Method and apparatus for obtaining potential function model for conductor / insulator interface simulation

Abstract

The application provides a potential function model acquisition method and device for conductor / insulator interface simulation, and the method comprises the following steps: the dielectric response of the conductor and the insulator is respectively processed in two sub-frames and then combined into the overall dielectric response of the system. Based on the hybrid description of the dielectric response, the charge distribution of the system under the given external electric field boundary condition can be obtained. Then, the long-range and short-range interaction separation is adopted to describe the potential energy surface of the system, and finally, the machine learning potential energy surface model containing the dielectric response of the conductor / insulator interface is obtained. The application can simultaneously describe the dielectric response of the conductor and the insulator.

Description

Technical Field

[0001] This invention relates to the fields of computational chemistry and physics, and in particular to a method and apparatus for obtaining potential function models for conductor / insulator interface simulation. Background Technology

[0002] Computational simulations provide microscopic information about electrochemical interfaces, serving as a crucial research tool for understanding the characteristics and mechanisms of interfacial reactions. However, achieving accurate simulations of electrochemical interfaces requires simultaneously satisfying two conditions: sufficient sampling and inclusion of electronic structure information. In principle, first-principles molecular simulation methods, such as first-principles molecular dynamics, can be used to simulate the system to meet these two conditions. However, due to computational limitations, first-principles molecular simulations can only be applied to simple model systems with thousands of atoms, simulating processes lasting hundreds of picoseconds. In reality, electrochemical interfaces typically involve much larger time and spatial scales, posing a significant challenge to existing computational simulation methods. In recent years, the rise of artificial intelligence (AI) methods has become a breakthrough in solving this problem, improving computational efficiency by millions of times while maintaining computational accuracy, making it possible to simulate millions of atoms at the nanosecond level while maintaining first-principles accuracy. However, existing AI methods still face two major challenges in simulating electrochemical interfaces: describing long-range interactions within the system and the system's dielectric response. Currently widely used machine learning potential functions often employ descriptors based on local environments to directly predict the system's potential energy surface. This makes it impossible to describe long-range interactions using machine learning potential functions. The lack of long-range interactions can cause significant errors in simulation results when ions are present in the system, therefore, they need to be included in the simulation of electrochemical interfaces. Existing methods that include long-range interactions in the system mainly fall into two categories: separating long-range and short-range interactions in the system, using machine learning models to predict the system's charge distribution, calculating long-range electrostatic interactions based on the charge distribution using classical electrostatic algorithms, and calculating short-range interactions based on short-range machine learning potential functions; or constructing descriptors containing long-range information to describe the system's potential energy surface. These two types of methods do not differ significantly in accuracy and efficiency. Furthermore, many currently widely used machine learning potential functions do not include the system's dielectric response, making them unsuitable for simulations under given external electric field boundary conditions. Even the few existing potential function models that include the system's dielectric response cannot simultaneously describe the dielectric responses of conductors and insulators, thus they cannot be applied to simulations of metal / insulator interfaces, such as electrochemical interfaces. Summary of the Invention

[0003] To address the aforementioned technical problems, the technical solution adopted by this invention is as follows:

[0004] According to one aspect of the present invention, a method for obtaining a potential function model for conductor / insulator interface simulation is provided, the method comprising the following steps:

[0005] S100, based on the initial atomic structure of a given specified system, perform molecular simulation on the specified system to obtain n atomic structures, and extract m atomic structures from the n atomic structures as the initial dataset. The specified system includes conductors and insulators in contact with each other, where n > m > 1.

[0006] S200, using first principles, calculate the total energy, total atomic force information, and maximum localized Vonnière function center for each atomic structure corresponding to the current dataset, and obtain a maximum localized Vonnière function center machine learning model based on the maximum localized Vonnière function center; the initial value of the current dataset is the initial dataset; the total atomic force information includes the total atomic force for all atoms corresponding to the corresponding atomic structure.

[0007] S300, based on the maximally localized Vanier function center machine learning model, obtain the charge space distribution corresponding to the insulator in each atomic structure corresponding to the current dataset, and apply the same external electric field as the external electric field applied in the first-principles calculation to each atomic structure corresponding to the current dataset, and obtain the charge space distribution corresponding to the conductor in the atomic structure with the applied external electric field based on the polarizable electrode model; thus obtaining the overall charge space distribution corresponding to each atomic structure corresponding to the current dataset.

[0008] S400 constructs a long-range machine learning potential function for the system based on the overall charge spatial distribution of all atomic structures corresponding to the current dataset, and uses the constructed long-range machine learning potential function to obtain the long-range energy and long-range atomic force information for each atomic structure corresponding to the current dataset; the long-range atomic force information includes the long-range atomic forces for all atoms corresponding to the corresponding atomic structure.

