Crystal structure prediction method and system based on mixed potential energy deep learning model

Abstract

The invention discloses a crystal structure prediction method and system based on a mixed potential energy deep learning model, and relates to the field of material structures, and the method comprises the steps: inputting basic information of a to-be-predicted material, obtaining a reasonable evaluation result, and obtaining a structure feature value of each crystal structure; performing corresponding feature learning on a local environment between atoms of each crystal structure, and outputting initial energy of each crystal structure; and obtaining each stable crystal structure of the material to be predicted, calculating short-distance interaction information between atoms, and processing to obtain the empirical potential of each stable crystal structure. According to the method, the local structure and interaction between atoms are accurately described by combining the deep learning model MEGNet and the Buckingham potential, so that the precision of energy prediction of the crystal structure is remarkably improved. The search process of the crystal structure is optimized, the possible local minimum problem is avoided, and it is ensured that a globally optimal or nearly optimal stable structure is found.

Description

Technical Field

[0001] The present invention relates to the field of material structure, and specifically to a crystal structure prediction method and system based on a hybrid potential energy deep learning model. Background Art

[0002] Accurate prediction of crystal structure is one of the core tasks in material design and new material discovery, especially when looking for new materials that are stable and have excellent performance. The crystal structure of a material determines its physical, chemical, and mechanical properties. Therefore, accurate prediction of the crystal structure can not only accelerate the material screening process, but also provide a theoretical basis for the practical application of the material. Traditional computational methods, such as density functional theory (DFT), can provide very high computational accuracy due to its rigorous quantum mechanical foundation, and are therefore widely used in energy calculations and structural optimization in materials science. However, the DFT method has extremely high computational overhead when dealing with large-scale material systems, especially when conducting large-scale material screening. Its computational cost and time cost greatly limit the scope of application of this method.

[0003] In order to solve this problem, in recent years, data-driven machine learning methods, especially graph neural networks, have gradually emerged in the field of materials science. Compared with traditional physical model-based methods, GNN methods can effectively capture the local environmental characteristics of crystal structures by simulating the complex relationships between atoms, thereby predicting the properties of materials. Especially when dealing with molecules and crystal systems with highly non-Euclidean structures, graph neural networks can model directly in graph structures, which gives them significant advantages in molecular and crystal structure prediction.

[0004] The prior art, such as the invention patent with announcement number: CN117393089B, is a crystal evolution simulation method based on a single-mode Bessel crystal phase field model, which relates to the field of phase field model technology and includes the following steps: constructing a single-mode Bessel crystal phase field model through Bessel function and classical density functional model; substituting density distribution functions of different phases into the single-mode Bessel crystal phase field model respectively, solving parameters at the minimum free energy of different phases; constructing a crystal phase diagram according to parameters at the minimum free energy of different phases through the common tangent method; discretizing the crystal phase field evolution equation through the Fourier spectrum method to obtain the crystal phase field iteration equation; simulating the crystal evolution process according to the crystal phase diagram and the crystal phase field iteration equation. Based on the Bessel function fitting, a two-point correlation function is obtained and based on this, combined with the classical density functional model, a new single-mode crystal phase field model is constructed.

[0005] The prior art, such as the invention patent with announcement number: CN112466411B, is a crystal structure prediction method based on crystal topology theory, which involves the field of material structure technology. The method includes: using crystal topology methods to systematically and comprehensively search for energy-stable crystal structures; obtaining the mechanical stability of the research object; obtaining the mechanical properties of the research object. The structural characteristics of the research object are analyzed, and crystal structures of different topological types are obtained. The results of the crystal structure performance analysis of the carbon system are obtained based on the energy stability, mechanical stability, and mechanical properties of the research object.

[0006] Based on the above scheme, it can be seen that the existing technology often uses models or algorithms that consume a lot of computing power when analyzing and simulating crystal structures, and is only applicable to the phase change process of specific materials, lacking a certain degree of versatility. When dealing with complex interactions between atoms, traditional machine learning methods usually simplify the modeling of these interactions, which leads to low prediction accuracy of machine learning methods in the process of energy prediction and structural optimization. Especially in the modeling of short-range interactions, it may not be able to accurately reflect the repulsive and attractive forces between atoms, affecting the prediction accuracy of the overall energy. At the same time, the existing technology pays more attention to the geometric structure of the crystal and ignores the influence of the electronic structure of the crystal. Therefore, there is a problem of inaccurate prediction in the current crystal structure prediction due to lack of in-depth consideration. Summary of the invention

[0007] In view of the deficiencies of the prior art, the present invention provides a crystal structure prediction method and system based on a hybrid potential energy deep learning model. To achieve the above objectives, the present invention is implemented through the following technical solutions: A crystal structure prediction method based on a hybrid potential energy deep learning model, comprising: Input the basic information of the material to be predicted into the structure prediction platform, generate the basic information characteristic values ​​of the material to be predicted and conduct a rationality assessment. After obtaining the rationality assessment results, based on the basic information characteristic values ​​of the material to be predicted, retrieve the various crystal structure data sets that meet the material to be predicted from the structure prediction platform and merge them to obtain the structural characteristic values ​​of each crystal structure.

[0008] Based on the structural eigenvalues ​​of each crystal structure, the corresponding feature learning is performed on the local environment between atoms of each crystal structure, and the preliminary energy of each crystal structure is output, recorded as the neural network potential of each crystal structure. The neural network potential evaluation results of each crystal structure are obtained by analysis and processing.

[0009] Based on the neural network potential evaluation results of each crystal structure, the stable crystal structures of the material to be predicted are screened and the short-range interaction information between atoms of each stable crystal structure is calculated. The empirical potential of each stable crystal structure is obtained by cumulative processing.

[0010] The neural network potential and empirical potential of each stable crystal structure are comprehensively analyzed and processed to obtain the total energy potential of each stable crystal structure, which is then imported into the optimization algorithm for iterative update. The stable crystal structure with the lowest total energy potential is output as the target crystal structure of the material to be predicted.

[0011] As a preferred technical solution, the basic information of the material to be predicted is input into the structure prediction platform, the basic information characteristic value of the material to be predicted is generated and the rationality evaluation is performed. The specific process is as follows: The basic information of the material to be predicted includes the total number of atomic types of the material to be predicted, the number of atoms of each type, the charge number of each type of atoms, the total number of chemical bond types, the number of chemical bonds, the bond lengths of each type of chemical bonds, and the bond angles of each type of chemical bonds.

[0012] The reference bond angles and reference bond lengths corresponding to various types of chemical bonds are extracted from the structure prediction platform, and are jointly analyzed and processed with the basic information of the material to be predicted to generate the basic information characteristic value of the material to be predicted, and compared with the rationality assessment threshold to obtain the rationality assessment result of the material to be predicted. If the basic information characteristic value of the material to be predicted is greater than or equal to the rationality assessment threshold, the rationality assessment result of the material to be predicted is determined to be reasonable; if the basic information characteristic value of the material to be predicted is less than the rationality assessment threshold, the rationality assessment result of the material to be predicted is determined to be unreasonable, and a reminder message is sent to the structure prediction platform.

