NC State
BioResources
Kong, Y., and Fu, S. (2022). "Computer simulation in lignocellulosic biomass conversion processes: A review," BioResources 17(4), Page numbers to be added.

Abstract

Much attention has been paid to the solubilization and conversion processes of cellulose, hemicellulose, and lignin to produce high value-added chemicals and fuels. Computer simulation shows great potential in understanding the conversion process of plant biomass. This paper reviewed the solubilization processes, catalytic conversion, and pyrolysis of cellulose, hemicellulose, and lignin by density functional theory and molecular dynamics. The authors also provide prospects for the challenges and future developments in this area.


Download PDF

Full Article

Computer Simulation in Lignocellulosic Biomass Conversion Processes: A Review

Yi Kong,a and Shiyu Fu a,b,*

Much attention has been paid to the solubilization and conversion processes of cellulose, hemicellulose, and lignin to produce high value-added chemicals and fuels. Computer simulation shows great potential in understanding the conversion process of plant biomass. This paper reviewed the solubilization processes, catalytic conversion, and pyrolysis of cellulose, hemicellulose, and lignin by density functional theory and molecular dynamics. The authors also provide prospects for the challenges and future developments in this area.

DOI: 10.15376/biores.17.4.Kong

Keywords: Lignocellulosic biomass; Density functional theory; Molecular dynamics; Biomass conversion

Contact information: a: State Key Laboratory of Pulp and Paper Engineering, South China University of Technology, Guangzhou 510640, PR China; b: South China University of Technology-Zhuhai Institute of Modern Industrial Innovation, Zhuhai 519175, PR China;*Corresponding author: shyfu@scut.edu.cn

INTRODUCTION

With the continuous consumption of fossil fuels and consequent environmental problems, there is an urgent need for a green, renewable energy source. Biomass, also known as lignocellulosic biomass, is a widely renewable bioenergy source considered a potential alternative to conventional energy sources to alleviate human energy shortages and environmental problems (Xu et al. 2019). Lignocellulosic biomass is derived from numerous species, such as trees (pine, poplar, and eucalyptus), crops (corn, wheat, and rice), grasses, and its main components are usually cellulose, hemicellulose, and lignin (Zakzeski et al. 2010; Zhou et al. 2014, 2015). Figure 1 depicts the composition of lignocellulosic biomass from different sources (Xu et al. 2020a).

Agriculture-based biomass materials could be applied in wide areas. It is a hot topic that biomass is converted into liquid biofuels and chemicals. The conversion processes of biomass, including catalytic depolymerization, oxidation, hydrogenation, pyrolysis, and dissolution, involve complex chemical reactions and to some extent, lack of insight into the conversion mechanism. The limitations of traditional experimental methods in studying plant biomass conversion have made computer simulations a promising tool. Computer simulations can identify the best complex transformation pathways of plant biomass and understand the transformation mechanisms at the molecular and atomic levels (Buehler and Ackbarow 2007). In addition, computer simulations allow experiments to be inexpensively guided by models. For instance, computer simulations can help gain insight into the dissolution mechanisms of existing solvents, providing a basis for designing solvents with high selectivity and high dissolution efficiency. Similarly, computer simulations are applied to selecting catalysts in the catalytic conversion of biomass and the relationship between temperature and product in the pyrolysis process.

Fig. 1. Composition of biomass raw materials (Xu et al. 2020a; Reprinted with permission from Elsevier)

There is a conflict between the model’s accuracy and the computational time in the simulation calculation. A high-precision model matching the actual situation requires more computation power or computation time. At the same time, these methods result in errors by simplifying the model or using more approximations. Two common computational methods, Density Functional Theory (DFT) and Molecular Dynamics (MD), have gained widespread attention in converting components of plant biomass (Sangha et al. 2012).

To improve the understanding of biomass conversion, this paper focuses on the conversion pathways and mechanisms of solubilization, catalytic conversion, and pyrolysis processes among cellulose, hemicellulose, and lignin using MD and DFT methods, including the structure of lignocellulosic materials, computational methods, and biomass conversion processes, which may provide a theoretical basis for the study of plant biomass. Finally, the challenges and future developments in biomass conversion are presented. Computational models and chemicals involved in this paper are organized in Table 1 in the order of appearance.

LIGNOCELLULOSIC MATERIAL COMPONENTS

Cellulose

Cellulose is the main component of lignocellulosic biomass and the most distributed and abundant homopolysaccharide in nature (Mood et al. 2013). Its molecule consists of glucose monomers linked by β(1-4) bonds, and the adjacent glucose monomers are alternately rotated by 180°. The cellulose chains are arranged to form elementary fibrils with a diameter of about 3 to 4 nm, while the crystalline and non-crystalline regions of cellulose are present in the elementary fibrils. Elementary fibrils are further organized into microfibrils with a diameter of 10 to 20 nm. At the same time, microfibrils, together with the surrounding matrix, such as hemicellulose and lignin, are combined to form macrofibril of the cell wall. (Gibson 2012).