[0009] S500: Based on the short-range energy dataset and the short-range atomic force dataset, obtain N short-range machine learning potential functions for the system, and sum the N short-range machine learning potential functions for the system with the long-range machine learning potential functions for the system to obtain N system machine learning potential functions; wherein, each energy in the short-range energy dataset is equal to the difference between the total energy corresponding to the corresponding atomic structure and the long-range energy, and the force of each atom in the short-range atomic force dataset is formed by the difference between the total atomic force information corresponding to the corresponding atomic structure and the atomic force information in the long-range atomic force information, N≥2.

[0010] S600 uses one of the N system machine learning potential functions as a reference system machine learning potential function to perform molecular simulations on the current k source atomic structures by applying a corresponding set external electric field, resulting in k*n1 atomic structures, which are used as the current test atomic structures; n1>1; the initial values ​​of the current k source atomic structures are the specified k initial atomic structures.

[0011] S700: Based on the potential functions of N systems, obtain the k*n1 maximum atomic force deviations corresponding to the current test atomic structure. If there is a maximum atomic force deviation within the preset error range among the k*n1 maximum atomic force deviations, then the corresponding atomic structure is used as a candidate dataset, and m1 atomic structures are obtained from the candidate dataset and added to the current dataset. Additionally, k atomic structures are obtained from the m1 atomic structures added to the dataset as the current k source atomic structures. Then execute S200. Otherwise, execute S800.

[0012] S800, the current dataset is used as the target dataset, and the target machine learning potential function is obtained based on the target dataset as a potential function model for conductor / insulator interface simulation.

[0013] According to a second aspect of the present invention, a potential function model acquisition device for conductor / insulator interface simulation is provided, the device comprising:

[0014] The first molecular simulation module is used to perform molecular simulation on a specified system based on the initial atomic structure of the given specified system, to obtain n atomic structures, and to extract m atomic structures from the n atomic structures as the initial dataset. The specified system includes conductors and insulators in contact with each other, where n > m > 1.

[0015] The first calculation module is used to calculate the total energy, total atomic force information, and maximum localized Vonnière function center for each atomic structure corresponding to the current dataset using first principles, and to obtain a maximum localized Vonnière function center machine learning model based on the maximum localized Vonnière function center; the initial value of the current dataset is the initial dataset; the total atomic force information includes the total atomic force for all atoms corresponding to the corresponding atomic structure.

[0016] The charge space distribution acquisition module is used to acquire the charge space distribution of insulators in each atomic structure corresponding to the current dataset based on the maximum localized Wannier function center machine learning model, and to apply an external electric field to each atomic structure corresponding to the current dataset, the same external electric field applied during the first-principles calculation, and to acquire the charge space distribution of conductors in the atomic structure with the applied external electric field based on the polarizable electrode model; thus obtaining the overall charge space distribution of each atomic structure corresponding to the current dataset.

[0017] The second calculation module is used to construct a long-range machine learning potential function for the system based on the overall charge spatial distribution of all atomic structures corresponding to the current dataset, and to use the constructed long-range machine learning potential function to obtain the long-range energy and long-range atomic force information for each atomic structure corresponding to the current dataset; the long-range atomic force information includes the long-range atomic forces for all atoms corresponding to the corresponding atomic structure.

[0018] The machine learning potential function acquisition module is used to acquire N short-range machine learning potential functions of the system based on the short-range energy dataset and the short-range atomic force dataset, and to sum the N short-range machine learning potential functions of the system with the long-range machine learning potential function of the system to obtain N system machine learning potential functions; wherein, each energy in the short-range energy dataset is equal to the difference between the total energy corresponding to the corresponding atomic structure and the long-range energy, and the force of each atom in the short-range atomic force dataset is formed by the difference between the total atomic force information corresponding to the corresponding atomic structure and the atomic force information in the long-range atomic force information, and N≥2.

[0019] The second molecular simulation module is used to perform molecular simulations on the current k source atomic structures by applying a corresponding set external electric field based on one of the N system machine learning potential functions as the reference system machine learning potential function, resulting in k*n1 atomic structures, which are used as the current test atomic structures; n1>1; the initial values ​​of the current k source atomic structures are the specified k initial atomic structures.

[0020] The judgment module is used to obtain the k*n1 maximum atomic force deviations corresponding to the current test atomic structure based on the machine learning potential functions of N systems. If there is a maximum atomic force deviation within the preset error range among the k*n1 maximum atomic force deviations, the corresponding atomic structure is used as a candidate dataset, and m1 atomic structures are obtained from the candidate dataset and added to the current dataset. Additionally, k atomic structures are obtained from the m1 atomic structures added to the dataset as the current k source atomic structures. Otherwise, the current dataset is used as the target dataset.

