Method, system, equipment and medium for screening inorganic sodium ion solid electrolyte material

Abstract

The invention relates to an inorganic sodium ion solid electrolyte material screening method, system, equipment and medium, sodium-based inorganic compound data is extracted from an open source material database, and a candidate library is formed through preliminary screening; constructing a regression model by using training data of the dynamic solid electrolyte database, and optimizing and selecting an optimal machine learning model; inputting the preliminary candidate library into the model, and predicting and screening out a preliminary screening material of which the ionic conductivity is higher than a first threshold value; performing simulation calculation on the primarily screened material based on a first principle, simulating sodium ion diffusion tracks at different temperatures by using de novo molecular dynamics, calculating a diffusion coefficient, and extrapolating the diffusion coefficient to ionic conductivity at room temperature; and comparing the electrical conductivity predicted by machine learning and the electrical conductivity simulated by the first principle, and selecting a material of which the two electrical conductivity are both higher than a second threshold value and the migration energy barrier is relatively low as the optimal sodium ion solid electrolyte. Compared with the prior art, the method can efficiently and accurately screen out the high-performance sodium ion solid electrolyte material.

Description

Technical Field

[0001] The present invention relates to the field of materials informatics technology, and in particular to a method, system, equipment, and medium for screening inorganic sodium ion solid electrolyte materials based on collaborative optimization of machine learning and first principles. Background Art

[0002] Sodium-ion batteries are considered a potential alternative to lithium-ion batteries due to their abundant resources and low cost. Solid-state electrolyte materials are superior to traditional liquid electrolytes in terms of safety, thermal stability, and electrochemical stability. Therefore, the development of high-performance inorganic sodium-ion solid electrolytes is of great significance. Currently, screening of such materials mainly relies on experimental methods or first-principles calculations, which are inefficient and difficult to cope with the rapid screening of large-scale materials. With the development of material databases and computing technology, machine learning has gradually been applied to material property prediction, improving screening efficiency.

[0003] CN116364212A proposes a method for accelerating the screening and optimization of fast ion conductor materials using a machine learning model. Although it achieves efficient screening of lithium ion conductor materials, it focuses on lithium-based materials with specific anions, and its applicability to sodium ion solid electrolytes is unclear. It also fails to comprehensively consider the thermodynamic stability and ion migration performance of the materials.

[0004] CN118942592A discloses a prediction model for the properties of ordered metal phases based on machine learning. It uses machine learning combined with first-principles high-throughput calculations to screen metal materials, but focuses on the formation energy and mechanical properties of ordered metal phases. It does not directly involve the optimization of key properties such as the ionic conductivity of sodium ion solid electrolytes. The feature engineering processing is complex and the data dependence is high, and the scope of application of the model is limited.

[0005] CN119418832A introduces a method for preparing garnet-type solid electrolytes assisted by a machine learning model. It focuses on optimizing the ionic conductivity of garnet-type solid electrolytes through machine learning. However, the screening scope is limited to lithium-ion solid electrolyte materials with specific structures. It is difficult to apply to diverse sodium-ion solid electrolyte material systems, and insufficient attention is paid to the thermodynamic stability of the materials.

[0006] In summary, the existing technology has problems such as low efficiency, high cost and narrow screening range in the screening and optimization of sodium ion solid electrolyte materials. There is an urgent need for a method that can efficiently and accurately screen high-performance sodium ion solid electrolyte materials. Summary of the Invention

[0007] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a method, system, equipment, and medium for screening inorganic sodium ion solid electrolyte materials based on the collaborative optimization of machine learning and first principles, which can efficiently and accurately screen out high-performance sodium ion solid electrolyte materials, overcoming the problems of low efficiency, high cost, and narrow screening range in the prior art.

[0008] The purpose of the present invention can be achieved by the following technical solutions:

[0009] The first aspect of the present invention provides a method for screening inorganic sodium ion solid electrolyte materials based on collaborative optimization of machine learning and first principles, comprising the following steps:

[0010] S1: Extract sodium-based inorganic compound data from the open source material database and perform preliminary screening based on thermodynamic stability parameters to form a preliminary candidate library;

[0011] S2: Based on the training data of the dynamic solid electrolyte database, Magpie software is used to generate multidimensional structural descriptors, build regression models, and select the machine learning model with the best prediction performance through hyperparameter optimization and cross-validation;

[0012] S3: Inputting the preliminary candidate library obtained in S1 into the machine learning model selected in S2, predicting the ionic conductivity of each material in the preliminary candidate library, and screening out preliminary screening materials whose predicted ionic conductivity values ​​are higher than a first preset threshold;

[0013] S4: Perform simulation calculations on the primary screening materials obtained in S3 based on first principles, including using ab initio molecular dynamics to simulate the sodium ion diffusion trajectory at different temperatures, calculate the diffusion coefficient, and then extrapolate it to the ionic conductivity at room temperature;

[0014] S5: Compare the conductivity predicted by the machine learning model in S3 with the conductivity obtained by the first-principles simulation in S4, and select the material whose predicted and simulated conductivity are both higher than the second preset threshold and whose migration energy barrier is lower as the optimal sodium ion solid electrolyte.