[0013] As a preferred technical solution, based on the basic information characteristic value of the material to be predicted, the crystal structure data sets that meet the material to be predicted are retrieved from the structure prediction platform and merged to obtain the structural characteristic value of each crystal structure, specifically including: Based on the basic information characteristic values ​​of the material to be predicted, mapping and matching are performed with the crystal structure data sets corresponding to the basic information characteristic value intervals pre-stored in the structure prediction platform to obtain the crystal structure data sets of the material to be predicted. The crystal structure data sets of the material to be predicted include the lengths of the sides of the unit cell, the unit cell angles, the space group symbols and the atomic coordinates of each crystal structure.

[0014] Based on the space group symbol of each crystal structure, a mapping match is performed with the symmetry coefficient corresponding to each space group symbol pre-stored in the structure prediction platform to obtain the symmetry coefficient of each crystal structure, and the symmetry coefficient is used to characterize the space group symmetry characteristics of the crystal structure.

[0015] Based on the coordinates of each atom in each crystal structure, the correlation coefficient of each crystal structure is obtained through analysis and processing, and the correlation coefficient of each crystal structure is used to characterize the order degree of each crystal structure.

[0016] The lengths of the sides of the unit cells of each crystal structure, the unit cell angles, the symmetry coefficients and the correlation coefficients are processed and analyzed to obtain the structural characteristic values ​​of each crystal structure, and the structural characteristic values ​​are used to describe the crystal structure characteristics.

[0017] As a preferred technical solution, the corresponding feature learning of the local environment between atoms in each crystal structure specifically includes: Based on the structural characteristic values ​​of each crystal structure, mapping and matching are performed with the local segmentation ratios corresponding to the structural characteristic value intervals pre-stored in the structure prediction platform to obtain the local segmentation ratios of each crystal structure.

[0018] Based on the local segmentation ratio of each crystal structure, each crystal structure is locally segmented accordingly to obtain each local environment of each crystal structure.

[0019] Each crystal structure is converted into a graph structure and input into the MEGNet model, and each atom of each crystal structure is recorded as a node. The local environment interaction information and the overall environment interaction information of each node of each crystal structure are obtained. Each node is updated layer by layer through the message passing mechanism, and the feature representation vector of each node of each crystal structure is output.

[0020] The interactions between nodes are recorded as edge vectors. Based on the feature representation vectors of the two nodes connected by each edge vector, they are iteratively updated through the graph convolution layer to output the feature representation vectors of each edge of each crystal structure.

[0021] As a preferred technical solution, the outputting of the preliminary energy of each crystal structure specifically includes: Based on the feature representation vector of each node of each crystal structure and the feature representation vector of each edge of each crystal structure, the feature representation value of each node of each crystal structure is obtained after processing by the graph convolution layer. The feature representation value of each node is used to describe the local environment information and interaction information of each node in the corresponding crystal structure.

[0022] The characteristic representation value of each node of each crystal structure is input into the neural network model, and the preliminary energy of each crystal structure is obtained through processing and analysis, which is recorded as the neural network potential of each crystal structure.

[0023] As a preferred technical solution, the analysis and processing to obtain the neural network potential evaluation results of each crystal structure includes: A neural network potential threshold is extracted from the structure prediction platform. If the neural network potential of a crystal structure is greater than or equal to the neural network potential threshold, the neural network potential evaluation result of the crystal structure is determined to be unstable.

[0024] If the neural network potential of a certain crystal structure is less than the neural network potential threshold, the neural network potential evaluation result of the crystal structure is determined to be stable, and the crystal structures whose neural network potential evaluation results are stable are extracted and recorded as stable crystal structures.

[0025] As a preferred technical solution, the calculation of the short-range interaction information between atoms of each stable crystal structure and obtaining the empirical potential of each stable crystal structure by accumulation processing specifically includes: The coordinates of each atom in each stable crystal structure are counted, and the spatial distances between atoms are obtained through analysis and processing. The short-range interaction information between atoms in each stable crystal structure is calculated using the Buckingham potential model. The short-range interaction information of each atomic pair is accumulated and the result is recorded as the empirical potential of each stable crystal structure.

[0026] As a preferred technical solution, the neural network potential and empirical potential of each stable crystal structure are comprehensively analyzed and processed to obtain the total energy potential of each stable crystal structure, and the total energy potential is imported into the optimization algorithm for iterative update, and the stable crystal structure with the lowest total energy potential is output as the target crystal structure of the material to be predicted, which specifically includes: Based on the neural network potential and empirical potential of each stable crystal structure, the total energy potential of each stable crystal structure is obtained through analysis and processing, and compared and analyzed with the total energy potential threshold preset in the structure prediction platform. The stable crystal structures with total energy potential lower than the total energy potential threshold are counted and recorded as the predicted crystal structures.

[0027] The total energy potential of each predicted crystal structure is introduced into the Bayesian optimization algorithm for optimization. When the Bayesian optimization algorithm converges, the predicted crystal structure with the lowest total energy potential is output as the target crystal structure of the material to be predicted.

[0028] As a preferred technical solution, the analysis process obtains the total energy potential of each stable crystal structure, and the specific processing conditions are: ; in, For the The total energy potential of a stable crystal structure, For the A neural network potential with a stable crystal structure, For the The empirical potential of a stable crystal structure, is the energy balance coefficient, is the number of stable crystal structures, , is the total number of stable crystal structures.

[0029] In addition, the present invention also provides a crystal structure prediction system based on a hybrid potential energy deep learning model, characterized in that it includes; The crystal structure characteristic module is used to input the basic information of the material to be predicted on the structure prediction platform, generate the basic information characteristic values ​​of the material to be predicted and perform rationality evaluation. After obtaining the rationality evaluation results, based on the basic information characteristic values ​​of the material to be predicted, various crystal structure data sets that meet the material to be predicted are retrieved from the structure prediction platform and merged to obtain the structural characteristic values ​​of each crystal structure.

[0030] The neural network potential calculation module is used to perform corresponding feature learning on the local environment between atoms of each crystal structure based on the structural characteristic values ​​of each crystal structure, output the preliminary energy of each crystal structure, record it as the neural network potential of each crystal structure, and analyze and process to obtain the neural network potential evaluation results of each crystal structure.

[0031] The empirical potential calculation module is used to screen out the stable crystal structures of the materials to be predicted based on the neural network potential evaluation results of each crystal structure, calculate the short-range interaction information between atoms of each stable crystal structure, and obtain the empirical potential of each stable crystal structure through cumulative processing.

[0032] The target crystal structure output module is used to comprehensively analyze and process the neural network potential and empirical potential of each stable crystal structure to obtain the total energy potential of each stable crystal structure, and import it into the optimization algorithm for iterative update. The stable crystal structure with the lowest total energy potential is output as the target crystal structure of the material to be predicted.

[0033] Compared with the prior art, the embodiments of the present invention have at least the following beneficial effects: (1) The present invention provides a crystal structure prediction method and system based on a hybrid potential energy deep learning model, which combines the deep learning model MEGNet with the Buckingham potential, and can accurately describe the local structure and interaction between atoms, thereby significantly improving the accuracy of crystal structure energy prediction. By introducing the Bayesian optimization algorithm, the search process of the crystal structure is optimized, which not only improves the computational efficiency, but also avoids the local minimum problem that traditional methods may fall into, ensuring that the global optimal or near-optimal stable structure can be found. The method of the present invention has significant advantages in large-scale material screening and new material discovery, and can maintain high-precision prediction capabilities while reducing computational costs. In addition, the technical solution of the present invention has wide applicability and can be applied to the crystal structure prediction of a variety of material systems.