[0021] The target machine learning potential function acquisition module is used to acquire the target machine learning potential function based on the target dataset, as a potential function model for conductor / insulator interface simulation.

[0022] The present invention has at least the following beneficial effects:

[0023] The technical solution provided by this invention starts from a single structure, automatically explores the workflow of the system's potential energy surface and the collection structure, and trains a machine learning potential function that can be used to simulate the conductor / insulator interface under a given external electric field boundary condition, which can accelerate the electrochemical interface simulation.

[0024] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 A flowchart for obtaining the potential function model for conductor / insulator interface simulation provided in an embodiment of the present invention;

[0027] Figures 2a to 3 This is a schematic diagram illustrating the experimental results of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0030] It should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the steps as sequential processes, many of these steps can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the steps can be rearranged. A process can be terminated when its operation is complete, but it may also have additional steps not included in the figures. A process can correspond to a method, function, procedure, subroutine, subroutine, etc.

[0031] Example 1:

[0032] This embodiment provides a method for obtaining a potential function model for conductor / insulator interface simulation. This method addresses the limitation of existing machine learning potential function methods that cannot simultaneously describe the dielectric responses of conductors and insulators. It processes the dielectric responses of the conductor and insulator separately within two sub-frames and then merges them into the overall dielectric response of the system. Based on this hybrid description of the dielectric response, the charge distribution of the system under a given external electric field boundary condition can be obtained. Furthermore, a long-range and short-range action separation approach is adopted to describe the potential energy surface of the system, ultimately obtaining a machine learning potential energy surface model that includes the dielectric response of the conductor / insulator interface. Specifically, as shown... Figure 1 As shown, the method may include the following steps:

[0033] S100: Based on the initial atomic structure of a given specified system, perform molecular simulation on the specified system to obtain n atomic structures, and extract m atomic structures from the n atomic structures as the initial dataset. The specified system includes conductors and insulators in contact with each other, i.e., a conductor / insulator interface system. Wherein, n > m > 1.

[0034] In this embodiment of the invention, the specified system can be set based on actual needs and can be any system used for molecular simulation. In this embodiment of the invention, the specific objects corresponding to conductors and insulators can be determined based on actual needs.

[0035] In this embodiment of the invention, molecular simulation can be performed based on molecular dynamics simulation or Monte Carlo simulation. Specifically, a calculation configuration file for molecular dynamics simulation or Monte Carlo simulation can be configured in the molecular dynamics simulation software based on the atomic structure of the specified system, and the simulation is executed to obtain the corresponding simulation trajectory. n can be set according to actual needs; in one illustrative embodiment, n can be tens of thousands, meaning tens of thousands of atomic movements are simulated. m can be set according to actual needs, on the order of hundreds, and the m atomic structures can be user-specified structures or randomly selected structures.

[0036] As is known to those skilled in the art, atomic structure can include atomic types and atomic positions.

[0037] S200, using first principles, calculate the total energy, total atomic force information, and the maximum localized Vonnière function center for each atomic structure corresponding to the current dataset, and obtain a maximum localized Vonnière function center machine learning model based on the maximum localized Vonnière function center; the initial value of the current dataset is the initial dataset; the total atomic force information includes the total atomic force of all atoms corresponding to the corresponding atomic structure, that is, the total atomic force information of a certain atomic structure includes the atomic force of each atom in that atomic structure.

[0038] Those skilled in the art will understand that any method that uses first principles to calculate the total energy, total atomic force information, and the maximum localized Vonnière function center for each atomic structure corresponding to the current dataset is within the scope of protection of this invention.

[0039] In this embodiment of the invention, the maximally localized Magnitude function center machine learning model can be obtained based on the DeepMD-kit open-source software. Specifically, the current dataset and the corresponding maximally localized Magnitude function center can be used as the training set to train the initial maximally localized Magnitude function center machine learning model, thus obtaining the maximally localized Magnitude function center machine learning model. That is, the current dataset is used as the model's input, and the maximally localized Magnitude function center is used as the model's output to train the model. The initial maximally localized Magnitude function center machine learning model can be a deep learning neural network model. It is understood by those skilled in the art that any method of using the current dataset and the corresponding maximally localized Magnitude function center as the training set to train the initial maximally localized Magnitude function center machine learning model to obtain the maximally localized Magnitude function center machine learning model falls within the protection scope of this invention.

[0040] S300, based on the maximally localized Vanier function center machine learning model, obtain the charge space distribution corresponding to the insulator in each atomic structure corresponding to the current dataset, and apply the same external electric field as the external electric field applied in the first-principles calculation to each atomic structure corresponding to the current dataset, and obtain the charge space distribution corresponding to the conductor in the atomic structure with the applied external electric field based on the polarizable electrode model; thus obtaining the overall charge space distribution corresponding to each atomic structure corresponding to the current dataset.

