Cellulose composition and method for producing the same, and ionic liquid used to dissolve cellulose and / or its derivatives

Abstract

The present invention provides a novel cellulose composition and method for producing the same, comprising cellulose and / or its derivatives and an ionic liquid explored by predetermined molecular dynamics simulations, as well as an ionic liquid used for dissolving cellulose and / or its derivatives. [Solution] An ionic liquid search system for realizing a cellulose composition and a method for producing the same, which provides a novel cellulose composition, (1) generates training data using molecular dynamics (hereinafter referred to as "MD (Molecular Dynamics) method"), (2) creates a predictive model using the generated training data, (3) searches for a solvent using the generated predictive model, and (4) extracts a medium with a good predicted value by repeating the processes from (1) to (3).

Description

Technical Field

[0001] The present disclosure relates to cellulose compositions, methods for producing the same, and ionic liquids used for dissolving cellulose and / or derivatives thereof.

Background Art

[0002] Conventionally, ionic liquids are known as solvents for dissolving cellulose.

[0003] For example, Non-Patent Document 1 can be cited as a study on the dissolution mechanism. In this study, for example, as a result of analyzing the dissolution of a model cellulose crystal structure in an imidazolium-based ionic liquid by a molecular dynamics approach, it is said that the dissolution involves the cleavage of hydrogen bonds between cellulose molecular chains due to the penetration of the ionic liquid. Also, in ionic liquids having high solubility, both anions and cations contribute to the cleavage of intermolecular hydrogen bonds, but in ionic liquids with low solubility, this cleavage does not occur sufficiently.

Prior Art Documents

Non-Patent Documents

[0004]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] Conventionally, it has been difficult to efficiently search for ionic liquids suitable for dissolving cellulose and / or its derivatives. In other words, determining the solubility of cellulose and / or its derivatives in various ionic liquids using molecular dynamics simulations requires an extremely long time.

[0006] Under these circumstances, the primary objective of this disclosure is to provide a novel cellulose composition comprising cellulose and / or its derivatives and an ionic liquid explored by a predetermined molecular dynamics simulation. Furthermore, the disclosure also aims to provide a method for producing the cellulose composition and an ionic liquid explored by a predetermined molecular dynamics simulation used to dissolve the cellulose and / or its derivatives. [Means for solving the problem]

[0007] The inventors of this disclosure have diligently studied to solve the above-mentioned problems. As a result, they have found a solution comprising cellulose and / or its derivatives and an ionic liquid, wherein the ionic liquid has an anion or cation self-diffusion coefficient calculated by a predetermined ionic liquid search system of 0.5 × 10⁻⁶ -6 cm 2 We found that in cellulose compositions with a concentration of 1 / s or higher, cellulose and / or its derivatives dissolve suitably in ionic liquids.

[0008] This disclosure is the result of further consideration based on these findings. Specifically, this disclosure provides inventions in the following embodiments.

[0009] Item 1. comprising cellulose and / or its derivatives and an ionic liquid, The aforementioned ionic liquid has an anion or cation self-diffusion coefficient of 0.5 × 10, calculated using the ionic liquid search system described below. -6 cm 2 A cellulose composition with a density of / s or greater. (Ionic liquid discovery system) To derive the self-diffusion coefficient of cellulose relative to ionic liquids using molecular dynamics (MD) simulations, a system composed of ionic liquids (cation-anion pairs) and a mixed system of cellulose crystal fragments and ionic liquids are constructed using packmol software. GAFF is used for the force fields of each ionic liquid, and the atomic point charge is set to a 0.8 scale value of the RESP charge. A crystal model (10 molecular chains × 10 residues) representing cellulose fiber fragments is constructed from crystal structure analysis data of cellulose Iβ type. MD calculations are performed under constant temperature (400K), constant pressure (1 bar), and time (250ns) conditions for a system in which imidazolium-type ionic liquids are randomly placed around the crystal model. Glycam06 is applied to the carbohydrate molecular force field parameters, and a modified GAFF is applied to the ionic liquid force field parameters. Item 2. The index value of the ionic liquid search system includes the interaction free energy (PMF1) obtained from the radial distribution function between anions and cations of the ionic liquid to be searched. The cellulose composition according to item 1, wherein the ionic liquid has an interaction free energy (PMF1) of -1.0 to 0.5 kcal / mol / ion-pair. Item 3. The index value of the ionic liquid search system includes the interaction free energy (PMF2) obtained from the radial distribution function between the anions of the ionic liquid to be searched and the cellulose and / or its derivatives. The cellulose composition according to claim 1 or 2, wherein the ionic liquid has an interaction free energy (PMF2) of -1.5 kcal / mol / ion-pair or less. Item 4. The cellulose composition according to any one of Items 1 to 3, wherein, as an index value for the ionic liquid search system, the ratio of the number of hydrogen bonds (HB) between cellulose and / or its derivatives in the structure after the initial crystal structure has been unraveled to the number of hydrogen bonds (HB0) between cellulose and / or its derivatives in the initial crystal structure is calculated as (HB / HB0) × 100, and the ratio is 45% or less. Item 5. The index value of the ionic liquid exploration system includes the interaction energy (ΔE3) between cellulose and / or its derivatives. The cellulose composition according to any one of claims 1 to 4, wherein the interaction energy (ΔE3) is -200.0 kcal / mol / 10mer or greater. Item 6. The index value of the ionic liquid search system includes the interaction energy (ΔE2) between cellulose and / or its derivatives and anions. The cellulose composition according to any one of claims 1 to 5, wherein the interaction energy (ΔE2) is -2.0 kcal / mol / ion-pair or less. Item 7. The cellulose composition according to any one of items 1 to 6, wherein the ionic liquid is an imidazole derivative. Item 8. The cellulose composition according to Item 7, wherein the imidazole derivative has at least one functional group selected from the group consisting of hydrocarbons, fluorine, alkoxymethyl, alkoxyallyl (alkene), ether, vinyl, epoxy, methylthio, and sulfide. Item 9. Ionic liquids used to dissolve cellulose and / or its derivatives, The aforementioned ionic liquid has an anion or cation self-diffusion coefficient of 0.5 × 10, calculated using the ionic liquid search system described below. -6 cm 2 Ionic liquids with a temperature of 1 / s or higher. (Ionic liquid discovery system) To derive the self-diffusion coefficient of cellulose relative to ionic liquids using molecular dynamics (MD) simulations, a system composed of ionic liquids (cation-anion pairs) and a mixed system of cellulose crystal fragments and ionic liquids are constructed using packmol software. GAFF is used for the force fields of each ionic liquid, and the atomic point charge is set to a 0.8 scale value of the RESP charge. A crystal model (10 molecular chains × 10 residues) representing cellulose fiber fragments is constructed from crystal structure analysis data of cellulose Iβ type. MD calculations are performed under constant temperature (400K), constant pressure (1 bar), and time (250ns) conditions for a system in which imidazolium-type ionic liquids are randomly placed around the crystal model. Glycam06 is applied to the carbohydrate molecular force field parameters, and a modified GAFF is applied to the ionic liquid force field parameters. Item 10. The index value of the ionic liquid search system includes the interaction free energy (PMF1) obtained from the radial distribution function between anions and cations of the ionic liquid to be searched. The ionic liquid is the ionic liquid according to item 9, wherein the interaction free energy (PMF1) is -1.0 to 0.5 kcal / mol / ion-pair. Item 11. The index value of the ionic liquid search system includes the interaction free energy (PMF2) obtained from the radial distribution function between the anions of the ionic liquid to be searched and the cellulose and / or its derivatives. The ionic liquid is the ionic liquid according to item 9 or 10, wherein the interaction free energy (PMF2) is -1.5 kcal / mol / ion-pair or less. Item 12. As an index value for the ionic liquid search system, when the ratio of the number of hydrogen bonds between cellulose and / or its derivatives (HB) in the structure after the initial crystal structure has been unraveled to the number of hydrogen bonds between cellulose and / or its derivatives (HB0) in the initial crystal structure is calculated as (HB / HB0) × 100, An ionic liquid according to any one of items 9 to 11, wherein the aforementioned proportion is 45% or less. Item 13. The index value of the ionic liquid exploration system includes the interaction energy (ΔE3) between cellulose and / or its derivatives. The ionic liquid according to any one of items 9 to 11, wherein the interaction energy (ΔE3) is -200.0 kcal / mol / 10mer or greater. Item 14. The index value of the ionic liquid search system includes the interaction energy (ΔE2) between cellulose and / or its derivatives and anions. The ionic liquid according to any one of items 9 to 11, wherein the interaction energy (ΔE2) is -2.0 kcal / mol / ion-pair or less. Item 15. The ionic liquid according to item 9 or 10, wherein the ionic liquid is an imidazole derivative. Item 16. The ionic liquid according to item 15, wherein the imidazole derivative has at least one functional group selected from the group consisting of hydrocarbons, fluorine, alkoxymethyl, alkoxyallyl (alkene), ether, vinyl, epoxy, methylthio, and sulfide. Item 17. The process includes a step of mixing cellulose and / or its derivatives with an ionic liquid. The aforementioned ionic liquid has an anion or cation self-diffusion coefficient of 0.5 × 10, calculated using the ionic liquid search system described below. -6 cm 2 A method for producing a cellulose composition that is ≥ / s. (Ionic liquid discovery system) To derive the self-diffusion coefficient of cellulose relative to ionic liquids using molecular dynamics (MD) simulations, a system composed of ionic liquids (cation-anion pairs) and a mixed system of cellulose crystal fragments and ionic liquids are constructed using packmol software. GAFF is used for the force fields of each ionic liquid, and the atomic point charge is set to a 0.8 scale value of the RESP charge. A crystal model (10 molecular chains × 10 residues) representing cellulose fiber fragments is constructed from crystal structure analysis data of cellulose Iβ type. MD calculations are performed under constant temperature (400K), constant pressure (1 bar), and time (250ns) conditions for a system in which imidazolium-type ionic liquids are randomly placed around the crystal model. Glycam06 is applied to the carbohydrate molecular force field parameters, and a modified GAFF is applied to the ionic liquid force field parameters. [Effects of the Invention]

