Method, system, terminal and medium for generating first-principles computing data set

Abstract

The present application is applicable to the field of battery material calculation, and provides a method, system, terminal and medium for generating a first-principles calculation data set, wherein the method includes: constructing a supercell structure based on a specific positive electrode unit cell structure, wherein the specific positive electrode unit cell structure is a unit cell structure in which the load ions are not separated; adjusting the atomic positions and lattice vectors based on the supercell structure, and obtaining a structure to be calculated in which the atomic force values ​​belong to a preset force range and the structure is electrically neutral; determining the magnetic moment parameters of all transition metal atoms in the structure to be calculated; performing first-principles calculations based on the structure to be calculated and the magnetic moment parameters of each transition metal atom in the structure to be calculated, and obtaining a first-principles calculation data set whose convergence meets the set convergence criteria. This scheme improves the convergence of the first-principles calculation of positive electrode materials, and provides a first-principles calculation data set with good convergence for the training of machine learning force fields for positive electrode materials.

Description

Technical Field

[0001] The present application belongs to the field of battery material calculation, and in particular, relates to a method, system, terminal and medium for generating a first-principles calculation data set. Background Art

[0002] In the field of battery materials, the microscopic physical and chemical properties of battery materials can be predicted through first-principles calculation methods and machine learning force fields. Among them, the prediction efficiency of machine learning force fields is better than that of first-principles calculation methods, but training excellent machine learning force fields cannot be separated from the support of first-principles calculation data sets.

[0003] Currently, in the field of battery material calculations, there are machine learning force fields trained to predict the microscopic physical and chemical properties of negative electrode materials and electrolytes, but there is a lack of machine learning force fields for accurately predicting the microscopic physical and chemical properties of positive electrode materials. Battery positive electrode materials are mostly oxide materials or oxygen-containing acid salt materials containing transition metal atoms such as nickel, cobalt, manganese, and iron. The transition metal atoms in the positive electrode materials present different oxidation states during the battery charging and discharging process, and have complex charge and spin magnetic moment arrangements. When performing first-principles calculations on positive electrode material structures with complex charge and magnetic moment arrangements, the calculations are often difficult to converge, and therefore it is impossible to provide a first-principles calculation data set with good convergence effect for the machine learning force field of the positive electrode material for training. Summary of the invention

[0004] The embodiments of the present application provide a method, system, terminal and medium for generating a first-principles computing data set to solve the problem in the prior art that the first-principles computing of positive electrode materials is difficult to converge and it is difficult to generate the first-principles computing data set required for training the machine learning force field of the positive electrode materials.

[0005] A first aspect of an embodiment of the present application provides a method for generating a first principles calculation data set, comprising:

[0006] Based on a specific positive electrode unit cell structure, a supercell structure is constructed, wherein the specific positive electrode unit cell structure is a unit cell structure in which the load ions are not separated;

[0007] Based on the supercell structure, the atomic positions and lattice vectors are adjusted to obtain a structure to be calculated in which the atomic force values ​​belong to a preset force range and the structure is electrically neutral;

[0008] Determining the magnetic moment parameters of all transition metal atoms in the structure to be calculated;

[0009] A first-principles calculation is performed based on the structure to be calculated and the magnetic moment parameters of each transition metal atom in the structure to be calculated to obtain a first-principles calculation data set whose convergence meets a set convergence standard.

[0010] A second aspect of an embodiment of the present application provides a system for generating a first principles computing data set, comprising:

[0011] A construction module, for constructing a supercell structure based on a specific positive electrode unit cell structure, wherein the specific positive electrode unit cell structure is a unit cell structure in which the load ions are not separated;

[0012] An adjustment module, used for adjusting the atomic positions and lattice vectors based on the supercell structure, to obtain a structure to be calculated in which the atomic force value belongs to a preset force range and the structure is electrically neutral;

[0013] A determination module, used to determine the magnetic moment parameters of all transition metal atoms in the structure to be calculated;

[0014] A calculation module is used to perform first-principles calculation based on the structure to be calculated and the magnetic moment parameters of each transition metal atom in the structure to be calculated, so as to obtain a first-principles calculation data set whose convergence meets a set convergence standard.

[0015] A third aspect of an embodiment of the present application provides a terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in the first aspect when executing the computer program.

[0016] A fourth aspect of an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0017] A fifth aspect of the present application provides a computer program product, which, when executed on a terminal, enables the terminal to execute the steps of the method described in the first aspect.

[0018] As can be seen from the above, the present application is based on a specific positive electrode unit cell structure, i.e., a unit cell structure from which the load ions are not separated, to construct a supercell structure, expand the material scale, and more comprehensively reflect the microstructural characteristics of the material. Subsequently, the atomic position and lattice vector are adjusted based on the constructed supercell structure, so that the stress state of all atoms in the structure reaches a preset ideal equilibrium value, promotes the convergence of the calculation, and ensures that the overall structure remains electrically neutral, thereby obtaining a stable and ideal structure to be calculated. This step not only improves the physical rationality of the structure, but also lays a solid foundation for subsequent calculations. There are transition metal atoms in the structure to be calculated, and the interaction between their electrons is relatively complex. Determining the magnetic moment parameters of all transition metal atoms in the structure to be calculated as the input of the first principles calculation can not only promote the convergence of the calculation, reduce the number of iterations, and improve the calculation efficiency, but also significantly improve the accuracy of the calculation and the reliability of the calculation results. Finally, the first principles calculation is performed based on the magnetic moment parameters of the structure to be calculated and the structure to be calculated. After calculation iterations, a first principles calculation data set whose convergence meets the set convergence criteria is obtained. This application improves the convergence success rate of first-principles calculations of battery positive electrode materials, provides a first-principles calculation data set with good convergence for the training of machine learning force fields of positive electrode materials, and avoids the waste of time and resources caused by a large number of failed calculations, thereby improving the generation efficiency of first-principles calculation data sets. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] Figure 1 It is a flowchart of a method for generating a first principles calculation data set provided in an embodiment of the present application;