[0041] Since the type of each atom in the insulator is known, the magnitude of the charge of each atom can be determined. Furthermore, since the position of each atom is known, the spatial distribution of the insulator charge in each atomic structure can be determined; that is, the spatial distribution of charge includes both the magnitude and position of the charge. It is understood by those skilled in the art that any method for obtaining the spatial distribution of the charge corresponding to the insulator in each atomic structure of the current dataset based on the aforementioned maximally localized Vanier function center machine learning model falls within the scope of protection of this invention.

[0042] In this embodiment of the invention, the polarizable electrode model in the ELECTRODE module of the LAMMPS open-source software can be used to obtain the charge spatial distribution corresponding to the conductor in the atomic structure to which an external electric field has been applied. The specific method for obtaining this information can be the existing technology.

[0043] S400 constructs a long-range machine learning potential function for the system based on the overall charge spatial distribution of all atomic structures corresponding to the current dataset, and uses the constructed long-range machine learning potential function to obtain the long-range energy and long-range atomic force information for each atomic structure corresponding to the current dataset; the long-range atomic force information includes the long-range atomic forces for all atoms corresponding to the corresponding atomic structure.

[0044] In this embodiment of the invention, the long-range machine learning potential function of the system is constructed based on a classical electrostatic interaction algorithm, such as the Ewald summation algorithm. Those skilled in the art will understand that any method for constructing a long-range machine learning potential function of a system based on the overall charge spatial distribution corresponding to all atomic structures in the current dataset, and for using a classical electrostatic interaction algorithm to obtain the corresponding overall charge spatial distribution to construct the long-range machine learning potential function of the system, falls within the protection scope of this invention.

[0045] Those skilled in the art will understand that any method that utilizes the constructed system's long-range machine learning potential function to obtain the long-range energy and long-range atomic force information corresponding to each atomic structure in the current dataset falls within the protection scope of this invention.

[0046] S500: Based on the short-range energy dataset and the short-range atomic force dataset, obtain N system short-range machine learning potential functions, and sum the N system short-range machine learning potential functions with the system long-range machine learning potential functions, that is, sum the system short-range machine learning potential functions and the system long-range machine learning potential functions to obtain N system machine learning potential functions; wherein, each energy in the short-range energy dataset is equal to the difference between the total energy and the long-range energy corresponding to the corresponding atomic structure, and the force of each atom in the short-range atomic force dataset is formed by the difference between the atomic force information of the total atomic force information of the corresponding atomic structure and the atomic force information of the long-range atomic force information, that is, the atomic force difference of each atom in each atomic structure is equal to the difference between the corresponding total atomic force and the corresponding long-range atomic force, that is, the atomic force difference ΔP corresponding to the r-th atom in a certain atomic structure. r =(PC) r -PL r PC r PL represents the total atomic force corresponding to the r-th atom in this atomic structure, and PL represents the long-range atomic force corresponding to the r-th atom in this atomic structure.

[0047] In this embodiment of the invention, N ≥ 2. The specific value of N can be determined based on actual needs.

[0048] In this embodiment of the invention, the short-range machine learning potential function of the system can be obtained based on the DeepMD-kit open-source software. Specifically, the current dataset and the corresponding energy and atomic force information can be used as a training set to train the initial short-range machine learning potential function model of the system, thereby obtaining the short-range machine learning potential function of the system. That is, the current dataset is used as the input of the model, and the energy and atomic force information is used as the output of the model to train the model. It is known to those skilled in the art that any method that uses the current dataset and the corresponding energy and atomic force information as a training set to train the initial short-range machine learning potential function of the system to obtain the short-range machine learning potential function of the system falls within the protection scope of this invention.

[0049] S600 uses one of the N system machine learning potential functions as a reference system machine learning potential function to perform molecular simulations on the current k source atomic structures by applying a corresponding set external electric field, explores the system potential energy surface, generates the corresponding simulation trajectory, and obtains k*n1 atomic structures as the current test atomic structures; n1>1; the initial values ​​of the current k source atomic structures are the specified k initial atomic structures.

[0050] In this embodiment of the invention, the machine learning potential function of the reference system can be obtained randomly.

[0051] In this embodiment of the invention, the value of k can be set based on actual needs, where 1 < k < m. n1 can be equal to n or not equal to n.

[0052] In this embodiment of the invention, the initial atomic structure may belong to the n atomic structures obtained in S100, or it may not belong to any of them, and it may be an atomic structure specified by the user.