[0015] Furthermore, S1 specifically includes the following steps:

[0016] Extract sodium-containing inorganic compound data from the open-source materials databases Materials Project and / or ICSD;

[0017] collecting key physical parameters in inorganic compound data, the key physical parameters including band gap and energy above convex hull;

[0018] Conduct preliminary screening of materials based on thermodynamic stability parameters;

[0019] Based on the collected key physical parameters, the materials are preliminarily screened using thermodynamic stability parameters to remove materials that are thermodynamically unstable or do not meet basic performance requirements, thus forming a preliminary candidate library containing potentially stable materials.

[0020] Furthermore, S2 specifically includes the following steps:

[0021] Extract the material structure and performance data from the dynamic solid electrolyte database as a training set;

[0022] Magpie software was used to analyze the crystal structure and chemical composition of the materials in the training set, generating multidimensional structural descriptors including the number of atoms, bond length distribution, and element electronegativity;

[0023] Then, based on the multidimensional structural descriptor, a regression model is constructed using a machine learning algorithm in the scikit-learn library;

[0024] The optimal parameter combination of the model is determined through hyperparameter optimization, and the predictive performance of the model is evaluated using cross-validation;

[0025] Select R 2 The model with the highest value and the smallest prediction error is selected as the optimal machine learning model and used for subsequent material property prediction.

[0026] Furthermore, S3 specifically includes the following steps:

[0027] The material data in the preliminary candidate library obtained in S1 is used as input and imported into the machine learning model selected in S2. Based on the selected machine learning model, the ionic conductivity of each material in the preliminary candidate library is predicted;

[0028] Materials with predicted ionic conductivity values ​​higher than a first preset threshold are screened out from the ionic conductivity prediction results to form a primary screening material library.

[0029] Furthermore, S4 specifically includes the following steps:

[0030] The structure of the primary screening materials obtained in S3 was optimized using VASP software, including the projected augmented wave method to describe the interaction between ions and electrons, and the Perdew-Burke-Ernzerhof generalized gradient approximation as the electron exchange correlation functional;

[0031] Based on the initially screened materials with completed structural optimization, ab initio molecular dynamics simulations were performed at multiple preset high temperatures. The mean square displacement at different temperatures was calculated based on the simulation trajectory, and the diffusion coefficient was then obtained. The ionic conductivity at room temperature was extrapolated using the Arrhenius formula.

[0032] Furthermore, S4 also includes the calculation of the diffusion energy barrier, which specifically includes the following process:

[0033] Based on the sodium ion diffusion trajectories obtained from ab initio molecular dynamics simulations, possible sodium ion migration pathways were determined;

[0034] The climbing image nudging elastic band method is used to construct multiple interpolated transition states on the determined migration path;

[0035] Perform structural optimization for each transition state and calculate the system energy;

[0036] The energy of each transition state is corrected by vibrational free energy to eliminate the effects of zero-point energy and temperature on energy;

[0037] Subtracting the corrected energies of adjacent transition states yields the diffusion energy barrier for sodium ions along that path.

[0038] Furthermore, S5 specifically includes the following steps:

[0039] Comparing the conductivity predicted by the machine learning model in S3 with the conductivity obtained by the first principles simulation in S4, the material whose conductivity is higher than the second preset threshold is selected as the optimal candidate material;

[0040] Based on the migration energy barrier obtained in S4, the material with the lowest migration energy barrier is selected from the optimal candidate materials as the optimal sodium ion solid electrolyte.

[0041] The second aspect of the present invention provides an inorganic sodium ion solid electrolyte material screening system based on machine learning and first principles collaborative optimization, comprising:

[0042] Data acquisition and preliminary screening module: extracts sodium-based inorganic compound data from the open source material database, performs preliminary screening based on thermodynamic stability parameters, and forms a preliminary candidate library;

[0043] Machine learning model building module: Extract sodium-based inorganic compound data from open source material databases, conduct preliminary training data based on dynamic solid electrolyte databases based on thermodynamic stability parameters, use Magpie software to generate multidimensional structural descriptors, build regression models, and select the machine learning model with the best prediction performance through hyperparameter optimization and cross-validation;

[0044] High-throughput prediction module: Extracts sodium-based inorganic compound data from an open-source materials database and performs preliminary screening based on thermodynamic stability parameters. The resulting preliminary candidate library is input into a selected machine learning model to predict the ionic conductivity of each material in the preliminary candidate library, and screens out preliminary screening materials with predicted ionic conductivity values ​​above a first preset threshold.

