Functional material designing or screening method and device and storage medium
Through atomic neighbor relationship judgment algorithm and self-supervised machine learning, and the material structure is described in combination with graph theory, the subjectivity and inaccuracy problems when screening battery materials in the existing technology are solved, and the accurate screening and performance improvement of battery materials is achieved. .
Patent Information
- Application Number
- CN202410821083.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-24
- Publication Date
- 2024-11-19
AI Technical Summary
The prior art cannot accurately extract the topological connection relationship between atoms, resulting in subjectivity and inaccuracy when screening potential battery materials similar to known fast ion conductor materials in the material structure database, and cannot meet the growing fast ion conductor materials of battery. R&D needs.
Atomic neighbor relationship judgment algorithm and self-supervised machine learning clustering algorithm are used to describe the material structure through graph theory, extract the unique sub-graph representation of key functional groups, and perform isomorphic bijective matching of sub-graphs, and filter the material database with Materials of the same or similar topological structure.
It realizes accurate extraction of topological connection relationships between atoms and objective and efficient screening of local atomic environments, quickly and accurately obtains potential battery materials, meets the research and development needs of fast ion conductor materials for battery and improves ion migration performance.
Smart Images

Figure CN118983024A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of functional material design or screening, and in particular to a method, device and storage medium for functional material design or screening. Background Art
[0002] Under the data-driven materials research paradigm, analyzing and utilizing large amounts of material data to reveal the underlying connections between material properties provides an efficient and low-cost approach to the design and discovery of new materials. The core advantage of the data-driven paradigm is its ability to process and analyze large data sets obtained from experiments and theoretical predictions, thereby revealing the complex nonlinear relationship between material structure and performance. This methodology is expected to accelerate the discovery of new materials by quickly predicting the properties of unknown materials, opening up new possibilities for the research and application of materials science.
[0003] In the above context, the long-standing relationship between material structure and performance in materials science determines that compounds with structures similar to known fast ion materials are more likely to exhibit high ionic conductivity. Therefore, efficiently extracting the key structural elements and their connection relationships of reference materials and quickly screening them in the material structure database is an effective method to obtain potential battery fast ion conductor materials. At present, traditional material structure feature extraction methods usually consider matching algorithms for structural descriptions such as crystal space symmetry, atomic coordination number and Wyckoff site. This method achieves structural matching by comparing the atomic positions in the two structures atom by atom after standard unit cell transformation, but this global structural comparison cannot accurately describe the similarity of local atomic environments, and the matching process is often subjective.
[0004] To overcome these limitations, it is urgent to develop a method that can accurately extract the topological connection relationship between atoms and objectively and efficiently screen potential battery materials similar to known fast ion conductor materials based on the topological structure characteristics of the local atomic environment, so as to meet the growing needs of research and development of battery fast ion conductor materials. Summary of the invention
[0005] The purpose of this application is to provide a new method, device and storage medium for designing or screening functional materials.
[0006] This application adopts the following technical solutions:
[0007] The first aspect of the present application discloses a method for designing or screening a functional material, comprising the following steps:
[0008] The feature extraction step includes extracting key structural primitives for realizing the specified function of the functional material to be designed, as well as the connection relationship of the structural primitives; wherein the functional material to be designed can select a similar known material as a reference material, extract the key structural primitives for the specified function of the reference material, and the connection relationship of the structural primitives for material design or screening;
[0009] The screening step includes screening materials having the same or similar key structural primitives and structural primitive connection relationships from a material database according to the key structural primitives and structural primitive connection relationships of the functional material to be designed, as the functional material designed or screened in the present application;
[0010] The material database contains a number of known or unknown materials for designing or screening required functional materials; and the description method of the key structural elements and the connection relationship of the structural elements of the functional materials to be designed is the same as the description method of the structural elements and the connection relationship of the structural elements of each material in the material database.
[0011] It should be noted that the functional material design or screening method of the present application can accurately extract the topological connection relationship between atoms, and objectively and efficiently screen the required functional materials based on the topological structure characteristics of the local atomic environment, for example, potential battery materials similar to known fast ion conductor materials, to meet the growing needs of research and development of battery fast ion conductor materials.
[0012] In one implementation of the present application, the key structural elements include key functional groups of the functional materials to be designed.
[0013] In one implementation of the present application, the key functional group uses an atomic neighbor relationship judgment algorithm to obtain the connection relationship between each atom, and represents it through a type of adjacency matrix for the local structure.
[0014] In one implementation of the present application, an adjacency matrix of multiple key functional groups is used to represent key structural primitives and structural primitive connection relationships of the combinatorial design.
[0015] In one implementation of the present application, the algorithm for determining the atomic neighbor relationship includes obtaining the connection relationship determination results of each atom structure by structure based on the atomic intrinsic radius attribute and the self-supervised machine learning clustering algorithm.
[0016] In one implementation of the present application, the functional material is a battery material.
[0017] (1) For inorganic crystal materials:
[0018] The feature extraction step includes extracting a unique subgraph representation of key functional groups of the battery material to be designed described by graph theory;
[0019] The screening step includes constructing a subgraph isomorphic bijective matching to screen potential battery materials with similar overall topological structures from the material database;
[0020] (2) For amorphous materials:
[0021] The feature extraction step includes taking the groups that determine the ion migration performance and the short-range ordered primitives of the amorphous materials as the key functional groups for the battery materials to be designed described by graph theory, and extracting the unique subgraph representation of the key functional groups;
[0022] The screening step includes constructing subgraph isomorphism bijective matching to screen potential battery materials with similar overall topological structures from the material database.