[0010] The present disclosure provides a novel cellulose composition comprising cellulose and / or its derivatives and an ionic liquid explored by a predetermined molecular dynamics simulation. Furthermore, the present disclosure provides a method for producing the cellulose composition and an ionic liquid explored by a predetermined molecular dynamics simulation used to dissolve the cellulose and / or its derivatives. [Brief explanation of the drawing]

[0011] [Figure 1] Figure 1 is a diagram illustrating the overview of the ionic liquid discovery system of this disclosure. [Figure 2] Figure 2 shows an example of a system for performing ionic liquid discovery. [Figure 3] Figure 3 is a processing flow diagram showing an example of the training data generation process. [Figure 4] Figure 4 shows an example of data in PDB (Protein Data Bank) format. [Figure 5] Figure 5 is a processing flow diagram showing an example of the process for creating a predictive model. [Figure 6] Figure 6 is a process flow diagram showing an example of a solvent discovery process. [Figure 7] Figure 7 is a diagram illustrating a typical ionic liquid. [Figure 8] Figure 8 shows a crystal model (10 molecular chains × 10 residues) based on crystal structure analysis data of cellulose type Iβ, assuming a cellulose fiber fragment. [Figure 9] Figure 9 shows a series of cycles for verification using MD calculations in the ionic liquid exploration system of the example. [Figure 10] Figure 10 is a plot of the data obtained from MD calculations and the predicted data from the model trained with GcNN in the example. [Modes for carrying out the invention]

[0012] Each configuration and its combination in each embodiment is an example, and additions, omissions, substitutions, and other modifications can be made as appropriate without departing from the spirit of this disclosure. This disclosure is not limited by the embodiments.

[0013] In the numerical ranges described step by step in the present disclosure, the upper limit value or the lower limit value described in a certain numerical range may be replaced with the upper limit value or the lower limit value of the numerical range described in other step-by-step descriptions. Also, the upper limit value and the upper limit value, the upper limit value and the lower limit value, or the lower limit value and the lower limit value described separately may be combined to form a numerical range, respectively. Further, in the numerical range described in the present disclosure, the upper limit value or the lower limit value described in a certain numerical range may be replaced with the value shown in the examples.

[0014] [Cellulose composition] The cellulose composition of the present disclosure contains at least cellulose and / or its derivative and an ionic liquid, and the self-diffusion coefficient of the anion or cation calculated by the ionic liquid search system described later for the ionic liquid is 0.5×10 -6 cm 2 / s or more. That is, the cellulose composition of the present disclosure is a composition in which at least cellulose and / or its derivative and an ionic liquid having a self-diffusion coefficient of 0.5×10 -6 cm 2 / s or more explored by using a predetermined ionic liquid search system are combined, and it is a novel cellulose composition. Hereinafter, the cellulose composition of the present disclosure will be described in detail.

[0015] Cellulose and / or its derivative is a crystalline polymer, poorly soluble in water and common organic solvents, and has poor processability. The cellulose raw material is not particularly limited, and examples include pulp, cotton linter, and lignocellulosic biomass etc.

[0016] In the cellulose composition of the present disclosure, it is preferable that at least a part of cellulose and / or its derivative is dissolved in the ionic liquid. Cellulose can be used alone or in combination of two or more kinds.

[0017] Examples of cellulose derivatives include cellulose esters, cellulose carbamates, and cellulose ethers. Cellulose derivatives can be used individually or in combination of two or more. Among cellulose derivatives, cellulose esters are preferred.

[0018] Cellulose esters include, for example, cellulose organic acid esters, cellulose organic acid ester ethers, cellulose inorganic acid esters, and mixed cellulose organic acid / inorganic acid esters.

[0019] Examples of cellulose organic acid esters include cellulose acylates [cellulose acetates such as cellulose diacetate (DAC) and cellulose triacetate (TAC); cellulose C3-6 acylates such as cellulose propionate and cellulose butyrate; cellulose acetate-C3-6 acylates such as cellulose acetate propionate (CAP) and cellulose acetate butyrate (CAB)], and aromatic organic acid esters (cellulose C7-12 aromatic carboxylic acid esters such as cellulose phthalate and cellulose benzoate).

[0020] Examples of cellulose organic acid esters and ethers include C2-6 acylcellulose C1-6 alkyl ethers such as acetylmethylcellulose, acetylethylcellulose, and acetylpropylcellulose; and C2-6 acylcellulose hydroxy C2-6 alkyl ethers such as acetylhydroxyethylcellulose and acetylhydroxypropylcellulose.

[0021] Examples of cellulose inorganic acid esters include cellulose nitrate, cellulose sulfate, and cellulose phosphate.

[0022] Examples of cellulose organic acid / inorganic acid mixed esters include cellulose acetate nitrate.

[0023] The content of cellulose and / or its derivatives in the cellulose composition of this disclosure is not particularly limited, but examples include about 2% by mass or more, about 5% by mass or more, about 8% by mass or more, and also examples include about 99% by mass or less, 95% by mass or less, 90% by mass or less, and preferred ranges include about 2-99% by mass, 2-95% by mass, 2-90% by mass, about 5-100% by mass, 5-95% by mass, 5-90% by mass, about 10-99% by mass, 10-95% by mass, and 10-90% by mass.

[0024] The ionic liquid contained in the cellulose composition disclosed herein has an anion or cation self-diffusion coefficient of 0.5 × 10, calculated using the ionic liquid search system described below. -6 cm 2 It is / s or greater.

[0025] (Ionic liquid discovery system) To derive the self-diffusion coefficient of cellulose relative to ionic liquids using molecular dynamics (MD) simulations, a system composed of ionic liquids (cation-anion pairs) and a mixed system of cellulose crystal fragments and ionic liquids are constructed using packmol software. GAFF is used for the force fields of each ionic liquid, and the atomic point charge is set to a 0.8 scale value of the RESP charge. A crystal model (10 molecular chains × 10 residues) representing cellulose fiber fragments is constructed from crystal structure analysis data of cellulose Iβ type. MD calculations are performed under constant temperature (400K), constant pressure (1 bar), and time (250ns) conditions for a system in which imidazolium-type ionic liquids are randomly placed around the crystal model. Glycam06 is applied to the carbohydrate molecular force field parameters, and a modified GAFF is applied to the ionic liquid force field parameters.

[0026] For a more specific method for the ionic liquid exploration system, the method described in the examples will be adopted. As described later, cellulose exists as a highly crystalline fiber in which molecular chains are assembled by complex and strong intermolecular interactions (such as hydrogen bonds and dispersion forces). Therefore, it has the problem of being poorly soluble in water or common organic solvents and having poor processability. In this disclosure, as an ionic liquid exploration system, we extended the relationship between the amount of intermolecular hydrogen bonds and solubility obtained by conventional molecular dynamics (MD) calculations, and designed a solvent with high solubility by fusing high-throughput MD calculations and machine learning (ML). We found evaluation indicators similar to those for hydrogen bonds, such as the change in free energy of cation-anion interactions in ionic liquids. As a result, we found that the balance of these values ​​determines the solubility of cellulose in ionic liquids.

[0027] In the ionic liquid search system of this disclosure, it is preferable to create a predictive model for predicting the solubility of polymers in a solvent using machine learning, and to use the predictive model to search for solvents in which polymers have good solubility. Specifically, training data is generated using molecular dynamics (MD) (Figure 1: (1)), a predictive model is created using the training data (Figure 1: (2)), and solvents are searched using the predictive model (Figure 1: (3)). Furthermore, training data is created using the found solvents, and the predictive model is reconstructed using the training data. In other words, the processes from (1) to (3) above are repeated (Figure 1: (4)).

