Method for predicting lithium ion conductivity of solid electrolyte

Abstract

The invention discloses a method for rapidly and accurately predicting the lithium ion conductivity of a solid electrolyte. The method may include simulating a crystal structure of the solid electrolyte and creating a training set for machine learning based on the crystal structure; calculating specific potential energy of the simulated crystal structure through machine learning by using the training set; and predicting the lithium ion conductivity of the solid electrolyte according to the potential energy by using molecular dynamics simulation.

Description

Technical Field

[0001] The present disclosure relates to a method for rapidly and accurately predicting the lithium ion conductivity of solid electrolytes. Background Art

[0002] Nowadays, secondary batteries are widely used in many devices ranging from large devices such as vehicles, power storage systems, etc. to small devices such as mobile phones, video cameras, notebook computers, etc.

[0003] As the application fields of secondary batteries expand, the demand for improving the safety and high performance of secondary batteries is also increasing. As a type of secondary battery, lithium secondary batteries have advantages such as high energy density and large capacity per unit area compared to nickel-manganese batteries or nickel-cadmium batteries.

[0004] However, most electrolytes used in conventional lithium secondary batteries are liquid electrolytes, such as organic solvents, and therefore, safety issues such as electrolyte leakage and the resulting fire risk are constantly being raised.

[0005] Therefore, in order to improve the safety of secondary batteries, interest in all-solid-state batteries using solid electrolytes rather than liquid electrolytes is increasing. All-solid-state batteries using solid electrolytes instead of liquid electrolytes have advantages in safety because all components of all-solid-state batteries (including electrodes and electrolytes) are solid. In addition, all-solid-state batteries can use lithium metal or lithium alloy as anode materials, so it is well known that they have advantages in battery performance, namely high energy density, high output and long life, so research on all-solid-state batteries is underway, especially research on solid electrolytes with argyrodite-type crystal structures is actively underway.

[0006] In order to develop a solid electrolyte with high lithium ion conductivity, a lot of time and effort is required to study the synthesis of solid electrolytes with various crystal structures and compositions. In order to overcome such shortcomings, many attempts have been made to predict the lithium ion conductivity of solid electrolytes by calculation.

[0007] Recently, most studies on the calculation of lithium-ion conductivity of solid electrolytes have used the ab initio molecular dynamics (AIMD) method of density functional theory (DFT). However, the disadvantages of the AIMD method are that there is a large difference between the experimental values ​​and the predicted values, the predicted solid electrolyte crystal structure is small, the lithium-ion conductivity is calculated at an assumed high temperature instead of room temperature, and the calculation cost is high.

[0008] Another approach is molecular dynamics simulation, which can solve most of the above problems, but has the disadvantage that calculations cannot be performed unless there is a potential energy corresponding to the structure and composition to be calculated.

[0009] The above information disclosed in this Background section is only for enhancement of understanding of the background of the invention and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art. Summary of the invention

[0010] The present disclosure is directed to solving the above-mentioned problems associated with the prior art, and an object of the present disclosure is to provide a method for predicting lithium ion conductivity of solid electrolytes, which uses machine learning to solve the shortcomings of density functional theory (DFT) and molecular dynamics simulation.

[0011] Another object of the present disclosure is to provide a method for quickly and accurately calculating the lithium ion conductivity of a solid electrolyte.

[0012] In an exemplary embodiment, a method for predicting lithium ion conductivity of a solid electrolyte is provided. The method may include: simulating a crystal structure of the solid electrolyte and creating a training set based on the crystal structure for machine learning; using the training set to calculate potential energy specific to the simulated crystal structure through machine learning; and using molecular dynamics simulation to predict the lithium ion conductivity of the solid electrolyte based on the potential energy.

[0013] The training set can be created using the ab initio molecular dynamics (AIMD) method.

[0014] The potential energy specific to the simulated crystal structure may be a moment tensor potential (MTP). The MTP may be used to calculate lithium ion diffusivity, which is then used to predict lithium ion conductivity.

[0015] In one exemplary embodiment, a method for predicting lithium ion conductivity of a solid electrolyte having an argyrodite-type crystal structure and represented as Li 6-a PS 5-a X 1+a , where 0≤a≤1, and X=Cl, Br or I. The method may include: simulating a plurality of compounds, wherein the element X of the plurality of compounds occupies the 4c ​​site contained in the argyrodite-type crystal structure in different proportions; assuming a virtual space comprising n×m×p units, wherein n, m and p are each integers in the range of 1 to 3, and differentially assigning the plurality of compounds to the units based on the thermodynamic stability of the plurality of compounds to simulate a random structure; creating a training set for machine learning based on the random structure; calculating the potential energy specific to the random structure through machine learning using the training set; and predicting the lithium ion conductivity of the solid electrolyte based on the potential energy using molecular dynamics simulation.