[0034] (2) The present invention comprehensively considers the rationality of the atomic composition and the rationality of the chemical bond composition of the material to be predicted to identify whether the material to be predicted exists, and then predicts the crystal structure of the material to be predicted, thereby improving the accuracy of the prediction, reducing unnecessary waste of resources, and improving the prediction efficiency.

[0035] (3) The present invention provides a standardized method to describe and compare different crystal structures by obtaining the structural characteristic values ​​of each crystal structure, thereby quickly determining the position of the crystal structure in the material space to obtain more detailed local environment information, helping the model to better understand and predict the interactions and information in the crystal structure.

[0036] Of course, any product implementing the present invention does not necessarily need to achieve all of the above advantages at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart of the crystal structure prediction method based on the hybrid potential energy deep learning model of the present invention.

[0038] Figure 2 This is a comparison chart of the MEGNet-Buckingham algorithm, DFT predicted energy, and DFT-PSO method.

[0039] Figure 3 This is the crystal structure of CaS corresponding to the number mp-22862 in the Materials Project material library.

[0040] Figure 4 This is the crystal structure with the lowest energy searched by the method proposed in the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] In the description of the present invention, it should be understood that the terms "opening", "upper", "lower", "thickness", "top", "middle", "length", "inside", "all around" and the like indicating orientation or positional relationship are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the components or elements referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0043] See also Figure 1 As shown, an embodiment of the present invention provides a crystal structure prediction method based on a hybrid potential energy deep learning model, comprising: Input the basic information of the material to be predicted into the structure prediction platform, generate the basic information characteristic values ​​of the material to be predicted and conduct a rationality assessment. After obtaining the rationality assessment results, based on the basic information characteristic values ​​of the material to be predicted, retrieve the various crystal structure data sets that meet the material to be predicted from the structure prediction platform and merge them to obtain the structural characteristic values ​​of each crystal structure.

[0044] The basic information of the material to be predicted is input into the structure prediction platform, the basic information characteristic value of the material to be predicted is generated and the rationality evaluation is performed. The specific process is as follows: The basic information of the material to be predicted includes the total number of atomic types of the material to be predicted, the number of atoms of each type, the charge number of each type of atoms, the total number of chemical bond types, the number of chemical bonds, the bond lengths of each type of chemical bonds, and the bond angles of each type of chemical bonds. In the embodiment of the present invention, the basic information is obtained through a chemical resource library, a chemical material library, or user input and upload.

[0045] It should be noted that atomic species refers to the types of different elements present in a material. Atomic species determines the electronic structure and potential chemical properties of the material. Different elements have different chemical activities, valence electron numbers, and atomic radii, which will affect the overall properties of the material.

[0046] The number of atoms of each type refers to the number of atoms of each element in the material. The number of atoms determines the stoichiometric ratio and affects the crystal structure and phase state of the material. For example, different proportions of the same element can form different compounds.

[0047] The charge number of each type of atom refers to the charge state of each atom in the material (usually considering the oxidation state). The charge number affects the chemical reactivity and electronic configuration of the atom, which in turn affects the electronic properties, conductivity and magnetism of the material.

[0048] Chemical bond types refer to different types of chemical bonds formed between atoms, such as covalent bonds, ionic bonds, metallic bonds, and hydrogen bonds. The type of chemical bond directly affects the mechanical strength, melting point, electrical conductivity, and solubility of the material.

[0049] Bond count refers to the total number of each type of chemical bond in a material. The number of bonds is related to the stability and structural integrity of the material. For example, more covalent bonds generally means higher hardness and melting point.

[0050] The bond length of various chemical bonds refers to the average distance between two bonding atomic nuclei. The bond length affects the density and lattice parameters of the material and is a key factor in determining the mechanical and electronic properties of the material.

[0051] The bond angle of various chemical bonds refers to the angle between two adjacent chemical bonds in a molecule. The bond angle affects the geometric shape of the molecule, which in turn affects the physical and chemical properties of the molecule, such as polarity, reactivity, and intermolecular forces.

[0052] The reference bond angles and reference bond lengths corresponding to various types of chemical bonds are extracted from the structure prediction platform, and analyzed and processed together with the basic information of the material to be predicted to generate the basic information characteristic values ​​of the material to be predicted, including: ; in, is the basic information characteristic value of the material to be predicted, For the The number of atoms, For the The charge number of the atom, For the The length of the chemical bond, For the The bond angle of a chemical bond, For the The reference bond length of the chemical bond, For the The reference bond angle of the chemical bond, is the atomic species number, , is the total number of atomic species, is the number of chemical bond types, , is the total number of chemical bond types, is the unit weight factor of the electroneutrality characteristic, is the unit weight factor of the comprehensive characteristics of bond length and bond angle.

[0053] It should be noted that the unit weight factor of the electroneutrality characteristic and the unit weight factor of the bond length and bond angle comprehensive characteristic are both in the range of 0 to 1 and satisfy The unit weight factor of the electric neutrality feature is an influence factor corresponding to the basic information characteristic value of the material to be predicted preset in the structure prediction platform, and represents a numerical value indicating the degree of influence of the electric neutrality feature on the basic information characteristic value of the material to be predicted. The unit weight factor of the bond length and bond angle comprehensive feature is an influence factor corresponding to the basic information characteristic value of the material to be predicted preset in the structure prediction platform, and represents a numerical value indicating the degree of influence of the bond length and bond angle comprehensive feature on the basic information characteristic value of the material to be predicted. When used, the unit weight factor of the electric neutrality feature and the unit weight factor of the bond length and bond angle comprehensive feature can be directly obtained from the structure prediction platform, and the corresponding relationship is a pre-set mapping relationship. For example, the basic information of the material to be predicted involved in the embodiment of the present invention, such as the total number of atomic types, the number of atoms of various types, the charge number of atoms of various types, the total number of chemical bond types, the number of chemical bonds, the bond lengths of various types of chemical bonds, and the bond angles of various types of chemical bonds of the material to be predicted, is input into the mapping set for mapping matching to obtain the unit weight factor of the electric neutrality feature and the unit weight factor of the bond length and bond angle comprehensive feature of the material to be predicted involved in the embodiment of the present invention, and the mapping relationship is one-to-one correspondence.

[0054] It should also be noted that the electrical neutrality characteristics of the material to be predicted and the comprehensive characteristics of bond length and bond angle have a certain correlation. The electrical neutrality characteristics refer to the balance state of positive and negative charges in the material, which is crucial to the stability and chemical properties of the material. The comprehensive characteristics of bond length and bond angle describe the spatial arrangement and connection between atoms in the material. These characteristics directly affect the physical and chemical properties of the material.