[0023] In one implementation of the present application, subgraph isomorphism bijective matching includes ensuring that the unique subgraph representations of the two structures are in one-to-one correspondence, and that graph nodes and edges between corresponding subgraph representations match each other.
[0024] In one implementation of the present application, the calculation method of the overall topological structure similarity includes:
[0025] (1) For the calculation of the overall topological environment similarity between the atomic structures of different materials, the atomic element information is ignored, and a unique subgraph representation is extracted based on the local structural topological environment corresponding to all atoms, and a subgraph isomorphic bijective matching is performed between the two structures to be calculated;
[0026] (2) Calculate the similarity of the local structural topological environment corresponding to each element in the atomic structure of different materials, extract the corresponding unique subgraph representation according to different element types, and perform isomorphic bijective matching of subgraphs element by element;
[0027] (3) For the calculation of framework topological environment similarity between atomic structures of different materials, some atomic types are excluded, and a unique subgraph representation is extracted based on the remaining atoms, and subgraph isomorphism bijective matching is performed.
[0028] The second aspect of the present application discloses a device for designing or screening functional materials, comprising a feature extraction module and a screening module; wherein the feature extraction module includes a function for extracting key structural primitives and the connection relationship between structural primitives for realizing the specified function of the functional material to be designed; the screening module includes a function for screening materials having the same or similar key structural primitives and the connection relationship between structural primitives from a material database according to the key structural primitives and the connection relationship between structural primitives of the functional material to be designed as the functional material required for the present application; the material database contains a number of known or unknown materials for designing or screening the required functional materials; and the description method of the key structural primitives and the connection relationship between structural primitives of the functional material to be designed is the same as the description method of the structural primitives and the connection relationship between structural primitives of each material in the material database.
[0029] It should be noted that the device for designing or screening functional materials of the present application actually implements each step in the method for designing or screening functional materials of the present application through each module; therefore, the specific limitations of each module can refer to the method of the present application and will not be repeated here.
[0030] The third aspect of the present application discloses a device, which includes a memory and a processor; the memory includes a device for storing programs; the processor includes a device for implementing the method of functional material design or screening of the present application by executing the program stored in the memory.
[0031] The fourth aspect of the present application discloses a computer-readable storage medium, in which a program is stored. The program can be executed by a processor to implement the method for designing or screening functional materials of the present application.
[0032] The fifth aspect of the present application discloses the application of materials obtained by the method for designing or screening functional materials of the present application in the preparation of battery materials.
[0033] In one implementation of the present application, the screened materials are solid electrolyte materials or electrode materials, such as the 104 materials shown in Table 1 of the present embodiment.
[0034] It can be understood that based on the inventive concept of the present application, other fast ion conductors can be obtained by using different reference materials and screening in different material databases, and are not limited to the 104 materials recorded in the embodiments of the present application.
[0035] In one implementation of the present application, the key structural elements of the electrode material are composed of a framework group, an energy storage group, and a transport group.
[0036] In one implementation of the present application, the key structural unit of the solid electrolyte material is composed of a framework group and a transport group.
[0037] In one implementation of the present application, the framework group of the electrode material includes a polyanion group.
[0038] In one implementation of the present application, the polyanion group includes at least one of sulfate, phosphate and arsenate.
[0039] In one implementation of the present application, the framework group of the solid electrolyte material comprises a transition metal ion polyhedral group that does not participate in valence change.
[0040] In one implementation of the present application, the invariant valence transition metal ions include at least one of Mg, Al, Ti, Zr and La.
[0041] In one implementation of the present application, the energy storage group comprises a transition metal ion coordination polyhedron that participates in the valence change.
[0042] In one implementation of the present application, the variable valence transition metal ions include at least one of iron, cobalt, nickel, manganese, copper, tin, titanium, and vanadium.
[0043] In one implementation of the present application, the transport group includes a carrier coordination polyhedron of lithium, sodium or magnesium in the battery material.
[0044] It should be noted that sulfate, phosphate and arsenate are a class of low-cost polyanion groups that can provide good migration channels for ion transport; it is understandable that other ionic groups that can provide good migration channels for ion transport can also be applied to this application, not limited to sulfate, phosphate and arsenate. It is understandable that the polyhedral center transition metal ions that participate in the valence change during the operation of the battery can be used as energy storage groups of electrode materials, and the polyhedral center transition metal ions that do not change the valence can be used as energy storage groups of solid electrolyte materials. The transport performance of carriers in the battery material structure is manifested as the ionic conductivity of the material, which determines the key performance of the battery such as the overall cycle rate and fast charge.
[0045] In one implementation of the present application, the solid electrolyte material is at least one of LiSn2(PO4)3, Li4NbIn3(PO4)6, Li3In2(PO4)3, Li2Mn2(SO4)3, Li5Mn2(PO4)3, Li2Mg2(PO4)3, LiTi2(PO4)3, LiZr2(AsO4)3, Li2 Zr4(PO4)3(AsO4)3, Li3Bi2(PO4)3, Li3Sc2(PO4)3 and Li3 Zr2(PO4)3.
[0046] It can be understood that the series of high-performance functional groups contained in the fast ion conductors and solid electrolyte materials screened and determined in this application can be used for the screening and design of inorganic crystalline materials and amorphous materials, and battery materials such as solid electrolytes and electrodes with specific functions can be customized by combining the above three types of groups.