[0016] The training set can be created using the ab initio molecular dynamics (AIMD) method.

[0017] The above compound may include: the proportion of element X is 0% and the space group is The first compound; the second compound in which the proportion of element X is 25% and the space group is R3m; the proportion of element X is 50% and the space group is P2 1 22; a third compound in which the ratio of element X is 50% and the space group is P2mm; a fourth compound in which the ratio of element X is 75% and the space group is R3m; and a fifth compound in which the ratio of element X is 100% and the space group is The size of the unit can be about ~Approx.

[0018] When simulating a random structure, multiple compounds can be differentially allocated by positioning any one compound in any unit. When simulating a random structure, among multiple compounds, a compound with higher thermodynamic stability can be allocated to a larger number of units, thereby increasing the proportion of the random structure occupied by the compound.

[0019] The van der Waals (vdW) modified semilocal xc functional (optB88) was used to calculate the potential energy of random structures.

[0020] The potential energy of the random structure can be a moment tensor potential (MTP).

[0021] In an exemplary embodiment, the present disclosure provides a solid electrolyte having an argyrodite-type crystal structure and represented by Li 6-a PS 5-a X 1+a (0≤a≤1, and X=Cl, Br or I).

[0022] In another exemplary embodiment, the present disclosure provides a method for predicting the lithium ion conductivity of a solid electrolyte, comprising simulating a plurality of compounds in which an element X occupies a 4c site contained in an argyrodite-type crystal structure in different proportions, assuming a virtual space comprising n×m×p units, wherein n, m, and p are each integers in the range of 1 to 3, and differentially assigning the plurality of compounds to the units based on the thermodynamic stability of the plurality of compounds to simulate a random structure, calculating a potential energy specific to the random structure; and predicting the lithium ion conductivity of the solid electrolyte based on the potential energy using molecular dynamics simulation.

[0023] In a preferred embodiment, the compound may include: the proportion of element X is 0% and the space group is The first compound has a ratio of 25% of the element X and a space group of R3m, and the second compound has a ratio of 50% of the element X and a space group of P21 22, a third compound in which the proportion of element X is 50% and the space group is P2mm, a fourth compound in which the proportion of element X is 75% and the space group is R3m, and a fifth compound in which the proportion of element X is 100% and the space group is The sixth compound.

[0024] In another preferred embodiment, the size of the unit can be about ~Approx.

[0025] In yet another preferred embodiment, when simulating a random structure, multiple compounds can be differentially distributed by positioning any one compound in any one unit.

[0026] In yet another preferred embodiment, when simulating a random structure, among a plurality of compounds, a compound with higher thermodynamic stability may be distributed to a larger number of units, thereby increasing the proportion of the random structure occupied by the compound.

[0027] In yet another preferred embodiment, when calculating the potential energy of a random structure, a van der Waals (vdW) modified semi-local xc functional (optB88) may be used to calculate the potential energy of the random structure.

[0028] In another preferred embodiment, the potential energy of the random structure may be a moment tensor potential (MTP).

[0029] In another preferred embodiment, when predicting the lithium ion conductivity, the lithium ion conductivity of the solid electrolyte at a temperature of about 300K or higher may be predicted.

[0030] In another preferred embodiment, when predicting lithium ion conductivity, the crystallinity (χ c ) is the lithium ion conductivity of the solid electrolyte of 0.7 to 0.8.

[0031] Other embodiments and preferred embodiments of the present invention are discussed below.

[0032] The above and other features of the present invention are discussed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The above-mentioned and other features of the present disclosure will now be described in detail with reference to certain exemplary embodiments shown in the accompanying drawings, which are given hereinafter by way of illustration only and thus do not limit the present disclosure, and in which:

[0034] Figure 1 showing the argyrodite-type crystal structure of the solid electrolyte according to the present disclosure;

[0035] Figure 2showing the crystal structures of various compounds according to the present disclosure;

[0036] Figure 3 displaying random structures according to the present disclosure; and

[0037] Figure 4 Shown are the results of measuring ion diffusion rate in the process of predicting lithium ion conductivity of a solid electrolyte according to the present disclosure.

[0038] It should be understood that the accompanying drawings are not necessarily drawn to scale, but rather present a somewhat simplified representation of various preferred features that illustrate the basic principles of the invention. The specific design features of the present disclosure disclosed herein, including, for example, specific dimensions, orientations, locations, and shapes, will depend in part on the specific intended application and use environment.

[0039] In the drawings, reference numerals refer to the same or equivalent parts of the present disclosure throughout the several figures of the drawing. DETAILED DESCRIPTION

[0040] The above-mentioned purpose, other purposes, advantages and features of the present disclosure will become apparent from the description of the embodiments given below with reference to the accompanying drawings. However, the present disclosure is not limited to the embodiments disclosed herein and can be implemented in various different forms. The embodiments are provided to make the description of the present disclosure more thorough and to fully convey the scope of the present disclosure to those skilled in the art.