[0055] The charge number of an atom affects the distribution of electron clouds between atoms, which in turn affects the bond length of a chemical bond. For example, atoms with higher charge numbers may attract electron clouds more tightly, resulting in shorter bond lengths. The bond angles of electrically neutral materials are usually determined by the charge balance between atoms. If the charge distribution is uneven, it may cause a distortion of the bond angle, which will affect the structure of the material. Different types of chemical bonds have different requirements for electrical neutrality. For example, ionic bonds require strict charge balance, while covalent bonds may allow a certain charge offset. Electrical neutrality is the basis of material stability, while bond length and bond angle determine the geometric structure of the material. A stable material usually has a balanced charge distribution and optimized bond length and bond angle. The type and charge number of atoms determine the difference in electronegativity between atoms, which in turn affects the bond length and bond angle. The type and number of chemical bonds determine the overall structure of the material and the distribution of bond lengths and bond angles. Bond lengths and bond angles directly reflect the spatial arrangement and interaction between atoms, which are all related to the charge state of atoms. In short, the correlation between the characteristics of electrical neutrality and the comprehensive characteristics of bond lengths and bond angles is reflected in the fact that they jointly determine the stability and properties of the material.

[0056] The basic information characteristic value of the material to be predicted is compared with the rationality assessment threshold to obtain the rationality assessment result of the material to be predicted. If the basic information characteristic value of the material to be predicted is greater than or equal to the rationality assessment threshold, the rationality assessment result of the material to be predicted is determined to be reasonable. If the basic information characteristic value of the material to be predicted is less than the rationality assessment threshold, the rationality assessment result of the material to be predicted is determined to be unreasonable, and a reminder message is sent to the structure prediction platform.

[0057] The rationality assessment threshold is processed as follows: In a specific embodiment, the rationality assessment threshold is set directly in the structure prediction platform during the development of the crystal structure prediction system based on the hybrid potential energy deep learning model involved in the embodiment of the present invention. There are many methods for setting the rationality assessment threshold, such as obtaining it through statistical analysis, including counting the basic information characteristic values ​​under the reasonable state of historical materials (for example, when the electrical neutrality is 0), and averaging the basic information characteristic values ​​obtained multiple times to obtain the rationality assessment threshold.

[0058] The method of retrieving various crystal structure data sets that meet the material to be predicted from the structure prediction platform based on the basic information characteristic value of the material to be predicted and merging them to obtain the structural characteristic value of each crystal structure specifically includes: Based on the basic information of the material to be predicted, it is matched with the crystal structure library corresponding to the basic information pre-stored in the structure prediction platform, and then mapped and matched with the crystal structure data sets in the corresponding crystal structure library according to the characteristic values ​​of the basic information of the material to be predicted, so as to obtain the crystal structure data sets of the material to be predicted, wherein the crystal structure data sets of the material to be predicted include the lengths of the sides of the unit cell, the unit cell angles, the space group symbols and the atomic coordinates of each crystal structure.

[0059] It should be noted that the unit cell is the smallest repeating unit in the crystal structure, and the length of each side of the unit cell refers to the length of the unit cell in three spatial dimensions. The length of the unit cell side determines the linear size and proportion of the crystal, which is the basis for calculating the crystal density, lattice energy and other physical properties.

[0060] The unit cell angles are the angles between the three sides of the unit cell, which describe the spatial orientation of the unit cell. The unit cell angles affect the shape and symmetry of the crystal and are an important factor in determining the spatial structure of the crystal.

[0061] The space group symbol is a classification of the symmetry of the arrangement of atoms in a crystal. It describes all possible symmetry operations in the crystal, such as rotation, reflection, and inversion. The space group symbol provides a complete description of the symmetry of the crystal, which determines the possible positions of atoms in the crystal and the physical properties of the crystal, such as optical and magnetic properties. For example, the space group symbol of a crystal is P21 / c, where P indicates that the space group belongs to the Parai class (P class), that is, there is no reflection symmetry. 21 means that the space group has a two-fold helical axis with the direction of the axis being [1 0 0], and the angle of the helical rotation is 180 degrees divided by 2, that is, 90 degrees. c means that the space group has a reflection plane perpendicular to the helical axis.

[0062] Atomic coordinates refer to the position of each atom in the unit cell described by a set of coordinates, which are normalized relative to the starting point and side length of the unit cell. Atomic coordinates are a direct way to describe the arrangement of atoms inside a crystal, and are essential for understanding crystal structure, calculating interatomic distances and angles, and predicting crystal properties.

[0063] These parameters are the basic descriptors of crystal structure, which can accurately define a crystal structure and make it uniquely determined in three-dimensional space. Through the space group symbol, we can understand the symmetry of the crystal, and the unit cell parameters and atomic coordinates are the basis for calculating the physical properties of the crystal.

[0064] Based on the space group symbol of each crystal structure, a mapping match is performed with the symmetry coefficient corresponding to each space group symbol pre-stored in the structure prediction platform to obtain the symmetry coefficient of each crystal structure, and the symmetry coefficient is used to characterize the space group symmetry characteristics of the crystal structure.

[0065] Based on the atomic coordinates in each crystal structure, the correlation coefficient of each crystal structure is obtained by analysis and processing. The correlation coefficient of each crystal structure is used to characterize the order degree of each crystal structure. The specific calculation formula is: ; ; in, For the The correlation coefficient of the crystal structure is For the The crystal structure of The spatial horizontal coordinates of atoms, For the The crystal structure of The spatial front-back coordinates of atoms, For the The crystal structure of The spatial coordinates of the atoms, For the The average spatial coordinate value of the atoms in the crystal structure, For the The average spatial front-to-back coordinates of atoms in the crystal structure, For the The average of the spatial upper and lower coordinates of the atoms in the crystal structure, For the The covariance matrix of the crystal structure, To calculate the The correlation coefficient of each atom in the crystal structure, is the normalization factor, Number the crystal structures. , is the total number of crystal structures, Number the atoms, , is the atomic species number, , is the total number of atomic species.

[0066] The lengths of the sides of the unit cells of each crystal structure, the unit cell angles, the symmetry coefficients and the correlation coefficients are processed and analyzed to obtain the structural characteristic values ​​of each crystal structure, and the structural characteristic values ​​are used to describe the crystal structure characteristics.

[0067] ; ; in, For the The structural eigenvalues ​​of the crystal structure, For the The unit cell volume of a crystal structure, For the The length of the side a of the unit cell of the crystal structure is For the The length of the side b of the unit cell of the crystal structure is For the The length of the side c of the unit cell of the crystal structure is For the The angle between the unit cell b side and the unit cell c side of the crystal structure is For the The angle between the unit cell a side and the unit cell c side of a crystal structure is For the The angle between the unit cell side a and the unit cell side b of a crystal structure is For the The symmetry coefficient of the crystal structure is For the The correlation coefficient of the crystal structure is is the characteristic factor per unit volume of the unit cell of the crystal structure, Number the crystal structures. , is the total number of crystal structures, is a natural constant.

[0068] It should be noted that the characteristic factor of the unit volume of the unit cell can be directly extracted from the structure prediction platform when used. In the structure prediction platform, the unit cell volume and the characteristic factor of the unit volume of the unit cell form a preset mapping set, and the mapping relationship is one-to-one. The unit cell volume involved in the embodiment of the present invention is input into the mapping set for corresponding mapping matching to obtain the characteristic factor of the unit volume of the unit cell involved in the embodiment of the present invention.