[0053] In this embodiment of the invention, the external electric field corresponding to the current k source atom structures is greater than or equal to the external electric field corresponding to the previous k source atom structures, that is, the external electric field is applied in an incremental manner. Specifically, the external electric field strength is extrapolated from small to large during the exploration process, and the external electric field conditions can be achieved by controlling the external electric field or controlling the amount of electricity.

[0054] In this embodiment of the invention, the k*n1 atomic structure can be achieved by modifying and integrating the ELECTRODE module in the open-source software LAMMPS with the LAMMPS interface in the open-source software DeepMD-kit. The LAMMPS interface in DeepMD-kit is used to perform molecular dynamics simulations, and the ELECTRODE module in LAMMPS applies the polarizable electrode method to the molecular dynamics simulations. Those skilled in the art will recognize that the method of modifying and integrating the ELECTRODE module in the open-source software LAMMPS with the LAMMPS interface in the open-source software DeepMD-kit is prior art.

[0055] S700: Based on the machine learning potential functions of N systems, obtain the k*n1 maximum atomic force deviations corresponding to the current test atomic structure. If there is a maximum atomic force deviation within the preset error range among the k*n1 maximum atomic force deviations, then the corresponding atomic structure is used as a candidate dataset, and m1 atomic structures are obtained from the candidate dataset and added to the current dataset. Additionally, k atomic structures are obtained from the m1 atomic structures added to the dataset as the current k source atomic structures, and S200 is executed. Otherwise, if there is no maximum atomic force deviation within the preset error range, S800 is executed.

[0056] In this embodiment of the invention, the preset error range can be [a, b], where a is a lower limit value, b is a higher limit value, b = f * a, and f ≥ 2. b can be an empirical value; in one illustrative embodiment, b can be 0.2 millielectronvolts per angstrom. In one illustrative embodiment, 2 ≤ f ≤ 3.

[0057] Furthermore, in this embodiment of the invention, in S700, the maximum atomic force deviation corresponding to each atomic structure of the currently tested atomic structure satisfies the following condition:

[0058] Pf=max(Pf1,Pf2,...,Pf i ..., Pf h ), Pf i The maximum atomic force deviation of the i-th atom in the corresponding atomic structure is given by the value of i, which ranges from 1 to h, where h is the number of atoms in the corresponding atomic structure. P ij To determine the atomic force of the i-th atom obtained from the machine learning potential function of the j-th system, The maximum value is the average force on the i-th atom obtained from the machine learning potential function of N systems. `max()` represents taking the maximum value.

[0059] In this embodiment of the invention, if the maximum atomic force deviation is less than b, it indicates that the structure is one that machine learning can accurately predict. Usually, the training error of the machine learning potential function is taken. If it is greater than a, it indicates that the structure has an excessively large machine learning prediction error and may be an unreasonable structure. If it falls between a and b, it is a structure with a moderate machine learning prediction error and needs to continue learning.

[0060] In this embodiment of the invention, m1 can be set according to actual needs; m1 can be equal to m or not equal to m. Those skilled in the art will know that if the number of atomic structures in the candidate dataset is less than or equal to m1, all atomic structures in the candidate dataset can be added to the current dataset to update the current dataset.

[0061] S800, the current dataset is used as the target dataset, and the target machine learning potential function is obtained based on the target dataset as a potential function model for conductor / insulator interface simulation.

[0062] Furthermore, in this embodiment of the invention, S800 may specifically include:

[0063] S810 uses first-principles calculations to determine the total energy, total atomic force information, and maximum localized Vonnière function center for each atomic structure in the target dataset, and obtains a maximum localized Vonnière function center machine learning model based on the maximum localized Vonnière function center.

[0064] S820: Based on the maximally localized Vanier function center machine learning model in S810, obtain the charge space distribution corresponding to the insulator in each atomic structure corresponding to the target dataset, and apply the same external electric field as the external electric field applied in the first-principles calculation process in S810 to each atomic structure corresponding to the target dataset, and obtain the charge space distribution corresponding to the conductor in the atomic structure with the applied external electric field based on the polarizable electrode model; thus obtaining the overall charge space distribution corresponding to each atomic structure corresponding to the target dataset.

[0065] S830 constructs a long-range machine learning potential function for the system based on the overall charge spatial distribution of all atomic structures corresponding to the target dataset, and uses the constructed long-range machine learning potential function to obtain the long-range energy and long-range atomic force information for each atomic structure corresponding to the target dataset.

[0066] S840, based on the short-range energy dataset and the short-range atomic force dataset, obtains a short-range machine learning potential function for the system, and sums the short-range machine learning potential function and the long-range machine learning potential function of the system to obtain the target system machine learning potential function, which serves as the potential function model for conductor / insulator interface simulation.

[0067] The specific implementations of S810 to S840 are basically the same as those of S200 to S500. The difference is that the number of steps set in S810 to S840 during the training process using the dataset is greater than the number of steps set in S200 to S500.