[0041] In the following description of the embodiments, even if depicted in different drawings, the same elements are represented by the same reference numerals. In the accompanying drawings, the size of the structure may be exaggerated than its actual size for clarity of description. In the following description of the embodiments, terms such as "first" and "second" can be used to describe various elements, but these elements are not limited. These terms are only used to distinguish one element from other elements. For example, without departing from the scope and spirit of the present invention, the first element can be named as the second element, and similarly, the second element can be named as the first element. Singular expressions can cover plural expressions unless they have significantly different contextual meanings.

[0042] In the following description of the embodiments, terms such as "comprises", "includes" and "having" should be interpreted as indicating the presence of the features, numbers, steps, operations, elements or components, or combinations thereof described in the description, and do not exclude the presence of one or more other features, numbers, steps, operations, elements, components, or combinations thereof, or the possibility of increasing the same features, numbers, steps, operations, elements, components, or combinations thereof. In addition, it should be understood that when a component (such as a layer, film, region or sheet) is referred to as being "on" another component, the component can be "directly" "on" the other component, or other components can be inserted between the two components. In the same way, it should be understood that when a component (such as a layer, film, region or sheet) is referred to as being "below" another component, the component can be "directly" "below" the other component, or other components can be inserted between the two components.

[0043] All numbers, values ​​and / or expressions used in the specification that represent the amount of components, reaction conditions, polymer compositions and blends are approximate values, which reflect the various uncertainties of measurement generated when these values ​​are obtained from essentially different things, so it should be understood that, unless otherwise stated, they are all modified by the term "about". Unless specifically stated or obvious from the context, the term "about" used herein should be understood to be within the normal tolerance range of the field, such as within 2 standard deviations of the mean value. "About" can be understood as 10%, 9%, 8%, 7%, 6%, 5%, 4%, 3%, 2%, 1%, 0.5%, 0.1%, 0.05% or 0.01% of the value. Unless the context clearly indicates otherwise, all numerical values ​​provided herein are modified by the term "about". In addition, it should be understood that if a numerical range is disclosed in the specification, unless otherwise stated, the range includes all continuous values ​​from the minimum value of the range to the maximum value. In addition, if the range refers to an integer, unless otherwise stated, the range includes all integers from the minimum integer to the maximum integer of the range.

[0044] The terms used herein are only used to describe specific embodiments and are not intended to limit the present disclosure. As used herein, the singular forms "one", "an" and "this" are also intended to include plural forms, unless the context clearly indicates otherwise. These terms are only intended to be used to distinguish one component from another component, and these terms do not limit the nature, order or sequence of the constituent components. It will also be understood that the terms "including" and / or "comprising" used in this specification indicate the presence of the features, integers, steps, operations, elements and / or components described, but do not exclude the presence or addition of one or more of the features, integers, steps, operations, elements, components, and / or combinations thereof. As used herein, the term "and / or" includes any and all combinations of one or more related listed items. Throughout the specification, unless there is a clear contrary description, the term "comprise" and its variations (comprises, comprising) should be understood to include the elements described without excluding any other elements. In addition, the terms "unit", "-er, -or" and "module" described in the specification refer to units for processing at least one function and operation, and can be implemented by hardware components or software components and combinations thereof.

[0045] Although the exemplary embodiments are described as using multiple units to perform the exemplary processes, it should be understood that the exemplary processes can also be performed by one or more modules. In addition, it should be understood that the term controller / control unit refers to a hardware device that includes a memory and a processor, and the hardware device is specifically programmed to perform the processes described herein. The memory is configured to store the modules, and the processor is specifically configured to execute the modules to perform one or more processes described further below.

[0046] In addition, the control logic of the present disclosure may be implemented as a non-transitory computer-readable medium on a computer-readable medium containing executable program instructions executed by a processor, controller, etc. Examples of computer-readable media include, but are not limited to, ROM, RAM, compact disk (CD)-ROM, magnetic tape, floppy disk, flash drive, smart card, and optical data storage device. The computer-readable medium may also be distributed in a network-coupled computer system so that the computer-readable medium is stored and executed in a distributed manner, such as by a telematics server or a controller area network (CAN).

[0047] The method for predicting the lithium ion conductivity of a solid electrolyte according to the present disclosure may be a method for predicting the lithium ion conductivity of a solid electrolyte having an argyrodite-type crystal structure and represented by Li 6-a PS 5- a X1+a , where 0≤a≤1, and X=Cl, Br or I.

[0048] The above method may include simulating multiple compounds, wherein the element X of the multiple compounds occupies the 4c ​​site contained in the argyrodite-type crystal structure in different proportions (S10), assuming a virtual space containing n×m×p units, wherein n, m and p are each integers in the range of 1 to 3, and differentially assigning the multiple compounds to the units based on the thermodynamic stability of the multiple compounds to simulate a random structure (S20), calculating the potential energy of the random structure (S30), and using molecular dynamics simulation to predict the lithium ion conductivity of the solid electrolyte based on the potential energy.