[0069] It should also be noted that there is a certain correlation between the unit cell volume of each crystal structure, the symmetry coefficient of each crystal structure and the correlation coefficient of each crystal structure. The unit cell volume refers to the volume of the smallest repeating unit (unit cell) in the crystal structure. It is determined by the three side lengths of the unit cell and reflects the density and space occupancy of the atomic or molecular arrangement in the crystal. The symmetry coefficient describes the richness of symmetry operations in the crystal structure. The higher the crystal symmetry, the higher its symmetry coefficient. Symmetry includes point symmetry, plane symmetry, axisymmetry, etc. The correlation coefficient is usually used to describe the positional relationship between different atoms or molecules in the crystal structure, or to describe the similarity between different crystal structures. The unit cell volume may affect the symmetry of the crystal structure. For example, a change in volume may cause a change in the unit cell parameters, thereby changing the symmetry of the crystal. In some cases, an increase or decrease in volume may result in a decrease or increase in symmetry. The unit cell volume is related to the crystal density, and changes in density may affect the stability of the crystal structure, thereby affecting the symmetry. The symmetry of the crystal structure may affect the correlation of its physical properties. For example, a highly symmetrical crystal structure may have a more regular distribution of properties, resulting in a higher correlation coefficient. Crystal structures with high symmetry tend to have simpler correlations because symmetry can simplify the parameters that describe the crystal structure. Changes in the unit cell volume may affect the physical and chemical properties of the material, and changes in these properties can be quantified by correlation coefficients.

[0070] Based on the structural eigenvalues ​​of each crystal structure, the corresponding feature learning is performed on the local environment between atoms of each crystal structure, and the preliminary energy of each crystal structure is output, recorded as the neural network potential of each crystal structure. The neural network potential evaluation results of each crystal structure are obtained by analysis and processing.

[0071] The corresponding feature learning of the local environment between atoms in each crystal structure specifically includes: Based on the structural characteristic value of each crystal structure, a mapping match is performed with the local segmentation ratio corresponding to each structural characteristic value interval pre-stored in the structure prediction platform to obtain the local segmentation ratio of each crystal structure. In an embodiment of the present invention, the larger the structural characteristic value, the smaller the local segmentation ratio.

[0072] Based on the local segmentation ratio of each crystal structure, each crystal structure is locally segmented accordingly to obtain each local environment of each crystal structure.

[0073] Based on the coordinates of each atom, each crystal structure is converted into a graph structure and input into the MEGNet model, and each atom of each crystal structure is recorded as a node. The local environment interaction information and the overall environment interaction information of each node of each crystal structure are obtained. Each node is updated layer by layer through the message passing mechanism, and the feature representation vector of each node of each crystal structure is output.

[0074] Local environment interaction information refers to the interaction between each node and its immediate neighbors. In a graph structure, this is equivalent to the node's adjacency information, including the type, distance (bond length), angle (bond angle), and charge distribution of neighboring atoms. Local environment interaction information is crucial to understanding the chemical properties of atoms because it reflects the direct chemical environment of atoms, which directly affects the electronic state and reactivity of atoms. The type of neighboring atoms refers to the type of other atoms that are directly connected to the target atom. For example, in a molecule, a carbon atom may be connected to hydrogen atoms and oxygen atoms. The distance (bond length) refers to the straight-line distance between the target atom and its neighboring atoms, which usually corresponds to the length of the chemical bond. The bond length can affect the chemical properties and reactivity of atoms. The angle (bond angle) refers to the angle formed by the target atom and its two neighboring atoms, which describes the spatial structure in a molecule or crystal. The bond angle has an important influence on the geometry and properties of molecules. The charge distribution refers to the distribution of electrons around a chemical bond, which affects the polarity of atoms and the interaction between molecules.

[0075] Global environmental interaction information involves the combined effect of the entire crystal structure on a single atom. It includes not only local interactions, but also indirect effects of atoms farther away in the crystal on specific atoms, as well as the effects of crystal symmetry and periodicity on atoms. Global environmental interaction information helps capture long-range order and periodic features in the crystal structure. Global environmental interaction information includes local interactions, indirect effects, crystal symmetry and periodicity, and long-range order. Although local interactions overlap with local environmental interaction information, they are considered as part of larger-scale interactions in the global environment. Indirect effects refer to the effects of atoms farther away in the crystal on the electronic state and chemical properties of the target atom. These effects may be transmitted through diffusion of electron clouds, modulation of crystal fields, etc. Crystal symmetry and periodicity This refers to the regular arrangement and repetitive patterns in the crystal structure, which have long-range effects on the electronic state and physical properties of atoms. For example, some atoms in a crystal may have a specific energy level splitting due to symmetry. Long-range order refers to the regularity and consistency of atomic arrangement throughout the crystal, which affects the macroscopic properties of the material, such as conductivity and magnetism.

[0076] The message passing mechanism refers to an algorithmic process in graph neural networks where each node updates its own features based on the features of its neighbor nodes and the features of the edges. In the MEGNet model, this mechanism is implemented by propagating and aggregating information in the network. Each node receives "messages" from its neighbor nodes, which contain the status information of the neighbor nodes.

[0077] Layer-by-layer updating refers to the repetition of the message passing process in multiple layers of the network. In each layer, the feature vector of a node is updated based on the messages received from neighboring nodes. This updating is iterative, and each layer may capture higher-level abstract features. For example, the first layer may only capture information about direct neighbors, while subsequent layers may be able to capture more distant interactions or more complex structural patterns.

[0078] After multiple layers of message passing and updating, each node will have a feature representation vector that encodes the local and global environment information of the node in the crystal structure.

[0079] The interactions between nodes are recorded as edge vectors. Based on the feature representation vectors of the two nodes connected by each edge vector, they are iteratively updated through the graph convolution layer to output the feature representation vectors of each edge of each crystal structure.

[0080] In the graph structure, the edge vector represents the interaction between two nodes (atoms). This vector contains a variety of information, such as the type of chemical bond between the two atoms (covalent bond, ionic bond, etc.), bond length, bond angle, charge transfer, etc. Edge vectors are important elements used to represent the relationship between nodes in graph neural networks. They carry detailed information about the interaction between nodes.

[0081] The graph convolution layer is the core component of the graph neural network, which simulates the convolution operation of the traditional convolutional neural network in image processing, but is applicable to graph structured data. In the graph convolution layer, the feature vector of each node is updated according to the feature vectors of its neighboring nodes and the edge vectors connecting them. This process is iterative, which means that in each layer of the network, the feature vector of the node is updated. During the iterative update process of the graph convolution layer, not only the feature vector of the node is updated, but the edge vector is also updated accordingly according to the update of the node. In the end, each edge will have an updated feature representation vector, which reflects how the interaction of the edge affects the entire graph structure after considering the information of the surrounding nodes.

[0082] Each node (atom) has a feature vector that encodes the properties of the node, such as atomic type, charge state, position, etc.

[0083] The MEGNet (Material Graphs Network) model is a deep learning framework for material property prediction. The MEGNet model combines Graph Neural Networks (GNNs) and physics-inspired methods and is specifically designed to process data in materials science.

[0084] MEGNet uses graph neural networks to represent the crystal structure of materials. In graph neural networks, atoms are regarded as nodes and chemical bonds are regarded as edges. This representation method can effectively capture the three-dimensional structural information of materials. The MEGNet model not only considers the local atomic environment in the crystal structure, but also integrates global properties such as unit cell parameters and space group information. This multi-scale representation helps to understand the properties of materials more comprehensively. Some physical information such as atomic charge, bond length and bond angle are embedded in the model, which helps the model better understand the electronic structure and other relevant properties of the material.