[0028] (Machine Learning (ML)) In other words, in this disclosure, machine learning (ML) preferably includes the following processes: A training data generation process that calculates an index value representing the solubility of cellulose and / or its derivatives in the ionic liquid to be studied, A model creation process that creates a predictive model that learns the relationship between feature quantities based on the chemical structure of the ionic liquid to be learned and index values ​​representing the solubility of cellulose and / or its derivatives using machine learning, and outputs predicted values ​​representing the solubility of cellulose and / or its derivatives in the ionic liquid to be predicted, A search process that repeatedly determines the ionic liquid to be predicted based on a predetermined algorithm, and outputs the predicted value using the feature quantities based on the chemical structure of the ionic liquid to be predicted and the prediction model, This includes one or more computers running the program. In this disclosure, the index value includes at least the self-diffusion coefficient of the ionic liquid being studied.

[0029] Cellulose and its derivatives are poorly soluble in water and common organic solvents and have poor processability; therefore, it is extremely useful to search for ionic liquids that exhibit excellent solubility and apply them as solvents for cellulose compositions.

[0030] Ionic liquids are salts of cations and anions that are molten salts at room temperature and are in a liquid state at approximately 100°C or below. Ionic liquids are structurally diverse, and extensive exploration using experimental methods such as MD is time-consuming. Therefore, an efficient method for exploring ionic liquids would be particularly useful. The cellulose composition of this disclosure may contain neutral molecules such as organic solvents in addition to ionic liquids (a mixed solvent of ionic liquid and organic solvent, etc.), or the solvent contained in the cellulose composition of this disclosure may consist only of ionic liquids. Furthermore, the solvent contained in the cellulose composition of this disclosure may have an anion or cation self-diffusion coefficient of 0.5 × 10⁻⁶ as calculated by the ionic liquid exploration system of this disclosure. -6 cm 2 It may consist only of ionic liquids with a self-diffusion coefficient of 0.5 × 10⁻¹⁰ or higher, or the self-diffusion coefficient of the liquid may be 0.5 × 10⁻¹⁰. -6 cm 2 It may also be a mixed liquid with an ionic liquid with a saturation level of less than / s.

[0031] In generating training data, an index value representing the solubility of a solvent (ionic liquid) for a polymer (cellulose) is determined by molecular dynamics simulation (MD simulation) (also called "solubility parameters"). MD simulation can be performed using existing methods. That is, the force fields of the ionic liquid and cellulose are defined, and the simulation is performed with cellulose placed in the ionic liquid. Examples of index values ​​representing solubility include parameters representing the interaction between cellulose molecules, the self-diffusion coefficient of the ionic liquid, the mean square displacement of the polymer in the solvent, the peak value of the radial distribution function between the solvent and the polymer, the change in free energy relative to that peak value, the peak value of the radial distribution function between the two components in the solvent, and the change in free energy relative to that peak value.

[0032] In creating the predictive model, the relationship between the features of ionic liquids and an index value representing solubility is learned through machine learning. Machine learning can be performed using existing methods such as deep learning, linear regression, and decision tree regression. The explanatory variables, which are the features of ionic liquids, can be represented by, for example, molecular descriptor-based models or graph-based models. A molecular descriptor-based model may be, for example, a molecular fingerprint that patterns the molecular structure of ionic liquids. A graph-based model may be, for example, a graph convolutional neural network that quantifies the features of the molecular structure. The target variable, the index value representing solubility, is a value calculated through the training data generation described above.

[0033] In ionic liquid search, a predicted value representing the solubility of a given polymer in a target ionic liquid is obtained using a created prediction model. The target liquid is generated based on a predetermined algorithm. This algorithm generates the target solvent by rearranging the side chains attached to the initial structure of the compound skeleton. Multiple target solvents are generated, and a predicted value is calculated for each. By repeating this process, a solvent with a good predicted value can be searched for. For example, solvents whose predicted value exceeds a predetermined threshold may be extracted. Alternatively, a preferred solvent may be searched for based on evaluation values ​​that include multiple predicted values ​​and scores other than the predicted values.

[0034] By repeating the above process, the prediction accuracy of the prediction model can be improved. Specifically, for solvents whose predicted values ​​are deemed good, a force field is created based on their structure, and an index value representing solubility is calculated using MD simulation. The calculated index value is then used as training data. Furthermore, the solvents extracted in the solvent search can be evaluated based on the calculated index value.

[0035] Figure 2 is a diagram showing an example of a system for performing ionic liquid search. System 100 includes a computer 1. Computer 1 also includes a processor 11, a storage device 12, and a user interface (UI) 13. The processor 11 is an arithmetic processing unit such as a CPU (Central Processing Unit) and performs each process according to the embodiment by executing a program. The storage device 12 is at least one of a main memory such as RAM (Random Access Memory) or ROM (Read Only Memory), and an auxiliary storage device (secondary storage device) such as an HDD (Hard-disk Drive), SSD (Solid State Drive), or flash memory. The main memory temporarily stores programs read by the processor 11 and reserves a workspace for the processor 11. The auxiliary storage device stores programs executed by the processor 11 and other data. The storage device 12 is assumed to have pre-stored identification information for identifying solvents and predetermined polymers used to generate learning data. The UI 13 is an input / output device such as a touch panel, keyboard, or pointing device. The UI 13 accepts input through user operation and outputs information to the user.

[0036] <Generating training data> Figure 3 is a processing flow diagram showing an example of the training data generation process. The training data generation process is started at any time based on user input. It is also assumed that the storage device 12 already contains information defining a predetermined solvent and a predetermined polymer to be used in the simulation. The solvent and polymer can be stored in a database or text data, for example, as strings expressed using SMILES (Simplified Molecular Input Line Entry System) notation. For example, 1-ethyl-3-methylimidazolim can be represented by the following string using SMILES notation. CCn1cc[n+](C)c1

[0037] When the training data generation process begins, the processor 11 of computer 1 creates force fields for the solvent and polymer (Figure 3: S1). In this step, the string representing the solvent, written in SMILES notation, is converted into the three-dimensional structure of the molecule using existing methods. The three-dimensional structure can be represented as data in formats such as XYZ coordinates (orthogonal coordinates) or Z-matrix. The aforementioned 1-ethyl-3-methylimidazolim, when represented in XYZ coordinates, looks like Figure 4, for example. Figure 4 is an example of PDB (Protein Data Bank) format data containing a molecular structure represented in XYZ coordinates. The PDB format is a file consisting of fixed-length records of 80 characters per line. Records beginning with the identifier "ATOM" or "HETATM" contain the X, Y, and Z coordinates (angstroms) of the atom. Figure 4 shows an example of an excerpt of the HETATM record, which describes atomic coordinates, from the PDB format data. The characters from 31 to 54, enclosed by the dashed line, contain the X, Y, and Z coordinates of the atom.

[0038] Furthermore, a force field can be created for each of the solvents and polymers based on their three-dimensional structure. Force field E total The potential function representing this is defined, for example, by equation (1) below. E total =E bounded +E nonbounded ...(1) Note E bounded E is a bond term that represents the chemical bond strength of a covalent bond, and is calculated by integrating terms based on the bond length within the molecule, the bond angle (offset), and the bond rotation angle (dihedral angle). nonboundedThe force field is a non-bonding term representing electrostatic and intermolecular forces, calculated by integrating terms based on intramolecular and intermolecular van der Waals forces and terms based on electrostatic interactions (electrostatic forces). The terms based on intramolecular bond length, bond angle, bond rotation angle, van der Waals forces, and electrostatic interactions represent intramolecular interactions. The terms based on van der Waals forces and electrostatic interactions represent intermolecular interactions. Furthermore, the force field can be created using functions in existing software. For example, the GAFF (general AMBER force field) within the MD simulation software Amber Tools can be used to calculate the chemical bond force of covalent bonds and the van der Waals forces among the non-bonding terms. Additionally, electrostatic forces can be calculated using existing quantum chemistry calculation software.

[0039] After S1, processor 11 performs an MD simulation on the solvent (Figure 3: S2). The MD simulation can also be performed using existing methods. In the MD simulation, the equations of motion for individual atoms are created using the force field described above, and the time evolution of the molecular structure (i.e., the position and velocity of each atom over time) is simulated by integrating the equations of motion. In this step, the simulation is performed on a unit cell composed only of the solvent, at atmospheric pressure and at a temperature that allows sufficient solvent diffusion.

[0040] After S2, the processor 11 performs MD simulations on the solvent and polymer (Figure 3: S3). In this step, for example, the polymer is placed in the solvent and the simulation is performed in the same way as in S2.