[0068] The potential function model for simulating conductor / insulator interfaces provided in this invention allows users to control the electric field or electric charge of the simulation system according to the actual research problem.

[0069] Example 2:

[0070] This embodiment provides a potential function model acquisition device for conductor / insulator interface simulation, the device comprising:

[0071] The first molecular simulation module is used to perform molecular simulation on a specified system based on the initial atomic structure of the given specified system, to obtain n atomic structures, and to extract m atomic structures from the n atomic structures as the initial dataset. The specified system includes conductors and insulators in contact with each other, where n > m > 1.

[0072] The first calculation module is used to calculate the total energy, total atomic force information, and maximum localized Vonnière function center for each atomic structure corresponding to the current dataset using first principles, and to obtain a maximum localized Vonnière function center machine learning model based on the maximum localized Vonnière function center; the initial value of the current dataset is the initial dataset; the total atomic force information includes the total atomic force for all atoms corresponding to the corresponding atomic structure;

[0073] The charge space distribution acquisition module is used to acquire the charge space distribution of insulators in each atomic structure corresponding to the current dataset based on the maximum localized Wannier function center machine learning model, and to apply an external electric field to each atomic structure corresponding to the current dataset, the same external electric field applied during the first-principles calculation, and to acquire the charge space distribution of conductors in the atomic structure with the applied external electric field based on the polarizable electrode model; thus obtaining the overall charge space distribution of each atomic structure corresponding to the current dataset.

[0074] The second calculation module is used to construct a long-range machine learning potential function for the system based on the overall charge spatial distribution of all atomic structures corresponding to the current dataset, and to use the constructed long-range machine learning potential function to obtain the long-range energy and long-range atomic force information for each atomic structure corresponding to the current dataset; the long-range atomic force information includes the long-range atomic forces for all atoms corresponding to the corresponding atomic structure.

[0075] The machine learning potential function acquisition module is used to acquire N short-range machine learning potential functions of the system based on the short-range energy dataset and the short-range atomic force dataset, and to sum the N short-range machine learning potential functions of the system with the long-range machine learning potential function of the system to obtain N system machine learning potential functions; wherein, each energy in the short-range energy dataset is equal to the difference between the total energy corresponding to the corresponding atomic structure and the long-range energy, and each atomic force in the short-range atomic force dataset is equal to the difference between the total atomic force information corresponding to the corresponding atomic structure and the long-range atomic force information, and N≥2.

[0076] The second molecular simulation module is used to perform molecular simulations on the current k source atomic structures by applying a corresponding set external electric field based on one of the N system machine learning potential functions as the reference system machine learning potential function, resulting in k*n1 atomic structures, which are used as the current test atomic structures; n1>1; the initial values ​​of the current k source atomic structures are the specified k initial atomic structures.

[0077] The judgment module is used to obtain the k*n1 maximum atomic force deviations corresponding to the current test atomic structure based on the machine learning potential functions of N systems. If there is a maximum atomic force deviation within the preset error range among the k*n1 maximum atomic force deviations, the corresponding atomic structure is used as a candidate dataset, and m1 atomic structures are obtained from the candidate dataset and added to the current dataset. Additionally, k atomic structures are obtained from the m1 atomic structures added to the dataset as the current k source atomic structures. Otherwise, the current dataset is used as the target dataset.

[0078] The target machine learning potential function acquisition module is used to acquire the target machine learning potential function based on the target dataset, as a potential function model for conductor / insulator interface simulation.

[0079] This device can be used to perform Figure 1 The method shown in the illustrated embodiment is relevant here; therefore, the functions that each functional module of the device can achieve can be referred to. Figure 1 The embodiments shown are described in detail below.

[0080] In one specific application of this embodiment, the conductor / insulator interface system can be a Pt(111) / KF solution interface system. Taking the Pt(111) / KF solution interface system as an example, this invention uses the method provided by this invention to obtain the corresponding machine learning potential energy model. The error of the obtained machine learning potential energy model can be as follows: Figures 2a to 2b As shown. Figures 2a to 2b As shown, the maximum localized root mean square error of the Wannier function center on the machine learning potential energy model on both the training and test datasets is... Order of magnitude, such as Figure 2a and Figure 2b As shown; the root mean square error of atomic force of the machine learning potential energy model on the training and test datasets is... Order of magnitude, such as Figure 2c and Figure 2d As shown in the figure. This demonstrates that the machine learning potential model obtained by the method of the present invention has high accuracy.