[0049] Each operation will be described in detail below.

[0050] Producing compound (S10)

[0051] Figure 1 The argyrodite crystal structure of the solid electrolyte according to the present disclosure is shown. The argyrodite crystal structure may include a 4b site, a 48h site, a 24g site, a 4c site, and a 4a site. According to the present disclosure, Li 6-a PS 5-a X 1+a The solid electrolyte represented by (0≤a≤1, and X=Cl, Br or I) can be represented as [Li + 6 ][PS 4 ] 3- [S 2- X - ].

[0052] PS with polyhedral shape 4 3- Can be located at site 4b. + All 24g sites and some 48h sites can be occupied, thus forming a cage shape.

[0053] Anion S 2- and X - Each can be located at the 4a site or the 4c ​​site, thereby forming a shape that confines these anions in a cage formed by lithium. The lithium ion conductivity of the solid electrolyte can be determined according to S 2- and X - The positioning of sites 4a and 4c varies.

[0054] First, the basic framework structure can be obtained by forming lithium sites using the enumlib code (http: / / github.com / msg-byu / enumlib). Then, compounds with element X occupying 0%, 25%, 50%, 75% and 100% of the 4c ​​site can be designed. Figure 2 The crystal structures of various compounds according to the present disclosure are shown. Specifically, these compounds may include an element X having a proportion of 0% and a space group of The first compound has a ratio of 25% of the element X and a space group of R3m, and the second compound has a ratio of 50% of the element X and a space group of P2 1 22, a third compound in which the proportion of element X is 50% and the space group is P2mm, a fourth compound in which the proportion of element X is 75% and the space group is R3m, and a fifth compound in which the proportion of element X is 100% and the space group is The sixth compound.

[0055] Generate random structure (S20)

[0056] When a plurality of compounds are generated as described above, in a conventional method for predicting the lithium ion conductivity of a solid electrolyte, the lithium ion conductivity of the plurality of compounds is calculated, and then the lithium ion conductivity of the solid electrolyte in a bulk state is predicted based on the calculation results, so the conventional method is highly scalable, requires a long time, and is inaccurate.

[0057] The present disclosure is dedicated to solving the problems existing in the above-mentioned conventional methods, and is characterized in that a random structure including a plurality of compounds in a specific ratio is generated, the potential energy of the random structure is calculated, and the lithium ion conductivity of the solid electrolyte is predicted based on the calculated potential energy.

[0058] Figure 3 Shows a random structure according to the present disclosure.

[0059] First, a super unit structure including n×m×p units (n, m, and p are each an integer in the range of 1 to 3) can be modeled. Figure 3 A super unit structure including a 3×3×3 arrangement is exemplarily shown.

[0060] The size of the unit is not limited to a specific size and may be, for example, approximately ~Approx.

[0061] The random structure can simulate the actual solid electrolyte in the bulk state. The compound with the most thermodynamically stable crystal structure can occupy the largest part of the solid electrolyte configuration in the bulk state. That is, a plurality of compounds can be differentially assigned to the unit according to the thermodynamic stability of each compound, thereby simulating the actual solid electrolyte in the bulk state as a random structure. Here, only one compound can be assigned to one unit.

[0062] How many units a particular compound is assigned to can be calculated by the following formula.

[0063] Number of sites = P i (E)*n s / P i (E) Sum

[0064] Here, n s can be the number of units that form a random structure. Figure 3 When modeling the random structure shown, n s It can be 27.

[0065] P i (E) may represent the thermodynamic contribution and may be calculated by the following formula.

[0066]

[0067] Here, ΔE may represent the relative thermodynamic stability of each component relative to the most stable component using density functional theory (DFT). k may represent the Boltzmann constant. T may represent temperature.

[0068] Therefore, when simulating random structures, among multiple compounds, compounds with higher thermodynamic stability can be assigned to a larger number of units, thereby increasing the proportion of the compounds occupying the random structure.

[0069] Calculate the potential energy of random structures (S30)

[0070] In order to calculate the lithium ion conductivity of the solid electrolyte using molecular dynamics simulation, the potential energy of the random structure may be required. The potential energy may refer to the interaction force between atoms forming the corresponding solid electrolyte. The potential energy of the random structure may be a moment tensor potential (MTP).

[0071] As functionals for calculating MTP, there are Perdew-Burke-Ernzerhof (PBE) functional (hereinafter referred to as "PBE"), PBE functional with DFT-D3 (hereinafter referred to as "DFT-D3"), van der Waals (vdW) corrected semilocal xc functional (hereinafter referred to as "optB88"), etc. The present disclosure is characterized in that the potential energy of a random structure is calculated using optB88.

