Method for simulating polycrystalline solid electrolyte and lithium metal interface based on machine learning potential

Abstract

The invention discloses a method for simulating a polycrystalline solid electrolyte and lithium metal interface based on a machine learning potential. The method comprises the following steps: firstly, constructing a plurality of groups of interface structures consisting of a plurality of typical crystal faces of the solid electrolyte and a lithium crystal face; then, aiming at different crystal face combinations, sampling by adopting a self-adaptive sampling strategy; screening out a reasonable structure sample; structural energy and atomic stress data are obtained through calculation according to a first principle; circularly sampling and training a machine learning potential function model; and finally, carrying out large-scale molecular dynamics simulation based on the trained potential function, and revealing interface structure evolution and an SEI formation mechanism. According to the method, systematicness and universality of polycrystal face sampling modeling are highlighted, generalization ability and simulation precision of a potential function are improved, and an efficient and extensible method is provided for research on stability of a polycrystal solid electrolyte / lithium metal interface.

Description

Technical Field

[0001] This invention relates to the field of solid electrolyte interface design and molecular dynamics simulation technology, and particularly to a method for constructing a machine learning potential function for the interface between a solid electrolyte and lithium metal, and a multi-scale simulation method for the properties of the interface between a solid electrolyte and lithium metal. Background Technology

[0002] With the rapid development of portable electronic devices, electric vehicles, and large-scale energy storage systems, the demand for high-energy-density and high-safety lithium-ion batteries is increasing. Existing commercial liquid electrolyte batteries pose safety hazards such as flammability, leakage, and lithium dendrite puncture. All-solid-state lithium metal batteries, which use a solid electrolyte instead of a liquid electrolyte and lithium metal as the negative electrode material, theoretically possess higher energy density and superior safety, and are widely considered the core development direction for next-generation battery systems.

[0003] However, the interfacial stability between the solid electrolyte and lithium metal severely restricts the development of all-solid-state lithium metal batteries. Due to the extremely strong reducing properties of lithium metal, the solid electrolyte is prone to interfacial decomposition, generating a series of electronically insulating but ionicly conductive byproducts. These products accumulate to form the solid electrolyte interface (SEI), which has a complex structure and unstable evolution, easily causing problems such as lithium dendrite puncture, interfacial contact degradation, and increased impedance, leading to rapid capacity decay and deterioration of cycle performance. Therefore, a deep understanding and regulation of the structural evolution process of the solid electrolyte-lithium metal interface is a key technical challenge for promoting the development of solid-state batteries.

[0004] In terms of research methods, traditional experimental techniques (such as NMR, TOF-SIMS, RIM, etc.) can obtain information on the composition and distribution of interface elements, but are limited by spatial resolution, temporal resolution, and operational conditions, making it difficult to directly observe the evolution process at the atomic scale of the interface. Meanwhile, while first-principles molecular dynamics (AIMD) can provide precise descriptions at the electronic structure level, its computational resources are extremely high, limiting it to simulations with a few hundred atoms and picosecond timescales, making it unsuitable for studying the evolution of SEI at the nanoscale and nanosecond timescales in real battery systems. Therefore, the machine learning-based interatomic potential (MLIP) method, which has emerged in recent years, offers a new direction for solving this problem. MLIP establishes regression models of interatomic interactions by learning from first-principles datasets, combining near-DFT accuracy with high computational efficiency, and can support molecular dynamics simulations involving tens of thousands of atoms and nanosecond or even longer timescales. Various MLIP models, such as moment tensor potential, depth potential, and Gaussian potential, have been used for modeling complex material systems and have proven excellent performance in studies of crystal defects, interfaces, and phase transitions.