Table 1. Names and Abbreviations of Computational Models and Chemicals

The following types of cellulose are currently known: cellulose I, II, III, IV, and their sub-types. Naturally occurring cellulose is defined as cellulose I, which together with cellulose II are the most common and widely used cellulose models in computational simulations (O’Sullivan 1997). Figure 2 shows the 36-chain model of crystalline cellulose Iβ microfibril. The dense crystalline structure of the cellulose crystalline region provides the plant high mechanical strength, but severely hinders the access of chemical reagents and enzymes to the cellulose, leading to difficulties in the degradation of natural cellulose. Therefore, it is necessary to understand its structure-property and other important information to use cellulose materials better.

Fig. 2. A 36-chain 3D model of crystalline cellulose Iβ microfibril: a) View along the c direction; b) view along the b direction of the cellulose microfibril segment (10 glucopyranose units along the c direction). Tan: carbon atoms; red: oxygen atoms; white: hydrogen atoms.

Hemicellulose

Hemicellulose, another polysaccharide abundant in nature, is a homopolymer or heteropolymer composed of main and branched chains (Rivas et al. 2016). The type of monosaccharides, the type and location of chemical bonds, and the distribution of side chains of hemicellulose vary considerably with the source of biomass. The composition of hemicellulose and the monomer species after hydrolysis were described in detail by Girio and colleagues (Girio et al. 2010). Different monosaccharides are present in the short side chain branches of hemicellulose, including pentoses (xylose, rhamnose, and arabinose), hexoses (glucose, mannose, and galactose), and glyoxylates (glucuronides and galacturonic acids) (Terrett and Dupree 2019).

Typical representatives of hemicelluloses are glucomannan and xylan, the most abundant components in the secondary cell walls of herbs and hardwoods. Xylan is mainly chosen as a hemicellulose model in the simulation of biomass conversion processes while decorated with different side chains to meet various experimental objectives. The typical structure of hemicellulose substituted with acetate, ferulate, and glucuronide is shown in Fig. 3.

Fig. 3. Typical structure of hemicellulose substituted by acetyl (pink), arabian furan (green), ferulate (blue), and 4-O-methyl gluconic acid (red)

Lignin

Lignin is an amorphous and complex natural polymer. It plays a supporting role in plants and enhances mechanical strength and stiffness by binding to cellulose fibers. Lignin is composed of three basic structural units: the para-hydroxyphenyl lignin (H unit), the guaiacyl lignin (G unit), and the syringyl lignin (S unit). The content of each unit varies considerably among the different sources of lignin. For example, the major component of cork lignin is the G-unit with small amounts of S and H-units, while the lignin of deciduous trees contains approximately equal parts of S and G-units and less of H-units. The ratio of G: S: H in monocot grasses is about 70:25:5 (Feofilova and Mysyakina 2016). The major chemical bonds between these units are β-O-4, β-5, β-β, 5-5, 4-O-5, and β-1 (Garcia Calvo-Flores and Dobado 2010; Ralph et al. 2019). Lignin contains many functional groups, including methoxy, phenolic hydroxyl, aliphatic hydroxyl, carboxyl, and carbonyl groups. The presence of functional groups contributes to the extremely complex lignin structure. It gives the cell wall of plants the ability to resist attack by external environments (e.g., enzymes, chemicals, and microorganisms) (Lancefield and Westwood 2015). The model structures of lignin are shown in Fig. 4.

Fig. 4. Model structures of hardwood and softwood lignin (Lancefield and Westwood 2015; Reprinted under Creative Commons Attribution 3.0 from RSC)

COMPUTATIONAL SIMULATION METHODS

With the rapid development of computers, various theoretical computational methods emerged, involving computational models that can be built at multilength and time scales. The length scales can be scaled up to the atomic level and the macroscopic scale. Simulations are possible on time scales ranging from picoseconds (ps) to seconds (s) (Zhang et al. 2019c; Zhou et al. 2021). Among the many theoretical computational methods, DFT and MD methods are widely used by researchers because of their excellent performance in terms of conformity to experimental data (Petridis et al. 2011; Sangha et al. 2012).

DFT

The method DFT studies the electronic structure of multi-electron systems by electron density and is widely used in condensed matter physics, materials science, quantum chemistry, and life sciences (Grimme 2011). The basic concept of DFT is to reduce intractable problems (electron-electron interactions) to interaction-free issues through various approximations and to integrate all errors in one term for analysis (Thomas 1927). The method began with the Thomas-Fermi model (Thomas 1927), which introduced the concept of density generalization and simplified the computational form and process, but did not take into account the interaction between electrons and did not provide an accurate description of the kinetic energy term, and therefore was not applicable in many systems. Until the 1960s, the Hohenberg-Kohn theorem and the Kohn-Sham equation were proposed to improve the content of DFT and finally established a strictly meaningful density functional computational theory. (Hohenberg and Kohn 1964; Kohn and Sham 1965; Kim et al. 2013). The most widely used functionals in DFT are M06-2x, B3LYP, X3LYP, and B3PW91, among which the most widely used and popular functional in plant biomass conversion simulation is B3LYP (Qi et al. 2016). These functionals (PBE, PW91, BLYP, and B3LYP) are deficient in describing dispersion interactions and examining electrostatically dominated weak interactions such as hydrogen bonding (Grimme 2011). To address the problem, Grimme proposed the dispersion-corrected DFT functionals (DFT-D), which had four versions, namely DFT-D1 (Grimme 2004), D2 (Grimme 2006), D3 (Grimme et al. 2010), and D4 (Caldeweyher et al. 2017). Subsequently, Grimme demonstrated through an extensive test set that the DFT-D3 functional had an excellent performance in describing non-covalent interactions of non-polar moieties or neutral systems (Goerigk and Grimme 2011). The DFT-D3 version has high accuracy and a small increase in time consumption. Moreover, it has a high degree of fit in the adaptation of mainstream functionals and mainstream quantitative programs.