[0047] In one implementation of the present application, the screened materials are modified by element replacement or element doping to improve the ion migration performance of the battery materials.
[0048] It should be noted that the present application is the first to use the materials obtained by the method of the present application for battery materials, which can improve the ion migration performance of battery materials. In addition, the materials obtained by the screening of the present application can also be element-replaced or element-doped to further improve the ion migration performance of battery materials. It can be understood that the key to the present application is to apply the materials obtained by the screening of the present application to battery materials. The present application is the first to use these materials as battery materials, thereby improving the ion migration performance of battery materials; as for element replacement or element doping, reference can be made to the prior art, and no specific limitation is made here.
[0049] The beneficial effects of this application are:
[0050] The method for designing or screening functional materials of the present application can accurately extract the topological connection relationship between atoms, and objectively and efficiently screen the required functional materials based on the topological structure characteristics of the local atomic environment. When applied to the design and screening of fast ion conductor materials, it can quickly and accurately obtain potential battery materials to meet the growing needs of research and development of battery fast ion conductor materials. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] Figure 1 It is a schematic diagram of converting the reference material LiTi2(PO4)3 into a sub-diagram representation in the embodiment of the present application;
[0052] Figure 2 This is a schematic diagram of the battery material research and development strategy process for inorganic crystal material types in the embodiments of the present application;
[0053] Figure 3 It is the atomic environment and sub-graph representation corresponding to four structural types similar to the LiTi2(PO4)3 reference structure obtained by structural screening in the embodiment of the present application;
[0054] Figure 4 1 is the CI-NEB calculation result of the lithium ion migration channel of structure types I, II and IV in the embodiment of the present application, corresponding to (a), (b) and (c) respectively; (d) is the minimum migration energy barrier of the lithium percolation transmission path;
[0055] Figure 5The lithium ion conductivity and migration energy barrier of the candidate fast ion conductor material in the embodiment of the present application are represented by a frequency distribution histogram to show the data distribution of the lithium ion conductivity and the migration energy barrier;
[0056] Figure 6 It is a potential solid electrolyte material screened and obtained in the examples of this application. DETAILED DESCRIPTION
[0057] The existing traditional material structure feature extraction methods have the limitation that global structure comparison cannot accurately describe the similarity of local atomic environments, and the matching process is often subjective.
[0058] The method for designing or screening functional materials of the present application is based on the basic idea of extracting key structural primitives and their connection relationships to achieve the specified function, and screening functional materials with similar structures to the reference materials based on the basic principle that similar structures determine similar performances. The extracted key functional groups can be used as basic structural fragments for designing functional materials. Among them, the key functional groups can use the atomic neighbor relationship judgment algorithm to obtain the atomic connection relationship of each atom in the structure, and are represented by a type of adjacency matrix for local structures.
[0059] It should be noted that the adjacency matrix representation of the present application can extract key information about the interaction between atoms from the crystal structure, thereby more effectively helping the model learn the relationship between the crystal structure and the material properties. The present application method is based on chemical bond theory, and explicitly describes the interaction between atoms as chemical bonds, including metallic bonds, covalent bonds, ionic bonds, etc., accompanied by the sharing or transfer of electrons between atoms. On this basis, the method can extract the chemical bond connection relationship between atoms, while ignoring some minor information, such as slight atomic position distortion, thereby avoiding the description changes caused by slight symmetry breaking in the traditional crystallographic structure description. The method further abstracts the chemical bond connection relationship through graph theory. When considering the connection relationship between atoms, graph theory is an intuitive representation method. In graph theory, a graph is a set of nodes and edges connecting these nodes. These nodes can represent various entities, while edges represent the relationship or interaction between entities. By treating the atoms in the crystal structure as points and the interaction relationship between atoms as edges, the entire crystal structure can be easily converted into a graph representation.
[0060] The difference from the traditional graph representation is that the graph representation based on the atomic spatial structure needs to consider the periodic boundary conditions of the unit cell. On the basis of the above method, a subgraph representation method based on the local environment of atoms is adopted to deal with periodic boundary conditions. A subgraph consists of a set of nodes of the original graph and a set of edges between these nodes, which allows researchers to focus on a specific part of the graph to analyze and understand the local properties of the graph. In the subgraph representation used in this application, the algorithm first traverses all atoms in the unit cell, obtains the local subgraph representation corresponding to each atom, and uses the graph isomorphism algorithm to compare all subgraphs, excludes duplicate subgraphs according to the graph connection relationship, and obtains a unique subgraph representation of the corresponding structure. These unique subgraph representations are not affected by different unit cell methods, and can be further used as evaluation indicators for structural screening, or as input features of machine learning models. In addition, the scope of the local environment description contained in the subgraph representation during the subgraph construction process is affected by the graph representation distance, where the graph representation distance of 1 corresponds to the nearest neighbor connection relationship of the central atom, and the graph representation distance of 2 corresponds to the next nearest neighbor connection relationship of the central atom, and so on.