[0045] First-principles simulation module: This module simulates the initial screening materials based on first-principles calculations, including using ab initio molecular dynamics to simulate the sodium ion diffusion trajectory at different temperatures, calculate the diffusion coefficient, and then extrapolate it to the ionic conductivity at room temperature;

[0046] Comprehensive evaluation module: Compare the conductivity predicted by the machine learning model with the conductivity obtained by first-principles simulation, and select materials with both predicted and simulated conductivity higher than the second preset threshold and a low migration energy barrier as the optimal sodium ion solid electrolyte.

[0047] The third aspect of the present invention provides an electronic device comprising a memory and a processor, wherein the processor is used to execute a program in the memory, thereby realizing the above-mentioned method for screening inorganic sodium ion solid electrolyte materials based on collaborative optimization of machine learning and first principles.

[0048] A fourth aspect of the present invention provides a storage medium comprising computer-executable instructions, which, when executed by a computer processor, is used to execute the above-mentioned inorganic sodium ion solid electrolyte material screening method based on collaborative optimization of machine learning and first principles.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] 1. This invention combines machine learning methods with first-principles calculations. During the material screening process, the trained machine learning model is first used to quickly predict the key performance indicators of the material. The prediction results are then carefully verified using first-principles methods, thereby significantly improving the screening efficiency while ensuring calculation accuracy, overcoming the limitations of traditional calculation methods such as large amount of calculation and long cycle.

[0051] 2. This paper constructs a machine learning model based on the high-quality Dynamic Solid Electrolyte Database (DDSE). Magpie is used to generate structural descriptors, and multiple regression algorithms, including random forest, XGBoost, CatBoost, and K-nearest neighbor, are combined for training and parameter adjustment on the scikit-learn platform. Cross-validation is used to evaluate model performance, ensuring the model's practicality in terms of predictive accuracy and applicability.

[0052] 3. The present invention pre-screens the preliminary database based on key physical parameters such as band gap and thermodynamic stability (E_hull). On this basis, a machine learning model is used to achieve high-throughput performance prediction, quickly identifying potential sodium ion solid electrolyte materials, effectively narrowing the material screening range and improving the screening efficiency.

[0053] 4. After screening, the present invention introduces first-principles calculations to further verify the material's ionic conductivity and migration and diffusion energy barriers. The AIMD method is used to obtain the diffusion coefficient at different temperatures, and the ionic conductivity at room temperature is calculated by combining the Arrhenius relationship and the Nernst-Einstein equation. The CI-NEB method is also used to accurately assess the ion migration energy barrier, forming a multi-level, multi-dimensional verification mechanism to improve the accuracy and reliability of the screening results.

[0054] 5. The screening process proposed in this paper is highly scalable and applicable to the development of other types of solid electrolytes or functional materials. By constructing a universal predictive model and computational process, it provides an efficient and feasible technical path for the rapid discovery of new energy storage materials. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 Schematic diagram of the process of screening inorganic sodium ion solid electrolyte materials based on collaborative optimization of machine learning and first principles in the present invention. DETAILED DESCRIPTION

[0056] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. Component models, material names, connection structures, circuit structures, control methods, algorithms, and other features not explicitly described in this technical solution are considered common technical features disclosed in the prior art.

[0057] Example 1

[0058] This embodiment provides a method for screening inorganic sodium ion solid electrolyte materials based on collaborative optimization of machine learning and first principles. Figure 1 , including the following steps:

[0059] S1: Extract sodium-based inorganic compound data from the open source material database and perform preliminary screening based on thermodynamic stability parameters to form a preliminary candidate library;

[0060] S1 specifically includes the following steps:

[0061] Extract sodium-containing inorganic compound data from the open-source materials databases Materials Project and / or ICSD;

[0062] collecting key physical parameters in inorganic compound data, the key physical parameters including band gap and energy above convex hull;

[0063] Conduct preliminary screening of materials based on thermodynamic stability parameters;

[0064] Based on the collected key physical parameters, the materials are preliminarily screened using thermodynamic stability parameters to remove materials that are thermodynamically unstable or do not meet basic performance requirements, thus forming a preliminary candidate library containing potentially stable materials.

[0065] The S1 step aims to efficiently screen sodium-based inorganic compounds with potential application value from massive open-source material databases, laying the foundation for the subsequent development of high-performance sodium-ion solid electrolyte materials. Open-source material databases (such as the Materials Project and / or ICSD) provide a rich material data resource, covering the structure and property information of various known materials. By extracting sodium-containing inorganic compound data from these databases, a basic data set can be obtained for further analysis.