MD

Although DFT could accurately describe the relevant properties of electrons, it cannot deal with larger systems due to a large amount of computation required. The method MD was undoubtedly an excellent choice when dealing with complex biological systems. Through analyzing atoms and molecules as a whole particle to analyze their motions, while the details inside the particle are simplified or disregarded, thus greatly reducing the computational requirements of each particle, MD simulation is the most suitable choice to balance basic accuracy and computational efficiency (Dong et al. 2017; Zhou et al. 2021).

The MD can be used to calculate macroscopic properties of substances or systems, such as density, solubility, melting and boiling points, enthalpy of evaporation, dielectric constant, and viscosity (Jung et al. 2018, 2019; Petridis et al. 2011). It is also used to view the microscopic properties, structure, and characteristics of a system, such as the spatial distribution of atoms or molecules, the trajectory of particles, the ratio of conformational distribution, diffusion dissolution, and adsorption processes (Yui and Hayashi 2007). The freedom of the studied system is very high. The simulation environment (temperature, pressure, external electric field, and external forces) could be set for the purpose, and all information inside the system is unambiguous. The range of systems studied by MD simulations is enormous, with common systems, such as proteins, nucleic acids, polysaccharides, polymers, metals, inorganic materials, and various small molecules, among others. The molecular force field, one of the important components of MD theory, is applied to describe the interaction potential between atoms. Common force fields include AMBER (Damm et al. 1997) (Assisted Model Building with Energy Refinement), GAFF (General AMBER Force Field), GLYCAM (Kirschner et al. 2008) (Glycans and Glycoconjugates in AMBER), OPLS (Damm et al. 1997) (Optimized Potentials for Liquid Simulations), GROMOS (Oostenbrink et al. 2004) (Groningen Molecular Simulation), CHARMM (Brooks et al. 1983) (Chemistry at Harvard Macromolecular Mechanics), CGenFF, COMPASS, GROMOS series (Hermans et al. 1984), PCFF, etc. In addition, chemical reaction processes could be simulated with ReaxFF (van Duin et al. 2001) (reaction force fields), called ReaxFF Reaction Molecular Dynamics (ReaxFF-RMD) (Xu et al. 2021a).

DISSOLUTION MECHANISM

Efficient, non-polluting dissolution systems are essential for the widespread use of plant biomass feedstocks. The screening of solvents by experimental methods is quite complicated, time-consuming, and costly. To better guide experiments and design solvents, researchers relied on computer simulations and modeling to gain insight into the dissolution mechanism. In this section, the dissolution processes and mechanisms of cellulose, hemicellulose, and lignin substrates are investigated by the MD method mainly and the DFT method as a supplement.

Cellulose Dissolution Mechanism

The interaction between cellulose and solvents is a hot topic of research on cellulose. Water can swell cellulose materials but not dissolve in water. Mazeau and Rivet used MD to simulate the wetting of cellulose (110) and (100) planes by water. The simulation results showed that the (100) plane dissolved more slowly than the (110) plane. The cellulose crystal structure of both planes remained unchanged (Mazeau and Rivet 2008). Bergenstråhle and co-workers simulated the dissolution behavior of cellulose crystals in aqueous solutions. The results showed that both the accumulation of hydrophobic groups and the hydrogen bonding in the cellulose chains inhibited the dissolution of cellulose crystals (Bergenstråhle et al. 2010). Similarly, Malaspina and Faraudo investigated the interaction of different molecules with the (100) and (010) planes of cellulose. The (100) and (010) planes are both hydrophilic and lipophilic, as indicated by the contact angles of water and tetradecane on the Iβ cellulose plane. The wetting of cellulose planes under water and organic solvent conditions is strongly heterogeneous on the molecular scale. Both hydrophilic and hydrophobic ions interacted weakly with the nanocellulose surface, although the interaction of the hydrophobic ions with the cellulose surface seems to be stronger than that of the hydrophilic ions. This result could be attributed to the fact that the cellulose surface had both hydrophilic and lipophilic behavior. (Malaspina and Faraudo 2019). Gross and co-workers simulated the dissolution of cellulose in four solvents, LiCl/ N,N-dimethylacetamide (DMA), water, LiCl/water, and DMA. It was found that LiCl/DMA had the best solubility for dextran chains. Due to the low solubility of dextran and LiCl in DMA, Li+ and Cl could effectively interact with glucose, resulting in the dissolution of cellulose. The small size of the cation preferentially interacts strongly with cellulose, which is considered the main driving force responsible for cellulose dissolution (Gross et al. 2013).