[0072] Recently, machine learning potentials are being studied as a way to replace the relatively accurate but computationally expensive density functional theory (DFT). Among the many machine learning potentials such as Neural Network Potential (NNP), Gaussian Approximation Potential (GAP), Spectral Neighbor Analysis Potential (SNAP), etc., the present disclosure is characterized by the use of moment tensor potential (MTP).

[0073] Moment tensor potential (MTP) can refer to the potential energy of a crystal structure in machine learning potential energy. Potential energy can refer to the sum of the energy possessed by each atom taking into account its local atomic environment, and the calculation formula is as follows.

[0074]

[0075] Here, n i The local atomic environment of the ith atom can be represented by the atomic type and position of the ith atom, as well as the atomic type and position of the jth atom around the ith atom.

[0076] V(n i ) can be expressed by the following formula.

[0077]

[0078] Here, α It can be a constant that matches the training process in the Machine-Learning Interatomic Potential (MLIP) software package. α can be the sum of the basis functions of each atom. B α It can be defined by the following formula.

[0079] levM μ,v =2+4μ+v

[0080] Here, μ and ν can be coefficient terms in the machine learning interatomic potential (MLIP) software package. As the values ​​of μ and ν increase, the computational accuracy may improve, but the computational cost may increase significantly. μ,ν can be a moment tensor descriptor and can be expressed by the following formula.

[0081]

[0082] Here, z i and z jcan represent the atomic type of the i-th and j-th atoms, and r ij It can represent the vector from the jth atom to the ith atom. The above formula can be divided into μ The radial part and In the radial part, the result depends on the interatomic distance, which is used to specify the cutoff radius R including the distances of the neighboring atoms based on the i-th atom. cut , and the atomic type can be obtained.

[0083] In the present disclosure, the hyperparameters used for machine learning are set as follows. Specifically, by setting the cutoff radius R cut for lev max The potential energy is fitted with a value of 8, a weight ratio of energy to force of 100:1, and a ratio of training set to validation set of 9:1 to obtain the moment tensor potential (MTP) of the random structure.

[0084] Predicting lithium ion conductivity (S40)

[0085] Molecular dynamics simulations can be performed using the moment tensor potential (MTP) generated in the previous operation. Molecular dynamics simulations can be performed using the Large-scale Atomic / Molecular Massively Parallel Simulator (LAMMPS) software package.

[0086] According to the present disclosure, the lithium ion conductivity of a solid electrolyte at a room temperature of about 300 K can be predicted. Since the lithium ion conductivity of a solid electrolyte at room temperature is low, in conventional methods, the lithium ion conductivity of a solid electrolyte at a high temperature of about 700 K is predicted, but the present disclosure is characterized in that the lithium ion conductivity of a solid electrolyte at a room temperature of about 300 K is predicted by extrapolating the lithium ion conductivity of the solid electrolyte at a high temperature using Arrhenius fitting. Specifically, in the present disclosure, the lithium ion conductivity of a solid electrolyte at a temperature of about 300 K to about 1800 K can be predicted.

[0087] To improve the accuracy of the prediction, the molecular dynamics simulation can be performed at seven temperatures, that is, in the temperature range of about 350 K to about 500 K, at intervals of 25 K. To obtain an average value, the calculation can be repeated twice under the same conditions.

[0088] At temperatures of about 350K, about 375K, etc., the diffusion rate of lithium ions is low, so the migration of lithium ions at this temperature may not be sufficient. At low temperatures, the rate of change of the mean square displacement (MSD) of lithium ions over time is very low. Therefore, when calculating the lithium ion conductivity, the temperature at which the rate of change of the mean square displacement (MSD) of lithium ions over time is low is excluded, and additional temperatures added at intervals of about 25K starting from about 500K can be included to obtain the diffusion rate at higher temperatures. In molecular dynamics simulations, NVT simulations can be performed from about 10K to the target temperature with a time step of 1fs to achieve a system equilibrium of 100ps. Thereafter, NPT simulations of 10ns or more can be performed at the target temperature. The diffusion rate can be obtained from the MSD by two repeated NPT simulations, and the lithium ion conductivity of the solid electrolyte at about 300K can be obtained by fitting the diffusion rate obtained at the corresponding temperature using the Arrhenius equation. The calculation method of lithium ion conductivity will be described in detail below.

[0089] In order to obtain the diffusion rate of lithium ions, the mean square displacement (MSD) representing the particle position deviation (change in time relative to a reference position) should be calculated. The diffusion rate of lithium ions can be calculated from the mean square displacement (MSD) using the following formula when using the Large Atomic / Molecular Massively Parallel Simulator (LAMMPS) software package for calculations.