[0021] Figure 2 It is a schematic diagram of a process of generating a first principles calculation data set provided in an embodiment of the present application;

[0022] Figure 3 is a comparative schematic diagram of the atomic magnetic moments of transition metal atoms predicted by different methods provided in the embodiments of the present application;

[0023] Figure 4 It is a schematic diagram comparing the total energy predicted by a fine-tuned and trained positive electrode material machine learning force field and a force field that has not been fine-tuned and trained, provided in an embodiment of the present application;

[0024] Figure 5It is a schematic diagram comparing the atomic forces predicted by a fine-tuned and trained positive electrode material machine learning force field and a force field that has not been fine-tuned and trained, provided in an embodiment of the present application;

[0025] Figure 6 It is a schematic diagram of the comparison between the atomic magnetic moment predicted by a fine-tuned machine learning force field for a positive electrode material and a force field that has not been fine-tuned and trained, provided in an embodiment of the present application;

[0026] Figure 7 It is a structural diagram of a system for generating a first principles calculation data set provided in an embodiment of the present application;

[0027] Figure 8 It is a structural diagram of a terminal provided in an embodiment of the present application. DETAILED DESCRIPTION

[0028] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0029] It should be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0030] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0031] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0032] As used in this specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0033] In a specific implementation, the terminal described in the embodiments of the present application includes, but is not limited to, other portable devices such as mobile phones, laptop computers, or tablet computers with touch-sensitive surfaces (e.g., touch screen displays and / or touch pads). It should also be understood that in some embodiments, the device is not a portable communication device, but a desktop computer with a touch-sensitive surface (e.g., touch screen displays and / or touch pads).

[0034] In the following discussion, a terminal including a display and a touch-sensitive surface is described. However, it should be understood that the terminal may include one or more other physical user interface devices such as a physical keyboard, mouse and / or joystick.

[0035] The terminal supports various applications, such as one or more of the following: a drawing application, a presentation application, a word processing application, a website creation application, a disk burning application, a spreadsheet application, a game application, a telephone application, a video conferencing application, an email application, an instant messaging application, a workout support application, a photo management application, a digital camera application, a digital camcorder application, a web browsing application, a digital music player application, and / or a digital video player application.

[0036] Various applications that can be executed on the terminal can use at least one common physical user interface device such as a touch-sensitive surface. One or more functions of the touch-sensitive surface and corresponding information displayed on the terminal can be adjusted and / or changed between applications and / or within corresponding applications. In this way, the common physical architecture of the terminal (e.g., the touch-sensitive surface) can support various applications with user interfaces that are intuitive and transparent to the user.

[0037] It should be understood that the size of the serial numbers of the steps in this embodiment does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present application.

[0038] When studying battery materials, the microscopic physical and chemical properties of battery materials are the focus of researchers. First-principles calculation methods represented by density functional theory methods can predict the microscopic physical and chemical properties of battery materials without being restricted by experimental conditions, and have become one of the important tools for battery material research.

[0039] In recent years, with the rapid development of artificial intelligence technology, machine learning methods have also entered the field of battery material research. Through the calculation results provided by the first-principles calculation method, the machine learning force field used to predict the microscopic physical and chemical properties of battery materials is trained. The trained machine learning force field successfully retains the high accuracy of the first-principles calculation method in predicting the microscopic physical and chemical properties of materials. At the same time, it avoids the complex calculation process such as quantum chemical calculations in the first-principles calculation method, improves the prediction efficiency, and saves the time for predicting the microscopic physical and chemical properties of materials. Therefore, machine learning force fields are widely used in the prediction of microscopic physical and chemical properties of battery materials.

[0040] However, currently in the field of battery material calculation, there are machine learning force fields trained to predict negative electrode materials and electrolytes, but there is a lack of machine learning force fields for accurately predicting the material properties of positive electrode materials. The reason is that most battery positive electrode materials are transition metal oxide materials containing one or more transition metal atoms such as iron (Fe), cobalt (Co), manganese (Mn), nickel (Ni), or oxygen-containing acid salt materials containing one or more transition metal atoms such as iron (Fe), cobalt (Co), manganese (Mn), nickel (Ni), etc. The transition metal atoms in them present different oxidation states during the battery charging and discharging process, and have complex charge and spin magnetic moment arrangements, which makes it difficult to converge the first-principles calculation, and it is impossible to provide a large number of first-principles calculation data sets with good convergence for the training of the machine learning force field of the positive electrode material, so it is impossible to train a machine learning force field for the positive electrode material to predict the microscopic physical and chemical properties of the positive electrode material.

[0041] The present application provides a method, system, terminal and medium for generating a first-principles calculation data set, which performs structural transformation on the structure of positive electrode materials, strictly controls the structure and magnetic moment state of the material, promotes calculation convergence, and provides a first-principles calculation data set with good convergence for the training of the positive electrode material machine learning force field, so as to efficiently and accurately predict the microscopic physical and chemical properties of the positive electrode material according to the trained positive electrode material machine learning force field, and provide strong support for the material research of the positive electrode material.

[0042] In order to illustrate the technical solution described in this application, a specific embodiment is provided below for illustration.

[0043] See also Figure 1 , Figure 1: is a flow chart of a method for generating a first principles calculation data set provided in an embodiment of the present application. Figure 1 As shown, a method for generating a first-principles calculation data set comprises the following steps:

[0044] Step 101 : constructing a supercell structure based on a specific positive electrode unit cell structure, wherein the specific positive electrode unit cell structure is a unit cell structure in which the load ions are not separated.

[0045] The positive electrode material of the battery is one of the key factors that determine the performance and characteristics of the battery. It has a unique unit cell structure. These structures accommodate load ions (such as lithium ions or sodium ions) and one or more transition metal atoms (such as iron, cobalt, manganese, nickel, etc.) to achieve the transmission and storage of charge inside the battery.