[0005] The design of sampling strategies is crucial in training MLIP models. While traditional equal-interval sampling and perturbation sampling can cover part of the configuration space, they often suffer from high data redundancy and low efficiency. Furthermore, significant differences in atomic arrangement, density distribution, and contact configuration between different crystal planes can affect the interfacial reaction rate and SEI evolution path. Therefore, establishing a reusable, high-precision interface modeling methodology applicable to multi-faceted combinations, combined with efficient sampling strategies and high-quality potential function construction processes, is of great significance for interface design and stability assessment in all-solid-state batteries. Summary of the Invention

[0006] This invention addresses the shortcomings of existing technologies by providing a simulation method for the interface between a polycrystalline solid-state electrolyte and lithium metal based on machine learning potentials. By combining polycrystalline structure construction, machine learning potential function training, and large-scale molecular dynamics simulations, the evolution process of the heterogeneous interface can be accurately captured.

[0007] To achieve the above-mentioned objectives, the technical solution adopted by the present invention is as follows:

[0008] A simulation method for the interface between a polycrystalline solid electrolyte and lithium metal based on machine learning potentials, characterized by the following steps:

[0009] S1 Polycrystalline Solid Electrolyte and Lithium Metal Interface Structure Model Construction: Multiple heterojunction interface models were constructed by combining several common crystal planes of solid electrolyte with lithium metal crystal planes as initial structural inputs;

[0010] S2 Data Preparation: For the constructed polycrystalline solid electrolyte-lithium metal interface structure model, first-principles molecular dynamics calculations were used to collect structural, force, and energy information of several interfaces;

[0011] The collected data is preprocessed, and the preprocessing operations include one or more of the following: interval sampling, data cleaning, data standardization, and removal of non-converged outliers.

[0012] The initial potential function is obtained by pre-training using a pre-processed dataset;

[0013] S3 Structural Sampling: Using the initial potential function obtained from the pre-training in step S2, the polycrystalline solid electrolyte and lithium metal interface structure model in S1 is simulated using molecular dynamics simulation software, and a dual adaptive sampling method is used to obtain representative structural samples.

[0014] S4 Structure Screening: Further screening of representative structure samples obtained in S3. The screening process includes: using one or both of the D-optimality method or the atomic distance judgment method to screen out unreasonable structures and form data to be labeled.

[0015] S5 First Principles Annotation: The energy and interatomic force information of the data to be annotated obtained from S4 screening are calculated using density functional theory.

[0016] Data processing of labeled data: Remove data from labeled data that did not converge in first-principles calculations;

[0017] The structural sample data that has been calculated, labeled and processed is added to the dataset in S2 to form a new dataset;

[0018] S6 Potential Function Training: Using the new dataset in S5, train the selected machine learning atomic potential function model.

[0019] S7: Repeat the S3-S6 process. When the trained potential function meets the termination condition, select a new heterogeneous interface model from the multiple heterogeneous interface models constructed in S1 and repeat the S3-S6 process again. When all heterogeneous interfaces in S1 have been used and the termination condition is met, the potential function training ends. The termination condition is to reach the maximum number of iterations and to reach the potential function convergence criterion.

[0020] S8 Interface Evolution Simulation: Combining molecular dynamics simulation software, based on a trained potential energy function, it performs large-scale, long-term molecular dynamics simulations, considering scales from the molecular scale to the mesoscopic scale; simulating forces, atomic diffusion, etc., specifically including system energy, interatomic forces, and diffusion at the interface;

[0021] S9 Result Analysis: After the simulation is completed, the simulation results are analyzed to extract information such as atomic information, atomic position information, and atomic migration rate, and to obtain information including ion diffusion rate, elemental distribution, and structural distribution.

[0022] Preferably, in step S1, the multiple common crystal planes of the solid electrolyte include low-index crystal planes and crystal planes corresponding to the three strongest peaks in the X-ray diffraction pattern.

[0023] Preferably, in step S2, first-principles molecular dynamics calculations are used to collect structural, force, and energy information of several interfaces. The first-principles molecular dynamics calculation software includes VASP, QuantumESPRESSO, and GPW.

[0024] Preferably, the molecular dynamics simulation software used in step S3 is Lammps.