[0085] MEGNet uses a message passing mechanism to update the states of nodes (atoms) and edges (chemical bonds). In this process, the feature vector of each atom is updated according to the feature vectors of its neighboring atoms.

[0086] In graph neural networks, the readout layer is used to extract global features from the graph to generate a representation of the entire material. In MEGNet, this layer is used to generate the final feature vector for predicting material properties.

[0087] The advantage of the MEGNet model is that it can learn highly abstract features from complex material structures, which are closely related to the physical and chemical properties of the materials.

[0088] The outputting of the preliminary energy of each crystal structure specifically includes: Based on the feature representation vector of each node of each crystal structure and the feature representation vector of each edge of each crystal structure, the feature representation value of each node of each crystal structure is obtained after processing by the graph convolution layer. The feature representation value of each node is used to describe the local environment information and interaction information of each node in the corresponding crystal structure.

[0089] The characteristic representation value of each node of each crystal structure is input into the neural network model, and the preliminary energy of each crystal structure is obtained through processing and analysis, which is recorded as the neural network potential of each crystal structure.

[0090] ; in, For the A neural network potential with a crystal structure, For the The crystal structure of The feature representation value of each node, are neural network parameters, is the neural network model expression, Number the crystal structures. , is the total number of crystal structures, is the node number, , is the total number of nodes, is the atomic species number, , is the total number of atomic species.

[0091] It should be noted that the neural network parameters refer to the coefficients connecting the neurons in the neural network, which determine the influence of the input features on the output results, that is, the importance of each feature in the prediction.

[0092] The analysis and processing obtains the neural network potential evaluation results of each crystal structure, including: A neural network potential threshold is extracted from the structure prediction platform. If the neural network potential of a crystal structure is greater than or equal to the neural network potential threshold, the neural network potential evaluation result of the crystal structure is determined to be unstable.

[0093] If the neural network potential of a certain crystal structure is less than the neural network potential threshold, the neural network potential evaluation result of the crystal structure is determined to be stable, and the crystal structures whose neural network potential evaluation results are stable are extracted and recorded as stable crystal structures.

[0094] It should be noted that the neural network potential threshold is processed as follows: in a specific embodiment, the neural network potential threshold is set directly in the structure prediction platform during the development of the crystal structure prediction system based on the hybrid potential energy deep learning model involved in the embodiment of the present invention. There are many methods for setting the neural network potential threshold, such as obtaining it through statistical analysis, including counting the neural network potential in historical stable crystal structures, and averaging the neural network potentials obtained multiple times to obtain the neural network potential threshold.

[0095] Based on the neural network potential evaluation results of each crystal structure, the stable crystal structures of the material to be predicted are screened and obtained, and the short-range interactions between atoms of each stable crystal structure are calculated. The empirical potential of each stable crystal structure is obtained by cumulative processing.

[0096] The calculating of the short-range interactions between atoms of each stable crystal structure, performing accumulation processing and recording as the empirical potential of each stable crystal structure specifically includes: The short-range interactions include repulsive terms, attractive terms, and exponential decay.

[0097] The repulsive term refers to the fact that when atoms are very close, a strong repulsive force will be generated between atoms due to the repulsion of the electron cloud. This term describes this repulsive effect.

[0098] The attraction term means that as the distance between atoms increases, there will be van der Waals force or other attractive forces between atoms, and this term describes this attractive force.

[0099] Exponential decay refers to this term indicating that the interaction force decreases rapidly as the distance between atoms increases.

[0100] The coordinates of each atom in each stable crystal structure are counted, and the spatial distances between atoms are obtained by analysis and processing. The short-range interactions between atoms in each stable crystal structure are calculated using the Buckingham potential model. The short-range interactions of each atomic pair are accumulated and the results are recorded as the empirical potential of each stable crystal structure. Specifically, it includes: ; ; in, For the The empirical potential of a stable crystal structure, Indicates The crystal structure of Nodes and The distance between the nodes, A is the strength coefficient of the repulsion term, B is the constant parameter that controls the attenuation rate of the repulsion term, and C is the strength coefficient of the attraction term. and are node numbers, and d, , , is the number of stable crystal structures, , is the total number of stable crystal structures, .

[0101] It should be noted that the intensity coefficient of the repulsive term, the constant parameter controlling the decay rate of the repulsive term, and the intensity coefficient of the attractive term are calculated by density functional theory DFT.

[0102] It should also be noted that the Buckingham potential model is an empirical potential energy function widely used in molecular dynamics and materials science to describe short-range interactions between atoms. This model takes into account the repulsive and attractive forces between atoms. When the atoms are very close, the repulsive force becomes very strong, which prevents the atoms from overlapping each other. When there is a certain distance between atoms, the attractive force plays a major role, which reflects the van der Waals force or London dispersion force between atoms. The Buckingham potential model mainly describes short-range interactions, that is, the forces when the atoms are close.

[0103] The neural network potential and empirical potential of each stable crystal structure are comprehensively analyzed and processed to obtain the total energy potential of each stable crystal structure, which is then imported into the optimization algorithm for iterative update. The stable crystal structure with the lowest total energy potential is output as the target crystal structure of the material to be predicted.

[0104] The neural network potential and empirical potential of each stable crystal structure are comprehensively analyzed and processed to obtain the total energy potential of each stable crystal structure, and the total energy potential of each stable crystal structure is imported into the optimization algorithm for iterative update, and the stable crystal structure with the lowest total energy potential is output as the target crystal structure of the material to be predicted, specifically including: The analysis process obtains the total energy potential of each stable crystal structure, and the specific processing conditions are: ; in, For the The total energy potential of a stable crystal structure, For the A neural network potential with a stable crystal structure, For the The empirical potential of a stable crystal structure, is the energy balance coefficient, is the number of stable crystal structures, , is the total number of stable crystal structures.

[0105] It should be noted that, in the embodiment of the present invention, the energy balance coefficient is obtained by selecting a group of materials with known structures and calculating the energy of the group of materials using high-precision first-principles calculations, and recorded as benchmark data. The energy balance coefficient in the hybrid model is adjusted using the benchmark data, and the optimal energy balance coefficient is obtained by minimizing the error function.

[0106] Based on the neural network potential and empirical potential of each stable crystal structure, the total energy potential of each stable crystal structure is obtained through analysis and processing, and compared and analyzed with the total energy potential threshold preset in the structure prediction platform. The stable crystal structures with total energy potential lower than the total energy potential threshold are counted and recorded as the predicted crystal structures.

[0107] like Figure 2 The figure shows the comparison of the energy calculation results using the hybrid potential energy model and DFT, as well as the comparison with the DFT-PSO algorithm time. The energy predicted by the method of the embodiment of the present invention is close to the DFT value, and the time for searching the crystal structure is better than the DFT-PSO algorithm, which is less than 3 orders of magnitude. It should be noted that the Chinese names of the chemical formulas in the horizontal axis of the figure are sodium fluoride, potassium fluoride, rubidium fluoride, cesium fluoride, lithium chloride, sodium chloride, potassium chloride, rubidium chloride, cesium chloride, calcium oxide, strontium oxide, cadmium oxide, cobalt sulfide, zinc sulfide, carbon, silicon, gallium arsenide and cadmium telluride.

