvent system was optimized by DFT methodology. Evaluation of the sigma potential and sigma surface analysis showed that benzene and ethyl acetate have better extraction and selectivity to HMF. Moreover, FDCA indicated ideal behavior in DMF and DMSO solvents. GC-MS data revealed that the polar molecular solvent (ethanol-water) extracted more carboxylic acids and alcohols, while n-hexane the middle the solvents for phenolic compounds. It is worth noting that the higher aromatics in HMF Thus,


Introduction
Recently, bioplastics have become promising feedstocks for value-added products, which have the potential to replace fossil fuel-based petrochemicals. Petrochemicals pose environmental and climate-related challenges to our society.
Environmental concerns increase when chemical manufacturers expand their product portfolio and increase production due to the continuously increasing demands for polymeric materials. More than 8000 Mt of plastics have been consumed annually, roughly 80% of which have ended up in land lls and oceans (Geyer et al. 2017). This is because as consumers, we are becoming more addicted to single-use disposable plastics. The U.S. Department of Energy (DOE) predicted that plastic manufacturers will consume 20% of all petroleum materials with a share of 15% annual global carbon emissions by the year 2050 (Perlack et al. 2011). The global warning that arises from plastic production, consumption, and waste management requires renewable chemical strategies (Erickson and Winters 2012; Zheng and Suh 2019).
Biore nery is recognized as a valuable route of using biomass as a feedstock that can reduce our dependence on fossil fuels (Stuart and El-Halwagi 2012). Biomass is an abundant feedstock with great potential to produce biofuels, biochemical, and bioproducts (Ghorbannezhad et al. 2020;Bilal et al. 2021). Today, more than 100 billion tons of biomass are available around the world, most of which end up as waste due to ineffective technologies (Dutta et al. 2014; Searle and Malins 2015). Biomass can completely boost the bioeconomy by rebuilding fuel and chemical manufacturing processes. Lignocellulosic biomass is made up of carbohydrate polymers such as cellulose (40-50%) and hemicellulose (20-40%) and aromatic polymers such as lignin (20-30%). These chemical components are the main building blocks of many biochemical platforms ( Figure 1). 5-hydroxymethylfurfural (HMF) is recognized as a high-value building block compound of cellulose that can be transformed into a variety of value-added chemicals (Kuster 1990;Rosatella et al. 2011;Davidson et al. 2021). HMF is recognized as the "top 10" chemical platform in the circular bioeconomy by DOE. It is synthesized by the dehydration of mono and polysaccharides present in lignocellulosic biomass. Furandicarboxylic acid (FDCA) is one of the most valuable biochemicals that can be derived from HMF by biological or chemical oxidation (Arikan et al. 2021). During conversion, the oxidation of the aldehyde of HMF produces 5-hydroxymethyl-2-furan carboxylic acid, which is an intermediate compound. Further oxidation of 5-hydroxymethyl-2-furan carboxylic acid produces furan carboxylic acid (FFCA), which nally turns into FDCA. The schematic of cellulose conversion is shown in Figure 2.