In addition, solvents, such as aqueous sodium hydroxide (Miyamoto et al. 2015), sodium hydroxide/urea system (Liu et al. 2018), N-Methylmorpholine-N-Oxide (NMMO) (Rabideau and Ismail 2015), and ionic liquids (ILs) (Li et al. 2015b; Yuan and Cheng 2015; Li et al. 2017b), are used to dissolve cellulose. Cellulose solubilization in aqueous sodium hydroxide solution is almost unrelated to the direct interaction of OH with cellulose. Water molecules and Na+ can penetrate cellulose crystals from the interlayer space in the (1-10) plane, so that hydrophobic sheets are dissociated into microcrystals, which then become independent chains (Miyamoto et al. 2015). Wang and co-workers investigated the dissolution process of cellulose in aqueous solutions of five quaternary ammonium hydroxides and found that the hydrophobicity of cations was the main driving force for cellulose dissolution. The hydrophobic cations aggregated at the cellulose interface, reducing the surface tension and facilitating cellulose dispersion (Wang et al. 2018).

Fig. 5. Crystalline and dissociated states of the 15 wt% cellulose solution. Reprinted with permission from (Rabideau and Ismail 2015). Copyright 2015 American Chemical Society.

Using MD, cellulose dissolution was investigated in aqueous NaOH/urea solutions, and showed that the cellulose inclusion complex was anisotropic. In this system, Na+ and OH are located in the hydroxyl and hydroxymethyl regions of the cellulose molecule, and the urea molecule occupies the surface of the hydrophobic pyranose ring (Liu et al. 2018). The role of urea dissolving cellulose is that it preferentially absorbs the hydrophobic surface of the glucose ring and gradually weakens the hydrophobic effect of the cellulose chain (Wernersson et al. 2015). The NMMO was also a good solvent for solubilizing cellulose. Rabideau and Ismail constructed the model shown in Fig. 5 and found that cellulose dissolved spontaneously when the ratio of NMMO to water was close to 1:1. The formation of NMMO-cellulose and water-cellulose hydrogen bonds promote the dissolution of cellulose (Rabideau and Ismail 2015).

Recently, the dissolution of lignocellulosic materials (especially cellulose) in IL is excellent, but the dissolution mechanism of IL is still not perfect. In 2007, Liu and co-workers performed MD calculations for cellobiose/[Bmim]Cl at different temperatures. The anion was closer to the cellobiose than the cation, and the moderate to strong hydrogen bonds formed the anion with the hydroxyl groups of the cellobiose in a ratio close to 1:1. The imidazole cation is involved in the π-π stacking interaction (Liu et al. 2007). Yao and co-workers also applied cellobiose as a model and investigated the solubilization mechanism of cellulose in [Bmim]Cl, [Emim]Cl, and [Emim]OAc by the DFT method. They indicated that anions played an essential role in the reaction process while cations had little influence (Yao et al. 2015).

A MD simulation of cellulose distribution in [Bmim]Cl using 36 cellulose chains with a degree of polymerization of 16 showed that the anions tended to form hydrogen bonds with glucose, while the cations were closer to the cellulose chains in the axial direction (Gross et al. 2011). The interaction of cellulose molecules was investigated with several of the most common anions in imidazolyl ILs by DFT calculations. The anion forming hydrogen bonds with hydroxyl protons of cellulose is a key factor in solubilizing cellulose. The strength of the anion interactions with cellulose was in the following order: Cl > OAc > (MeO)2PO2 > BF4 > PF6, SCN > PF6 (Guo et al. 2010; Zhao et al. 2013). The solubilization effect of 13 ILs with fixed anions (OAc) on cellulose showed that the acidic protons on the cationic heterocycles were necessary for cellulose solubilization, while the Van der Waals interactions between the cations and cellulose were not important. These cations with highly electronegative atoms in the main chain or large size groups in the alkyl chain would reduce cellulose solubility (Lu et al. 2014). The ILs were diverse and easy to design, and after the initial understanding of the mechanism of cellulose solubilization by ILs, researchers placed more attention on developing and designing efficient ILs.

Steric hindrance is the key to developing a preferred solvent for cellulose. The volume and flexibility of the cationic structure must be considered while maintaining the anionic polarity. Two ILs, [C8mim]OAc and OE2imC3C, were able to solubilize cellulose but not xylan within 60 min. The MD calculations showed that the polysaccharide formed a stable hydrogen bonding network in [Emim]OAc, but a relatively unstable hydrogen bonding structure in OE2imC3C. [Emim]OAc and OE2imC3C had similar hydrogen-bonding networks, and the difference in the ability to solubilize cellulose and xylan might be attributed to spatial site resistance (Kadokawa et al. 2021). Meng and co-workers designed 20 ILs based on tetraalkylammonium cations and OAcfor their ability to solubilize cellulose and found that the solubilization effect was highly dependent on the size of the ions, and the most effective ILs were less viscous (Meng et al. 2017).

Researchers found that ILs containing entraining agents could also solubilize cellulose, sometimes with better solubilization than pure IL systems. Dimethyl sulfoxide as an auxiliary reduced the viscosity of ionic liquids, thus improving the solventization of ionic liquids (Meng et al. 2017). The MD results indicate that DMSO does not interact specifically with glucose and would not significantly affect the specific interactions between cations and anions or between ionic liquids and polymers (Andanson et al. 2014; Velioglu et al. 2014). Compared to DMSO, water molecules might be closer to the glucose ring, thus interfering with the IL-glucose interaction, which might be the reason for the unsuitability of water as a coagent in ILs of solubilized cellulose (Huo et al. 2013; Andanson et al. 2014). Due to this property of water, it could be used as a regenerative solvent for cellulose (Gupta et al. 2013). The MD further showed that the anti-solvent had a strong affinity for ILs, and the H-bond between cellulose and ILs was broken, promoting the regeneration of cellulose (Ju et al. 2022).