[0061] Since the subgraph representation method based on neighbor relationship only relies on the neighbor relationship between atoms to describe the crystal structure, whether the neighbor relationship between atoms can be accurately judged determines the quality of the structural information provided by the structural subgraph representation. Traditional judgment of neighbor relationship between atoms usually relies on threshold method and Voronoi method. The threshold method determines whether atoms are bonded by setting a fixed distance threshold. If the distance between two atoms is less than the threshold, the two atoms are considered to be neighbors. This method is simple and intuitive, but its disadvantage is that the selection of threshold is often empirical and may not be applicable to all types of atomic pairs, especially in the presence of local structural distortion caused by atomic doping or defects. The Voronoi method uses the Voronoi diagram to analyze the polyhedral region of atoms in space. The Voronoi polyhedron of each atom is defined by its neighboring atoms. If the Voronoi polyhedrons of two atoms share a common face, the two atoms are considered to be neighbors. The Voronoi method considers the spatial distribution more finely, but it is too sensitive to slight changes in the position of atoms, resulting in unstable judgment of neighbor relationships. Especially when there are defects in the structure, the simple Voronoi method is prone to mistakenly consider the atoms at both ends of the defect as neighbors. In actual use, both methods are difficult to ensure accurate judgment of the neighbor relationship between atoms. Therefore, it is necessary to develop a new method that can efficiently and accurately judge the connection relationship between atoms.
[0062] In an improved method of the present application, a method for judging the neighbor relationship between atoms based on Voronoi area-distance is proposed, that is, a list of neighbor atoms is adaptively identified according to a self-supervised clustering algorithm. First, in order to avoid the influence of periodic boundary conditions, a Voronoi diagram is generated in a 2×2×2 extended unit cell, and the area of the Voronoi polygons between all atoms is calculated. These areas are normalized by the total area and form data pairs with the corresponding relative atomic distances normalized by the covalent radius. Since the Voronoi method may capture next-nearest neighbor atoms when lithium vacancies are included, the current method uses a Gaussian Mixture Model (GMM) to cluster the above data pairs and classify the data pairs into several clusters in a two-dimensional plane. Among them, the class of atoms with the smallest normalized atomic distance between the cluster center is used as the final list of neighbor atoms. Compared with the traditional K-means model, a circle (or a high-dimensional sphere) is placed in the center of each cluster, and its radius is defined by the point in the cluster that is farthest from the center of the circle (center of the sphere); the Gaussian mixture model is good at handling cluster distributions with non-circular (spherical) distributions through the superposition of multiple Gaussian distributions, and shows better universality in the area-distance classification problem. Furthermore, in order to avoid the influence of the preset number of clusters on the clustering results, the number of classification clusters is iterated from 1 to 3 during the fitting process of the GMM model, and the model with the lowest Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) scores is selected to determine the final number of classification clusters. In general, the above method analyzes each structure independently to ensure the accuracy of its neighbor relationship judgment.
[0063] In a specific application of the present application, it is used for battery material research and development. For known battery fast ion conductor materials used as reference materials, the structural screening method mines the local structural environment in the material database structure that is conducive to ion conduction. The research and development strategy includes:
[0064] (1) For inorganic crystalline material types, based on the unique subgraph representation of key functional groups, a subgraph isomorphic bijective matching is constructed between the reference material and the material to be mined, and a class of potential battery materials with similar overall topological structures to the reference material is screened;
[0065] (2) For amorphous material types, the key functional groups that determine the ion migration performance of the reference material are extracted, and the short-range ordered elements of the amorphous material are designed.
[0066] Taking the research and development strategy of inorganic crystal material types as an example, for two identical crystal structures in a two-dimensional plane using different primitive cell methods, the subgraphs they construct are essentially the same. The subgraph isomorphism comparison algorithm will compare the unique subgraph representations of the two crystal structures one by one, and determine that all unique subgraph representations of the two crystals are isomorphic; for two different crystal structures, the bijective mapping between the corresponding subgraphs of different graph representation distances will indicate the degree of similarity between these structures. In the actual data mining process based on structural subgraph isomorphism matching, subgraph generation, unique subgraph screening, and subgraph bijective matching can be performed in parallel to obtain good structural mining efficiency.
[0067] In another application of the present application, a material design method based on subgraph splicing is disclosed. According to the material properties to which it belongs, a suitable material functional group is selected. The adjacency matrix representation of each functional group can be set as a building block for material design assembly. By reasonably splicing the adjacency matrices of different functional groups, a new material structure can be designed. It should be noted that the splicing process needs to ensure chemical stability and structural continuity, which usually involves the control of the formation and breaking of chemical bonds. The advantage of this method is that it can systematically and accurately control the structure of the material from the atomic level, providing a powerful design tool for the development of new materials with characteristic functions.
[0068] The present application is further described in detail below through specific examples and drawings. The following examples are only used to further illustrate the present application and should not be construed as limiting the present application.
[0069] Example
[0070] In this case, the inorganic crystal material type battery material is used as an example to design and screen functional materials. Specifically, the inorganic crystal material LiTi2(PO4)3 is selected as the reference material to explore potential fast ion conductor materials. LiTi2(PO4)3 has a NASICON structure, which is a fast ion conductor material with good ion migration performance. In this material structure, the coordination environment framework composed of octahedron (Oct) and tetrahedron (Tet) is crucial for constructing the lithium migration path. Convert the reference material LiTi2(PO4)3 into a schematic diagram represented by a subgraph, as shown Figure 1 As shown. The schematic diagram of the battery material R&D strategy process of inorganic crystal material type is as follows Figure 2 shown.
[0071] Using the above unique subgraph representation, a search was performed in the inorganic crystal material database consisting of several material databases including Inorganic Crystal Structure Database (ICSD), Open Quantum Materials Database (OQMD), Materials Project Database, Crystallography Open Database and AFLOW, and a subgraph isomorphic bijective matching was constructed to screen potential battery materials with similar overall topological structures from the material database.