[0108] DFT-PSO is a method that combines Density Functional Theory with the Particle Swarm Optimization algorithm. DFT is a quantum mechanical method that is widely used to calculate the electronic structure of atoms, molecules, and solid materials. DFT approximately describes the behavior of multi-electron systems by solving the Kohn-Sham equation. PSO is a computational method in which each "particle" represents a candidate solution in the problem space and moves through the solution space through an iterative process to find the optimal solution. Particles adjust their positions based on their own experience and the experience of their neighboring particles. Combining DFT with PSO, i.e. DFT-PSO, means using the PSO algorithm to optimize the process of material structure, while DFT is used to evaluate the energy of each candidate structure.

[0109] It should be noted that the total energy potential threshold is processed as follows: in a specific embodiment, the total energy potential threshold is set directly in the structure prediction platform during the development of the crystal structure prediction system based on the hybrid potential energy deep learning model involved in the embodiment of the present invention. There are many methods for setting the total energy potential threshold, such as obtaining it through statistical analysis, including statistically analyzing the total energy potential of historical stable crystal structures, and averaging the total energy potentials obtained multiple times to obtain the total energy potential threshold.

[0110] The total energy potential of each predicted crystal structure is introduced into the Bayesian optimization algorithm for optimization. When the Bayesian optimization algorithm converges, the predicted crystal structure with the lowest total energy potential is output as the target crystal structure of the material to be predicted.

[0111] Bayesian optimization is an algorithm for global optimization that is suitable for situations where the objective function is computationally expensive and poorly defined. In materials design, the objective function is often a predictive model of material properties, such as energy.

[0112] The Bayesian optimization algorithm seeks the input value that minimizes the objective function by iteratively updating the model of the objective function. In an embodiment of the present invention, the goal is to find the crystal structure that minimizes the total energy potential. When the Bayesian optimization algorithm reaches a certain stopping condition, the algorithm is considered to converge. When the algorithm converges, the crystal structure with the lowest predicted total energy is output.

[0113] It needs to be explained that the Bayesian optimization algorithm is an optimization method based on a probability model, which mainly consists of the following steps: Surrogate Model: The algorithm first builds a probabilistic model to approximate the target function. Gaussian Processes (GPs) are the most commonly used probabilistic model.

[0114] Acquisition Function: The acquisition function is used to determine the next sampling point. Based on the probability model, sampling is performed in the unknown region while sampling in the known optimal region. Common acquisition functions include Expected Improvement (EI), Probability of Improvement (PI), and Upper Confidence Bound (UCB).

[0115] In each iteration, the next sampling point is selected using the acquisition function. The total energy of the crystal structure is calculated at the selected sampling point. The new sampling point is added to the data set and the probability model is updated. The above iterative process is repeated until the stopping condition is met, such as reaching the preset algorithm convergence.

[0116] like Figure 3 and Figure 4 As shown, Figure 3 This is the crystal structure of CaS corresponding to the number mp-22862 in the Materials Project material library. Figure 4 This is the crystal structure with the lowest energy searched by the method proposed in the present invention.

[0117] In this embodiment, the present invention provides a crystal structure prediction system based on a hybrid potential energy deep learning model, comprising: The crystal structure characteristic module is used to input the basic information of the material to be predicted on the structure prediction platform, generate the basic information characteristic values ​​of the material to be predicted and perform rationality evaluation. After obtaining the rationality evaluation results, based on the basic information characteristic values ​​of the material to be predicted, various crystal structure data sets that meet the material to be predicted are retrieved from the structure prediction platform and merged to obtain the structural characteristic values ​​of each crystal structure.

[0118] The neural network potential calculation module is used to perform corresponding feature learning on the local environment between atoms of each crystal structure based on the structural characteristic values ​​of each crystal structure, output the preliminary energy of each crystal structure, record it as the neural network potential of each crystal structure, and analyze and process to obtain the neural network potential evaluation results of each crystal structure.

[0119] The empirical potential calculation module is used to screen out the stable crystal structures of the materials to be predicted based on the neural network potential evaluation results of each crystal structure, calculate the short-range interactions between atoms of each stable crystal structure, and obtain the empirical potential of each stable crystal structure through cumulative processing.

[0120] The target crystal structure output module is used to comprehensively analyze and process the neural network potential and empirical potential of each stable crystal structure to obtain the total energy potential of each stable crystal structure, and import it into the optimization algorithm for iterative update. The stable crystal structure with the lowest total energy potential is output as the target crystal structure of the material to be predicted.

[0121] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0122] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that technicians in the relevant technical field can understand and use the present invention well. As long as they do not deviate from the structure of the present invention or exceed the scope defined by the present invention, they should all belong to the protection scope of the present invention.

Claims

1. A crystal structure prediction method based on a hybrid potential energy deep learning model, characterized in that: include: Input the basic information of the material to be predicted into the structure prediction platform, generate the basic information characteristic value of the material to be predicted and conduct a rationality evaluation, and after obtaining the rationality evaluation result, retrieve various crystal structure data sets that meet the material to be predicted from the structure prediction platform based on the basic information characteristic value of the material to be predicted, and merge to obtain the structural characteristic value of each crystal structure; Based on the structural characteristic values ​​of each crystal structure, the corresponding characteristic learning is performed on the local environment between atoms of each crystal structure, and the preliminary energy of each crystal structure is output, which is recorded as the neural network potential of each crystal structure, and the neural network potential evaluation results of each crystal structure are obtained by analysis and processing; Based on the neural network potential evaluation results of each crystal structure, the stable crystal structures of the material to be predicted are screened, and the short-range interaction information between atoms of each stable crystal structure is calculated. The empirical potential of each stable crystal structure is obtained through accumulation processing; The neural network potential and empirical potential of each stable crystal structure are comprehensively analyzed and processed to obtain the total energy potential of each stable crystal structure, which is then imported into the optimization algorithm for iterative update. The stable crystal structure with the lowest total energy potential is output as the target crystal structure of the material to be predicted.

2. The crystal structure prediction method based on the hybrid potential energy deep learning model according to claim 1, characterized in that: The basic information of the material to be predicted is input into the structure prediction platform, the basic information characteristic value of the material to be predicted is generated and the rationality evaluation is performed. The specific process is as follows: The basic information of the material to be predicted includes the total number of atomic species, the number of atoms of each type, the charge number of atoms of each type, the total number of chemical bond species, the number of chemical bonds, the bond lengths of chemical bonds of each type, and the bond angles of chemical bonds of each type. The reference bond angles and reference bond lengths corresponding to various types of chemical bonds are extracted from the structure prediction platform, and are jointly analyzed and processed with the basic information of the material to be predicted to generate the basic information characteristic value of the material to be predicted, and compared with the rationality assessment threshold to obtain the rationality assessment result of the material to be predicted. If the basic information characteristic value of the material to be predicted is greater than or equal to the rationality assessment threshold, the rationality assessment result of the material to be predicted is determined to be reasonable; if the basic information characteristic value of the material to be predicted is less than the rationality assessment threshold, the rationality assessment result of the material to be predicted is determined to be unreasonable, and a reminder message is sent to the structure prediction platform.