[0090]

[0091] Here, D may be the diffusion rate of lithium ions. Since the ion diffusion rate in a solid without phase change follows the following Arrhenius equation, the lithium ion diffusion rate at a specific temperature can be calculated. That is, the diffusion rate obtained by simulation at high temperature (usually about 600K to 1200K) can be used to infer the diffusion rate of lithium ions at about 300K through Arrhenius fitting.

[0092]

[0093] Here, D 0 It can represent the diffusion rate of lithium ions at the starting temperature. a It can be the diffusion activation energy. K can be the Boltzmann constant. T can represent the temperature. E a It can be calculated by performing a linear regression on a plot between log D and 1 / T.

[0094] By substituting the above-calculated diffusion rate at about 300K into the following Einstein diffusion equation, the lithium ion conductivity of the solid electrolyte at about 300K can be finally calculated.

[0095]

[0096] Here, p may be the density of diffuse ions in a unit cell, z may be the charge of the diffuse ions, F may be the Faraday constant, and R may be the gas constant.

[0097] The present disclosure will be described in more detail below through the following test examples. The following test examples are only used to exemplarily describe the present disclosure and are not intended to limit the scope of the present invention.

[0098] The results of the comparative analysis of the accuracy of PBE, DFT-D3, and optB88 are described below.

[0099] First, 6 compounds were generated in the same manner as in operation S10. In order to cover a variety of cases in the training set, additional modeling was performed on the structures obtained by applying ±5% strain in all lattice directions to each compound, resulting in a total of 18 training sets. In addition, the temperature for calculating the diffusivity was set to approximately 300K, 600K, 900K, and 1200K, and a total of 7200 training sets were completed by selecting 100 structures at each temperature.

[0100] It has a argyrodite crystal structure and is represented by Li 6 PS 5 The solid electrolyte of X (X is Cl, Br or I) has a structural change depending on the ratio of halogen elements occupying 4a sites and 4c sites. In the ordered structure with halogen elements occupying 0% and 100% of the 4c ​​sites, the solid electrolyte can exhibit very low lithium ion conductivity, while in the disordered structure with halogen elements occupying 25%, 50% and 75% of the 4c ​​sites, the solid electrolyte can exhibit very high lithium ion conductivity. When the potential energies calculated by functionals such as PBE, DFT-D3 and optB88 are reasonable, these potential energies should be able to clearly represent this difference.

[0101] First, a training set with halogen elements occupying 100% and 50% of the 4c ​​site was generated, and the Li 6 PS 5 Lithium ion conductivity σ of Cl structure RT The results are shown in Table 1 and Figure 4 shown.

[0102] [Table 1]

[0103]

[0104] Refer to Table 1 and Figure 4The trend of lithium ion conductivity calculated based on MTP-PBE is consistent with the trend of lithium ion conductivity calculated based on DFT-PBE. However, in the structure where the proportion of halogen elements occupying the 4c ​​site is 100%, the lithium ion conductivity calculated based on MTP-PBE is significantly different from the experimental value (i.e., 1 mS / cm), and is higher than the lithium ion conductivity calculated based on DFT-PBE. Therefore, it can be confirmed that the moment tensor potential (MTP) calculated by PBE is not suitable for predicting the lithium ion conductivity of solid electrolytes with argyrodite-type crystal structure.

[0105] Afterwards, the potential energy was calculated using optB88 and DFT-D3 using the 7200 training sets generated above, and the lithium ion conductivity σ at room temperature (about 300K) was predicted based on the potential energy. RT , and compared with the values ​​calculated by the AIMD method. The prediction results are shown in Table 2 below.

[0106] [Table 2]

[0107]

[0108] In Table 2, E rel represents the relative energy of each structure relative to the most stable structure obtained by comparing the DFT optimized energy and is expressed as the same value as in the AIMD and MTP methods. 块状 It is a value obtained by statistically analyzing the thermodynamic and kinetic contributions of six compounds in which halogen elements occupy the 4c ​​site in different proportions, so that the lithium ion conductivity of the solid electrolyte in the bulk state can be represented by the lithium ion conductivity of each compound. 80％ It is assumed that the crystallinity of the solid electrolyte c The lithium ion conductivity of the solid electrolyte when the crystallinity of the solid electrolyte is 0.8 (80%). In general, assuming that the crystallinity of the solid electrolyte predicted by simulation is 80%, the lithium ion conductivity of the solid electrolyte can be almost consistent with the actual lithium ion conductivity of the solid electrolyte. In the present disclosure, the crystallinity χ c The lithium ion conductivity of the solid electrolyte is 0.7 to 0.8.