[0072] The core goal of subgraph isomorphism bijective matching is to ensure that the unique subgraph representations of the two structures are one-to-one corresponding, and that the graph nodes and edges between the corresponding subgraph representations match each other. Specifically, it includes:
[0073] (1) For the calculation of the overall topological environment similarity between the atomic structures of different materials, the atomic element information is ignored, and a unique subgraph representation is extracted based on the local structural topological environment corresponding to all atoms, and a subgraph isomorphic bijective matching is performed between the two structures to be calculated;
[0074] (2) Calculate the similarity of the local structural topological environment corresponding to each element in the atomic structure of different materials, extract the corresponding unique subgraph representation according to different element types, and perform isomorphic bijective matching of subgraphs element by element;
[0075] (3) For the calculation of framework topological environment similarity between atomic structures of different materials, some atomic types are excluded, and a unique subgraph representation is extracted based on the remaining atoms, and subgraph isomorphism bijective matching is performed.
[0076] First-principles calculations further demonstrate that they have similar lithium migration activation barriers to LiTi2(PO4)3. Detailed comparison of lithium migration pathways between these different structural types shows that the octahedral-octahedral (Oct-Oct) migration pathway can achieve a lower lithium migration barrier than the NASICON reference phase. This has identified a series of potential solid electrolyte materials with room temperature ionic conductivity of up to 10 -2 S / cm level, and the electrochemical stability window range exceeds 2V.
[0077] Through the subgraph isomorphism algorithm, four structural types highly similar to LiTi2(PO4)3 were found in this example. The results are as follows Figure 3As shown. For the convenience of analysis, only the subgraph with Ti as the central node is described in the figure. Structural type I, corresponding to the P21 / c space group, has a [MO6]2[XO4]3 structure (M represents a metal atom and X represents a main group element) consistent with the NASICON reference phase, but exhibits a different arrangement in the c-axis direction. The NASICON reference phase exhibits a linear arrangement, while structural type I exhibits a staggered arrangement. This structural type has been reported in previous studies to have a stable framework and good ion mobility, so it can be used as a battery positive electrode and solid electrolyte material. Structural type II belongs to the Pbca space group, which has two different [MO6]2[XO4]3 arrangements, one corresponding to the reference phase and the other corresponding to structural type I. Structural type III belongs to the Ia3d space group, which is consistent with Li3Ln3Te2O 12 and Li5Ln3Nb2O 12 The structures of garnet-type solid electrolytes are very similar. In these structures, lithium ions migrate through Oct and Tet sites in sequence, forming a flat energy landscape and showing good ionic conductivity. Structure type III is similar to Li3Ln3Te2O 12 It has the same space group, but lithium occupies different sites, forming a spiral ion transport channel in the structural framework. Structure type IV belongs to the C2 / c space group, which has the structural characteristics of a tubular channel, and the lithium ions are in the tubular channel that provides a migration path. It should be noted that since structure type II is actually an intermediate phase between the reference phase and structure type I, this means that the lithium coordination environment in structure type II is similar to that in the reference phase and structure type I, and the lithium migration channel is mainly provided by the two original phases. Therefore, structure type II is no longer considered in the following discussion.
[0078] In order to fairly compare the lithium migration capabilities of these five different structural types, density functional theory calculations were performed using the Climbing Image Nudged Elastic Band (CI-NEB) method to determine the corresponding lithium migration energy barriers. The CI-NEB method is an efficient technique for searching for the minimum energy path of the reaction path, which is used to determine the most likely transition state between the initial state and the final state of the system. Specifically, the transition state is first determined by constructing a series of intermediate configurations between the initial state and the final state (through methods such as linear interpolation), and then adding spring forces between adjacent configurations to ensure the continuity of the path, thereby simulating an elastic band connecting the initial and final states. Afterwards, the force minimization relaxation of the elastic band and the configuration makes these configurations converge to the minimum energy path. In addition, in order to avoid the conflict between the real force and the spring force during the configuration optimization process, the NEB method only considers the projection component of the real force perpendicular to the energy path and the projection component of the spring force parallel to the energy path during the calculation process. After several iterations of the conventional NEB method, the CI-NEB method modified the force on the intermediate configuration with the highest energy and no longer relied on elastic force, causing the intermediate configuration with the highest energy to converge to the saddle point in the energy path. The CI-NEB calculation process used the same chemical composition as the LiTi2(PO4)3 structure, thus ensuring that the comparison results would not be affected by differences in the types of constituent elements. The CI-NEB calculation results of the lithium ion migration channel are shown in Figure 2. Figure 4 shown.