[0025] Preferably, in step S4, the atomic distance judgment method selects a standard where the atomic distance is too close.

[0026] Preferably, in step S4, the parameter γ ∈ (2, 10) in the D-optimality method.

[0027] Preferably, in step S5, density functional theory (DFT) is used to calculate its energy and interatomic force information, and the first-principles molecular dynamics calculation software includes VASP, QuantumESPRESSO, and GPW.

[0028] Preferably, the termination condition in step S7 is that the potential function convergence criterion is reached, and the potential function convergence criterion is that no new structure is generated in the dual adaptive sampling method.

[0029] Preferably, the large-scale, long-term molecular dynamics simulation in step S8 refers to a large-scale simulation involving 20,000 to 100,000 atoms.

[0030] Preferably, the large-scale, long-term molecular dynamics simulation in step S8 is a simulation time of 1ns-10ns.

[0031] Preferably, in step S9, the simulation includes the following:

[0032] a. Simulate the interaction of forces between atoms, i.e., the forces acting on atoms;

[0033] b. Simulate and calculate the system's energy;

[0034] c. Atomic diffusion processes at heterogeneous interfaces;

[0035] d. Nucleation and evolution of SEI on heterogeneous interfaces;

[0036] e. SEI composition and elemental distribution at heterogeneous interfaces.

[0037] Compared with the prior art, the advantages of the present invention are as follows:

[0038] 1. Overcoming the bottleneck of polycrystalline interface modeling

[0039] This study achieves, for the first time, a systematic modeling capability for the interface between polycrystalline solid electrolytes and lithium metal, overcoming the limitations of traditional methods in simulating single crystal planes. By integrating multiple typical crystal plane combinations (including low-index crystal planes and XRD characteristic crystal planes), a structurally representative heterostructure interface model is constructed, significantly improving the accuracy of predicting the interface evolution behavior of real polycrystalline materials.

[0040] 2. Improve the efficiency and accuracy of potential function training.

[0041] This innovative approach integrates a dual adaptive sampling strategy with an active learning mechanism to dynamically capture key configuration samples of interface responses. By combining the D-optimality method and atomic distance screening technology, redundant data is effectively eliminated, reducing the potential function training cycle to less than 30% of that of traditional methods while maintaining quantum computing accuracy.

[0042] 3. Achieve large-scale interface evolution analysis

[0043] Leveraging the computational efficiency advantages of machine learning potential functions, this technology supports molecular dynamics simulations of systems at the nanosecond timescale, down to the tens of thousands of atoms. Breaking through the spatial and temporal limitations of first-principles calculations, it reveals for the first time at the atomic scale the entire process of SEI nucleation, growth, and structural evolution at the polycrystalline solid electrolyte interface.

[0044] 4. Revealing the evolution law of crystal face dependence

[0045] By comparing simulation results of different crystal plane combinations (such as Li6PS5Cl(100) / (222) / (311) / (220) crystal planes), the regulatory mechanism of crystal plane orientation on SEI layer thickness, ion diffusion rate, and interface stability was clarified. This provides a theoretical basis for the directional design of highly stable electrolyte interfaces.

[0046] 5. Enhance the universality of industrial applications

[0047] It is compatible with mainstream solid-state electrolyte systems such as sulfides and oxides, and supports simulation of multiple metal anode interfaces, including lithium, sodium, and potassium. Its modular design can be directly embedded into the solid-state battery R&D process, significantly reducing experimental trial-and-error costs and accelerating the development of high-performance all-solid-state batteries.