[0066] As one of the key physical parameters of a material, the band gap reflects the minimum energy required for electrons in the material to jump from the valence band to the conduction band, and is closely related to the electrical and optical properties of the material. A smaller band gap usually means that the material has better electrical conductivity, which is crucial for sodium-ion solid electrolyte materials because electrolytes need to have efficient ion transport capabilities. On the other hand, the energy above the convex hull is a key indicator for evaluating the thermodynamic stability of a material. The convex hull is a convex envelope composed of a series of enthalpy-composition points of stable materials. Materials above the convex hull are generally considered to be thermodynamically unstable, while materials below or close to the convex hull are more likely to be thermodynamically stable. By calculating and analyzing the energy above the convex hull of a material, the relative stability of the material can be quickly judged.

[0067] In step S1, data on sodium-containing inorganic compounds are first extracted from the open source materials database, and then the key physical parameters of these materials are collected, including band gaps and energies above the convex hull. Using these key physical parameters, the materials are preliminarily screened based on the principle of thermodynamic stability. Specifically, by setting thresholds for band gaps and energies above the convex hull, materials that are thermodynamically unstable or do not meet basic performance requirements are removed. For example, a lower threshold for the band gap can be set to ensure that the material has sufficient electrical conductivity, while an upper threshold for energies above the convex hull can be set to exclude thermodynamically unstable materials. In this way, sodium-based inorganic compounds with potential stability can be screened out from a large number of candidate materials to form a preliminary candidate library. This step not only narrows the scope of subsequent research, but also improves screening efficiency, reduces unnecessary computational and experimental costs, and provides a high-quality material data set for subsequent further screening based on machine learning and first principles.

[0068] S2: Based on the training data of the dynamic solid electrolyte database, Magpie software is used to generate multidimensional structural descriptors, build regression models, and select the machine learning model with the best prediction performance through hyperparameter optimization and cross-validation;

[0069] In the specific implementation, S2 specifically includes the following steps:

[0070] Extract the material structure and performance data from the dynamic solid electrolyte database as a training set;

[0071] Magpie software was used to analyze the crystal structure and chemical composition of the materials in the training set, generating multidimensional structural descriptors including the number of atoms, bond length distribution, and element electronegativity;

[0072] Then, based on the multidimensional structural descriptor, a regression model is constructed using a machine learning algorithm in the scikit-learn library;

[0073] The optimal parameter combination of the model is determined through hyperparameter optimization, and the predictive performance of the model is evaluated using cross-validation;

[0074] Select R 2 The model with the highest value and the smallest prediction error is selected as the optimal machine learning model and used for subsequent material property prediction.

[0075] In S2, the structural and performance data of the materials were extracted from the dynamic solid electrolyte database to construct a training set. Magpie software was used to analyze the crystal structure and chemical composition of the materials in the training set to generate multidimensional structural descriptors. These descriptors include key features such as the number of atoms, bond length distribution, and element electronegativity, which can characterize the properties of the materials at the atomic and crystal structure levels. Then, the machine learning algorithm in the scikit-learn library was used to build a regression model based on these multidimensional structural descriptors. The optimal parameter combination of the model was determined through hyperparameter optimization, and its predictive performance was evaluated using cross-validation. Finally, R was selected. 2 The model with the highest value and the smallest prediction error is selected as the optimal machine learning model. This process aims to establish a machine learning model that can accurately predict the key properties of sodium ion solid electrolyte materials (such as ionic conductivity) to support the subsequent efficient screening of high-performance materials.

[0076] S3: Inputting the preliminary candidate library obtained in S1 into the machine learning model selected in S2, predicting the ionic conductivity of each material in the preliminary candidate library, and screening out preliminary screening materials whose predicted ionic conductivity values ​​are higher than a first preset threshold;

[0077] In the specific implementation, S3 specifically includes the following steps:

[0078] The material data in the preliminary candidate library obtained in S1 is used as input and imported into the machine learning model selected in S2. Based on the selected machine learning model, the ionic conductivity of each material in the preliminary candidate library is predicted;

[0079] Materials with predicted ionic conductivity values ​​higher than a first preset threshold are screened out from the ionic conductivity prediction results to form a primary screening material library.

[0080] In S3, the material data from the preliminary candidate library obtained in S1 is input into the machine learning model selected in S2. Based on the relationship between characteristics and performance learned during the training phase, the model predicts the ionic conductivity of each material. By setting a first preset threshold for ionic conductivity, materials with ionic conductivity above this threshold are screened from the predicted results to form a preliminary screening material library. This step utilizes the machine learning model to quickly and efficiently identify materials with potential high ionic conductivity from a large number of materials, reducing reliance on experiments and complex computational simulations and accelerating the R&D process.

[0081] S4: Perform simulation calculations on the primary screening materials obtained in S3 based on first principles, including using ab initio molecular dynamics to simulate the sodium ion diffusion trajectory at different temperatures, calculate the diffusion coefficient, and then extrapolate it to the ionic conductivity at room temperature;

[0082] In the specific implementation, S4 specifically includes the following steps:

[0083] The structure of the primary screening materials obtained in S3 was optimized using VASP software, including the projected augmented wave method to describe the interaction between ions and electrons, and the Perdew-Burke-Ernzerhof generalized gradient approximation as the electron exchange correlation functional;

[0084] Based on the initially screened materials with completed structural optimization, ab initio molecular dynamics simulations were performed at multiple preset high temperatures. The mean square displacement at different temperatures was calculated based on the simulation trajectory, and the diffusion coefficient was then obtained. The ionic conductivity at room temperature was extrapolated using the Arrhenius formula.