[0079] Figure 4In the figure, the CI-NEB calculation results of lithium ion migration channels of I, II and IV are shown in (a), (b) and (c), respectively, and (d) shows the minimum migration energy barrier of lithium percolation transport path. All structural types present three-dimensional ion channels, and some lithium ion migration paths that run through the entire unit cell are composed of multiple interconnected channels, forming a complex lithium ion percolation network. For the type I structure, lithium ions are located in the 8d Wyckoff site with inversion symmetry, where the occupied sites are represented by light green and the unoccupied 8d sites are represented by dark green. As the lithium content increases, the remaining sites will be occupied and new metastable sites will appear, which may trigger the coordinated movement of the metastable sites with the remaining stable sites, thereby obtaining a lower lithium ion migration energy barrier. The type I structure has lithium ion migration channels along each axis. The lithium ion migration channels marked in red constitute a complete lithium ion migration path along a single unit cell axis, showing the lowest migration energy barrier, about 0.241eV. The channel marked in yellow constitutes another complete path with the highest ion migration energy barrier (about 0.618eV). Another dark blue marked channel connects with the above two channels to form a complete ion migration path, and the energy barrier of this channel is about 0.538eV. Structural type II is an intermediate phase between the reference phase and structural type I. It shares the ion migration channels of the two structural types and generates new channels at the junction of the two phases. The calculation results show that this channel has a high energy barrier of about 0.765eV, which makes it difficult for lithium ions to migrate along this channel. In the type III structure, lithium ions occupy the inversion symmetric 24c Wyckoff site, and a spiral ion migration path is formed along the axial direction by 4 identical channels, and its energy barrier value is about 0.140eV. However, lithium ions only occupy some sites marked in light green in the current LiTi2(PO4)3 composition, and the remaining sites marked in dark green can be further filled with lithium ions to form homogeneous paths orthogonal to the current ion migration path. These paths form a three-dimensional percolation network. The type IV structure contains two lithium ion migration paths, both of which show low energy barrier values. Among them, the path marked in red passes through the tubular pores in the structure and has the lowest energy barrier, which is about 0.068eV. At the same time, the path marked in dark blue allows ions to migrate between different tubular channels. It should be noted that both paths are composed of two channels, corresponding to different channel environments. In summary, the activation energy barriers for the migration of lithium ions along the lowest energy barrier path in type I to IV structures are all in the range of 0.1 to 0.4eV, which is comparable to the reference phase (0.39eV). This confirms that similar local atomic topological environments may lead to similar lithium ion migration energy barriers.
[0080] This example further screened a large database of crystal structures consisting of 544,145 inorganic compounds to find materials belonging to structural types I, III and IV to identify candidate materials suitable as solid-state electrolytes. Such materials should contain lithium, and the study was limited to compounds with less than 200 atoms in the unit cell to avoid excessive density functional calculation burden. In addition, compounds containing hydrogen atoms were ignored because database entries for these materials are usually difficult to contain accurate hydrogen coordinate information (some of which may even be missing). The above criteria further narrowed the range of candidate compounds (29,330 structures). Subsequently, a subgraph isomorphism matching algorithm was used for screening, and all duplicate entries were removed through the graph matching scheme proposed in previous work, and 104 materials were finally obtained, as shown in Table 1.
[0081] Table 1 Structural components of candidate materials corresponding to the four structural types screened from the database
[0082]
[0083] The lithium ion conductivity and migration energy barrier of these materials were simulated using Ab Initio Molecular Dynamics (AIMD). AIMD is a computational method that studies the time evolution of materials and molecular systems through numerical simulation. It simulates the motion of atoms and molecules based on Newton's second law. In AIMD simulations, the acceleration of each atom in the system is calculated based on the resultant force it is subject to, which is generated by the interaction between atoms. The steps of AIMD simulation usually include initializing the atomic positions and velocities of the system, and then iteratively calculating the position and velocity of each atom in a series of discrete time steps. This involves solving Newton's equations of motion, usually using numerical integration methods such as the Verlet algorithm and its variants. In each time step, the acceleration is recalculated to ensure an accurate description of the interaction forces. In this way, AIMD simulations can provide a dynamic trajectory of the system's evolution over time, allowing researchers to observe changes in macroscopic physical properties such as temperature, pressure, and energy distribution, as well as structural changes at the atomic and molecular levels. AIMD calculation results are shown in Figure 1. Figure 5 As shown, it shows that a large number of compounds have higher ionic conductivity than the LiTi2(PO4)3 reference phase, and more than three-quarters of the compounds show lithium migration barriers less than 0.5 eV. These results demonstrate the reliability of the subgraph isomorphism method to identify potential fast ion conductors.
[0084] In this example, the method for developing battery materials based on subgraph isomorphic matching uses an atomic neighbor relationship judgment algorithm to obtain the atomic connection relationship of the material structure, extracts the key local structural topological environment through subgraph representation, and uses subgraph isomorphic bijective matching to screen candidate materials with similar structures. The extracted local structural topological environment can be used as a key structural group for the design of functional materials. Through CI-NEB calculations and AIMD simulation results, it is verified that the candidate materials screened generally have good ionic conductivity, which illustrates the effectiveness and accuracy of the above-mentioned method for developing battery materials based on subgraph isomorphic matching.
[0085] When pursuing high ionic conductivity of battery materials, other basic properties should be further considered. For example, for solid electrolyte materials, they should have lower electronic conductivity and a wider material electrochemical stability window. The material electrochemical stability window refers to the redox potential range between the top of the valence band and the bottom of the conduction band of the material. The operating voltage of the solid electrolyte material needs to be within the electrochemical window to avoid irreversible chemical reactions. This example sets a standard that the electronic band gap should exceed 2eV to ensure that no electron transfer occurs in the material. All compounds belonging to structural types III and IV are filtered out because their band gaps are almost close to zero. In addition, it is stipulated that the electrochemical stability window should exceed 0.5V. Figure 6 As shown, 13 of the 104 candidate compounds meet the above criteria at the same time, namely LiSn2(PO4)3, Li4NbIn3(PO4)6, Li3In2(PO4)3, Li2Mn2(SO4)3, Li5Mn2(PO4)3, Li2Mg2(PO4)3, LiTi2(PO4)3, LiZr2(AsO4)3, Li2 Zr4(PO4)3(AsO4)3, Li3Bi2(PO4)3, Li3Sc2(PO4)3 and Li3 Zr2(PO4)3. Figure 6 As shown, these 13 compounds exhibit good electrochemical stability, low lithium migration energy barriers and high lithium ion conductivity. Figure 6 In the figure, lithium ion conductivity is represented by the color and size of the circle. It is noteworthy that LiZr2(AsO4)3 shows the widest electrochemical stability window and a lithium ion migration energy barrier comparable to that of the LiTi2(PO4)3 reference phase. The calculation results of Li2Zr4(PO4)3(AsO4)3 show that partial replacement of AsO4 with PO4 can further improve the ionic conductivity of the LiZr2(AsO4)3 structure. Li2Mg2(SO4)3 is the most cost-effective candidate material, which has both an appropriate electrochemical stability window and competitive ionic conductivity. The Li3Bi2(PO4)3 material exhibits the highest room-temperature extrapolated ionic conductivity (52mS / cm), and lithium ions migrate in the three-dimensional percolation network provided by the framework structure.