[0048] 6. Establish cross-scale linkage bridges

[0049] Bridging the gap between the precision of quantum computing and the observation of mesoscopic phenomena, this study aims to bridge the cognitive gap between atomic-scale interfacial reaction mechanisms and macroscopic battery performance. By quantifying key parameters such as the SEI layer structure and elemental distribution, it promotes the transformation of battery failure mechanism research from empirical speculation to model-driven approaches. Attached Figure Description

[0050] Figure 1 This is a flowchart of a simulation method for the interface between a polycrystalline solid electrolyte and lithium metal according to an embodiment of the present invention;

[0051] Figure 2 This is a comparison error diagram between the potential function and DFT calculation in an embodiment of the present invention;

[0052] Figure 3 This is the initial simulation diagram of Li6PS5Cl(100)|Li(001) in Embodiment 1 of the present invention;

[0053] Figure 4 This is an atomic distribution diagram of the simulated process of Li6PS5Cl(100)|Li(001) in Example 1 of the present invention;

[0054] Figure 5 This is the initial simulation diagram of Li6PS5Cl(222)|Li(001) in Embodiment 2 of the present invention;

[0055] Figure 6 This is an atomic distribution diagram of the simulated process of Li6PS5Cl(222)|Li(001) in Embodiment 2 of the present invention;

[0056] Figure 7 This is the initial simulation diagram of Li6PS5Cl(311)|Li(001) in Embodiment 3 of the present invention;

[0057] Figure 8 This is the atomic distribution diagram of the simulated process of Li6PS5Cl(311)|Li(001) in Example 3 of the present invention;

[0058] Figure 9 This is the initial simulation diagram of Li6PS5Cl(220)|Li(001) in Embodiment 4 of the present invention;

[0059] Figure 10 This is the atomic distribution diagram of the simulated process of Li6PS5Cl(220)|Li(001) in Example 4 of the present invention. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and examples.

[0061] like Figure 1 As shown, this invention provides a simulation method for the interface between a polycrystalline solid electrolyte and lithium metal based on machine learning potential, specifically including the following steps:

[0062] S1 Polycrystalline Solid Electrolyte and Lithium Metal Interface Structure Model Construction: Multiple heterojunction interface models were constructed by combining several common crystal planes of solid electrolyte with lithium metal crystal planes as initial structural inputs;

[0063] S2 Data Preparation: For the constructed polycrystalline solid electrolyte-lithium metal interface structure model, first-principles molecular dynamics calculations were used to collect structural, force, and energy information of several interfaces;

[0064] The collected data is preprocessed, and the preprocessing operations include one or more of the following: interval sampling, data cleaning, data standardization, and removal of non-converged outliers.

[0065] The initial potential function is obtained by pre-training using a pre-processed dataset;

[0066] S3 Structural Sampling: Using the initial potential function obtained from the pre-training in step S2, the polycrystalline solid electrolyte and lithium metal interface structure model in S1 is simulated using molecular dynamics simulation software, and a dual adaptive sampling method is used to obtain representative structural samples.

[0067] S4 Structure Screening: Further screening of representative structure samples obtained in S3. The screening process includes: using one or both of the D-optimality method or the atomic distance judgment method to screen out unreasonable structures and form data to be labeled.

[0068] S5 First Principles Annotation: The energy and interatomic force information of the data to be annotated obtained from S4 screening are calculated using density functional theory.

[0069] Data processing of labeled data: Remove data from labeled data that did not converge in first-principles calculations;

[0070] The structural sample data that has been calculated, labeled and processed is added to the dataset in S2 to form a new dataset;

[0071] S6 Potential Function Training: Using the new dataset in S5, train the selected machine learning atomic potential function model.

[0072] S7: Repeat the S3-S6 process. When the trained potential function meets the termination condition, select a new heterogeneous interface model from the multiple heterogeneous interface models constructed in S1 and repeat the S3-S6 process again. When all heterogeneous interfaces in S1 have been used and the termination condition is met, the potential function training ends. The termination condition is to reach the maximum number of iterations and to reach the potential function convergence criterion.