[0041] After S3, processor 11 obtains parameters related to solubility from the results of the MD simulation (Figure 3: S4). The particle arrangement in equilibrium state is created by the processing in S2, and in S4, the self-diffusion coefficient of the solvent is determined by analyzing the MD simulation results. That is, the self-diffusion coefficient can be determined from the mean square displacement by tracking the trajectory of the molecules. The mean square displacement can be calculated as the average value of the squares of the distances between given atoms. In particular, by performing simulations at a predetermined temperature range or a predetermined temperature and evaluating the dissolution phenomenon by the fluidity of the solvent (i.e., the self-diffusion coefficient), it is possible to generate training data for creating a predictive model suitable for the search for ionic liquids described above.

[0042] Furthermore, in S4, parameters related to solubility, such as parameters representing intermolecular interactions, can be determined by analyzing the data obtained from the processing in S3. Examples of parameters representing interactions between polymers include electrostatic force and intermolecular force. Electrostatic force and intermolecular force may be calculated based on the number or energy of hydrogen bonds. Assuming that hydrogen bonds occur between acceptor heavy atom A, donor hydrogen atom H, and donor heavy atom D, the number of hydrogen bonds can be determined from the simulation results. That is, hydrogen bonds may be determined to be formed when the distance from A to D is less than a distance threshold (e.g., 3 angstroms) and the bond angle of AHD is within a predetermined range (e.g., greater than 135 degrees and less than 180 degrees).

[0043] Furthermore, in S4, the mean square displacement of the polymer in the solvent may be determined as a parameter related to solubility.

[0044] Furthermore, in S4, as parameters related to solubility, at least one of the following may be determined: the peak value of the radial distribution function between the polymer and the solvent (the density of existence at the distance r where the probability of existence is highest) and the change in free energy relative to the peak value. Generally, the radial distribution function g(r) is defined as the probability of existence of other particles surrounding a given particle (atom or molecule) as a function of distance r. Note that the distance may be based on a group of atoms rather than a single particle. This distance can be the distance between the centroid (centroid calculated based on the mass and coordinates of each atom) or center (charge center calculated based on the atomic charge of each atom) of a group of atoms in the repeating unit of the polymer that appears to have strong interactions, and the centroid or center of a group of atoms in the solvent molecule that appears to have strong interactions. Note that a group of atoms that appears to have strong interactions can be appropriately selected based on interactions such as relatively high charge intensity or relatively strong hydrophobic interactions. Furthermore, if the LJ (Lennard-Jones) potential is relatively strong, it is preferable to use the centroid (in other words, the geometric center) of the relevant atoms. Furthermore, if the Coulomb force is relatively strong, it is preferable to use the charge centers of the relevant atoms.

[0045] Specifically, when a particle i is at the center of a spherical shell with inner radius r and outer radius r+dr (i.e., thickness dr), and n(r) is the number of other particles present within that shell, the density of particles within the spherical shell can be calculated using the following equation (2). n(r) / 4πr 2 dr ···(2) At this time, the radial distribution function g of a certain particle i i (r) can be found using the following equation (3). g i (r) = n(r) / 4πr 2 drρ ···(3) That is, the radial distribution function g of particle i i (r) is the density of particles within the spherical shell divided by the average number density ρ of the entire system. The radial distribution function g(r) of the system can then be calculated using the following equation (4). g(r) = <g i (r)>=<n(r)> / 4πr2 drρ ···(4) In other words, the radial distribution function g(r) of the system is g i This is the particle average of (r) and n(r).

[0046] This radial distribution function can be converted into an index called the free energy change F(r) using the following equation (5). F(r) = -kTlog(g(r)) ... (5) Here, k is Boltzmann's constant and T is temperature. As the distance r approaches infinity, the radial distribution function g(r) approaches 1, and therefore the change in free energy F(r) approaches 0. In other words, this free energy is a change with respect to infinity. If the minimum value of F(r) (in other words, the change in free energy relative to the peak value of the radial distribution function) is taken as a parameter related to solubility, it will correlate strongly with solubility.

[0047] For example, if the polymer is cellulose and the solvent is an ionic liquid with chloride ions as anions, the radial distribution function g(r) can be calculated by determining the distance r between the 2, 3, and 6 hydroxyl hydrogens of cellulose and the chloride ions based on the trajectory in the MD simulation. Furthermore, if the minimum value of the free energy change F(r) obtained from the radial distribution function g(r) is defined as a parameter related to solubility, it will correlate strongly with hydrogen bonding and solubility.

[0048] Furthermore, when the solvent is composed of two or more components, intermolecular interactions can be an important indicator. Therefore, in S4, the peak value of the radial distribution function between solvents, or the change in free energy with respect to it, may be determined as a parameter related to solubility. In this case, the peak value of the radial distribution function g(r), expressed using the distance r between the centroids or centers of atom groups that appear to have strong interactions within the solvent molecules, and the minimum value of the change in free energy F(r) are also important indicators for designing the solvent molecules. When the solvent is an ionic liquid, the peak value of the radial distribution function g(r) between the charge center of the cation and the charge center of the anion, and the minimum value of the change in free energy F(r) may be used as parameters related to solubility.

[0049] In particular, when the solvent is an ionic liquid composed of imidazole cations and chloride ions, the centroid between two nitrogen atoms in the imidazolium ring may be taken as the charge center of the cation, and the radial distribution function g(r) may be calculated based on the trajectory of the MD simulation for the distance r to the chloride ion. Furthermore, if the minimum value of F(r) calculated from the radial distribution function g(r) is defined as the change in free energy between the solvent molecule, especially between the anion and the cation, then although it does not strongly correlate with hydrogen bonding, the solubility increases within a certain range.

[0050] Furthermore, if the polymer contains cellulose and the solvent is an ionic liquid, the radial distribution function g(r) may be calculated based on the trajectory of the MD simulation for the distance r between the anion and the cellulose. If the minimum value of F(r) calculated from g(r) is defined as the change in free energy between the solvent molecule, particularly the anion, and the cellulose, it correlates to some extent with hydrogen bonding, and the solubility increases within a certain range.

[0051] Furthermore, parameters correlated with hydrogen bonding (or solubility) are generally useful as solubility indicators. When the polymer is cellulose, there is a strong correlation between the interaction energy between polymers and hydrogen bonding (or solubility), making the interaction energy between polymers a useful solubility indicator. Additionally, there is a strong correlation between the diffusion coefficient and hydrogen bonding (or solubility), making the diffusion coefficient a useful solubility indicator. There is also a strong correlation between the interaction energy between cellulose and the solvent and hydrogen bonding (or solubility), making the interaction energy between cellulose and the solvent a useful solubility indicator.

[0052] Furthermore, the parameters related to solubility may be determined according to the type of polymer being studied. If the user determines that the polymer has hydrogen bonds and that these are important in intermolecular interactions, hydrogen bonding can be used as one indicator. For general-purpose polymers, the mean square displacement can be used as an indicator. Alternatively, the peak value of the radial distribution function between the polymer and the solvent, or the change in free energy relative to the peak value (in other words, the minimum value of the change in free energy), may be used as an indicator. If the solvent consists of two or more components, the peak value of the radial distribution function between the solvents (between two components of the solvent), or the change in free energy relative to the peak value, may be used as an indicator. Note that in S4, multiple parameters may be determined.

[0053] As described above, the training data generation process allows for the calculation of an index value representing the solubility of cellulose, a polymer, in an ionic liquid, which is a solvent, through simulation. Furthermore, by repeating the training data generation process, index values ​​representing the solubility of a given polymer in multiple solvents can be calculated.

[0054] <Creating a predictive model> Figure 5 is a processing flow diagram showing an example of the predictive model creation process. The predictive model creation process is started at any time based on user input, for example, after the training data has been prepared by the training data generation process. It is also assumed that the storage device 12 has already stored an index value representing the solubility of a predetermined polymer in a solvent.

[0055] When the predictive model creation process begins, the processor 11 of computer 1 performs preprocessing (Figure 5: S11). In this step, the data to be used as explanatory variables and target variables (training data) is converted into a format suitable for machine learning. In this disclosure, the characteristics of a solvent are used as explanatory variables, and an index representing the solubility of a predetermined polymer in the solvent is used as the target variable for machine learning. In this disclosure, since the processing is performed with the polymer, which is the solute, fixed to one, the training data does not include the characteristics of the predetermined polymer. However, it is also possible to include the characteristics of the polymer, which is the solute, as explanatory variables and create a model that can predict the solubility of any polymer.