[0085] In specific implementation, S4 also includes the calculation of the diffusion energy barrier, which specifically includes the following process:

[0086] Based on the sodium ion diffusion trajectories obtained from ab initio molecular dynamics simulations, possible sodium ion migration paths are determined. Specifically, the following steps are performed: analyzing the trajectories obtained from ab initio molecular dynamics simulations to identify the movement trajectories of sodium ions in the material lattice and find their diffusion paths; calculating the energy changes of different paths to determine that the low-energy path is the possible migration path; and combining the crystal structure information of the material to determine the specific location and direction of sodium ion migration, ultimately determining the possible sodium ion migration path.

[0087] The climbing image nudging elastic band method is used to construct multiple interpolated transition states on the determined migration path;

[0088] Perform structural optimization for each transition state and calculate the system energy;

[0089] The energy of each transition state is corrected by vibrational free energy to eliminate the effects of zero-point energy and temperature on energy;

[0090] Subtracting the corrected energies of adjacent transition states yields the diffusion energy barrier for sodium ions along that path.

[0091] In S4, the primary screening material obtained in S3 is simulated and calculated based on the first principles. The specific process is as follows:

[0092] The structure of the prescreened materials was optimized using VASP software, employing the projected augmented wave method to describe ion-electron interactions and the Perdew-Burke-Ernzerhof generalized gradient approximation as the electron exchange-correlation functional. After structural optimization, ab initio molecular dynamics simulations were performed at multiple preset high temperatures to generate simulation trajectories. The mean square displacement at different temperatures was calculated from these trajectories, and the diffusion coefficient was then derived. The ionic conductivity at room temperature was then extrapolated using the Arrhenius formula. Furthermore, the diffusion energy barrier was calculated. First, possible migration paths were identified from the sodium ion diffusion trajectories obtained from the molecular dynamics simulations. Then, the climbing image nudging elastic band method was used to construct multiple interpolated transition states along these paths. For each transition state, the structure was optimized and the system energy was calculated. The energy of each transition state was corrected using the vibrational free energy to eliminate the effects of zero-point energy and temperature. Finally, the corrected energies of adjacent transition states were subtracted to obtain the sodium ion diffusion barrier along the path. The goal of these steps was to verify and refine the predictions of the machine learning model through precise quantum mechanical calculations, thereby screening materials with excellent ionic conductivity.

[0093] S5: Compare the conductivity predicted by the machine learning model in S3 with the conductivity obtained by the first-principles simulation in S4, and select the material whose predicted and simulated conductivity are both higher than the second preset threshold and whose migration energy barrier is lower as the optimal sodium ion solid electrolyte.

[0094] In the specific implementation, S5 specifically includes the following steps:

[0095] Comparing the conductivity predicted by the machine learning model in S3 with the conductivity obtained by the first principles simulation in S4, the material whose conductivity is higher than the second preset threshold is selected as the optimal candidate material;

[0096] Based on the migration energy barrier obtained in S4, the material with the lowest migration energy barrier is selected from the optimal candidate materials as the optimal sodium ion solid electrolyte.

[0097] This process, by combining the results of machine learning and first-principles calculations, ensures that the selected materials have not only high ionic conductivity but also low migration energy barriers, thus ensuring the high performance of the materials in practical applications. In this way, potential sodium-ion solid electrolyte materials can be more accurately evaluated and screened.

[0098] Example 2

[0099] This embodiment provides a method for high-throughput screening of potential inorganic sodium ion solid electrolyte materials from a large number of materials, such as Figure 1 As shown, it mainly includes the following steps:

[0100] S1: Dataset Construction and Preliminary Screening: Sodium-containing inorganic materials were collected from open-source databases such as the Materials Project and ICSD, and their basic properties, including key parameters such as band gap and energy above the convex hull (Ehull), were obtained to establish a preliminary materials database. Subsequently, the preliminary database was filtered based on thermodynamic stability indicators such as band gap range (e.g., >0.5 eV) and Ehull (e.g., <0.1 eV / atom), identifying potentially stable materials and constructing a candidate library of sodium-containing inorganic materials.