[0073] S8 Interface Evolution Simulation: Combining molecular dynamics simulation software, based on a trained potential energy function, it performs large-scale, long-term molecular dynamics simulations, considering scales from the molecular scale to the mesoscopic scale; simulating forces, atomic diffusion, etc., specifically including system energy, interatomic forces, and diffusion at the interface;

[0074] S9 Result Analysis: After the simulation is completed, the simulation results are analyzed to extract information such as atomic information, atomic position information, and atomic migration rate, and to obtain information including ion diffusion rate, elemental distribution, and structural distribution.

[0075] Example 1:

[0076] Simulation of the heterostructure interface between the (100) crystal plane of Li6PS5Cl solid electrolyte and the (001) plane of lithium metal;

[0077] S1 Polycrystalline Solid Electrolyte and Lithium Metal Interface Structure Model Construction: The (100), (222), (220), and (311) crystal planes of Li6PS5Cl were selected and combined with the (001) crystal plane of Li to construct four initial heterostructure interfaces.

[0078] S2 Data Preparation: For the constructed Li6PS5Cl(100) / Li(001) interface structure model, first-principles molecular dynamics was used to calculate and collect the structure, force and energy information of several interfaces; the calculated data were subjected to interval sampling, data cleaning and removal of non-converged abnormal results; then the initial potential function was obtained by pre-training using the pre-processed dataset.

[0079] S3 Structural Sampling: Using the initial potential function obtained from the pre-training in step S2, the Li6PS5Cl(100) / Li(001) interface structure model in S1 is simulated using molecular dynamics simulation software, and a dual adaptive sampling method is used to obtain representative structural samples.

[0080] S4 structure screening: Simultaneously employing the D-optimality method (where γ∈(2,10)) and the atomic distance judgment method, where... The representative Li6PS5Cl(100) / Li(001) interface structure samples obtained in S3 were further screened to form data to be labeled.

[0081] S5 First Principles Labeling: The energy and interatomic force information of the Li6PS5Cl(100) / Li(001) data to be labeled obtained from S4 were calculated using density functional theory VASP software; the structural sample data after calculation, labeling and processing were added to the dataset in S2 to form a new dataset.

[0082] S6 Potential Function Training: Using the new dataset from S5, the model is trained using the MTP potential function, where the cutoff radius is... The order of the moment tensor is 6.

[0083] S7: Repeat steps S3-S6 until no new structure can be selected from Li6PS5Cl(100) / Li(001). Then, replace Li6PS5Cl(100) / Li(001) with Li6PS5Cl(311) / Li(001) and continue repeating steps S3-S6 until no new structure can be selected. Similarly, replace Li6PS5Cl(220) / Li(001) with Li6PS5Cl(100) / Li(001).

[0084] The S3-S6 process is repeated with Li6PS5Cl(311) / Li(001) until no new structure can be selected. Finally, Li6PS5Cl(222) / Li(001) is used to replace Li6PS5Cl(220) / Li(001), and the S3-S6 process is repeated until no new structure can be selected. The potential function construction process ends. After training, the potential function is compared with the DFT to obtain the potential function error distribution map, as shown in the figure. Figure 2As shown, the mean absolute error of this potential function is It can reach a level very close to that of DFT.

[0085] S8 Interface Evolution Simulation: Using the molecular dynamics simulation software Lammps, large-scale, long-term molecular dynamics simulations of the Li6PS5Cl(100)|Li(001) interface were performed based on the potential energy function trained in step S7. The NPT ensemble was used, with a temperature of 300 K and a pressure of 100 Bar. The initial simulation model was as follows: Figure 3 As shown, the simulation has reached and exceeded the mesoscopic scale; the simulation system's energy, interatomic forces, and diffusion at the interface;

[0086] S9 Result Analysis: After the simulation of the Li6PS5Cl(100)|Li(001) interface was completed, the simulation results were analyzed, and information such as atomic information, atomic position information, and atomic migration rate were extracted. Information including ion diffusion rate, elemental distribution, and structural distribution was obtained. The atomic distribution during the simulation process is shown in the figure below. Figure 4 The simulation clearly shows the SEI structure, with an amorphous region of 2.98 nm, a crystalline region of 2.53 nm, and an SEI surface region of 0.5 nm.