[0109] As shown in Table 2, it can be confirmed that the lithium ion conductivity σ of the ordered structure with 0% and 100% halogen element occupying the 4c ​​site calculated based on optB88 is RT The predicted lithium ion conductivity of the ordered structure is very low, i.e., 0.008 and 0.01 mS / cm, which is the same as the result of the conventional DFT-AIMD method. In addition, the lithium ion conductivity σ of the solid electrolyte predicted by optB88 is 80％The result is 2.46 mS / cm, which is almost the same as the reported experimental value. Therefore, it can be confirmed that the use of optB88 for moment tensor potential (MTP) calculations and lithium ion conductivity predictions is very reasonable.

[0110] On the other hand, referring to the results using DFT-D3, the calculated lithium ion conductivity of the ordered structure in which the proportion of halogen elements occupying the 4c ​​site is 0% is very low, but the calculated lithium ion conductivity of the ordered structure in which the proportion of halogen elements occupying the 4c ​​site is 100% is high, i.e., 3.7 mS / cm. Therefore, it can be confirmed that DFT-D3 is not suitable for predicting the lithium ion conductivity of the solid electrolyte according to the present disclosure.

[0111] Using the same method as above, optB88 was used to predict Li 6 PS 5 The lithium ion conductivity of Br is shown in Table 3 below.

[0112] [Table 3]

[0113]

[0114]

[0115] With Li 6 PS 5 Cl is similar to Li 6 PS 5 Calculated lithium ion conductivity σ for ordered structures with 0% and 100% Br and halogen elements occupying 4c sites RT In addition, the lithium ion conductivity σ of the solid electrolyte predicted by optB88 is 80％ is 1.59 mS / cm, which is very similar to the experimental value. Therefore, it can be confirmed that the method using optB88 can be applied to Li 6 PS 5 Br is also very reasonable.

[0116] In addition, optB88 was used to predict Li 6 PS 5 The predicted results are shown in Table 4 below.

[0117] [Table 4]

[0118]

[0119] As we all know, Li 6 PS 5The reason why the lithium ion conductivity of I is very low is that the structure in which the halogen element occupies 0% of the 4c ​​site is thermodynamically stable, so the six structures are not mixed, and only the structure in which the halogen element accounts for 0% exists in the bulk structure. Therefore, the predicted results of the lithium ion conductivity of the structure in which the halogen element accounts for 0% are consistent with the experimental values, and therefore, it can be confirmed that the method using optB88 can be applied to Li 6 PS 5 I is also reasonable.

[0120] Thereafter, random structures were generated using multiple structures, the potential energy of the random structures was calculated according to optB88, and molecular dynamics simulations were used to predict the lithium ion conductivity of the solid electrolyte.

[0121] In Li 6 PS 5 Cl, Li 6 PS 5 Br and Li 6 PS 5 In the solid electrolyte represented by I, the above six compounds are produced by changing the ratio of halogen elements occupying the 4c ​​site. Figure 3 The super unit including 3×3×3 units shown is modeled, and the number of units occupied by each compound in each solid electrolyte is calculated based on the relative structural stability of each compound, as shown in Table 5 below.

[0122] [Table 5]

[0123]

[0124] See Table 5 above, in Li 6 PS 5 Cl, the compound with a halogen element ratio of 25% (R3m) and the compound with a halogen element ratio of 50% (P2 1 22) has a higher thermodynamic contribution P i (E), occupying 7 units out of a total of 27 units respectively, followed by a compound having a halogen element ratio of 50% (P2mm) and a compound having a halogen element ratio of 75% (R3m), occupying 5 units respectively.

[0125] Random structures were generated by assigning individual compounds to cells.

[0126] The moment tensor potential (MTP) of the random structure was calculated according to optB88, and its lithium ion conductivity was calculated by molecular dynamics simulation using the moment tensor potential (MTP). The calculation results are shown in Table 6 below.

[0127] [Table 6]

[0128]

[0129] Referring to Table 6, it can be confirmed that Li 6 PS 5 Cl and Li 6 PS 5 Lithium ion conductivity σ of Br 80％ With the experimental value σ 实验 Very similar. 6 PS 5 Lithium ion conductivity σ of I 80％ Slightly higher than the experimental value σ 实验 The reason for this difference was determined to be that the structure in which the proportion of the halogen element occupying the 4c ​​site is 0% is thermodynamically stable, so these six structures are not actually mixed, and only the structure in which the proportion of the halogen element is 0% exists in the actual structure.

[0130] Lithium ion conductivity σ of random structures calculated by DFT-D3 80％ The value is 0.62 mS / cm, which is significantly different from the experimental value. That is, as described above, it can be confirmed that DFT-D3 is not suitable for predicting the lithium ion conductivity of a solid electrolyte having an argyrodite-type crystal structure.

[0131] As is apparent from the above description, according to the present disclosure, the lithium ion conductivity of a solid electrolyte can be calculated rapidly and accurately.