Hemicellulose Dissolution Mechanism

Few simulation studies were performed for the hemicellulose dissolution process and mechanism. The DFT method is applied to calculate the solubilization mechanism of hemicellulose and cellulose in solvents ([C1mim]OAc, water, and methanol). It was found that the binding energy between xylan and IL was 20 kcal/mol stronger than that between cellulose and IL (Payal et al. 2012). Mohan and co-workers screened 1428 ILs by the COSMO-RS model. Subsequently, OAc and [Emim] are considered better candidate ions for solubilization of cellulose and hemicellulose by interaction energy calculations. The hydrogen bonding between the hydroxyl group and the anion of cellulose/hemicellulose was determined, the dissolution order of cellulose and hemicellulose in ionic liquids was tentatively determined, and the anion of IL was determined to play a dominant role in the dissolution process (Mohan et al. 2018). Moyer and co-workers studied the potential of ionic liquids to solubilize lignocellulosic biomass, and [Amim]formate proved to be the most effective, especially in solubilizing hemicellulose (Moyer et al. 2018).

Lignin Dissolution Mechanism

Lignin is a major natural source of aromatic compounds, but it is difficult to achieve high-value utilization from lignin because of its structural recalcitrance and complexity. Lignin is more soluble in organic solvents than cellulose, and lignin oligomers are even soluble in water. Three lignin-water systems were simulated by Vu and co-workers using the PCFF force field. The lignin models were decamers composed of only G, S, or H units connected only by β-O-4. The simulation results showed that the reduced mobility of water molecules in the hydroxyl and methoxy regions of lignin was due to the hydrogen bond formation and hydrophobic effects, respectively (Vu et al. 2002). The solubilization behavior of enzymatic lignin in different organic-water systems showed that enzymatic lignin dissolved in acetone-water mixtures with a volume ratio of 7:3, while aggregated in water, ethanol, acetone, and tetrahydrofuran. The hydrophobic backbone and hydrophilic groups of lignin were solventized by acetone and water molecules, respectively, leading to lignin solubilization. Conversely, only the hydrophobic backbone or the hydrophilic groups were solvated in organic solvent or water, respectively, inducing serious lignin aggregation (Wang et al. 2020).

Ionic liquids are good solvents not only for cellulose but also for lignin. The interaction of Veratrylglycerol-β-Guaiacyl ether (VG) with [Amim]Cl was studied with DFT and MD simulations. The VG monomer had strong intramolecular hydrogen bonding, and its dimer had π-π stacking and hydrogen bonding interactions. The anion of [Amim]Cl played a more important role in lignin solubilization than the cation (Zhu et al. 2017). A cork-like lignin polymer was modeled by Manna and co-workers using 61 G units as the starting structure. The MD simulations of lignin were performed in the [Emim]OAc-water solvent system. In 50% and 80% of [Emim]OAc, the lignin structure was significantly unstable, and the pure [Emim]OAc system disrupted the intra-chain hydrogen bonds. The anion interacted mainly with the alkyl chain and hydroxyl group of lignin. In addition, the cationic imidazole group contributed to the solventization of the methoxy and hydroxyl groups, while the cationic ethyl group interacted with the benzene ring (Manna et al. 2021). Shi and co-workers used an [Emim]OAc-water mixture to study the pretreatment of switchgrass at 160 °C. The chemical composition and crystallinity of the pretreated biomass and the dissolution and depolymerization of lignin depended on the [Emim][OAc] concentration (Shi et al. 2014). Janesko conducted a comprehensive study to elucidate the non-covalent interactions between various ILs and cellulose/lignin. The ILs with π-conjugated cations favored the solubility of lignin (Janesko 2011).

PYROLYSIS MECHANISM

Pyrolysis is the thermal or thermochemical degradation of biomass materials under conditions of air isolation or supply of small amounts of air. It has become one of the common ways to utilize biomass. This section elucidates the pyrolysis mechanisms of cellulose, hemicellulose, and lignin at different temperatures by DFT and MD.

Cellulose Pyrolysis Mechanism

The reaction pathway of β-D-glucopyranose pyrolysis was investigated to synthesize gas (CO, H2) using the DFT method. The involvement of reactive hydrogen atoms could lower the reaction energy barrier. The CO was mainly from the decomposition of aldehyde groups, and H2 was mainly from the dehydrogenation process (Huang et al. 2013). Murillo and co-workers used MD to study the differences between maltose and cellobiose. The better thermal stability of fibrous disaccharides (gg and gt conformations) was attributed to the formation of stable intra- and inter-ring hydrogen-bonding networks (Murillo et al. 2015). In addition, the initial decomposition of cellulose pyrolysis was simulated at 600.15 K and 873.15 K using Car-Parrinello MD. The simulated cells and the main pyrolysis products are shown in Fig. 6. The pyrolysis products were mainly pre-LGA with by-products such as HMF and formic acid. At 873.15 K, the products’ energy barriers were: pre-LGA < pre-HMF < formic acid, which could prove that LGA was the main product of the rapid pyrolysis of cellulose (Agarwal et al. 2012). Yang and co-workers investigated the formation mechanisms of the three most abundant cellulose pyrolysis components, LGA, HAA, and PA. The LGA was likely formed by a synergistic mechanism; HAA could be formed from C1-C2 and C5-C6 by ring-opening and dehydration reactions, while PA was formed by ring-opening reactions followed by dehydration and transaldol reactions. The formation of HAA and PA competes with the formation of LGA (Yang et al. 2020).