[0046] The positive electrode material can be an oxide or an oxyacid salt containing load ions and transition metal atoms. It can be divided into two categories according to the number of transition metal atoms contained: one is the positive electrode material containing multiple transition metal atoms, such as the nickel-cobalt-manganese ternary positive electrode LiNi x Co y Mn z O2(x+y+z=1)(NCM), nickel-cobalt-aluminum ternary positive electrode LiNi x Co y Al z O2(x+y+z=1)(NCA), LiMn x Fe y PO4(x+y=1)(LMFP), lithium manganese-rich based materials xLi2MnO3·yLiMO2 (M is Ni, Co, Mn or a binary or ternary mixture of Ni, Co, Mn), etc.; the other type is the positive electrode material containing only one transition metal atom, such as lithium cobalt oxide LiCoO2 (LCO), lithium iron phosphate LiFePO4 (LFP), lithium manganese oxide LiMn2O4 (LMO), etc.

[0047] Correspondingly, the specific positive electrode unit cell structure is a unit cell structure corresponding to the positive electrode material of the battery, which contains load ions and one or more transition metal atoms and the load ions are not separated. The load ions are used to transfer charges between the positive electrode and the negative electrode of the battery to achieve the storage and release of electrical energy.

[0048] In the process of constructing a supercell structure based on a specific cathode unit cell structure, the construction process is divided into two categories according to the number of transition metal atoms in the unit cell: one is the construction of a supercell structure for a specific cathode unit cell structure containing only one transition metal atom; the other is the construction of a supercell structure for a complex unit cell structure containing multiple transition metal atoms. That is, the unit cell structures containing different numbers of transition metal atoms have different steps for constructing supercell structures.

[0049] In some embodiments, three different states of supercell structures are constructed according to the detachment state of the load ions. The first state is the state where the load ions are not detached, that is, the state of full embedding; the second state is the state where the load ions are completely detached, that is, the state of full de-embedding; and the third state is the state where the load ions are partially detached, that is, the state of partial de-embedding.

[0050] In some embodiments, the specific positive electrode unit cell structure contains a plurality of the transition metal atoms, and constructing a supercell structure based on the specific positive electrode unit cell structure includes: based on the specific positive electrode unit cell structure, constructing a supercell structure in which the plurality of the transition metal atoms are mixed and arranged and the load ions are not detached, a supercell structure in which the plurality of the transition metal atoms are mixed and arranged and the load ions are completely detached, and a supercell structure in which the plurality of the transition metal atoms are mixed and arranged and the load ions are partially detached.

[0051] In some embodiments, the construction process of the supercell structure of three load ion detachment states corresponding to the specific positive electrode unit cell structure containing multiple transition metal atoms is as follows:

[0052] According to the set expansion parameters, the specific positive electrode unit cell structure is expanded to obtain an expanded structure;

[0053] Rearranging multiple transition metal atoms in the extended structure to obtain a supercell structure in which multiple transition metal atoms are mixed and arranged and the load ions are not separated; and,

[0054] Removing all the load ions in the extended structure to obtain a deionized extended structure; rearranging the multiple transition metal atoms in the deionized extended structure to obtain a supercell structure in which multiple transition metal atoms are mixed and arranged and the load ions are completely separated; and,

[0055] The multiple transition metal atoms in the extended structure are rearranged, and some of the load ions in the extended structure are removed, so as to obtain a supercell structure in which multiple transition metal atoms are mixedly arranged and the load ions are partially separated.

[0056] Wherein, the setting of the expansion parameters is a pre-set parameter. Setting the expansion parameters refers to the ratio of the size parameters of the expanded structure to the size parameters of the original structure in the a-axis, b-axis and c-axis directions, and these ratios represent the multiples of the expansion of the original structure in these three directions. That is, setting the expansion parameters refers to the magnification of the expanded structure relative to the original structure in the a-axis, b-axis and c-axis directions. For example, if the expansion parameters are set to 1×1×2, it means that when expanding, the size of the original structure needs to be doubled in the c-axis direction, that is, the length of the expanded structure in the c-axis direction is twice that of the original structure, while in the a-axis and b-axis directions, the size of the original structure remains unchanged and no size adjustment is performed.

[0057] Rearrangement refers to replacing the original transition metal atoms at certain transition metal sites with another transition metal atom.

[0058] Removing the charged ions refers to removing the charged ions from certain charged ion sites.

[0059] In some embodiments, all load ions in a specific positive electrode unit cell structure are removed to obtain a deionized unit cell structure; the deionized unit cell structure is then expanded according to set expansion parameters to obtain a deionized extended structure; and multiple transition metal atoms in the deionized extended structure are rearranged to obtain a supercell structure in which multiple transition metal atoms are mixed and arranged and the load ions are completely separated.

[0060] Lithium iron manganese phosphate LiMn x Fe y Taking PO4 (x+y=1) as an example material for the positive electrode, the construction process of the supercell structure in different load ion detachment states is introduced below.

[0061] In some embodiments, the specific positive electrode unit cell structure is a primitive cell of a given positive electrode material. Starting from the primitive cell of LiFePO4, the structure is expanded according to the set expansion parameters 1×1×2, and the corresponding extended structure is obtained. The extended structure contains 8 transition metal sites, generating a uniformly distributed random integer n(Mn) between 0 and 8, which is used to define the number of Mn atoms. The number of Mn atoms n(Mn) is the number of Fe atoms that need to be replaced in the extended structure, and the number of Mn atoms n(Mn) is the number of replacement sites. From all Fe sites, n(Mn) different sites are randomly selected for atomic replacement. At each selected site, Fe atoms are replaced with Mn atoms to achieve the rearrangement of transition metal atoms, thereby generating a Fe / Mn mixed structure with n(Mn) Mn atoms. According to the set number, the above n(Mn) value determination, site selection and atomic replacement process are repeated to obtain a supercell structure in which a set number of multiple transition metal atoms are mixed and arranged and the load ions are not separated. For example, the number is set to 100, and the above process is repeated 100 times. Each time, a new random value n(Mn) is generated, and a new Fe / Mn mixed structure is created based on it. Finally, 100 Fe / Mn mixed structures with different Mn doping amounts and distribution patterns are obtained, that is, 100 supercell structures in which multiple transition metal atoms are mixed and arranged and the load ions are not separated.