[0101] S2: Build and train a machine learning model based on the screened sodium-containing materials, and use the data in the Dynamic Database of Solid Electrolytes (DDSE) to build a training set. The Magpie feature generator is used to extract descriptors such as the composition, crystal structure, and electronic structure of each sample. With the help of the scikit-learn library, a variety of mainstream regression models are trained, including random forest (RF), extreme gradient boosting (XGBoost), categorical feature boosting (CatBoost), and K-nearest neighbor (KNN) algorithms. The dataset is divided into a training set and a test set (ratio of 7:3), and the best hyperparameter combination is searched on the training set. The model performance is evaluated on the test set to determine the coefficient (R 2 ) as the evaluation index, and the model with the best performance is selected as the sodium ion conductivity prediction model.

[0102] S3: Machine learning prediction and high-throughput screening, the trained model is applied to the candidate library of sodium-containing materials to perform high-throughput prediction of their sodium ion conductivity and other properties. Combined with the set performance threshold (such as σ>1×10 -4 S / cm), screening out material samples with high conductivity potential.

[0103] S4: First-principles calculations are performed on some of the materials selected by high-throughput screening to verify their ion conductivity performance, including:

[0104] VASP software was used for structural optimization, the PAW method was adopted to describe the interaction between ions and electrons, and the PBE-GGA functional was selected to describe the exchange-correlation energy.

[0105] Set the K-point grid to Γ points, the plane wave cutoff energy to 500 eV, and the energy convergence criterion to 10 -5 eV.

[0106] Ab initio molecular dynamics (AIMD) simulations were performed at multiple temperature points, ranging from 300–1000 K, with a time scale of no less than 30 ps.

[0107] The diffusion trajectory of sodium ions is extracted based on the simulation trajectory, the mean square displacement at different temperatures is calculated, and the diffusion coefficient D is further calculated:

[0108]

[0109] Where N is the total number of diffusing particles, d is the dimension of the diffusion system, Δt is the time interval, r i (t) is the position of the i-th diffusion particle at time t, and the brackets <> t represents the average value over time t.

[0110] The diffusion coefficients at different temperatures were fitted using the Arrhenius formula:

[0111]

[0112] The diffusion coefficient at room temperature was obtained by extrapolation to 300 K, and the ionic conductivity was calculated according to the Nernst-Einstein equation:

[0113]

[0114] Where n is the density of the diffusing particles (i.e., sodium ions), q is the charge carried by the particles, and k B is the Boltzmann constant.

[0115] S5: Compare model predictions with AIMD validation results. Compare the AIMD simulation results of the screened samples with the machine learning model predictions and analyze the degree of agreement between the two. Prioritize materials that exhibit high ionic conductivity using both methods as the focus for the next phase of research.

[0116] S6: Analysis of Ion Migration Paths and Diffusion Energy Barriers. To further investigate the sodium ion migration mechanism of materials, representative high-performance materials were selected and the CI-NEB method was used to calculate the ion migration paths and their diffusion energy barriers. Using two adjacent stable sodium ion sites in the lattice (e.g., Na1 and Na2) as endpoints, multiple interpolated transition states were constructed and calculated using the hill-climbing image spring method to obtain the minimum energy path and its corresponding migration energy barrier. To improve accuracy, a vibrational free energy correction was introduced to the transition state structure. Ultimately, the diffusion barrier value was determined by the energy difference between adjacent transition states.

[0117] S7: Determine the optimal solid electrolyte material by integrating multi-dimensional performance indicators such as machine learning prediction values, AIMD simulated conductivity, and CI-NEB diffusion energy barrier to screen out the most promising inorganic sodium ion solid electrolyte materials, providing theoretical support for subsequent experimental verification and device development.

[0118] Example 3

[0119] This embodiment provides an inorganic sodium ion solid electrolyte material screening system based on machine learning and first principles collaborative optimization, including:

[0120] Data acquisition and preliminary screening module: extracts sodium-based inorganic compound data from the open source material database, performs preliminary screening based on thermodynamic stability parameters, and forms a preliminary candidate library;

[0121] Machine learning model building module: Extract sodium-based inorganic compound data from open source material databases, conduct preliminary training data based on dynamic solid electrolyte databases based on thermodynamic stability parameters, use Magpie software to generate multidimensional structural descriptors, build regression models, and select the machine learning model with the best prediction performance through hyperparameter optimization and cross-validation;

[0122] High-throughput prediction module: Extracts sodium-based inorganic compound data from an open-source materials database and performs preliminary screening based on thermodynamic stability parameters. The resulting preliminary candidate library is input into a selected machine learning model to predict the ionic conductivity of each material in the preliminary candidate library, and screens out preliminary screening materials with predicted ionic conductivity values ​​above a first preset threshold.

[0123] First-principles simulation module: This module simulates the initial screening materials based on first-principles calculations, including using ab initio molecular dynamics to simulate the sodium ion diffusion trajectory at different temperatures, calculate the diffusion coefficient, and then extrapolate it to the ionic conductivity at room temperature;

[0124] Comprehensive evaluation module: Compare the conductivity predicted by the machine learning model with the conductivity obtained by first-principles simulation, and select materials with both predicted and simulated conductivity higher than the second preset threshold and a low migration energy barrier as the optimal sodium ion solid electrolyte.