Cellulose pyrolysis at temperatures of 1000 to 3000 K using the ReaxFF-RMD method simulated the conversion pathways of seven major products and concluded that the C-O bond within the pyran ring was the most susceptible to breakage (Liu et al. 2021b). The effect of CaO on the pyrolysis of cellulose using ReaxFF-RMD simulations showed the addition of CaO significantly promoted the pyrolysis of cellulose and immobilized more oxygen-containing groups with the increase of Ca/C mass ratio by CaO. The total number of C-O bonds in the system after adding CaO was lower than in the pure cellulose pyrolysis system, and the number of CO and H2O molecules was reduced (Si et al. 2019). When co-pyrolysis of cellulose and polyethylene was conducted, the hydrocarbon and H radicals from polyethylene interacted with the alcoholic groups and furans, which contributed to producing alcohols and suppressed the generation of aldehydes and ketones. In the presence of polyethylene, hydrogenation reactions readily occurred, contributing to the production of furanol (Wang et al. 2021c).

Fig. 6. Nascent decomposition processes in cellulose pyrolysis. Reprinted with permission from (Agarwal et al. 2012). Copyright 2012 American Chemical Society.

Hemicellulose Pyrolysis Mechanism

Due to the complex composition and branching structure of hemicelluloses, initial simulation studies of hemicellulose pyrolysis mechanisms usually used monosaccharides or disaccharides as models. The pyrolysis processes of xylose and arabinose under the M06-2X/6-31++G(d,p) level using the DFT method indicated that there were eight possible pyrolysis pathways for xylose and five for arabinose (Huang et al. 2016, 2017). The potential reaction pathways of monosaccharides (xylose, mannose, and galactose) by DFT calculations concluded that FF was more easily produced during the pyrolysis of xylose than the other products.

In contrast, during the pyrolysis of mannose and galactose, the formation of HMF was more favorable because of the higher energy of hydroxymethyl dissociation. The lower energy potential of the O-acetyl group readily broke to form acetic acid (Wang et al. 2013a). The pyrolysis pathway of xylobiose was investigated using the DFT method. The ring-opening reaction was the most kinetically favorable pathway, followed by the cleavage of the β-(1,4) glycosidic bond and the dehydration reaction (Li et al. 2017c). In recent years, with the development of computers, many complex and accurate models have been used for calculations. A hemicellulose model as glucomannan with O-acetyl groups and galactose side chains were more favorable for the glycosidic bond breakage and O-acetyl dissociation than xylan by DFT calculations (Wang et al. 2021b). The most favorable initial reaction was the cleavage of the 4-O-methylglucuronic acid unit. Furfural could be obtained from the ring-opening of 4-O-methylglucuronide by three different pathways, while O-methyl is mainly responsible for the generation of CH3OH. In addition, the pathway for the generation of the special furan derivative 2-hydroxymethylene-tetrahydrofuran-3-one is verified for the first time by DFT calculations (Wu et al. 2019). The similar model (xylan containing multiple branched chains) for pyrolysis by DFT showed that during the initial pyrolysis of xylan, the branching broke easier than the depolymerization of the main chain, and the arabinose unit is more active than the 4-O-methylglucuronide unit (Hu et al. 2021a).

Lignin Pyrolysis Mechanism

M06-2X functional was used to study the pyrolysis process of PPE at the 6-31G* and 6-311++G** levels. The bond dissociation enthalpy (BDE) of β-O in PPE was 7.6 kcal/mol lower than that of Cα-Cβ (Beste and Buchanan 2009). Similarly, BDE was calculated for four prevalent linkage bonds (β-O-4, α-O-4, β-5, and 5-5) in 69 lignin models. The results showed that C-C bonds were more stable than ether bonds, and β-O-4 linkage bonds were more persistent than other ether bonds (Kim et al. 2011). The pyrolytic behavior of a lignin dimer model (1-(4-hydroxy-3-methoxyphenyl)-2-(2-methoxyphenoxy)-1-ethanol) was analyzed by DFT and the main bond BDE order was Cβ-O < Cα-Cβ < Cα-OH < Cα-Caromatic < Caromatic-O. Nine possible radical reaction pathways were proposed, and the reaction energy barriers were calculated for each path (Shen et al. 2021). The vanillin pyrolysis process was simulated with DFT method at the M06-2X/6-31 G+(d,p) level. CO, benzene, catechol, phenol, and cyclopentadiene were the main products of the secondary pyrolysis of vanillin. Formyl, hydroxyl, and methoxy all contribute to the formation of CO (Wang et al. 2016).