[0062] A supercell structure in which multiple transition metal atoms are mixed and arranged and the load ions are completely detached can be constructed by expansion, deionization, and rearrangement, or by deionization, expansion, and rearrangement.

[0063] In some embodiments, starting from the original cell of FePO4 after complete deionization of lithium ions, structural expansion, determination of the number of Mn atoms, site selection and atomic replacement are performed. The corresponding implementation process can refer to the above embodiments and will not be repeated here. Then, a supercell structure with a set number of multiple transition metal atoms mixed and arranged and the load ions completely separated is obtained.

[0064] In some embodiments, starting from the primitive cell of LiFePO4, the structure is expanded according to the set expansion parameters 1×1×2, and the corresponding extended structure is obtained. The extended structure contains 8 transition metal sites and 8 load ion sites, i.e., lithium sites, generating a uniformly distributed random integer n(Mn) between 0 and 8, and generating a uniformly distributed random integer n(Li) between 1 and 7, as the number of Mn atoms and the number of Li atoms, respectively. Among them, the number of load ions generated is any positive integer between 1 and the total number of load ion sites, which may be equal to 1, but cannot be equal to the total number of load ion sites. From all Fe sites, n(Mn) different sites are randomly selected for atomic replacement, that is, at each selected site, Fe atoms are replaced with Mn atoms. From all Li sites, 8-n(Li) different Li sites are randomly selected for lithium removal, that is, the load ions at the selected sites are removed. Finally, a supercell structure with mixed Fe / Mn and random deintercalation of load ions is obtained. Similarly, according to the set number, the above n(Mn) value determination, n(Li) value determination, site selection, atom replacement and lithium removal processes are repeatedly performed to obtain a supercell structure with a set number of multiple transition metal atoms mixed and arranged and the load ions partially detached.

[0065] In the above embodiments, starting from a basic single unit cell structure, multiple supercell structures are constructed through operations such as structural expansion, random mixing of transition metal atoms, and random deintercalation of load ions. According to the deintercalation state of the load ions, they are divided into a supercell structure in which multiple transition metal atoms are mixed and arranged and the load ions are not separated, a supercell structure in which multiple transition metal atoms are mixed and arranged and the load ions are completely separated, and a supercell structure in which multiple transition metal atoms are mixed and arranged and the load ions are partially separated.

[0066] In the process of constructing the supercell structure, the atomic replacement ratios of different transition metal atoms are random, the distribution sites are random, and the number and sites of the detached load ions are also random. The construction of the supercell structure of this application increases the diversity of the structure, and the first-principles calculation data set obtained from it can more accurately reflect the diversity and complexity of the positive electrode material structure in practical applications, and help to more comprehensively reflect the properties of the positive electrode material. The predictive ability of the machine learning force field of the positive electrode material trained in this way is greatly improved.

[0067] In some embodiments, the specific positive electrode unit cell structure contains a transition metal atom, and constructing a supercell structure based on the specific positive electrode unit cell structure includes: based on the specific positive electrode unit cell structure, constructing the supercell structure with a structural size of a preset size and from which the load ions are not detached, the supercell structure with a structural size of the preset size and from which the load ions are completely detached, and the supercell structure with a structural size of the preset size and from which the load ions are partially detached.

[0068] The preset size is a size parameter of the structure obtained by expanding the set expansion parameters.

[0069] In some embodiments, the construction process of the supercell structure of three load ion detachment states corresponding to a specific positive electrode unit cell structure containing a transition metal atom is as follows:

[0070] According to the set expansion parameters, the specific positive electrode unit cell structure is expanded to obtain an expanded structure, which is a supercell structure with a structure size of a preset size and from which the load ions are not separated; and,

[0071] All the load ions in the extended structure are removed to obtain a deionized unit cell structure, which is a supercell structure with a predetermined size and completely freed load ions; and,

[0072] Part of the load ions in the extended structure are removed to obtain a supercell structure with a preset size and partially detached load ions.

[0073] In some embodiments, all load ions in a specific positive electrode unit cell structure are removed to obtain a deionized unit cell structure; the deionized unit cell structure is expanded according to set expansion parameters to obtain a supercell structure with a preset size and complete detachment of load ions.

[0074] When the structure contains only one transition metal atom, there is no need to perform transition metal atom mixing, and the structure expansion and deionization operations can be performed.

[0075] When constructing a supercell structure, there are different construction methods corresponding to the types and number of transition metal atoms in a specific positive electrode unit cell structure. Regardless of the construction method, it is constructed in a hierarchical manner, and three supercell structures in which the load ions are separated are constructed, namely, a supercell structure in which the load ions are not separated, a supercell structure in which the load ions are completely separated, and a supercell structure in which the load ions are partially separated. Among them, the transition metal atoms in the supercell structure in which the load ions are not separated are in a reduced state; the transition metal atoms in the supercell structure in which the load ions are completely separated are in an oxidized state; some transition metal atoms in the supercell structure in which the load ions are partially separated are in a reduced state, and some transition metal atoms are in an oxidized state.

[0076] By constructing a supercell structure hierarchically, a rich and comprehensive state of cathode materials is provided, which helps to generate a more comprehensive first-principles calculation data set, and then train a machine learning force field for cathode materials with higher prediction accuracy and wider applicability.

[0077] Step 102, adjusting the atomic positions and lattice vectors based on the supercell structure to obtain a structure to be calculated in which the atomic force values ​​belong to a preset force range and the structure is electrically neutral.

[0078] The specific positive electrode unit cell structure, supercell structure, structure to be calculated and equilibrium structure in this application are collectively referred to as unit cell structure.

[0079] In the unit cell structure, the atomic positions and lattice vectors jointly define the geometric structure and periodicity of the crystal. The atomic positions refer to the specific positions of each atom inside the unit cell structure, which are the coordinates of the a-axis, b-axis, and c-axis of the unit cell. The lattice vectors are the basic vectors that define the size and shape of the unit cell, namely the three non-collinear vectors of the a-axis, b-axis, and c-axis.