[0087] Example 2

[0088] The MTP potential function obtained in Example 1 is applied to this example, i.e., steps S1-S7 in the example.

[0089] S8 Interface Evolution Simulation: Based on the MTP potential function obtained in Example 1, large-scale, long-term molecular dynamics simulations of the Li6PS5Cl(222)|Li(001) interface were performed using the molecular dynamics simulation software Lammps. The NPT ensemble was employed, with a temperature of 300 K and a pressure of 100 Bar. The initial simulation model was as follows: Figure 5 As shown, the simulation has reached and exceeded the mesoscopic scale; the simulation system's energy, interatomic forces, and diffusion at the interface;

[0090] S9 Result Analysis: After the simulation of the Li6PS5Cl(222)|Li(001) interface was completed, the simulation results were analyzed, and information such as atomic information, atomic position information, and atomic migration rate were extracted. Information including ion diffusion rate, elemental distribution, and structural distribution was obtained. Among these, the atomic distribution during the simulation process is shown in the figure below. Figure 6 As can be seen, the simulated SEI structure is clearly different from that of Example 1, with no amorphous region, a crystalline region of 2.85 nm, and an SEI surface region of 0.4 nm.

[0091] Example 3

[0092] The MTP potential function obtained in Example 1 is applied to this example, i.e., steps S1-S7 in the example.

[0093] S8 Interface Evolution Simulation: Based on the MTP potential function obtained in Example 1, large-scale, long-term molecular dynamics simulations of the Li6PS5Cl(311)|Li(001) interface were performed using the molecular dynamics simulation software Lammps. The NPT ensemble was employed, with a temperature of 300 K and a pressure of 100 Bar. The initial simulation model was as follows: Figure 7 As shown, the simulation has reached and exceeded the mesoscopic scale; the simulation system's energy, interatomic forces, and diffusion at the interface;

[0094] S9 Result Analysis: After the simulation of the Li6PS5Cl(311)|Li(001) interface was completed, the simulation results were analyzed, and information such as atomic information, atomic position information, and atomic migration rate were extracted. Information including ion diffusion rate, elemental distribution, and structural distribution was obtained. Among these, the atomic distribution during the simulation process is shown in the figure below. Figure 8 As can be seen, the simulated SEI structure is clearly different from that of Example 1, with the amorphous region being 1.36 nm, the crystalline region being 3.57 nm, and the SEI surface region being 0.5 nm.

[0095] Example 4

[0096] The MTP potential function obtained in Example 1 is applied to this example, i.e., steps S1-S7 in the example.

[0097] S8 Interface Evolution Simulation: Based on the MTP potential function obtained in Example 1, large-scale, long-term molecular dynamics simulations of the Li6PS5Cl(220)|Li(001) interface were performed using the molecular dynamics simulation software Lammps. The NPT ensemble was employed, with a temperature of 300 K and a pressure of 100 Bar. The initial simulation model was as follows: Figure 9 As shown, the simulation has reached and exceeded the mesoscopic scale; the simulation system's energy, interatomic forces, and diffusion at the interface;

[0098] S9 Result Analysis: After the simulation of the Li6PS5Cl(220)|Li(001) interface was completed, the simulation results were analyzed, and information such as atomic information, atomic position information, and atomic migration rate were extracted. Information including ion diffusion rate, elemental distribution, and structural distribution was obtained. Among these, the atomic distribution during the simulation process is shown in the figure below. Figure 10 As can be seen, the simulated SEI structure is clearly different from that of Example 1, with no amorphous region, a crystalline region of 4.68 nm, and an SEI surface region of 0.4 nm.