[0056] Solvent features are, for example, features based on the chemical structure of the solvent. Specifically, features may be described using molecular descriptor-based models such as molecular fingerprints. Molecular fingerprints represent the presence or absence of specific molecular bonds, functional groups, ring structures, etc., as binary vectors, or their occurrence counts as count vectors. For example, Morgan fingerprints obtained by existing software such as RDKit may be used, or descriptors based on physical features such as Mordred may be used. Alternatively, features representing the chemical structure of the solvent may be graph-based models, such as data that quantifies molecular structure features using a molecular graph convolutional neural network. Explanatory variables are standardized or normalized as appropriate.

[0057] Furthermore, the index value representing solubility, which is the objective variable, includes at least one of the following: the number or bond energy of hydrogen bonds formed within or between a given polymer; the self-diffusion coefficient of the solvent to be studied; the mean square displacement of a given polymer in the solvent to be studied; the peak value of the radial distribution function between the solvent and the polymer; the change in free energy relative to that peak value; the peak value of the radial distribution function between two components of the solvent; and the change in free energy relative to that peak value. The objective variable is also standardized or normalized as appropriate.

[0058] After S11, the processor 11 performs machine learning to create a predictive model (Figure 5: S12). The predictive model can employ deep learning models, linear regression models, decision tree regression models, etc., and is created by learning the relationship between features based on the chemical structure of the solvent to be learned and an index value representing the solubility of a given polymer in that solvent. If machine learning is performed on multiple index values, for example, a predictive model may be created for each index value. In addition, hyperparameters are determined appropriately for each model during machine learning. In this step, for example, the weights between nodes in a deep learning model are adjusted.

[0059] According to the prediction model creation process, by inputting the characteristic quantities of the solvent to be predicted, a prediction model can be created to calculate and output a predicted value for the solubility of a given polymer in that solvent.

[0060] <Solvent exploration> Figure 6 is a processing flow diagram showing an example of the solvent search process. The solvent search process is started at any time based on user input, for example, after a prediction model has been prepared by the prediction model creation process. It is also assumed that the prediction model created in the prediction model creation process is already stored in the storage device 12.

[0061] When the solvent search process is started, the processor 11 of computer 1 sets the initial chemical structure to be used for the search (Figure 6: S21). In this step, the basic structure of the solvent is input via UI 13, for example, based on user input, and stored in the memory device 12. If the solvent is a salt, for example, the following chemical structure is input as the basic structure. In this disclosure, the structure of the cation in the basic structure when the solvent is a salt, where the conditions are not changed, is also referred to as the cation basic skeleton. [ka] R is a hydrogen atom, a halogen atom, or a monovalent group, and multiple Rs may be the same or different from one another. The monovalent group may be, for example, an organic group or a functional group, as described later. Also, X is an anionic species. In subsequent processing, the types and combinations of R and anionic species are changed to create various solvents. Note that multiple initial chemical structures may be set.

[0062] After S21, the processor 11 sets the evaluation function to be used for the search (Figure 6: S22). The evaluation value V is defined, for example, by the following equation (6). V=Σ(α i M i )+βSA ···(6) M is the score of the predicted value from one or more prediction models. That is, an index value of 1 or more representing solubility predicted by the prediction model is substituted. SA is the SA (Synthetic Accessibility) score. The SA score is a score representing the synthesizability, which is an index of the complexity of the compound, and can be calculated using existing methods. α and β are weighting coefficients set according to the degree to which each score is given importance. In this way, in addition to the predicted value representing the solubility of a given polymer, the quality of the solvent may be comprehensively evaluated using a score representing the synthesizability of the solvent. In S22, for example, an evaluation function including weighting coefficients is set based on user operations. Note that the evaluation value V may be calculated without using the SA score. Also, if only one score is used in the evaluation function, it is not necessary to set weighting coefficients.

[0063] After S22, the processor 11 calculates an evaluation value V for the solvent to be predicted (Figure 6: S23). In this step, first, the solvent to be predicted is determined by combining R and anion species with the cation basic skeleton set in S21. For example, as the solvent to be predicted, R included in the above example basic structure may be an organic group composed of a carbon atom and at least some of halogen atoms such as hydrogen, oxygen, nitrogen, phosphorus, sulfur, and fluorine atoms, and some of the hydrogen atoms bonded to the carbon atom of the organic group may be substituted with other substituents. R may be a linear chain or include a branched structure. Also, Rs may bond to each other to create a solvent containing multiple ring structures. There are no particular limitations on the substituents that R may have, and examples include groups composed of at least some of hydrogen, carbon, oxygen, nitrogen, phosphorus, sulfur, and halogen atoms. Note that if R is a monovalent group, it may be a functional group such as a hydroxyl group, carboxyl group, amino group, phosphate group, thiol group, or sulfone group, rather than an organic group. Each R and anion species can be used to create various combinations through brute force, or a combination expected to yield a high evaluation score can be created using, for example, a genetic algorithm.

[0064] Figure 7 is a diagram illustrating a typical ionic liquid. Ionic liquids are room-temperature molten salts composed of anions and cations. As shown in Figure 7, basic structures with modified cations include imidazolium salts, pyrrolidinium salts, ammonium salts, pyridinium salts, piperidinium salts, phosphonium salts, etc. Anion species include Cl - , Br - , I - These are some examples. The solvent to be predicted to be created in S23 can be prepared by setting a basic skeleton based on a known cation in S21, and variations of the bonding structure and anion species may be stored in advance and used in S23, so that candidate ionic liquids can be prepared by combining a basic cation skeleton with a structure that can bond to the basic cation skeleton (hereinafter also referred to as the "bonding structure"). In the basic structure exemplified above, the bonding structure is R.

[0065] Furthermore, in S23, predicted values ​​are calculated for the solvent to be predicted using a prediction model. That is, the solvent to be predicted is subjected to the same pretreatment as in S11 in Figure 5 and input into the prediction model created in S12. As a result of the calculation by the prediction model, predicted values ​​of the index value representing solubility are calculated and output. If prediction models are created for multiple index values, multiple predicted values ​​will be output.

[0066] Then, in S23, the evaluation value of the solvent to be predicted is calculated using the evaluation function described above. In addition, solvents whose evaluation value V exceeds a predetermined threshold may be specifically extracted and output.

[0067] After S23, the processor 11 decides whether to terminate the process (Figure 6: S24). In this step, it decides whether to create further different target solvents by changing the binding structure and anion species for the cation basic skeleton described above. For example, it may decide to terminate the process when all combinations of predetermined binding structure and anion species variations have been created through brute force. Alternatively, when creating solvents using a genetic algorithm, it may decide to terminate the process when predetermined termination conditions are met. If it is decided not to terminate (S24: NO), the process returns to S23, and the processor 11 repeats the process for different target solvents. On the other hand, if it is decided to terminate (S24: YES), the solvent search process is terminated.

[0068] Solvent search processing allows for the exploration of candidate solvents that are likely to dissolve a given polymer with less computational cost than performing MD simulations.

[0069] <Repeat processing> Solvents with high evaluation values ​​found during the solvent search process may be used to verify their performance through MD simulation and as input for the training data generation process. Specifically, for solvents with high evaluation values ​​based on predictions, MD simulations may be performed based on their three-dimensional structure to calculate an index value representing solubility. The calculated index value can be used to verify the solvent's performance, and it can also be used as new training data to reconstruct the prediction model. Furthermore, especially when searching for ionic liquid solvents, the self-diffusion coefficient at a given temperature may be determined by MD simulation to confirm that crystallization has not occurred.

[0070] As described above, by repeatedly performing the training data generation process, the prediction model creation process, and the solvent search process, the accuracy of the prediction model is gradually improved, and solvents with high performance in dissolving a given polymer can be efficiently searched for.

[0071] <Variation> Each configuration and combination thereof in each embodiment is an example, and configurations may be added, omitted, replaced, and otherwise modified as appropriate, without departing from the spirit of this disclosure. This disclosure is not limited by the embodiments, but is limited only by the scope of the claims. Furthermore, each aspect disclosed herein can be combined with any other features disclosed herein.

[0072] The system described above may perform at least part of the following: training data generation, predictive model creation, and solvent search. For example, the predictive model creation process can create a predictive model that calculates and outputs a predicted value for the solubility of a given polymer in a given solvent by inputting the features of the solvent to be predicted. Furthermore, the solvent search process can search for candidate solvents that are likely to dissolve a given polymer with less computational cost than performing MD simulations.