[0080] After the supercell structure is obtained, the structure to be calculated is obtained by adjusting the atomic positions and the lattice vectors, wherein the atomic force values ​​of the structure to be calculated belong to a preset force range and the overall structure is electrically neutral.

[0081] In some embodiments, the adjustment of atomic positions and lattice vectors based on the supercell structure to obtain a structure to be calculated whose atomic force values ​​belong to a preset force range and whose structure is electrically neutral includes: adjusting the atomic positions and lattice vectors of the supercell structure to obtain a balanced structure whose atomic force values ​​belong to a first preset force range; selecting a structure to be disturbed whose structure is electrically neutral from the balanced structure; for each of the structures to be disturbed, repeating the steps of applying a first perturbation to the atomic positions of the structure to be disturbed and applying a second perturbation to the lattice vectors of the structure to be disturbed according to a set number of perturbations, to obtain the structure to be calculated whose atomic force values ​​belong to the second preset force range for each of the structures to be disturbed for the set number of perturbations; wherein the disturbance direction and disturbance amount corresponding to the first perturbation and the second perturbation are random values ​​within a preset range.

[0082] The adjustment of atomic positions and lattice vectors is divided into two stages.

[0083] The first adjustment stage: adjusting the atomic positions and lattice vectors of the supercell structure to obtain a balanced structure in which the atomic force values ​​belong to the first preset force range.

[0084] Among them, the equilibrium structure includes an equilibrium structure in which the load ions are not separated, an equilibrium structure in which the load ions are completely separated, and an equilibrium structure in which the load ions are partially separated.

[0085] This stage is the structural optimization stage, which aims to make the supercell structure close to the equilibrium position, eliminate the possible electronic state degeneracy problem in the structure, obtain a stable equilibrium structure, and thus promote the convergence of the first-principles calculation.

[0086] In an ideal state, the first preset force range is a single value of 0. However, in practical applications, considering the structural complexity, optimization cost and optimization efficiency of the unit cell structure, the first preset force range is set to a value range close to 0. For example, the first preset force range is

[0087] The second adjustment stage: for each structure to be disturbed that is electrically neutral and is selected from the equilibrium structure, the steps of applying the first perturbation to the atomic positions of the structure to be disturbed and applying the second perturbation to the lattice vectors of the structure to be disturbed are repeated according to the set number of perturbations, so as to obtain the structure to be calculated whose atomic force values ​​corresponding to the set number of perturbations for each structure to be disturbed belong to the second preset force range.

[0088] Among them, the structures to be disturbed include the structures to be disturbed from which the load ions have not been separated, the structures to be disturbed from which the load ions have been completely separated, and the structures to be disturbed from which the load ions have been partially separated. The structures to be calculated include the structures to be calculated from which the load ions have not been separated, the structures to be calculated from which the load ions have been completely separated, and the structures to be calculated from which the load ions have been partially separated.

[0089] In some embodiments, the first perturbation process is to apply a magnitude satisfying the mean value of The first perturbation is a chi-square distribution with 3 degrees of freedom and points to a random direction. The perturbation process of the second perturbation is to apply a second perturbation of 1% to the lattice vector of the structure to be perturbed. If the number of perturbations is set to 5 and the structure to be perturbed is 100, then 500 structures to be calculated are obtained.

[0090] In the second adjustment stage, two disturbance factors are introduced, the first disturbance and the second disturbance, which act together on the structure to be disturbed to change the force state of its atoms. These disturbances cause the atoms to deviate from their original equilibrium position, that is, the relative equilibrium position corresponding to the deviation from the equilibrium structure.

[0091] The first perturbation and the second perturbation each follow the preset rules, and the perturbation direction and perturbation amount are set to random values, and these values ​​are limited to a specific safety range to ensure the appropriateness of the perturbation intensity. Through perturbation, when the stress state is changed, it is avoided to cause excessive changes to the structure, which may lead to non-convergence of the first principle calculation.

[0092] For each structure to be disturbed, each time the operation of "applying a first disturbance to the atomic position and a second disturbance to the lattice vector" is performed, a structure to be calculated is obtained in which the atomic force value in the deviated equilibrium position belongs to the second preset force range. For each structure to be disturbed, the number of executions is the set number of disturbances. The second preset force range is a preset value range, and the value of the second preset force range is greater than the value of the first preset force range.

[0093] This dual perturbation strategy aims to fully simulate various tiny perturbations that may exist in the unit cell structure of actual materials and increase the data diversity of the atomic force label in order to provide a richer and more comprehensive first-principles calculation data set. Its role is even more prominent for specific cathode unit cell structures that contain only one transition metal atom. When constructing a supercell structure based on a specific cathode unit cell structure that contains only one transition metal atom, the number of supercell structures that can be generated is relatively limited because there is no possibility of mixing different types of transition metal atoms. The introduction of a perturbation step can overcome this limitation and increase the richness of the structure and data diversity.

[0094] In some embodiments, a machine learning force field is used to perform structural optimization on a supercell structure. The machine learning force field is a general pre-trained machine learning force field that has been represented and learned on a large number of computational data sets, and its predicted labels include atomic magnetic moments, such as a Crystal Hamiltonian Graph Neural Network (CHGNet) force field.

[0095] In some embodiments, the supercell structure is divided into three types according to the detachment state of the load ions. In the process of generating the first-principles calculation data set, the structure conversion and the generation of the first-principles calculation data set are strictly carried out in accordance with the three detachment states of the load ions. Therefore, the generated first-principles calculation data set is divided into three types according to the detachment state of the load ions, including a load ion non-detached data set, a load ion completely detached data set, and a load ion partially detached data set.

[0096] The oxidation state of transition metal atoms in the supercell structure from which the load ions have not been separated and the supercell structure from which the load ions have been completely separated is single, with the former having all transition metal atoms in a reduced state and the latter having all transition metal atoms in an oxidized state. The proportion of load ions separated from the supercell structure from which the load ions have been partially separated is determined by the charge and discharge state of the battery and cannot be predetermined, so the state of the transition metal atoms therein is relatively complex, with some transition metal atoms in a reduced state and some transition metal atoms in an oxidized state. Relatively speaking, the difficulty of generating a data set from which the load ions have not been separated and a data set from which the load ions have been completely separated is much lower than the difficulty of generating a data set from which the load ions have been partially separated. Accordingly, when generating the data set, the generation time of the data set from which the load ions have not been separated and the data set from which the load ions have been completely separated is earlier than the generation time of the data set from which the load ions have been partially separated.