[0099] Example 5 provides a simulation system for the interface between a polycrystalline solid electrolyte and lithium metal based on machine learning potential. This system can be used to implement the above-described simulation method for the interface between a polycrystalline solid electrolyte and lithium metal, specifically including:

[0100] Crystal plane combination modeling module: configured to automatically select multiple crystal planes of solid electrolyte, including low index crystal planes and crystal planes corresponding to the three strongest peaks in XRD diffraction patterns, and combinations with lithium metal crystal planes; construct multiple sets of initial structure models of heterointerfaces, and output optimized configurations that meet the condition of minimizing interfacial contact energy; the crystal plane selection criteria are based on crystal databases and first-principles surface energy calculation results.

[0101] Dynamic sampling engine module: integrates the Lammps molecular dynamics simulation interface, executes a dual adaptive sampling algorithm; monitors the rate of change of atomic configuration space in real time, and automatically terminates sampling when the rate of discovery of new configurations is <5%; outputs a representative set of structural samples containing energy / force mutation points.

[0102] Intelligent screening and verification module: configured to execute the D-optimality method and atomic distance judgment method in parallel; simultaneously verify the convergence of first-principles calculations and output a labeled dataset that meets the precision of quantum computing.

[0103] Moment Tensor Potential Training Core: Built-in Moment Tensor Potential (MTP) model with fixed parameter cutoff radius. Tensor order is 6th and non-adjustable; real-time monitoring of training error satisfies energy ≤ 10meV / atom. It supports relay training of multi-interface models and achieves minute-level iteration in a GPU-accelerated environment.

[0104] Mesoscale simulator: The hardware acceleration unit supports real-time calculations of systems with 20,000-100,000 atoms; preset NPT ensemble parameters: temperature 300K, pressure 100Bar, time scale 1-10ns; outputs atomic trajectory data and SEI hierarchical structure evolution parameters.

[0105] Visualization analysis terminal: generates thermal maps of lithium-ion migration paths, 3D cloud maps of elemental distribution, and quantitative maps of SEI crystal / amorphous region thickness; data interface supports exporting standard format analysis reports.

[0106] Example 6 provides a terminal device including a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve corresponding method flows or corresponding functions. The processor described in this embodiment can be used for the operation of a simulation method of the polycrystalline solid electrolyte-lithium metal interface.

[0107] Example 7 provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a terminal device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and extended storage media supported by the terminal device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device.

[0108] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the simulation method for the polycrystalline solid electrolyte-lithium metal interface in the above embodiments; one or more instructions in the computer-readable storage medium are loaded and executed by a processor.

[0109] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0110] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0111] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0112] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0113] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the implementation methods of the present invention, and should be understood that the scope of protection of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of the present invention.

Claims

1. A simulation method for the interface between a polycrystalline solid electrolyte and lithium metal based on machine learning potential, characterized in that, Specifically, the following steps are included: S1 Polycrystalline Solid Electrolyte and Lithium Metal Interface Structure Model Construction: Multiple heterojunction interface models were constructed by combining several common crystal planes of solid electrolyte with lithium metal crystal planes as initial structural inputs; S2 Data Preparation: For the constructed polycrystalline solid electrolyte-lithium metal interface structure model, first-principles molecular dynamics calculations were used to collect structural, force, and energy information of several interfaces; The collected data is preprocessed, and the preprocessing operations include one or more of the following: interval sampling, data cleaning, data standardization, and removal of non-converged outliers. The initial potential function is obtained by pre-training using a pre-processed dataset; S3 Structural Sampling: Using the initial potential function obtained from the pre-training in step S2, the polycrystalline solid electrolyte and lithium metal interface structure model in S1 is simulated using molecular dynamics simulation software, and a dual adaptive sampling method is used to obtain representative structural samples. S4 Structure Screening: Further screening of representative structure samples obtained in S3. The screening process includes: using one or both of the D-optimality method or the atomic distance judgment method to screen out unreasonable structures and form data to be labeled. S5 First Principles Annotation: The energy and interatomic force information of the data to be annotated obtained from S4 screening are calculated using density functional theory. Data processing of labeled data: Remove data from labeled data that did not converge in first-principles calculations; The structural sample data that has been calculated, labeled and processed is added to the dataset in S2 to form a new dataset; S6 Potential Function Training: Using the new dataset in S5, train the selected machine learning atomic potential function model. S7: Repeat S3-S6. When the trained potential function meets the termination condition, select a new heterogeneous interface model from the multiple heterogeneous interface models constructed in S1 and repeat S3-S6 again. When all heterogeneous interfaces in S1 have been used and the termination condition is met, the potential function training ends. The termination condition is to reach the maximum number of iterations and to reach the potential function convergence criterion. S8 Interface Evolution Simulation: Combining molecular dynamics simulation software, based on a trained potential energy function, it performs large-scale, long-term molecular dynamics simulations, considering scales from the molecular scale to the mesoscopic scale; simulating forces, atomic diffusion, etc., specifically including system energy, interatomic forces, and diffusion at the interface; S9 Result Analysis: After the simulation is completed, the simulation results are analyzed to extract information such as atomic information, atomic position information, and atomic migration rate, and to obtain information including ion diffusion rate, elemental distribution, and structural distribution.