3. The crystal structure prediction method based on the hybrid potential energy deep learning model according to claim 1, characterized in that: The method of retrieving various crystal structure data sets that meet the material to be predicted from the structure prediction platform based on the basic information characteristic value of the material to be predicted and merging them to obtain the structural characteristic value of each crystal structure specifically includes: Based on the basic information characteristic values ​​of the material to be predicted, mapping and matching are performed with each crystal structure data set corresponding to each basic information characteristic value interval pre-stored in the structure prediction platform to obtain each crystal structure data set of the material to be predicted, wherein each crystal structure data set of the material to be predicted includes each side length of the unit cell of each crystal structure, unit cell angle, space group symbol and each atomic coordinate; Based on the space group symbol of each crystal structure, mapping and matching are performed with the symmetry coefficients corresponding to each space group symbol pre-stored in the structure prediction platform to obtain the symmetry coefficient of each crystal structure, wherein the symmetry coefficient is used to characterize the space group symmetry characteristics of the crystal structure; Based on the coordinates of each atom in each crystal structure, the correlation coefficient of each crystal structure is obtained by analysis and processing, and the correlation coefficient of each crystal structure is used to characterize the order degree of each crystal structure; The lengths of the sides of the unit cells of each crystal structure, the unit cell angles, the symmetry coefficients and the correlation coefficients are processed and analyzed to obtain the structural characteristic values ​​of each crystal structure, and the structural characteristic values ​​are used to describe the crystal structure characteristics.

4. The crystal structure prediction method based on the hybrid potential energy deep learning model according to claim 1, characterized in that: The corresponding feature learning of the local environment between atoms in each crystal structure specifically includes: Based on the structural characteristic values ​​of each crystal structure, mapping and matching are performed with the local segmentation ratios corresponding to the structural characteristic value intervals pre-stored in the structure prediction platform to obtain the local segmentation ratios of each crystal structure; Performing corresponding local segmentation on each crystal structure based on the local segmentation ratio of each crystal structure to obtain each local environment of each crystal structure; Each crystal structure is converted into a graph structure and input into the MEGNet model, and each atom of each crystal structure is recorded as a node, and the local environment interaction information and the overall environment interaction information of each node of each crystal structure are obtained. Each node is updated layer by layer through the message passing mechanism, and the feature representation vector of each node of each crystal structure is output; The interactions between nodes are recorded as edge vectors. Based on the feature representation vectors of the two nodes connected by each edge vector, they are iteratively updated through the graph convolution layer to output the feature representation vectors of each edge of each crystal structure.

5. The crystal structure prediction method based on the hybrid potential energy deep learning model according to claim 4 is characterized in that: The outputting of the preliminary energy of each crystal structure specifically includes: Based on the feature representation vectors of each node of each crystal structure and the feature representation vectors of each edge of each crystal structure, a feature representation value of each node of each crystal structure is obtained after processing by a graph convolution layer, wherein the feature representation value of each node is used to describe the local environment information and interaction information of each node in the corresponding crystal structure; The characteristic representation value of each node of each crystal structure is input into the neural network model, and the preliminary energy of each crystal structure is obtained through processing and analysis, which is recorded as the neural network potential of each crystal structure.

6. The crystal structure prediction method based on the hybrid potential energy deep learning model according to claim 1, characterized in that: The analysis and processing obtains the neural network potential evaluation results of each crystal structure, including: Extracting a neural network potential threshold from the structure prediction platform, if the neural network potential of a certain crystal structure is greater than or equal to the neural network potential threshold, then determining that the neural network potential evaluation result of the crystal structure is unstable; If the neural network potential of a certain crystal structure is less than the neural network potential threshold, the neural network potential evaluation result of the crystal structure is determined to be stable, and the crystal structures whose neural network potential evaluation results are stable are extracted and recorded as stable crystal structures.

7. The crystal structure prediction method based on the hybrid potential energy deep learning model according to claim 1, characterized in that: The calculating of the short-range interaction information between atoms of each stable crystal structure and obtaining the empirical potential of each stable crystal structure by accumulation processing specifically includes: The coordinates of each atom in each stable crystal structure are counted, and the spatial distances between atoms are obtained through analysis and processing. The short-range interaction information between atoms in each stable crystal structure is calculated using the Buckingham potential model. The short-range interaction information of each atomic pair is accumulated and the result is recorded as the empirical potential of each stable crystal structure.

8. The crystal structure prediction method based on the hybrid potential energy deep learning model according to claim 1, characterized in that: The neural network potential and empirical potential of each stable crystal structure are comprehensively analyzed and processed to obtain the total energy potential of each stable crystal structure, and the total energy potential of each stable crystal structure is imported into the optimization algorithm for iterative update, and the stable crystal structure with the lowest total energy potential is output as the target crystal structure of the material to be predicted, specifically including: Based on the neural network potential and empirical potential of each stable crystal structure, the total energy potential of each stable crystal structure is obtained by analysis and processing, and compared and analyzed with the total energy potential threshold preset in the structure prediction platform, and the stable crystal structures with total energy potential lower than the total energy potential threshold are counted and recorded as each predicted crystal structure; The total energy potential of each predicted crystal structure is introduced into the Bayesian optimization algorithm for optimization. When the Bayesian optimization algorithm converges, the predicted crystal structure with the lowest total energy potential is output as the target crystal structure of the material to be predicted.

9. The crystal structure prediction method based on the hybrid potential energy deep learning model according to claim 7, characterized in that: The analysis process obtains the total energy potential of each stable crystal structure, and the specific processing conditions are: ； in, For the The total energy potential of a stable crystal structure, For the A neural network potential with a stable crystal structure, For the The empirical potential of a stable crystal structure, is the energy balance coefficient, is the number of stable crystal structures, , is the total number of stable crystal structures.

10. A crystal structure prediction system based on a hybrid potential energy deep learning model, characterized in that: include: A crystal structure characteristic module is used to input the basic information of the material to be predicted on the structure prediction platform, generate the basic information characteristic value of the material to be predicted and perform a rationality evaluation, and after obtaining the rationality evaluation result, retrieve various crystal structure data sets that meet the material to be predicted from the structure prediction platform based on the basic information characteristic value of the material to be predicted, and merge to obtain the structural characteristic value of each crystal structure; A neural network potential calculation module is used to perform corresponding feature learning on the local environment between atoms of each crystal structure based on the structural characteristic values ​​of each crystal structure, output the preliminary energy of each crystal structure, record it as the neural network potential of each crystal structure, and analyze and process to obtain the neural network potential evaluation results of each crystal structure; The empirical potential calculation module is used to screen out the stable crystal structures of the materials to be predicted based on the neural network potential evaluation results of each crystal structure, calculate the short-range interaction information between atoms of each stable crystal structure, and obtain the empirical potential of each stable crystal structure through accumulation processing; The target crystal structure output module is used to comprehensively analyze and process the neural network potential and empirical potential of each stable crystal structure to obtain the total energy potential of each stable crystal structure, and import it into the optimization algorithm for iterative update. The stable crystal structure with the lowest total energy potential is output as the target crystal structure of the material to be predicted.

Images