[0097] Since the supercell structures in different ion detachment states have different structural characteristics, different machine learning force fields are used for the supercell structures in different ion detachment states during structural optimization.

[0098] In some embodiments, the oxidation state of transition metal atoms in supercell structures from which the load ions have not been separated and in supercell structures from which the load ions have been completely separated is single, and a first structure optimization force field is used for structure optimization, and the first structure optimization force field may be an initial CHGNet force field.

[0099] In some embodiments, the oxidation state distribution of the transition metal atoms in the supercell structure from which the load ions are partially separated is relatively complex, and thus convergence during calculation is more difficult. The second structure optimization force field is used to optimize the structure of the supercell structure from which the load ions are partially separated. The second structure optimization force field is a force field trained on the first structure optimization force field based on the load ion non-detached data set and the load ion completely detached data set, that is, a force field trained by parameter fine-tuning, and accordingly, the second structure optimization force field can be a CHGNet force field trained by parameter fine-tuning. Through training, the second structure optimization force field simultaneously learns the coordination environment information of the transition metal atoms in the oxidized state and the transition metal atoms in the reduced state, and the structural optimization of the partial detachment of the load ions from the supercell structure is more accurate. In addition, in the subsequent steps, the second structure optimization force field is used to predict the magnetic moment, identify the oxidation state, and then set the magnetic moment parameters, which can avoid randomly assigning an erroneous high-energy electronic state and causing the calculation to not converge.

[0100] In some embodiments, the atomic positions and lattice vectors of the supercell structure are adjusted to obtain a balanced structure in which the atomic force values ​​belong to the first preset force range, including: adjusting the atomic positions and lattice vectors of the supercell structure from which the load ions have not been separated based on the first structure optimization force field, to obtain the balanced structure in which the load ions have not been separated and the atomic force values ​​belong to the first preset force range; and, adjusting the atomic positions and lattice vectors of the supercell structure from which the load ions have completely separated based on the first structure optimization force field, to obtain the balanced structure in which the load ions have completely separated and the atomic force values ​​belong to the first preset force range. That is, the first structure optimization force field is used to perform structural optimization on the supercell structure from which the load ions have not been separated and the supercell structure from which the load ions have completely separated, and the atomic positions and lattice vectors are relaxed simultaneously during the optimization process, thereby obtaining balanced structures in different load ion separation states.

[0101] In some embodiments, the convergence standard of the structure optimization of the first structure optimization force field is that the atomic force value is less than or equal to a preset force convergence standard value, such as Or, the number of iterations is less than or equal to a preset number of iterations, such as 1000, and the iteration is stopped when any criterion is reached.

[0102] In some embodiments, a first structure optimization force field is trained based on the data set of the load ions not detached and the data set of the load ions completely detached to obtain a second structure optimization force field. The second structure optimization force field after parameter fine-tuning training is used to perform structural optimization on the structure of the partial detachment of the load ions from the supercell structure, that is, the atomic positions and the lattice vectors of the supercell structure are adjusted to obtain a balanced structure in which the atomic force values ​​belong to the first preset force range, including: based on the second structure optimization force field, the atomic positions and the lattice vectors of the supercell structure in which the load ions are partially detached are adjusted to obtain the balanced structure in which the atomic force values ​​belong to the first preset force range in which the load ions are partially detached. The convergence standard of structural optimization of the second structure optimization force field can refer to the convergence standard of structural optimization of the first structure optimization force field.

[0103] In some embodiments, the maximum preset force value of the first preset force range is usually greater than the preset force convergence standard value. When the structure is optimized, the number of iterations is equal to the preset number of iterations and the atomic force value is greater than the preset force convergence standard value, and the iteration is stopped. At this time, the atomic force value of the balanced structure is greater than the preset force convergence standard value and less than or equal to the maximum preset force value of the first preset force range.

[0104] When performing first-principles calculations on cathode materials with periodic crystal structures, the calculation object should be electrically neutral. If it is not electrically neutral, the energy of the structure is divergent and the calculation is not feasible. Therefore, between the two adjustment stages, there is a process for determining the electrically neutral structure to be disturbed.

[0105] In some embodiments, the step of selecting an electrically neutral structure to be disturbed from the equilibrium structure includes: determining an equilibrium structure from which the load ions have not been separated as the structure to be disturbed, and determining an equilibrium structure from which the load ions have completely separated as the structure to be disturbed, which are respectively the structure to be disturbed from which the load ions have not been separated and the structure to be disturbed from which the load ions have completely separated.

[0106] Since each equilibrium structure where the load ions have not been separated and each equilibrium structure where the load ions are completely separated are both electrically neutral, the obtained equilibrium structures where the load ions have not been separated and the equilibrium structures where the load ions are completely separated can all be identified as structures to be disturbed, that is, the structures to be disturbed where the load ions have not been separated and the structures to be disturbed where the load ions are completely separated are obtained.

[0107] In some embodiments, before selecting the electrically neutral structure to be disturbed from the equilibrium structure, it also includes: predicting the predicted magnetic moment of each of the transition metal atoms in the equilibrium structure from which the load ion is partially detached based on the second structure optimization force field; determining the oxidation state identification result of each of the transition metal atoms in the equilibrium structure from which the load ion is partially detached based on the predicted magnetic moment; and determining the charge number of each of the equilibrium structures from which the load ion is partially detached based on the oxidation state identification result of each of the transition metal atoms.

[0108] The oxidation state identification of the equilibrium structure with complex oxidation state of the load ion partially detached is realized through the second structure optimization force field, thereby improving the accuracy of the oxidation state identification results.