[0125] The module division described above is merely a logical functional division. In actual implementation, other divisions may be employed. For example, multiple modules or components may be combined or integrated into another system, or some features may be omitted or not implemented. Furthermore, any coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection via some interface, device, or unit, and may be electrical, mechanical, or other.

[0126] Modules described as separate components may or may not be physically separate, and components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0127] Example 4

[0128] The present embodiment provides an electronic device, including a memory, a processor and a display unit, wherein the memory is used to store programs and data, and the processor is connected to the memory for calling and executing programs in the memory to implement the above-mentioned inorganic sodium ion solid electrolyte material screening method based on machine learning and first principle collaborative optimization. The electronic device can be a personal computer, a server or a mobile terminal, etc., with data processing and analysis capabilities, and can efficiently run machine learning models and algorithms related to first principle calculations. The display unit is used to visualize the key data and results in the screening process, such as the predicted value of the ionic conductivity of the material, the calculated results of the migration energy barrier, and the structural information of the optimal material, so that the user can intuitively understand the screening progress and results, thereby more efficiently carrying out the research and development of inorganic sodium ion solid electrolyte materials.

[0129] Example 5

[0130] The present embodiment provides a storage medium comprising computer executable instructions, which, when executed by a computer processor, are used to perform the inorganic sodium ion solid electrolyte material screening method based on machine learning and first principle collaborative optimization as described above. The storage medium can be a USB flash drive, a hard disk, a solid-state drive, an optical disc or a network attached storage device, etc., which can be read by a computer system and load the corresponding program. The user only needs to connect the storage medium to the computer, access and start the program through a standard file system, and then the method of the present invention can be used for material screening. The program provides a graphical user interface, and the user can easily input a material data set, set screening parameters, and view the predicted ionic conductivity, migration energy barrier and optimal material structure in real time, which greatly improves the R&D efficiency and provides strong support for the design and optimization of inorganic sodium ion solid electrolyte materials.

[0131] The above description of the embodiments is intended to facilitate understanding and use of the invention by those skilled in the art. It will be apparent that those skilled in the art can readily make various modifications to these embodiments and apply the general principles described herein to other embodiments without requiring inventive effort. Therefore, the present invention is not limited to the above-described embodiments. Improvements and modifications made by those skilled in the art based on the disclosure of the present invention, without departing from the scope of the present invention, should be within the scope of protection of the present invention.

Claims

1. A method for screening inorganic sodium ion solid electrolyte materials based on collaborative optimization of machine learning and first principles, characterized in that: The following steps are involved: S1: Extract sodium-based inorganic compound data from the open source material database and perform preliminary screening based on thermodynamic stability parameters to form a preliminary candidate library; S2: Based on the training data of the dynamic solid electrolyte database, Magpie software is used to generate multidimensional structural descriptors, build regression models, and select the machine learning model with the best prediction performance through hyperparameter optimization and cross-validation; S3: Inputting the preliminary candidate library obtained in S1 into the machine learning model selected in S2, predicting the ionic conductivity of each material in the preliminary candidate library, and screening out preliminary screening materials whose predicted ionic conductivity values ​​are higher than a first preset threshold; S4: Perform simulation calculations on the primary screening materials obtained in S3 based on first principles, including using ab initio molecular dynamics to simulate the sodium ion diffusion trajectory at different temperatures, calculate the diffusion coefficient, and then extrapolate it to the ionic conductivity at room temperature; S5: Compare the conductivity predicted by the machine learning model in S3 with the conductivity obtained by the first-principles simulation in S4, and select the material whose predicted and simulated conductivity are both higher than the second preset threshold and whose migration energy barrier is lower as the optimal sodium ion solid electrolyte.

2. The method for screening inorganic sodium ion solid electrolyte materials based on collaborative optimization of machine learning and first principles according to claim 1, characterized in that: S1 specifically includes the following steps: Extract sodium-containing inorganic compound data from the open-source materials databases Materials Project and / or ICSD; collecting key physical parameters in inorganic compound data, the key physical parameters including band gap and energy above convex hull; Conduct preliminary screening of materials based on thermodynamic stability parameters; Based on the collected key physical parameters, the materials are preliminarily screened using thermodynamic stability parameters to remove materials that are thermodynamically unstable or do not meet basic performance requirements, thus forming a preliminary candidate library containing potentially stable materials.

3. The method for screening inorganic sodium ion solid electrolyte materials based on collaborative optimization of machine learning and first principles according to claim 1, characterized in that: S2 specifically includes the following steps: Extract the material structure and performance data from the dynamic solid electrolyte database as a training set; Magpie software was used to analyze the crystal structure and chemical composition of the materials in the training set, generating multidimensional structural descriptors including the number of atoms, bond length distribution, and element electronegativity; Then, based on the multidimensional structural descriptor, a regression model is constructed using a machine learning algorithm in the scikit-learn library; The optimal parameter combination of the model is determined through hyperparameter optimization, and the predictive performance of the model is evaluated using cross-validation; Select R 2 The model with the highest value and the smallest prediction error is selected as the optimal machine learning model and used for subsequent material property prediction.