[0081] Figure 3This is a schematic diagram showing the calculated differential capacitance of the Pt(111) / KF solution interface. In this embodiment, the PBE functional and Grimme D3 correction are used to describe the potential energy surface. Using the machine learning potential energy function model obtained in this invention, the charge-potential relationship of the Pt(111) / KF solution interface can be obtained through constant charge simulation (curve 1). Taking the first derivative of the charge with respect to the potential yields the interface differential capacitance curve (curve 2). Figure 3 It can be seen that the calculated differential capacitance is quantitatively consistent with the data obtained from previous first-principles calculations and experiments. This indicates that the machine learning potential energy model obtained by the method of this invention has high accuracy.

[0082] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0083] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for obtaining a potential function model for conductor / insulator interface simulation, characterized in that, The method includes the following steps: S100, Based on the initial atomic structure of a given specified system, perform molecular simulation on the specified system to obtain n atomic structures, and extract m atomic structures from the n atomic structures as the initial dataset. The specified system includes conductors and insulators in contact with each other, where n > m > 1. S200, using first principles, calculate the total energy, total atomic force information, and maximum localized Vonnière function center for each atomic structure corresponding to the current dataset, and obtain a maximum localized Vonnière function center machine learning model based on the maximum localized Vonnière function center; the initial value of the current dataset is the initial dataset; The total atomic force information includes the total atomic force for all atoms corresponding to the corresponding atomic structure; S300, based on the maximally localized Wannier function center machine learning model, obtain the charge spatial distribution corresponding to the insulator in each atomic structure corresponding to the current dataset, and apply the same external electric field as the external electric field applied in the first-principles calculation to each atomic structure corresponding to the current dataset, and obtain the charge spatial distribution corresponding to the conductor in the atomic structure with the applied external electric field based on the polarizable electrode model. Obtain the overall charge spatial distribution for each atomic structure in the current dataset; S400 constructs a long-range machine learning potential function for the system based on the overall charge spatial distribution of all atomic structures in the current dataset, and uses the constructed long-range machine learning potential function to obtain the long-range energy and long-range atomic force information for each atomic structure in the current dataset. Long-range atomic force information includes the long-range atomic forces corresponding to all atoms in the corresponding atomic structure; S500: Based on the short-range energy dataset and the short-range atomic force dataset, obtain N short-range machine learning potential functions for the system, and sum the N short-range machine learning potential functions for the system with the long-range machine learning potential functions for the system to obtain N system machine learning potential functions; wherein, each energy in the short-range energy dataset is equal to the difference between the total energy corresponding to the corresponding atomic structure and the long-range energy, and the force of each atom in the short-range atomic force dataset is formed by the difference between the total atomic force information corresponding to the corresponding atomic structure and the atomic force information in the long-range atomic force information, N≥2; S600 uses one of the N system machine learning potential functions as a reference system machine learning potential function to perform molecular simulations on the current k source atomic structures by applying a corresponding set external electric field, resulting in k*n1 atomic structures, which are used as the current test atomic structures; n1>1; the initial values ​​of the current k source atomic structures are the specified k initial atomic structures; S700: Based on the machine learning potential functions of N systems, obtain the k*n1 maximum atomic force deviations corresponding to the current test atomic structure. If there is a maximum atomic force deviation within the preset error range among the k*n1 maximum atomic force deviations, then the corresponding atomic structure is used as a candidate dataset, and m1 atomic structures are obtained from the candidate dataset and added to the current dataset. Additionally, k atomic structures are obtained from the m1 atomic structures added to the dataset as the current k source atomic structures. Then, execute S200. Otherwise, execute S800. S800, the current dataset is used as the target dataset, and the target machine learning potential function is obtained based on the target dataset as a potential function model for conductor / insulator interface simulation.

2. The method according to claim 1, characterized in that, In S700, the maximum atomic force deviation corresponding to each atomic structure of the current test atomic structure satisfies the following condition: Pf=max(Pf1,Pf2,...,Pf i ..., Pf h ), Pf i The maximum atomic force deviation of the i-th atom in the corresponding atomic structure is given by the value of i, which ranges from 1 to h, where h is the number of atoms in the corresponding atomic structure. P ij To determine the atomic force of the i-th atom obtained from the machine learning potential function of the j-th system, The maximum value is the average force on the i-th atom obtained from the machine learning potential function of N systems. `max()` represents taking the maximum value.

3. The method according to claim 1, characterized in that, The preset error range is [a, b], where a is the lower limit and b is the upper limit, b = f * a, and f ≥ 2.

4. The method according to claim 1, characterized in that, The external electric field corresponding to the current k source atom structures is greater than or equal to the external electric field corresponding to the previous k source atom structures.