[0073] At least a portion of the functions of computer 1 may be distributed and implemented across multiple devices (in other words, multiple processors), or the same functions may be provided in parallel by multiple devices (in other words, multiple processors). Alternatively, one computer 1 may be configured to have multiple processors. That is, a system may be provided that includes one or more computers (or one or more processors) for executing the above-described processing. Computer 1 may also be one or more servers connected to terminals via a communication network. The servers receive requests from terminals operated by users and perform the processing according to the embodiments described above. The user's terminal may be a PC (Personal Computer), tablet, or other type of computer. The terminal also includes a processor, storage device, and an input / output interface for inputting and outputting information to and from the user. Furthermore, the servers and terminals are equipped with communication interfaces, which are wired or wireless communication modules, and are connected to each other via a network so that they can communicate with one another. The communication network includes, for example, an IP (Internet Protocol) network, and devices connected to the network can communicate based on a predetermined communication protocol. Part of the network may be a telephone network (fixed-line or mobile network), an ad hoc network, an intranet, a VPN (Virtual Private Network), a LAN (Local Area Network), a Wireless LAN, a WAN (Wide Area Network), or the Internet.

[0074] Furthermore, this disclosure includes a method for performing the above-described processing, a computer program, and a computer-readable recording medium on which the program is recorded. The recording medium on which the program is recorded enables the above-described processing by causing a computer to execute the program.

[0075] Here, a computer-readable recording medium refers to a recording medium that stores information such as data and programs through electrical, magnetic, optical, mechanical, or chemical means and can be read by a computer. Examples of such recording media that are removable from a computer include flexible disks, magneto-optical disks, optical disks, magnetic tapes, and memory cards. Examples of recording media that are fixed to a computer include HDDs, SSDs, and ROMs.

[0076] From the viewpoint of ensuring good solubility of cellulose and / or its derivatives in ionic liquids, and from the viewpoint of suitably exhibiting the effects of the present invention of cellulose, the self-diffusion coefficient of anions or cations calculated by the ionic liquid exploration system is preferably 0.3 × 10⁻⁶. -6 cm 2 / s or more, more preferably 0.4 × 10 -6 cm 2 / s or more, more preferably 0.5 × 10 -6 cm 2 It is 1.5 × 10⁻¹⁰ or more, and preferably 1.5 × 10⁻¹⁰ -6 cm 2 / s or less, more preferably 1.2 × 10 -6 cm 2 / s or less, more preferably 0.8 × 10 -6 cm 2 The value is less than or equal to / s, and a preferred range is 0.3 × 10 -6 cm 2 / s~1.5×10 -6 cm 2 / s, 0.3×10 -6 cm 2 / s~1.2×10 -6 cm 2 / s, 0.3×10 -6 cm 2 / s~0.8×10 -6 cm 2 / s, 0.4×10 -6 cm 2 / s~1.5×10 -6 cm 2 / s, 0.4×10 -6 cm 2 / s~1.2×10-6 cm 2 / s, 0.4×10 -6 cm 2 / s~0.8×10 -6 cm 2 / s, 0.5×10 -6 cm 2 / s~1.5×10 -6 cm 2 / s, 0.5×10 -6 cm 2 / s~1.2×10 -6 cm 2 / s, 0.5×10 -6 cm 2 / s~0.8×10 -6 cm 2 A possible range is around / s.

[0077] In this disclosure, from the viewpoint of good solubility of cellulose and / or its derivatives in ionic liquids, the index value of the ionic liquid search system of this disclosure (the index value of machine learning (ML)) includes the interaction free energy (PMF1) obtained from the radial distribution function between anions and cations of the ionic liquid to be searched, wherein the interaction free energy (PMF1) of the ionic liquid is preferably -1.0 kcal / mol / ion-pair to 0.5 kcal / mol / ion-pair, more preferably -0.9 kcal / mol / ion-pair to -0.6 kcal / mol / ion-pair, and even more preferably -1.0 kcal / mol / ion to -0.6 kcal / mol / ion-pair.

[0078] In this disclosure, from the viewpoint of good solubility of cellulose and / or its derivatives in ionic liquids, the index value of the ionic liquid search system of this disclosure (the index value of machine learning (ML)) includes the interaction free energy (PMF2) obtained from the radial distribution function between the anions of the ionic liquid to be searched and cellulose and / or its derivatives, wherein the interaction free energy (PMF2) of the ionic liquid is preferably -1.5 kcal / mol / ion-pair or less, more preferably -1.7 kcal / mol / ion-pair or less, and also preferably -3.0 kcal / mol / ion-pair or less. It is preferable that the value be above -2.2 kcal / mol / ion-pair, more preferably -2.0 kcal / mol / ion-pair or higher, and preferred ranges include approximately -3.0 to -1.5 kcal / mol / ion-pair, approximately -3.0 to -1.7 kcal / mol / ion-pair, approximately -2.2 to -1.5 kcal / mol / ion-pair, approximately -2.2 to -1.7 kcal / mol / ion-pair, approximately -2.0 to -1.5 kcal / mol / ion-pair, and approximately -2.0 to -1.7 kcal / mol / ion-pair.

[0079] In this disclosure, from the viewpoint of good solubility of cellulose and / or its derivatives in ionic liquids, the index value of the ionic liquid search system of this disclosure (the index value of machine learning (ML)) is calculated as (HB / HB0) × 100, where the ratio of the number of hydrogen bonds (HB) between cellulose and / or its derivatives in the structure after the initial crystal structure has been unraveled to the number of hydrogen bonds (HB0) between cellulose and / or its derivatives in the initial crystal structure is (HB / HB0) × 100, and the ratio is preferably 45% or less, more preferably 40% or less, and even more preferably 30% or less, with preferred ranges including approximately 0-45%, approximately 0-40%, and approximately 0-30%.

[0080] Furthermore, in this disclosure, from the viewpoint of good solubility of cellulose and / or its derivatives in ionic liquids, the index value of the ionic liquid search system of this disclosure includes the interaction energy (ΔE3) between cellulose and / or its derivatives, and the interaction energy (ΔE3) is preferably -200.0 kcal / mol / 10mer or more, more preferably -150.0 kcal / mol / 10mer or more, and even more preferably -100.0 kcal / mol / 10mer or more, and preferred ranges include approximately -200.0 to 0.0 kcal / mol / 10mer, approximately -150.0 to 0.0 kcal / mol / 10mer, and approximately -100.0 to 0.0 kcal / mol / 10mer.

[0081] Furthermore, in this disclosure, from the viewpoint of good solubility of cellulose and / or its derivatives in ionic liquids, the index value of the ionic liquid search system of this disclosure includes the interaction energy (ΔE2) between cellulose and / or its derivatives and anions, wherein the interaction energy (ΔE2) is preferably -2.0 kcal / mol / ion-pair or less, more preferably -2.5 kcal / mol / ion-pair or less, and even more preferably -2.8 kcal / mol / ion-pair or less, and preferred ranges include approximately -5.0 to -2.0 kcal / mol / ion-pair, approximately -5.0 to -2.5 kcal / mol / ion-pair, and approximately -5.0 to -2.8 kcal / mol / ion-pair.

[0082] In this disclosure, the specific chemical structure of the ionic liquid is determined by the self-diffusion coefficient of the anion or cation calculated by the ionic liquid discovery system of this disclosure, which is 0.5 × 10⁻⁶. -6 cm 2While not particularly limited to a maximum of / s or more, the ionic liquid is preferably an imidazole derivative from the viewpoint of good solubility of cellulose and / or its derivatives in the ionic liquid. Furthermore, from a similar viewpoint, it is preferable that the imidazole derivative has at least one functional group selected from the group consisting of hydrocarbons, fluorine, alkoxymethyl, alkoxyallyl (alkene), ether, vinyl, epoxy, methylthio, and sulfide.

[0083] Specific examples of ionic liquids include those specifically explored in the examples.

[0084] The saturation concentration when the cellulose composition of this disclosure dissolves in an ionic liquid is not particularly limited, but from the viewpoint of suitably dissolving cellulose and / or its derivatives, the mass ratio of cellulose and / or its derivatives per 100 parts by mass of ionic liquid can be, for example, about 2 parts by mass or more, about 5 parts by mass or more, about 8 parts by mass or more, and also, for example, about 34 parts by mass or less, 30 parts by mass or less, 25 parts by mass or less, and preferred ranges include about 2 to 34 parts by mass, about 2 to 30 parts by mass, about 2 to 25 parts by mass, about 5 to 34 parts by mass, about 5 to 30 parts by mass, about 5 to 25 parts by mass, about 8 to 34 parts by mass, about 8 to 30 parts by mass, and about 8 to 25 parts by mass.

[0085] In the cellulose composition of this disclosure, the total proportion of cellulose and / or its derivatives and the mixture with an ionic liquid / or organic solvent is, for example, 15% by mass or more, preferably 30% by mass or more, more preferably 50% by mass or more, and may be 90% by mass or more, 100% by mass or the like.

[0086] [Ionic liquid] The ionic liquids of this disclosure are ionic liquids used to dissolve cellulose and / or its derivatives. The ionic liquids of this disclosure have an anion or cation self-diffusion coefficient of 0.5 × 10⁻¹⁰ as calculated by the ionic liquid discovery system of this disclosure described above. -6 cm 2 It is / s or greater.