4. The method for screening inorganic sodium ion solid electrolyte materials based on collaborative optimization of machine learning and first principles according to claim 1, characterized in that: In S3, the following steps are specifically included: The material data in the preliminary candidate library obtained in S1 is used as input and imported into the machine learning model selected in S2. Based on the selected machine learning model, the ionic conductivity of each material in the preliminary candidate library is predicted; Materials with predicted ionic conductivity values ​​higher than a first preset threshold are screened out from the ionic conductivity prediction results to form a primary screening material library.

5. The method for screening inorganic sodium ion solid electrolyte materials based on collaborative optimization of machine learning and first principles according to claim 1, characterized in that: S4 specifically includes the following steps: The structure of the primary screening materials obtained in S3 was optimized using VASP software, including the projected augmented wave method to describe the interaction between ions and electrons, and the Perdew-Burke-Ernzerhof generalized gradient approximation as the electron exchange correlation functional; Based on the initially screened materials with completed structural optimization, ab initio molecular dynamics simulations were performed at multiple preset high temperatures. The mean square displacement at different temperatures was calculated based on the simulation trajectory, and the diffusion coefficient was then obtained. The ionic conductivity at room temperature was extrapolated using the Arrhenius formula.

6. The method for screening inorganic sodium ion solid electrolyte materials based on machine learning and first principles collaborative optimization according to claim 5, characterized in that: S4 also includes the calculation of the diffusion energy barrier, which specifically includes the following steps: Based on the sodium ion diffusion trajectories obtained from ab initio molecular dynamics simulations, possible sodium ion migration pathways were determined; The climbing image nudging elastic band method is used to construct multiple interpolated transition states on the determined migration path; Perform structural optimization for each transition state and calculate the system energy; The energy of each transition state is corrected by vibrational free energy to eliminate the effects of zero-point energy and temperature on energy; Subtracting the corrected energies of adjacent transition states yields the diffusion energy barrier for sodium ions along that path.

7. The method for screening inorganic sodium ion solid electrolyte materials based on collaborative optimization of machine learning and first principles according to claim 1, characterized in that: S5 specifically includes the following steps: Comparing the conductivity predicted by the machine learning model in S3 with the conductivity obtained by the first principles simulation in S4, the material whose conductivity is higher than the second preset threshold is selected as the optimal candidate material; Based on the migration energy barrier obtained in S4, the material with the lowest migration energy barrier is selected from the optimal candidate materials as the optimal sodium ion solid electrolyte.

8. An inorganic sodium ion solid electrolyte material screening system based on machine learning and first principles collaborative optimization, characterized in that: include: Data acquisition and preliminary screening module: extracts sodium-based inorganic compound data from the open source material database, performs preliminary screening based on thermodynamic stability parameters, and forms a preliminary candidate library; Machine learning model building module: Extract sodium-based inorganic compound data from open source material databases, conduct preliminary training data based on dynamic solid electrolyte databases based on thermodynamic stability parameters, use Magpie software to generate multidimensional structural descriptors, build regression models, and select the machine learning model with the best prediction performance through hyperparameter optimization and cross-validation; High-throughput prediction module: Extracts sodium-based inorganic compound data from an open-source materials database and performs preliminary screening based on thermodynamic stability parameters. The resulting preliminary candidate library is input into a selected machine learning model to predict the ionic conductivity of each material in the preliminary candidate library, and screens out preliminary screening materials with predicted ionic conductivity values ​​above a first preset threshold. First-principles simulation module: This module simulates the initial screening materials based on first-principles calculations, including using ab initio molecular dynamics to simulate the sodium ion diffusion trajectory at different temperatures, calculate the diffusion coefficient, and then extrapolate it to the ionic conductivity at room temperature; Comprehensive evaluation module: Compare the conductivity predicted by the machine learning model with the conductivity obtained by first-principles simulation, and select materials with both predicted and simulated conductivity higher than the second preset threshold and a low migration energy barrier as the optimal sodium ion solid electrolyte.

9. An electronic device comprising a memory and a processor, characterized in that: The processor is used to execute the program in the memory to implement the inorganic sodium ion solid electrolyte material screening method based on collaborative optimization of machine learning and first principles as described in any one of claims 1 to 7.

10. A storage medium containing computer-executable instructions, characterized in that: The storage medium of the computer-executable instructions, when executed by a computer processor, is used to execute the inorganic sodium ion solid electrolyte material screening method based on collaborative optimization of machine learning and first principles as described in any one of claims 1 to 7.

Images