[0109] Afterwards, the charges of each atom in the equilibrium structure from which each load ion is partially detached are added together, and the equilibrium structure from which each load ion is partially detached with a charge sum of 0 is determined as the electrically neutral structure to be disturbed, i.e., the structure to be disturbed from which each load ion is partially detached.

[0110] Since the equilibrium structure of partially detached load ions is not necessarily electrically neutral after the oxidation state of the transition metal atoms is identified by the second structure optimization force field, if a set number of structures to be disturbed with partially detached load ions are not obtained after selection, the construction of the supercell structure of partially detached load ions, structural optimization, and selection of structures to be disturbed with partially detached load ions are repeated until a set number of structures to be disturbed with partially detached load ions are obtained to meet the computing and training requirements.

[0111] Based on the supercell structure where the load ions have not been separated, the atomic positions and lattice vectors are adjusted to obtain a set number of structures to be calculated where the load ions have not been separated and the atomic force values ​​belong to the preset force range and the structure is electrically neutral.

[0112] Based on the supercell structure in which the load ions are completely detached, the atomic positions and lattice vectors are adjusted to obtain a structure to be calculated in which a set number of load ions are completely detached and the atomic force values ​​belong to a preset force range and the structure is electrically neutral.

[0113] Based on the supercell structure with partial detachment of the load ions, the atomic positions and lattice vectors are adjusted to obtain a structure to be calculated with a set number of partial detachments of the load ions, in which the atomic force values ​​belong to a preset force range and the structure is electrically neutral.

[0114] Step 103, determining the magnetic moment parameters of all transition metal atoms in the structure to be calculated.

[0115] The magnetic moment parameter in this step is the spin magnetic moment determined to be close to the real atomic magnetic moment of the transition metal atoms in the structure to be calculated. It is a key input parameter for first-principles calculations and is called the magnetic moment parameter, which is distinguished from the real atomic magnetic moment.

[0116] There are different methods for determining the magnetic moment parameters corresponding to the structures to be calculated in different load ion detachment states.

[0117] The magnetic moment parameters of each transition metal atom in the structure to be calculated without the detachment of the load ions are called the reduced state magnetic moment parameters, and the magnetic moment parameters of each transition metal atom in the structure to be calculated without the detachment of the load ions are called the oxidized state magnetic moment parameters. The difference between the number of spin-up electrons and the number of spin-down electrons in the d orbital of the transition metal atom in the structure to be calculated is calculated to obtain the reduced state magnetic moment parameters and the oxidized state magnetic moment parameters. The reduced state and oxidized state mentioned here are two states with lower and higher oxidation states, for example, Fe 2+ and Fe 3+ In comparison, Fe 2+ In the reduced state, Fe 3+ In oxidized state.

[0118] In some embodiments, before determining the magnetic moment parameters of all transition metal atoms in the structure to be calculated, the predicted magnetic moment of each transition metal atom in the equilibrium structure from which the load ion is partially detached is predicted based on the second structure optimization force field, and based on the predicted magnetic moment, the oxidation state identification result of each transition metal atom in the equilibrium structure from which the load ion is partially detached is determined.

[0119] Accordingly, the determining of the magnetic moment parameters of all transition metal atoms in the structure to be calculated includes: determining the magnetic moment parameters of each transition metal atom in the structure to be calculated from which the load ion is partially detached based on the oxidation state identification result.

[0120] The lithium iron manganese phosphate LiMn mentioned in step 101 x Fe y Take PO4 (x+y=1) as an example, the transition metal atoms therein are Mn and Fe.

[0121] In some embodiments, the Perdew-Burke-Ernzerhof (PBE) functional is selected to describe the exchange-correlation effect of electrons, and a U correction is applied to the electronic state of the d orbital of the transition metal atoms in the calculated structure, where the U values ​​corresponding to Mn-d and Fe-d are 4.5 eV and 4.3 eV, respectively. In the positive electrode material of lithium manganese iron phosphate, the 3d orbital electrons of Mn and Fe both adopt a high-spin arrangement, and the magnetic moment parameters can be determined according to the reduction state and oxidation state of Mn and Fe. The reduction state and oxidation state of Mn are Mn2+ and Mn 3+ , the number of 3d electrons is 5 and 4 respectively, and the corresponding calculated magnetic moment parameters are 5 and 4 respectively. The reduced state and oxidized state of Fe are Fe 2+ and Fe 3+ , the number of 3d electrons are 6 and 5 respectively, and the corresponding calculated magnetic moment parameters are 4 and 5 respectively.

[0122] In some embodiments, the second structure optimization force field is used to predict and determine the predicted magnetic moments of Mn and Fe in the disturbed structure where the load ions corresponding to the example positive electrode material are partially detached, and the oxidation state is identified to obtain the oxidation state identification results of the transition metal atoms. Among them, the critical value of the magnetic moment of the reduced state and the oxidized state of Mn is 4.3, and the critical value of the magnetic moment of the reduced state and the oxidized state of Fe is 4.1. The predicted magnetic moment of Mn is greater than 4.3, which is considered to be Mn in the reduced state. 2+ , less than 4.3 is considered as oxidized Mn 3+ The predicted magnetic moment of Fe is less than 4.1, which is considered to be reduced Fe. 2+ , greater than 4.1 is considered as oxidized Fe 3+ .

[0123] Afterwards, the magnetic moment parameters of Mn and Fe in the structure to be calculated in which the load ions corresponding to the example positive electrode material are partially separated are set according to the obtained oxidation state identification results. 2+ , Fe 3+ , Mn 2+ and Mn 3+ The magnetic moment parameters are set to 4, 5, 5 and 4 respectively.

[0124] This application uses the predicted magnetic moment of the transition metal atoms predicted by the second structure optimization force field to determine their oxidation state, and determines the magnetic moment parameters of the transition metal atoms in the structure to be calculated from which the load ions are partially detached according to the oxidation state of the transition metal atoms, so as to achieve the purpose of promoting the convergence of the first principles calculation.

[0125] Step 104 , performing first principles calculation based on the structure to be calculated and the magnetic moment parameters of each transition metal atom in the structure to be calculated, to obtain a first principles calculation data set whose convergence meets a set convergence standard.