5. The method according to claim 1, characterized in that, The S800 specifically includes: S810 uses first-principles calculations to determine the total energy, total atomic force information, and maximum localized Vonnier function center for each atomic structure in the target dataset, and obtains a maximum localized Vonnier function center machine learning model based on the maximum localized Vonnier function center. S820: Based on the maximally localized Vanier function center machine learning model in S810, obtain the charge space distribution corresponding to the insulator in each atomic structure of the target dataset, and apply the same external electric field as the one applied in the first-principles calculation process in S810 to each atomic structure of the target dataset, and obtain the charge space distribution corresponding to the conductor in the atomic structure with the applied external electric field based on the polarizable electrode model; thus obtaining the overall charge space distribution corresponding to each atomic structure of the target dataset. S830: Based on the overall charge spatial distribution of all atomic structures corresponding to the target dataset, a long-range machine learning potential function of the system is constructed, and the long-range energy and long-range atomic force information corresponding to each atomic structure of the target dataset are obtained by using the constructed long-range machine learning potential function of the system. S840, based on short-range energy dataset and short-range atomic force dataset, obtains a short-range machine learning potential function for the system, and sums the short-range machine learning potential function and the long-range machine learning potential function of the system to obtain the target system machine learning potential function, which serves as the potential function model for conductor / insulator interface simulation.

6. The method according to claim 1, characterized in that, In S100, molecular simulations are performed based on molecular dynamics simulations or Monte Carlo simulations.

7. The method according to claim 1, characterized in that, The maximally localized Wannier function center machine learning model and the system short-range machine learning potential function were obtained based on the DeepMD-kit open-source software.

8. The method according to claim 1, characterized in that, The long-range machine learning potential function of the system is constructed based on the classical electrostatic interaction algorithm.

9. A device for obtaining a potential function model for conductor / insulator interface simulation, characterized in that, The device includes: The first molecular simulation module is used to perform molecular simulation on a specified system based on the initial atomic structure of the given specified system, to obtain n atomic structures, and to extract m atomic structures from the n atomic structures as the initial dataset. The specified system includes conductors and insulators in contact with each other, where n > m > 1. The first calculation module is used to calculate the total energy, total atomic force information, and maximum localized Vonnière function center for each atomic structure corresponding to the current dataset using first principles, and to obtain a maximum localized Vonnière function center machine learning model based on the maximum localized Vonnière function center; the initial value of the current dataset is the initial dataset; the total atomic force information includes the total atomic force for all atoms corresponding to the corresponding atomic structure; The charge space distribution acquisition module is used to acquire the charge space distribution of insulators in each atomic structure of the current dataset based on the maximally localized Vanier function center machine learning model, and to apply an external electric field to each atomic structure of the current dataset that is the same as the external electric field applied in the first-principles calculation, and to acquire the charge space distribution of conductors in the atomic structure with the applied external electric field based on the polarizable electrode model; thus obtaining the overall charge space distribution of each atomic structure of the current dataset. The second calculation module is used to construct the long-range machine learning potential function of the system based on the overall charge spatial distribution of all atomic structures corresponding to the current dataset, and to use the constructed long-range machine learning potential function of the system to obtain the long-range energy and long-range atomic force information of each atomic structure corresponding to the current dataset; the long-range atomic force information includes the long-range atomic force of all atoms corresponding to the corresponding atomic structure; The machine learning potential function acquisition module is used to acquire N short-range machine learning potential functions of the system based on the short-range energy dataset and the short-range atomic force dataset, and to sum the N short-range machine learning potential functions of the system with the long-range machine learning potential function of the system to obtain N system machine learning potential functions; wherein, each energy in the short-range energy dataset is equal to the difference between the total energy corresponding to the corresponding atomic structure and the long-range energy, and the force of each atom in the short-range atomic force dataset is formed by the difference between the total atomic force information corresponding to the corresponding atomic structure and the atomic force information in the long-range atomic force information, N≥2; The second molecular simulation module is used to perform molecular simulations on the current k source atomic structures by applying a corresponding set external electric field based on one of the N system machine learning potential functions as the reference system machine learning potential function, resulting in k*n1 atomic structures, which are used as the current test atomic structures; n1>1; the initial values ​​of the current k source atomic structures are the specified k initial atomic structures; The judgment module is used to obtain the k*n1 maximum atomic force deviations corresponding to the current test atomic structure based on the machine learning potential functions of N systems. If there is a maximum atomic force deviation within the preset error range among the k*n1 maximum atomic force deviations, the corresponding atomic structure is used as a candidate dataset, and m1 atomic structures are obtained from the candidate dataset and added to the current dataset. Additionally, k atomic structures are obtained from the m1 atomic structures added to the dataset as the current k source atomic structures. Otherwise, the current dataset is used as the target dataset. The target machine learning potential function acquisition module is used to acquire the target machine learning potential function based on the target dataset, as a potential function model for conductor / insulator interface simulation.

Images