Therefore, one of the most cost-effective FDCA production routes could be the direct conversion of biomass into desired products on a commercial scale (Hwang et al. 2020; Arikan et al. 2021). In this regard, condensation reactions under acidic aqueous conditions must be suppressed by substituting with an organic solvent during the rehydration process (Bonner et al. 1960). Solvents are vital to obtain a higher yield of HMF and its derivatives. Currently, direct dehydration of glucose to produce HMF and its derivatives is limited by an ine cient separation process (Cai et Zunita et al. 2021). FDCA has shown low solubility in water and the problem of deactivation during the direct conversion of glucose to HMF in water as a solvent (Zhang et al. 2018;Esteban et al. 2020). The weak separation e ciency of short-chain solvents led to high volumetric consumption of organic solvents and salts to make a biphasic system. However, selecting an e cient solvent for puri cation and separation of products is a new research area for the development of biore nery processes. Methyl isobutyl ketone (MIBK), tetrahydrofuran (THF), and dimethyl sulfoxide (DMSO) have been used as moderate polar and high dielectric solvents to promote the dehydration reaction for the production of HMF (Weingarten et al. 2014). The authors indicated that the solvents increased the dehydration reaction but decreased the polymerization of HMF. Some studies revealed that the biphasic system aims to improve the HMF yield by in situ extraction of HMF with an organic phase. A higher partition coe cient of HMF can lead to increased productivity and make the process cost-effective (Blumenthal et al. 2016). The computational method applies a heuristic approach that estimates the activity coe cient to identify appropriate solvents. The combination of quantum mechanics and statistical mechanics presents multifunctional models to better describe molecular thermodynamics. Unlike empirical models such as UNIFAC, ASOG, etc., the conductor-like screening model for real solvents (COSMO-RS) can explain intermolecular interactions (Klamt 2011;Balchandani and Singh 2021). An ensemble of pairwise intermolecular interactions in COSMO-RS provides a signi cant deviation in phase behavior, leading to crucial insights into different solvent classes (Eckert and Klamt 2018). Density function theory (DFT) can be combined with COSMO-RS to optimize models (Momany and Schnupf 2014; Wang et al. 2020). This work aims to investigate and predict the thermodynamic properties of cellulose derivatives obtained from biomass. Five samples, including HMF, DFF, HMFCA, FDCA, and FFCA, were studied. Flash points, boiling points, Henry's constant, activity coe cients, vapor-liquid equilibrium (VLE) and liquid-liquid equilibrium (LLE) were obtained for various solvents and different dielectric constants. The products were identi ed and predicted via thermodynamic properties that have not been reported before.

Materials And Methods
The experimental pyrolysis of biomass was performed in Rx-300 TR, a tandem micropyrolyzer system from Frontier Laboratories, Japan. The system is classi ed into two parts: a) pyrolysis reactor with a steady amount of biomass of 1-2 mg and b) catalyst bed reactor. Hence, you do not use the catalysts; you can set the reactor temperature as the same as the pyrolysis reactor temperature. The milled biomass samples were placed in a stainless steel cup to drop into the rst reactor. A neutral carrier gas such as helium or nitrogen (1 ml/min) transfers the vaporized biomass compound from the thermal degradation reactor to the catalyst bed reactor. A temperature of 500°C is tted by a programmable temperature control to maintain a constant process condition. The hot vapor (syngas) is retained in the rst reactor for less than 2 seconds and then goes into the catalyst bed reactor, which is nally detected by GC-MS interfaced with the pyrolysis system. The percentage area peak of each compound could be determined to estimate the yield of products individually.

Computational Methods and Theory
COSMO-RS is the conductor-like screening model for real solvation that uses quantum chemistry (QC) to predict the thermodynamic properties of molecules (Klamt 1995;Marsh 2006). The molecules are embedded in a virtual conductor based on the polarization charge densities of the solute and solvent molecules.
DFT can be combined with the COSMO as it is available in most quantum chemical programs. The structure of the biomass used in this study was drawn manually using the Gauss view software, which was further optimized with the Gaussian 09 program (Keith and Millam 2016). From the optimized structure, COSMO, sigma pro le potential, and frequency les were generated. With the level of theory, Becke's three-parameter functional for exchange was combined with the nonlocal correlation potential of the Lee-Yang-Parr functional and the DGTZVP basis set (Godbout et al. 1992; Sosa et al. 1992). One where A i (σ) is the segment surface area that has charge density σ A i is the area of the whole surface cavity rooted in the medium.
where μ s σ ' is the chemical potential of a surface segment σ is the polarity of the surface under study.