2. The simulation method for the interface between a polycrystalline solid electrolyte and lithium metal based on machine learning potential according to claim 1, characterized in that, In step S1, the solid electrolyte has several common crystal planes, including low-index crystal planes and crystal planes corresponding to the three strongest peaks in the X-ray diffraction pattern.

3. The simulation method for the interface between a polycrystalline solid electrolyte and lithium metal based on machine learning potential according to claim 1, characterized in that, In step S2, first-principles molecular dynamics calculations are used to collect structural, force, and energy information of several interfaces. The first-principles molecular dynamics calculation software includes VASP, QuantumESPRESSO, and GPW.

4. The simulation method for the interface between a polycrystalline solid electrolyte and lithium metal based on machine learning potential according to claim 1, characterized in that, The molecular dynamics simulation software used in step S3 is Lammps.

5. The multi-scale simulation method for the interface properties of polycrystalline solid electrolyte and lithium metal based on machine learning potential according to claim 1, characterized in that, In step S4, the atomic distance judgment method selects the standard for excessively close atomic distances.

6. The simulation method for the interface between a polycrystalline solid electrolyte and lithium metal based on machine learning potential according to claim 1, characterized in that, In step S4, the parameter γ ∈ (2, 10) in the D-optimality method.

7. The simulation method for the interface between a polycrystalline solid electrolyte and lithium metal based on machine learning potential according to claim 1, characterized in that, In step S5, density functional theory (DFT) is used to calculate the energy and interatomic forces. First-principles molecular dynamics calculation software includes VASP, QuantumESPRESSO, and GPW.

8. The simulation method for the interface between a polycrystalline solid electrolyte and lithium metal based on machine learning potential according to claim 1, characterized in that, The termination condition in step S7 is that the potential function convergence criterion is reached, which is that no new structure is generated in the dual adaptive sampling method.

9. The simulation method for the interface between a polycrystalline solid electrolyte and lithium metal based on machine learning potential according to claim 1, characterized in that, The large-scale, long-term molecular dynamics simulation in step S8 refers to a simulation with 20,000 to 100,000 atoms.

10. The simulation method for the interface between a polycrystalline solid electrolyte and lithium metal based on machine learning potential according to claim 1, characterized in that, In step S8, the large-scale, long-term molecular dynamics simulation is performed, with the long duration being 1 ns to 10 ns.

11. The simulation method for the interface between a polycrystalline solid electrolyte and lithium metal based on machine learning potential according to claim 1, characterized in that, In step S9, the simulation includes the following: a. Simulate the interaction of forces between atoms, i.e., the forces acting on atoms; b. Simulate and calculate the system's energy; c. Atomic diffusion processes at heterogeneous interfaces; d. Nucleation and evolution of SEI on heterogeneous interfaces; e. SEI composition and elemental distribution at heterogeneous interfaces.

Images