CATALYTIC CONVERSION MECHANISM

Cellulose Catalytic Conversion Mechanism

The catalytic conversion of cellulose into high value-added chemicals and fuels is considered an effective way to utilize biomass energy. A novel cellulose conversion process using ILs as both solvent and catalyst has attracted much attention (Amarasekara and Owereh 2009; Jiang et al. 2011). [Bmim]Cl and [C4SO3Hmim]Cl were used as solvent and catalyst to study the conversion process of cellulose to glucose. The conversion included two basic steps: in the first step, the SO3H group protonated the glycosidic oxygen and Cl attacked the anomeric carbon leading to the breakage of the glycosidic bond to form a glycosyl chloride intermediate. In the second step, the oxygen atom of the water molecule attacked the anomeric carbon, accompanying deprotonation of the SO3 group and C-Cl bond breakage so that the catalytic cycle of glycosidic bond hydrolysis is completed (Li et al. 2015a). In the presence of this IL, glucose was further converted to the platform chemical HMF. The DFT calculations revealed the existence of two possible pathways to facilitate the formation of HMF, and identified that 1,2-enediol is a necessary intermediate for both paths.

Dehydration is regarded as one of the important reactions in the catalytic conversion of cellulose. Six chemical reaction pathways were proposed to decompose glucose in supercritical water at the B3LYP/AGU-cc-pVDZ level. Water molecules play a role in transferring hydrogen atoms during dehydration and keto-enol rearrangement, which significantly reduces the potential barrier to the reaction (Zhang et al. 2016). The selective conversion from cellulose to platform chemicals, such as HMF (van Putten et al. 2013; Li et al. 2018), LaA (Holm et al. 2010; Castillo Martinez et al. 2013), and LA (Lin et al. 2012; Li et al. 2014), was a major challenge. The mechanism of the cellulose-catalyzed conversion process was elaborated at the molecular level through DFT simulations to help better understand and guide the experiments. Currently, the conversion of glucose to HMF was divided into two steps: glucose was first isomerized to fructose in the presence of catalysts, such as enzymes, Lewis acid, or bases, and then fructose was dehydrated under acidic conditions to produce HMF. Sn-BEA zeolites could isomerize glucose to fructose by O2H deprotonation and hydrogen transfer from C2 to C1 to isomerize acyclic glucose into deprotonated fructose intermediate (Yang et al. 2013). The introduction of SO3H functional groups in ILs was a promising and versatile catalyst for the synthesis of HMF (Amarasekara and Owereh 2009). A series of Brønsted-Lewis acidic double cationic ILs could effectively convert glucose to HMF with a yield of up to 71% of HMF. The reaction mechanism is that AlCl4 facilitated the isomerization of glucose to fructose, while the SO3H group could form hydrogen bonds with fructose to promote dehydration (Guo et al. 2020). In addition, sulfonated poly(phenylene sulfide) (SPPS) could convert glucose and cellulose directly to HMF in ILs. The SO3H group of SPPS is both a proton donor (Brønsted acid) and a proton acceptor (conjugate base). The anions and cations of ILs in conjunction with SO3H-SPPS helped stabilize the reaction intermediates and transition states, thus allowing easy conversion of glucose to HMF (Li et al. 2018).

The catalytic production of LaA from cellulose/glucose has been widely studied as a high-value platform chemical for producing fine chemicals and biodegradable plastics. Wang and co-workers elucidated the mechanism of conversion from cellulose to LaA from theoretical and experimental studies by adding Pb(II), including hydrolysis of cellulose to glucose, isomerization of glucose to fructose, cleavage of fructose to trisaccharide, and conversion of trisaccharide to LaA. The role of Pb(II) was to convert the main product of the reaction to LaA, and in the absence of Pb(II) fructose would be dehydrated to HMF (Wang et al. 2013b). Y(III) was used as a catalyst to selectively convert hemicellulose, and cellulose fractions to LaA with a LaA yield of 66.3%. [Y(OH)2(H2O)2]+ significantly promoted the conversion of monosaccharides to LaA, especially facilitating the breakage of C3-C4 bonds of fructose. The visualization of the weak interactions (Fig. 7) showed that the initial intra- and interchain hydrogen bonds (blue part) in cellobiose were greatly weakened (green region) after the formation of new hydrogen bonds between H3O+-Cl and cellobiose (Xu et al. 2020b). The catalytic system formed by MgO and D-LaA had a significantly synergistic effect on the chemical catalytic conversion of tricarbonate sugars to D-LaA. Quantum chemical calculations showed that [Mg(OH)(H2O)3]+, as the active substance, was mainly involved in the conversion of DHA to PA. Pre-added D-LaA, assisted by undissolved MgO, mainly promoted the enantioselective formation of D-LaA from PA conversion (Xu et al. 2021b).

The conversion of cellulose to LA was usually divided into three steps: hydrolysis of cellulose to glucose, conversion of glucose to HMF, and hydration of HMF to LA and formic acid. Cellulose could be selectively converted to LA in the IL of [C3SO3Hmim]HSO4, and the specific conversion process from HMF to LA included (1) dehydration of C1 position and rehydration of C5 atom; (2) nucleophilic addition of water molecule to aldehyde group and protonation of C1 atom; (3) elimination of HCOOH, C4-C5 double bond electrophilic addition; and (4) ring-opening and subsequent enol-keto isomerization (Li et al. 2017a). A novel bifunctional catalyst, HScCl4, with high selectivity and efficiency was developed for converting HMF to LA. The DFT results showed that the synergistic catalytic effect generated by the acidic sites of HScCl4 significantly reduced the energy barriers of the reactants and intermediates, thus facilitating the conversion of HMF to LA (Liu et al. 2021a).