Table1. Properties, energy e cacy, and toxicity of studied solvent

Sigma Potential Analysis
The δ-potential describes the possible interaction of compounds with solvents according to polarity and hydrogen bonding. Molecular interactions between the solvents and HMF play a vital role in HMF solubility. The shape, size, and initial components of the molecule are essential for molecular interaction. Caloric pro les provide important information to predict molecular interactions in the uid phase. The hydrogen bond acceptor, hydrogen bond donor, and non-polar area associated with red, blue, and green, respectively, are shown in Figure 3.  Table 2 shows the chemical composition of cellulose in different solvents identi ed by GC-MS. The results showed that the DMF exhibits lower performance compared to ethyl acetate and hexane in water. More acid and esters (aliphatic compounds) were separated by ethanol-water solvents, whereas more aromatic compounds could be extracted by hexane and benzene as solvents. It is assumed that the low solubility of hydrocarbons in water is due to the reduction of the polarity differences. Typically, polar solvents enhance HMF and carboxylic acid extraction by 20%.
Cellulose is anticipated to reduce the solubility of phenolic compounds in water and affect the overall process separation. The free energy derivative validated the simulation of the interaction of FDCA separation in water and n-hexane that corresponds to the process. The interaction of FDCA separation in n-hexane showed a Gibbs free energy value of +0.8 kcal.mol −1 , whereas the same puri cation process in water is -0.9 kcal.mol −1 (Figure 4). However, n-hexane can be placed in the middle of the solvents, which theoretically could dissolve all compounds equally. Gibb's free energy is positive for n-hexane, illustrating low solubility limits for all concentration levels. The solubility of water in the organic phase of FDCA is lower compared to that of n-hexane. Thus, the Gibbs free energy presents a highly negative area in the water.
On the other hand, Gibbs energy of mixing is positive for an HMF-water system for the entire concentration range due to its very low mutual solubility limits. HMF-water is partially miscible, the solubility of water in the organic phase is higher as compared to the aqueous solubility. So the Excess Enthalpy of the mixing plot shows the negative region in the HMF region as shown in Figure 5. The HMF-water shows a maximum miscibility gap of a mole fraction of 0.9. Moreover, the Gibbs energy and excess enthalpy of mixing are negative for an FDCA-water system for the entire concentration range due to its mutual solubility limits. FDCA-water is partially miscible, the solubility of water in the organic phase is higher as compared to the aqueous solubility. Therefore, the Gibbs energy and excess enthalpy of the mixing plot shows the negative region in middle the FDCA-water region as shown in gure 4 (b).
The excess energies for HMF at solvent water / hexane at room temperature (298. It can be observed that the activity coe cient for the components HMF, FDCA gets very high for the case of in nite dilution at solvent hexane higher than solvent h20. This means that the activity coe cients and intermolecular interactions of Different solutes in solvents are very much dependent on the chemical structure of solute HMF and FDCA. This non-ideality is attributed to the degree of dissociation/reaction of the solute, to the solute-solvent interactions such as complex ion formation, and to the solute-solute interactions such as ion pairing. An activity coe cient incorporates the particle interactions into a single term that modi es the formal concentration to give an estimate of the effective concentration, or activity, of each ion. The partition coe cients (log P) of in nitely diluted solutes in a mixture of two immiscible solvents can be calculated with Partition Coe cients (LogP). We used for calculation of ethanol/water, benzene/water, ether/water, and hexane/Water partition coe cients. In the case of partly miscible liquids, like the ethanol-rich phase of ethanol and water, both components have nonzero mole fractions. The preset also gives a value for the molar volume quotient of the two solvents.

COSMO-RS Prediction Optimized by DFT
The sigma potential determines the pseudochemical potential of the molecular surface. The chemical potential of the surface segment was obtained by thermodynamics of molecular interactions. Using COSMO-RS theory, one can calculate the thermodynamic properties of uids from the 3D polarized charge distribution of individual molecules (σsurface) in terms of σ-pro le (2D histogram). Thus, valuable information regarding the polarity of a compound and its interaction with other surrounding molecules in the uid media can be deduced from these σ-pro le histograms. Figure   6 shows P X (σ); the charge density of these compounds in the polarized eld (σ). The following criteria can be applied implying an unfavorable solvation for HMF extraction. It is manifested that ethanol and methanol are miscible with water. It revealed that the HMF is a strong hydrogen bond donor because of the higher pro le area that it favors with water interaction. The hydroxylmethyl group of HMF is the hydrogen-bonding donor from aqueous media. On the contrary, the benzene and ethyl acetate solvents show a descending sigma potential sinking below -0.15 kcal mol −1 Å −2 for the surface charge density at 0.05 e Å −2 , where e cient HMF separation took place ( Figure 6).