[0087] For the details of the ionic liquid of the present disclosure, it is the same as the ionic liquid described in the item of [cellulose composition] mentioned above.

[0088] [Method for producing cellulose composition] The method for producing the cellulose composition of the present disclosure includes a step of mixing cellulose and / or its derivative with an ionic liquid (mixing step), and the ionic liquid has an anion or cation self-diffusion coefficient calculated by the ionic liquid exploration system of the present disclosure mentioned above of 0.5×10 -6 cm 2 / s or more.

[0089] For the details of cellulose and / or its derivative and the ionic liquid, they are respectively the same as the ionic liquid described in the item of [cellulose composition] mentioned above.

[0090] Also, the step of mixing cellulose and / or its derivative with the ionic liquid only requires that cellulose and / or its derivative and the ionic liquid are mixed. Preferably, it is mixed so that at least a part of cellulose and / or its derivative dissolves in the ionic liquid to produce a cellulose composition.

[0091] The temperature in the mixing step can be, for example, about 10 to 150°C. Also, the pressure can be, for example, about 1 to 10 atmospheres. Also, the mixing time can be, for example, about 5 minutes to 36 hours. The mixing can be carried out in atmospheric pressure, in an inert gas, etc.

Examples

[0092] Examples are shown below to more specifically explain the present disclosure.

[0093] Cellulose exists as a highly crystalline fiber formed by the aggregation of molecular chains through complex and strong intermolecular interactions (such as hydrogen bonds and dispersion forces). This results in poor solubility in water or common organic solvents and poor processability. In the embodiments of this disclosure, an ionic liquid exploration system was developed by extending the relationship between the amount of intermolecular hydrogen bonds and solubility, as determined by conventional molecular dynamics (MD) calculations. By integrating high-throughput MD calculations with machine learning (ML), a solvent with high solubility was designed, and evaluation metrics similar to those for hydrogen bonds, such as the change in free energy of cation-anion interactions in ionic liquids, were identified. As a result, it was found that the balance of these values ​​determines the solubility of cellulose in ionic liquids.

[0094] (Ionic liquid discovery system) In a system for searching for ionic liquids that suitably dissolve cellulose and / or its derivatives, a system composed of ionic liquids (cation-anion pairs) and a mixed system of cellulose crystal fragments and ionic liquids were constructed using packmol software to derive the self-diffusion coefficient of anions or cations. The force field of each ionic liquid was calculated using GAFF, and the atomic point charge was set to a 0.8 scale value of the RESP charge based on prior literature. Charge and constraint conditions are described in Uto, T., Yamamoto, K., Kadokawa, J., J. Phys. Chem. B, 122, 258-266 (2018). Furthermore, glucose in cellulose is realized by using constraint conditions. In addition, a crystal model (10 molecular chains × 10 residues) assuming cellulose fiber fragments was constructed from crystal structure analysis data of cellulose type Iβ. MD calculations were performed on a system in which imidazolium-type ionic liquids were randomly placed around the crystal model under constant temperature (400K), constant pressure (1 bar), and time (250ns). For the MD calculations, AMBER22 software was used, applying Glycam06 to the carbohydrate molecular force field parameters and a modified GAFF to the ionic liquid force field parameters. The same calculation method was used for systems of individual ionic liquids. Furthermore, the self-diffusion coefficient was derived using the cpptraj software from the ambertools package. Note that these calculations can also be performed with other software if the calculation method is equivalent. packmol is a program for randomly arranging molecular structures in space, amber is a program for performing molecular dynamics calculations, ambertools is a package that includes analysis of amber software, and cpptraj is the analysis program within it.

[0095] <Specific methods for searching for ionic liquids> A crystal model (10 molecular chains × 10 residues) representing cellulose fiber fragments was constructed from crystal structure analysis data of cellulose type Iβ (Figure 8). MD calculations were performed on a system with an imidazolium-type ionic liquid placed around the crystal model under constant temperature (400 K), constant pressure (1 bar), and time (250 ns). AMBER22 software was used for the MD calculations, with Glycam06 applied to the carbohydrate molecular force field parameters and a modified GAFF applied to the ionic liquid force field parameters.

[0096] From the obtained MD trajectories, the diffusion coefficient, number of intermolecular hydrogen bonds (HB), mean square displacement (RMSd), bias potential value, change in free energy (potential of mean force, PMF), and change in mean interaction energy were calculated. A graph convolutional neural network (GcNN) model using PyTorch geometric was constructed for these physical quantities, and machine learning (ML) was performed. Furthermore, inverse analysis was performed using a genetic algorithm by EvoMo to enumerate compounds that decreased HB. The anions constituting the ionic liquid were set to chloride ions, and the chemical structure of the cations was modified. By performing a series of cycles (Figure 9) of verification by MD calculations for the obtained candidates, we searched for ionic liquids that could be novel solvents not previously reported.

[0097] (Machine Learning (ML)) In this disclosure, machine learning (ML) includes the following processes: A training data generation process that calculates an index value representing the solubility of cellulose and / or its derivatives in the ionic liquid to be studied, A model creation process that creates a predictive model that learns the relationship between feature quantities based on the chemical structure of the ionic liquid to be learned and index values ​​representing the solubility of cellulose and / or its derivatives using machine learning, and outputs predicted values ​​representing the solubility of cellulose and / or its derivatives in the ionic liquid to be predicted, A search process that repeatedly determines the ionic liquid to be predicted based on a predetermined algorithm, and outputs the predicted value using the feature quantities based on the chemical structure of the ionic liquid to be predicted and the prediction model, This includes one or more computers running the program. The index value includes at least the self-diffusion coefficient of the ionic liquid being studied.

[0098] The specific machine learning techniques were carried out using the processing flow described in the <Training Data Generation> section above.

[0099] Figure 10 shows a plot of data obtained from MD calculations and predicted data from a GcNN-trained model. The coefficient of determination (R²) of the GcNN model with respect to the hydrogen bond number ratio was 0.64. As shown in Figure 10(a), the prediction accuracy decreased when the hydrogen bond number ratio fell below 50%, and no tendency for improvement was observed throughout the cycle. As quantities related to the hydrogen bond number ratio, the change in free energy of cation-anion interactions (PMF1) and the change in free energy of cellulose-anion interactions (PMF2) were calculated from the radial distribution function to serve as solubility indices. The R² values ​​for these were 0.92 (PMF1) and 0.88 (PMF2), respectively, indicating that a highly accurate GcNN model could be constructed (Figures 10(b), (c)). By repeating the MD-ML procedure for 7 cycles, it was possible to obtain a chemical structure that is estimated to have higher cellulose solubility than 1-ethyl-3-methylimidazolium chloride.

[0100] By using a machine learning model, it is possible to rank the synthesizable ionic liquids by evaluating any ionic liquid obtained using a common imidazolium salt type ionic liquid synthesis formula. Searching for raw materials from various databases of small molecules and listing the cation structures resulted in approximately 1 million entries.

[0101] Based on the molecular dynamics calculations and machine learning-based predictive model creation, and the generation of molecular structures based on the models, the proposed ionic liquid structures are then subjected to further molecular dynamics calculations for verification, completing a continuous cycle. Approximately 3000 different molecular structures have been examined. Finally, the molecular dynamics calculation results are reviewed, and suitable solvents for cellulose and / or its derivatives are ranked and proposed sequentially. Promising candidates are those expected to have higher solubility than the reference ionic liquids EmimCl and AmimCl. Tables 1 and 2 below summarize the analysis content used for evaluating physical properties. In Tables 1 and 2, "ns" means nanoseconds and represents the time tracked in molecular dynamics. The top 30 (Table 3 below) do not include any reference ionic liquids; only those judged to be superior solvents for cellulose and / or its derivatives are listed.

[0102] [Table 1]

[0103] [Table 2]

[0104] [Table 3]

[0105] According to Table 3 above, the chemical structures of promising compounds are shown as candidate numbers 1-30. From the chemical structures of candidate numbers 1-30, it was found that the cation represented by the following general formula is preferred as the cation structure of the ionic liquid that dissolves cellulose and / or its derivatives.

[0106] [ka]

[0107] R1 and R2 are independently a C1-C12 alkyl group or C1-C12 alkoxyalkyl group, a C1-C12 ether group, or a C1-C12 thioether group, respectively. If present, R3, R4, and R5 are independently hydrogen, a C1-C12 alkyl group, a C1-C12 alkoxyalkyl group or C1-C12 alkoxy group, a C1-C12 thioether group, or a halogen, respectively. The anions of the ionic liquid are halogens, pseudohalogens, or C1-C6 carboxylates.

[0108] The specific chemical structures for candidate numbers 1-30 are shown below.