[0086] Unlike the early methods of screening battery materials using global factors such as symmetry or empirical key local atomic environments (such as the coordination environment of lithium), the structural screening method included in this application can quickly screen ideal ion transport topological frameworks. Through subgraph isomorphism matching, crystal structures with topological frameworks similar to reference materials can be identified at a very low computational cost. Another advantage of the structural screening method included in this application is that it can be easily used for a wide range of battery reference materials. It is expected that when the structural screening method is used for other reference materials, such as garnet-type oxides, and some specific sulfide / halide solid electrolyte materials reported in recent years, more high ionic conductivity battery candidate materials can be found. It should be noted that component design can further improve the ionic conductivity of candidate materials because the structural screening method included in this application emphasizes the overall topological structure rather than the specific material composition. Specifically, lithium filling achieved by element doping is an effective component design strategy, which will cause the enhancement of Coulomb repulsion between ions, thereby activating the coordinated transmission of multiple lithium ions and significantly reducing the migration energy barrier of lithium ions. For a class of candidate materials with a higher lithium content, further introducing lithium vacancies in the framework structure by element doping, reducing the lithium vacancy formation energy, and improving the ion migration performance of the candidate materials. For example, through this composition design strategy, a class of LiTi2(PO4)3 obtained by aluminum doping based on the reference material LiTi2(PO4)3 1+x Al x Ti 2-x The (PO4)3 material has taken its ionic conductivity to a new level, verifying the effectiveness of the above-mentioned composition design strategy.
[0087] The above contents are further detailed descriptions of the present application in combination with specific implementation methods, and it cannot be determined that the specific implementation of the present application is limited to these descriptions. For ordinary technicians in the technical field to which the present application belongs, several simple deductions or substitutions can be made without departing from the concept of the present application, which should be deemed to fall within the scope of protection of the present application.
Claims
1. A method for designing or screening functional materials, characterized in that: The following steps are included: The feature extraction step includes extracting key structural primitives for realizing the specified functions of the functional material to be designed, as well as the connection relationship between the structural primitives; The screening step includes screening materials having the same or similar key structural primitives and structural primitive connection relationships from a material database as the functional material to be designed according to the key structural primitives and structural primitive connection relationships; The material database contains a number of known or unknown materials for designing or screening required functional materials; Furthermore, the description method of the key structural primitives and the connection relationship between the structural primitives of the functional material to be designed is the same as the description method of the structural primitives and the connection relationship between the structural primitives of each material in the material database.
2. The method according to claim 1, characterized in that: The key structural elements include key functional groups of the functional materials to be designed; Preferably, the key functional groups use an atom neighbor relationship judgment algorithm to obtain the connection relationship of each atom, and are represented by an adjacency matrix for a local structure; Preferably, the adjacency matrix of the plurality of key functional groups is used to represent the key structural primitives and structural primitive connection relationships of the combinatorial design; Preferably, the algorithm for determining the atomic neighbor relationship includes obtaining the connection relationship determination results of each atom structure by structure based on the atomic intrinsic radius attribute and a self-supervised machine learning clustering algorithm.
3. The method according to claim 1 or 2, characterized in that: The functional material is a battery material, (1) For inorganic crystalline materials: The feature extraction step includes extracting a unique subgraph representation of key functional groups of the battery material to be designed described by graph theory; The screening step includes constructing a subgraph isomorphic bijective matching to screen potential battery materials with similar overall topological structures from the material database; (2) For amorphous materials: The feature extraction step includes taking the groups that determine the ion migration performance and the short-range ordered primitives of the amorphous materials as the key functional groups for the battery materials to be designed described by graph theory, and extracting the unique subgraph representation of the key functional groups; The screening step includes constructing a subgraph isomorphic bijective matching to screen potential battery materials with similar overall topological structures from the material database; Preferably, the subgraph isomorphism bijective matching includes ensuring that the unique subgraph representations of the two structures are in one-to-one correspondence, and that the graph nodes and edges between the corresponding subgraph representations match each other; Preferably, the calculation method of the overall topological structure similarity includes: (1) For the calculation of the overall topological environment similarity between the atomic structures of different materials, the atomic element information is ignored, and a unique subgraph representation is extracted based on the local structural topological environment corresponding to all atoms, and a subgraph isomorphic bijective matching is performed between the two structures to be calculated; (2) Calculate the similarity of the local structural topological environment corresponding to each element in the atomic structure of different materials, extract the corresponding unique subgraph representation according to different element types, and perform isomorphic bijective matching of subgraphs element by element; (3) For the calculation of framework topological environment similarity between atomic structures of different materials, some atomic types are excluded, and a unique subgraph representation is extracted based on the remaining atoms, and subgraph isomorphism bijective matching is performed.