Fig. 7. The visualized weak interaction for the performance of H3O+-Cl on the breaking of inter- and intrachain hydrogen bonds in cellobiose. Reprinted with permission from (Xu et al. 2020b). Copyright 2020 American Chemical Society.

Hemicellulose Catalytic Conversion Mechanism

Due to the structural similarity of hemicellulose and cellulose, the related chemicals converted from cellulose could also be converted by hemicellulose (Liu et al. 2009; Mäki-Arvela et al. 2011; Xing et al. 2011; Liu et al. 2012). FF was an important derivative in the decomposition of xylose, and the formation mechanism of xylose to FF was studied by Wang and co-workers using the DFT method. As an intermediate conversion of xylose to FF, xylulose had a low energy barrier, 163.6 kJ/mol in the gas phase and 150.8 kJ/mol in water. Adding water molecules could significantly reduce the reaction barriers to FF formation resulting in higher FF yields in the aqueous environment (Wang et al. 2015). A butanone-water mixture solvent was used for the acid-catalyzed conversion of xylose to FF. The MD simulations confirmed the competition between butanone and water molecules around xylose and FF. The appropriate amount of butanone could replace most water molecules around xylose and FF. Thus, it promoted the dehydration reaction of xylose and blocked the consumption of FF (Zhao et al. 2019). Some ILs could also be used as reaction media for hemicellulose conversion. Xylose could be converted to HMF and FF by acid catalysis in [Bmim]Cl (Sievers et al. 2009).

In addition, metal chlorides (CrCl3, CuCl2-2H2O, CrCl3/LiCl, FeCl3-6H2O, LiCl, CuCl, and AlCl3) and inorganic acids (H2SO4, HCI, and H3PO4) were used as catalysts to catalyze the conversion of xylan to FF in [Bmim]Cl (Zhang et al. 2013). The [AlCln](n-3)- not only weakened the glycosidic bond leading to the hydrolysis of xylan to xylose but also caused the isomerization of xylose to form enol and the dehydration to furfural (Zhang and Zhao 2010).

Although computer simulations are less studied in the catalytic conversion mechanism of hemicellulose, a portion of simulation studies exist for the conversion of hemicellulose-based FF to other value-added chemicals. Kojcinovic and co-workers studied the catalytic hydrogenation of furfural on MoO2 and MoO3 and suggested that the most abundant end product was isopropyl levulinate. The specific reaction pathway was the hydrogenation of furfural to produce furfuryl alcohol, followed by etherification to make isopropyl furfuryl ether, and finally isopropyl levulinate by ring-opening. The DFT results indicated stronger surface adsorption with directly exposed Mo atoms (Kojcinovic et al. 2021). The DFT calculations showed furfural was bound to the bimetallic surface mainly through C=O bonds, with the furan ring tilted away from the surface, leading to 2-methylfuran via the hydrodeoxygenation (HDO) reaction, and furfuryl alcohol was identified as a possible intermediate. Similar interactions of C=O groups with furan rings were confirmed on SiO2-loaded FeNi bimetallic catalysts (Yu et al. 2014). In addition, the reaction pathways of furfural on Ni(111) monometallic and monolayer Fe/Ni(111) bimetallic surfaces were investigated. The DFT calculation showed that Fe/Ni(111) monolayer was an effective surface for furfural conversion and promoted the deoxidation of furfural by extending the C=O bond of the carbonyl group (Yu and Chen 2015).

Lignin Catalytic Conversion Mechanism

Obtaining high value-added products from lignin was challenging due to the recalcitrant nature of the lignin structure. With the increasing scientific understanding of lignin and the application of novel catalysts, many new lignin degradation strategies have been proposed in combination with computer simulations (Mensah et al. 2022). The DFT was applied to explore the effect of H protons on the catalytic process of β-O-4 type lignin dimer, and showed that the breakage of the Cβ-O bond was the main pathway for the cleavage of lignin macromolecules, and H protons weakened the Cβ-O bond (Chen et al. 2021). The formic acid-catalyzed depolymerization of lignin models to monoaromatic compounds using DFT is shown in Fig. 8. The process included a formylation phase, an elimination phase, and a hydrolysis phase. Additionally, a new mechanism (E1H-3c4e ) in the elimination phase was proposed, which had a lower potential barrier than the known E1 and E2 elimination (Qu et al. 2015).

The depolymerization of lignin was the essence of ILs as effective reagents for the biomass pretreatment process. [C3SO3HMIM][HSO4] was used to catalyze the depolymerization of lignin to obtain low molecular weight products. The electron density could be transferred from the substrate to the electron-deficient imidazole ring as demonstrated by both experimental and DFT theoretical calculations. The formation of strong hydrogen bonds between the anion and the substrate played an essential role in the catalytic process (Singh et al. 2018). There was consensus that the anions in ILs could form hydrogen-bonding interactions with lignin, while the mechanism of cations in lignin depolymerization remains to be added. Due to the large number of computational studies employing ILs for catalytic degradation of lignin, the highlights of these simulation studies are summarized in Table 2.