[0109] [ka]

[0110] [ka]

[0111] [ka]

[0112] [ka]

[0113] [ka]

[0114] [ka]

[0115] [ka]

[0116] [ka]

[0117]

change

[0118]

change

[0119]

change

[0120]

change

[0121]

change

[0122]

change

[0123]

change

[0124]

change

[0125]

change

[0126]

change

[0127]

change

[0128]

change

[0129]

change

[0130]

change

[0131]

change

[0132]

change

[0133]

change

[0134]

change

[0135]

change

[0136]

change

[0137]

change

[0138] [ka]

[0139] In this embodiment, functional group analysis including the chemical structures of lower-ranking candidates for the ionic liquid yielded the following findings.

[0140] In ionic liquids, imidazole derivatives that have hydrocarbons, fluorine, alkoxymethyl, alkoxyallyl (alkene), ether, vinyl, epoxy, methylthio group, sulfide, etc. as functional groups satisfy the above-described present invention and tend to function as effective solvents for cellulose and / or its derivatives.

[0141] On the other hand, in ionic liquids, if acetylene, alcohol, aldehyde, alkyne, amine, carboxylmethyl ester, carbonylmethyl, carboxyl group, carboxylic acid, cyano group, furan, hydrazine, nitro group, hydroxyl group, ketone, methoxylcarbonyl group, methylamino group, naphthol, nitrile, phenol ether, phenol, phenyl group, phosphine, primary amine, pyridine, lactone, hydroxyl group, t-butyl, thiocarbonyl, thioester, thiol, trifluoromethyl, etc. are present, cellulose and / or its derivatives tend to cease to be effective solvents. [Explanation of Symbols]

[0142] 100: System, 1: Computer, 11: Processor, 12: Storage device, 13: User interface

Claims

1. The mixture comprises cellulose and / or its derivatives, and an ionic liquid. The aforementioned ionic liquid has an anion or cation self-diffusion coefficient of 0.5 × 10, calculated using the ionic liquid search system described below. -6 cm 2 A cellulose composition having a density of / s or greater. (Ionic liquid exploration system) To derive the self-diffusion coefficient of cellulose relative to ionic liquids using molecular dynamics (MD) simulations, a system composed of ionic liquids (cation-anion pairs) and a mixed system of cellulose crystal fragments and ionic liquids are constructed using packmol software. GAFF is used for the force fields of each ionic liquid, and the atomic point charge is set to a 0.8 scale value of the RESP charge. A crystal model (10 molecular chains × 10 residues) representing cellulose fiber fragments is constructed from crystal structure analysis data of cellulose Iβ type. MD calculations are performed under constant temperature (400K), constant pressure (1 bar), and time (250 ns) conditions for a system in which imidazolium-type ionic liquids are randomly placed around the crystal model. Glycam06 is applied to the carbohydrate molecular force field parameters, and a modified GAFF is applied to the ionic liquid force field parameters.

2. The index value of the ionic liquid search system includes the interaction free energy (PMF1) obtained from the radial distribution function between anions and cations of the ionic liquid to be searched. The cellulose composition according to claim 1, wherein the ionic liquid has an interaction free energy (PMF1) of -1.0 to 0.5 kcal / mol / ion-pair.

3. The index value of the ionic liquid search system includes the interaction free energy (PMF2) obtained from the radial distribution function between the anions of the ionic liquid to be searched and the cellulose and / or its derivatives. The cellulose composition according to claim 1 or 2, wherein the ionic liquid has an interaction free energy (PMF2) of -1.5 kcal / mol / ion-pair or less.

4. The cellulose composition according to claim 1 or 2, wherein, as an index value for the ionic liquid search system, the ratio of the number of hydrogen bonds (HB) between cellulose and / or its derivatives in the structure after the initial crystal structure has been unraveled to the number of hydrogen bonds (HB0) between cellulose and / or its derivatives in the initial crystal structure is calculated as (HB / HB0) × 100, and the ratio is 45% or less.

5. The index value of the ionic liquid search system includes the interaction energy (ΔE3) between cellulose and / or its derivatives. The cellulose composition according to claim 1 or 2, wherein the interaction energy (ΔE3) is -200.0 kcal / mol / 10 mer or more.

6. The index value of the ionic liquid search system includes the interaction energy (ΔE2) between cellulose and / or its derivatives and anions. The cellulose composition according to claim 1 or 2, wherein the interaction energy (ΔE2) is -2.0 kcal / mol / ion-pair or less.

7. The cellulose composition according to claim 1 or 2, wherein the ionic liquid is an imidazole derivative.

8. The cellulose composition according to claim 7, wherein the imidazole derivative has at least one functional group selected from the group consisting of hydrocarbons, fluorine, alkoxymethyl, alkoxyallyl (alkene), ether, vinyl, epoxy, methylthio, and sulfide.

9. An ionic liquid used to dissolve cellulose and / or its derivatives, The aforementioned ionic liquid has an anion or cation self-diffusion coefficient of 0.5 × 10, calculated using the ionic liquid search system described below. -6 cm 2 An ionic liquid with a temperature of / s or higher. (Ionic liquid exploration system) To derive the self-diffusion coefficient of cellulose relative to ionic liquids using molecular dynamics (MD) simulations, a system composed of ionic liquids (cation-anion pairs) and a mixed system of cellulose crystal fragments and ionic liquids are constructed using packmol software. GAFF is used for the force fields of each ionic liquid, and the atomic point charge is set to a 0.8 scale value of the RESP charge. A crystal model (10 molecular chains × 10 residues) representing cellulose fiber fragments is constructed from crystal structure analysis data of cellulose Iβ type. MD calculations are performed under constant temperature (400K), constant pressure (1 bar), and time (250 ns) conditions for a system in which imidazolium-type ionic liquids are randomly placed around the crystal model. Glycam06 is applied to the carbohydrate molecular force field parameters, and a modified GAFF is applied to the ionic liquid force field parameters.

10. The index value of the ionic liquid search system includes the interaction free energy (PMF1) obtained from the radial distribution function between anions and cations of the ionic liquid to be searched. The ionic liquid according to claim 9, wherein the ionic liquid has an interaction free energy (PMF1) of -1.0 to 0.5 kcal / mol / ion-pair.

11. The index value of the ionic liquid search system includes the interaction free energy (PMF2) obtained from the radial distribution function between the anions of the ionic liquid to be searched and the cellulose and / or its derivatives. The ionic liquid according to claim 9 or 10, wherein the interaction free energy (PMF2) is -1.5 kcal / mol / ion-pair or less.

12. As an index value for the ionic liquid search system, when the ratio of the number of hydrogen bonds between cellulose and / or its derivatives in the structure after the initial crystal structure has been unraveled (HB) to the number of hydrogen bonds between cellulose and / or its derivatives in the initial crystal structure (HB0) is calculated as (HB / HB0) × 100, The ionic liquid according to claim 9 or 10, wherein the aforementioned proportion is 45% or less.

13. The index value of the ionic liquid search system includes the interaction energy (ΔE3) between cellulose and / or its derivatives. The ionic liquid according to claim 9 or 10, wherein the interaction energy (ΔE3) is -200.0 kcal / mol / 10 mer or greater.

14. The index value of the ionic liquid search system includes the interaction energy (ΔE2) between cellulose and / or its derivatives and anions. The ionic liquid according to claim 9 or 10, wherein the interaction energy (ΔE2) is -2.0 kcal / mol / ion-pair or less.

15. The ionic liquid according to claim 9 or 10, wherein the ionic liquid is an imidazole derivative.

16. The ionic liquid according to claim 15, wherein the imidazole derivative has at least one functional group selected from the group consisting of hydrocarbons, fluorine, alkoxymethyl, alkoxyallyl (alkene), ether, vinyl, epoxy, methylthio, and sulfide.

17. The process includes a step of mixing cellulose and / or its derivatives with an ionic liquid, The aforementioned ionic liquid has an anion or cation self-diffusion coefficient of 0.5 × 10, calculated using the ionic liquid search system described below. -6 cm 2 A method for producing a cellulose composition that is 1 / s or greater. (Ionic liquid exploration system) To derive the self-diffusion coefficient of cellulose relative to ionic liquids using molecular dynamics (MD) simulations, a system composed of ionic liquids (cation-anion pairs) and a mixed system of cellulose crystal fragments and ionic liquids are constructed using packmol software. GAFF is used for the force fields of each ionic liquid, and the atomic point charge is set to a 0.8 scale value of the RESP charge. A crystal model (10 molecular chains × 10 residues) representing cellulose fiber fragments is constructed from crystal structure analysis data of cellulose Iβ type. MD calculations are performed under constant temperature (400K), constant pressure (1 bar), and time (250 ns) conditions for a system in which imidazolium-type ionic liquids are randomly placed around the crystal model. Glycam06 is applied to the carbohydrate molecular force field parameters, and a modified GAFF is applied to the ionic liquid force field parameters.

Images