4. A device for designing or screening functional materials, characterized in that: Including feature extraction module and screening module, The feature extraction module includes a module for extracting key structural primitives for realizing the specified functions of the functional material to be designed, as well as the connection relationship between the structural primitives; The screening module includes a method for screening materials having the same or similar key structural primitives and structural primitive connection relationships from a material database as the functional material according to the key structural primitives and structural primitive connection relationships of the functional material to be designed as needed; The material database contains a number of known or unknown materials for designing or screening required functional materials; Furthermore, the description method of the key structural primitives and the connection relationship between the structural primitives of the functional material to be designed is the same as the description method of the structural primitives and the connection relationship between the structural primitives of each material in the material database.
5. The device according to claim 4, characterized in that: The key structural elements include key functional groups of the functional materials to be designed; Preferably, the key functional groups use an atom neighbor relationship judgment algorithm to obtain the connection relationship of each atom, and are represented by an adjacency matrix for a local structure; Preferably, the adjacency matrix of the plurality of key functional groups is used to represent the key structural primitives and structural primitive connection relationships of the combinatorial design; Preferably, the algorithm for determining the atomic neighbor relationship includes obtaining the connection relationship determination results of each atom structure by structure based on the atomic intrinsic radius attribute and a self-supervised machine learning clustering algorithm.
6. The device according to claim 4 or 5, characterized in that: The functional material is a battery material, (1) For inorganic crystal materials: The feature extraction module includes a unique subgraph representation for extracting key functional groups of battery materials to be designed described in a graph theory manner; The screening module includes a module for constructing a subgraph isomorphic bijective matching to screen potential battery materials with similar overall topological structures from a material database; (2) For amorphous materials: The feature extraction module includes extracting a unique subgraph representation of the key functional group from the battery material to be designed described by graph theory, taking the group that determines the ion migration performance and the short-range ordered primitive of the amorphous material as the key functional group; The screening module includes constructing a subgraph isomorphic bijective matching to screen potential battery materials with similar overall topological structures from a material database; Preferably, the subgraph isomorphism bijective matching includes ensuring that the unique subgraph representations of the two structures are in one-to-one correspondence, and that the graph nodes and edges between the corresponding subgraph representations match each other; Preferably, the calculation method of the overall topological structure similarity includes: (1) For the calculation of the overall topological environment similarity between the atomic structures of different materials, the atomic element information is ignored, and a unique subgraph representation is extracted based on the local structural topological environment corresponding to all atoms, and a subgraph isomorphic bijective matching is performed between the two structures to be calculated; (2) Calculate the similarity of the local structural topological environment corresponding to each element in the atomic structure of different materials, extract the corresponding unique subgraph representation according to different element types, and perform isomorphic bijective matching of subgraphs element by element; (3) For the calculation of framework topological environment similarity between atomic structures of different materials, some atomic types are excluded, and a unique subgraph representation is extracted based on the remaining atoms, and subgraph isomorphism bijective matching is performed.
7. A device for designing or screening functional materials, characterized in that: The device includes a memory and a processor; The memory includes a memory for storing programs; The processor includes a method for implementing the functional material design or screening method according to any one of claims 1 to 3 by executing the program stored in the memory.
8. A computer-readable storage medium, characterized in that: The storage medium stores a program, which can be executed by a processor to implement the method for designing or screening functional materials according to any one of claims 1 to 3.
9. Use of the material obtained by the method for designing or screening functional materials according to any one of claims 1 to 3 in the preparation of battery materials.
10. The use according to claim 9, characterized in that: The screened materials are solid electrolyte materials or electrode materials; Preferably, the key structural elements of the electrode material are composed of a framework group, an energy storage group, and a transport group; Preferably, the key structural unit of the solid electrolyte material consists of a framework group and a transport group; Preferably, the framework group of the electrode material includes a polyanion group; Preferably, the polyanion group includes at least one of sulfate, phosphate and arsenate; Preferably, the framework group of the solid electrolyte material comprises a transition metal ion polyhedral group that does not participate in the valence change; Preferably, the invariant transition metal ion comprises at least one of Mg, Al, Ti, Zr and La; Preferably, the energy storage group comprises a transition metal ion coordination polyhedron that participates in the valence change; Preferably, the variable valence transition metal ion includes at least one of iron, cobalt, nickel, manganese, copper, tin, titanium and vanadium; Preferably, the transport group includes a carrier coordination polyhedron of lithium, sodium or magnesium in the battery material; Preferably, the solid electrolyte material is at least one of LiSn2(PO4)3, Li4NbIn3(PO4)6, Li3In2(PO4)3, Li2Mn2(SO4)3, Li5Mn2(PO4)3, Li2Mg2(PO4)3, LiTi2(PO4)3, LiZr2(AsO4)3, Li2 Zr4(PO4)3(AsO4)3, Li3Bi2(PO4)3, Li3Sc2(PO4)3 and Li3 Zr2(PO4)3; Preferably, the screened materials are modified by element replacement or element doping to improve the ion migration performance of the battery materials.
Citation Information
Cited By
Method and device for predicting physical properties of electrolyte based on attention mechanism
CN120473021A
A method and device for predicting physical properties of electrolyte based on attention mechanism
CN120473021B