[0132] The present invention has been described in detail with reference to the preferred embodiments of the present invention. However, it will be appreciated by those skilled in the art that these embodiments may be modified without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A method for predicting the lithium ion conductivity of a solid electrolyte, the method include: Simulating a crystal structure of the solid electrolyte and creating a training set based on the crystal structure for machine learning; using the training set to calculate potential energies specific to the simulated crystal structure by machine learning; and The lithium ion conductivity of the solid electrolyte is predicted based on the potential energy using molecular dynamics simulation.

2. The method of claim 1, wherein the training set is created using an ab initio molecular dynamics (AIMD) method. The method according to claim 1 , wherein the potential energy specific to the simulated crystal structure is a moment tensor potential (MTP).

4. The method of claim 1, wherein the MTP is used to calculate lithium ion diffusion rate, which is then used to predict the lithium ion conductivity.

5. A method for predicting the lithium ion conductivity of a solid electrolyte having an argyrodite-type crystal structure and represented by Li 6-a PS 5-a X 1+a , wherein 0≤a≤1, and X=Cl, Br or I, the method include: Simulating a plurality of compounds, wherein the element X of the plurality of compounds occupies the 4c ​​site contained in the argyrodite-type crystal structure in different proportions; Assuming a virtual space comprising n×m×p units, wherein n, m and p are each integers ranging from 1 to 3, and differentially assigning the plurality of compounds to the units based on the thermodynamic stability of the plurality of compounds to simulate a random structure; Creating a training set based on the random structure for machine learning; calculating potential energy specific to the random structure by machine learning using the training set; and The lithium ion conductivity of the solid electrolyte is predicted based on the potential energy using molecular dynamics simulation.

6. The method of claim 5, wherein the training set is created using an ab initio molecular dynamics (AIMD) method.

7. The method according to claim 5, wherein the compound include: The proportion of element X is 0% and the space group is The first compound of A second compound in which the proportion of element X is 25% and the space group is R3m; The proportion of element X is 50% and the space group is P2 1 The third compound of 22; a fourth compound in which the proportion of element X is 50% and the space group is P2mm; a fifth compound in which the ratio of element X is 75% and the space group is R3m; and The proportion of element X is 100% and the space group is The sixth compound.

8. The method according to claim 5, wherein the size of the unit is 9. The method according to claim 5, in, When simulating the random structure, the plurality of compounds are differentially distributed by positioning any one of the compounds in any one of the units.

10. The method according to claim 5, in, When simulating the random structure, among the plurality of compounds, compounds having higher thermodynamic stability are allocated to a larger number of units, thereby increasing the proportion of the random structure occupied by the compounds.

11. The method according to claim 5, in, When calculating the potential energy of the random structure, the van der Waals vdW-corrected semi-local xc functional optB88 is used to calculate the potential energy of the random structure.

12. The method of claim 5, wherein the potential energy of the random structure is a moment tensor potential (MTP).

13. A method for predicting the lithium ion conductivity of a solid electrolyte having an argyrodite-type crystal structure and represented by Li 6-a PS 5-a X 1+a , wherein 0≤a≤1, and X=Cl, Br or I, the method include: Simulating a plurality of compounds, wherein the element X of the plurality of compounds occupies the 4c ​​site contained in the argyrodite-type crystal structure in different proportions; Assuming a virtual space comprising n×m×p units, wherein n, m and p are each integers ranging from 1 to 3, and differentially assigning the plurality of compounds to the units based on the thermodynamic stability of the plurality of compounds to simulate a random structure; calculating a potential energy specific to the random structure; and The lithium ion conductivity of the solid electrolyte is predicted based on the potential energy using molecular dynamics simulation.

14. The method of claim 13, wherein the compound comprises: The proportion of element X is 0% and the space group is The first compound of A second compound in which the proportion of element X is 25% and the space group is R3m; The proportion of element X is 50% and the space group is P2 1 The third compound of 22; a fourth compound in which the proportion of element X is 50% and the space group is P2mm; a fifth compound in which the ratio of element X is 75% and the space group is R3m; and The proportion of element X is 100% and the space group is The sixth compound.

15. The method according to claim 13, wherein the size of the unit is 16. The method according to claim 13, in, When simulating the random structure, the plurality of compounds are differentially distributed by positioning any one of the compounds in any one of the units.

17. The method according to claim 13, in, When simulating the random structure, among the plurality of compounds, compounds having higher thermodynamic stability are allocated to a larger number of units, thereby increasing the proportion of the random structure occupied by the compounds.

18. The method according to claim 13, in, When calculating the potential energy of the random structure, the van der Waals vdW-corrected semi-local xc functional optB88 is used to calculate the potential energy of the random structure. The method of claim 13 , wherein the potential energy of the random structure is a moment tensor potential (MTP).

20. The method according to claim 13, in, When predicting the lithium ion conductivity, it is predicted that at a temperature of 300K or higher and / or the crystallinity x c The lithium ion conductivity of the solid electrolyte is 0.7 to 0.8.

Images