The sigma potential pro les suggest DMF and DMSO as the alternative and most promising solvents for FDCA extraction. The hydrophobicity region at δ = 0 re ected the potential solubility difference of the solvents with water.
Benzene and ethyl acetate enable direct bonding to the OH group, resulting in an increase in the nonpolar alkyl group and thus of the hydrophobicity. On the other hand, DMF and DMSO replace the hydrophilic environment to improve the selectivity and overall yield of FCDA ( Figure 7). In fact, the low mutual solubility of DMF and DMSO makes them not eligible to interact on hydrogen bonding. Thus, DMF and DMSO exhibited a lower charge density in hydrogen bonding, which results in less interaction with water and better fractionation of the organic phase. Thus, nonpolar organic and hydrocarbon compounds with π-π bond are more favorable with DMF and DMSO solvents because of fewer interactions with the hydrogen bond. Nevertheless, it is interesting to note that there are clear differences in the interactions of solvents with HMF and FDCA.
Comparison of σ-pro le derived from HMF and FDCA shows that the charge density of the HMF and FDCA compounds can be found in the nonpolar region of σ-pro le, due to its heterocyclic ring (green in σ-surface). The functional groups (carboxylate, hydroxyl methyl) contain negative charge (red color in σ-surface) which is shifted towards the positive pole of the eld. In FDCA, these peaks were found at higher positive polarity, showing a stronger HB acceptor character of the carboxylate group rather than the hydroxyl methyl group in the homologous HMF. Thus, the different pro les from σ-surface are contributed independently of each atom related to its charge density, as well as the descriptor of each constituent in the solvents. Finally, with stronger HB donor and HB acceptor groups, COSMO-RS describes FDCA as more polar structure than those of the homolog HMF. Consequently, the results of this study aimed to develop a framework of models that could predict the quantum-chemical properties of solvents to screen the interactions between them. It is purely a prediction approach of COSMO-RS optimized by the DFT methodology. On the other hand, it provides an e cient nal product that is rich in HMF and FDCA and approves the emerging biobased molecules for commercializing green processes. Therefore, the e cient solvent design approach would be highly reliable with experimental data in the applicability domain for green and sustainable bioplastic manufacturing processes.

Conclusions
This study aims to evaluate the e ciency of different solvents for the desirable extraction of HMF and FDCA for future bioplastic industries. A multiscale COSMO-RS model on a desirable solvent system was optimized by DFT methodology. Evaluation of the sigma potential and sigma surface analysis showed that benzene and ethyl acetate have better extraction and selectivity to HMF. Moreover, FDCA indicated ideal behavior in DMF and DMSO solvents. GC-MS data revealed that the polar molecular solvent (ethanol-water) extracted more carboxylic acids and alcohols, while n-hexane lies in the middle of the solvents for phenolic compounds. It is worth noting that the higher aromatics in HMF promote the Diels-Alder reaction as well as the dehydration reaction. Thus, a promising route for the generation of FDCA from HMF could be through the use of strong concentrations of cellulose. In addition to that, DMF and DMSO improve the hydrogen bonding interaction between FDCA and solvent molecules, and thus lead to e cient separation.
However, the strategy employed in this study could be considered as a computational prediction to select a favorable solvent with e cient bioplastic manufacturing processes.
Declarations Figure 2 Overview of FDCA production pathways from cellulose  Gibbs tangent plane diagram for FDCA solvents (a: n-hexane; b: water) Figure 5 Gibb's tangent plane diagram for the HMF solvents (a: n-hexane; b: water)