[0126] First-principles calculations can be quantum chemical calculations of periodic solid materials based on density functional theory methods. The first-principles calculations can be performed by open source or commercial software commonly used in the field of material calculations, including but not limited to VASP, CP2K, Quantum ESPRESSO, Abinit, GPAW, WIEN2k, FHI-aims, CASTEP, Exciting, and MaterialsStudio.

[0127] The structure to be calculated and the magnetic moment parameters of the structure to be calculated are taken as input, and the first principles calculation method is used to obtain the first principles calculation results.

[0128] In order to improve the quality of the machine learning force field training set of cathode materials, the first-principles calculation results that meet the set convergence criteria are used as the first-principles calculation data set for training, and the load ion non-detached data set, load ion completely detached data set, and load ion partially detached data set that meet the set convergence criteria are obtained. The data set generation process is also the prediction process of the microscopic physical and chemical properties of cathode materials. Each data set contains microscopic physical and chemical data such as total energy, atomic force, atomic magnetic moment, and lattice stress.

[0129] The generation time of the data set of the undetached load ions and the data set of the completely detached load ions is earlier than the generation time of the data set of the partially detached load ions, that is, the data sets are generated in different levels according to the difficulty. In practical applications, the generation process of the first principle calculation data set of the positive electrode material can be found in Figure 2 .

[0130] like Figure 2 As shown, Figure 2 It is a schematic diagram of the generation process of a first principles calculation data set provided in an embodiment of the present application. Figure 2 The generation process of the first-principles calculation data set is summarized, and the generation process of the load ion non-detached data set, the load ion completely detached data set, and the load ion partially detached data set contained in the first-principles calculation data set is shown. As mentioned above, each load ion non-detached equilibrium structure and each load ion completely detached equilibrium structure are both electrically neutral, and the obtained load ion non-detached equilibrium structure and the load ion completely detached equilibrium structure can all be determined as structures to be disturbed, that is, the load ion non-detached structure to be disturbed and the load ion completely detached structure to be disturbed are obtained. Therefore, Figure 2 The step of "electrically neutral screening to obtain the structure to be disturbed in which the load ions are not detached" and "electrically neutral screening to obtain the structure to be disturbed in which the load ions are completely detached" can be skipped in the actual application process. Obtaining the equilibrium structure in which the load ions are not detached means obtaining the structure to be disturbed in which the load ions are not detached, and obtaining the equilibrium structure in which the load ions are completely detached means obtaining the structure to be disturbed in which the load ions are completely detached.

[0131] Lithium iron manganese phosphate LiMn x Fe y Taking PO4 (x+y=1) as an example of the positive electrode material, the convergence effect of the method described in the present application is explained.

[0132] The method described in the present application constructs a total of 500 structures to be calculated with undetached load ions, and obtains 372 converged calculation results, that is, a data set of undetached load ions containing 372 calculation results, with a convergence rate of 74%. A total of 500 structures to be calculated with completely detached load ions are constructed, and 500 converged calculation results are obtained, that is, a data set of completely detached load ions containing 500 calculation results are obtained, with a convergence rate of 100%. A total of 1,000 structures to be calculated with partially detached load ions are constructed, and 917 converged calculation results are obtained, that is, a data set of partially detached load ions containing 917 calculation results are obtained, with a convergence rate of 92%.

[0133] The prior art method constructs a completely random supercell structure, performs random delithiation, and randomly specifies the oxidized transition metal atoms, and then performs first-principles calculations. This method is recorded as a completely random method. The process is as follows:

[0134] (1) Starting from the LiFePO4 unit cell, a comparative extended structure is constructed according to the set expansion parameters 1×1×2, which contains 8 transition metal sites and 8 lithium sites.

[0135] (2) Generate random integers n(Mn) and n(Li) uniformly distributed between 0 and 8 as the number of Mn atoms and the number of Li atoms. Randomly select n(Mn) transition metal sites in the comparative extended structure, replace Fe at the selected sites with Mn, and randomly select 8-n(Li) lithium sites for delithiation, forming a comparative random supercell structure with both Fe / Mn mixing and delithiation.

[0136] (3) Based on the comparison of random supercell structures, the oxidation states of transition metal atoms are assigned. If n(Li) = 8, then all Fe atoms are Fe 2+ , all Mn are Mn 2+ ; If n(Li) = 0, then all Fe atoms are Fe 3+ , all Mn atoms are Mn 3+ In addition, 8-n (Li) sites were randomly selected from all Fe and Mn sites for oxidation. 2+ Oxidized to Fe 3 + , Mn 2+ Oxidized to Mn 3+ .

[0137] (4) Using the PBE functional and adding the U correction, the U values ​​corresponding to the d-electron states of Mn and Fe are 4.5 eV and 4.3 eV, respectively. The initial magnetic moments of Fe and Mn in the random supercell structure are compared according to the oxidation state assignment results in step (3). 2+ , Fe 3+ , Mn2+ and Mn 3+ The initial magnetic moments of the atoms are set to 4, 5, 5 and 4, respectively, and the initial magnetic moments of other atoms are set to 0. The first-principles calculations of the above-mentioned random supercell structures were performed using VASP software.

[0138] (5) Repeat steps (2) to (4) above 100 times and collect the converged calculation results into the calculation data set.

[0139] Finally, a total of 9 converged calculation results were obtained, and the calculation convergence ratio was 9%.

[0140] There are complex charge and magnetic moment arrangements in the unit cell structure of the positive electrode material. The completely random method randomly constructs the structure and specifies the charge and magnetic moment arrangements, which will cause a large deviation between the atomic position and the equilibrium position, resulting in a large number of nearly degenerate electronic states, making it difficult to converge the calculation of the positive electrode material. This application controls the adjustment of atomic positions and lattice vectors, and uses a machine learning force field with magnetic moment labels to predict the oxidation state of transition metal atoms, which can avoid the problem of electronic state degeneracy and promote the convergence of calculations to the correct electronic state.

[0141] By comparing the convergence ratios, it can be seen that the method described in the present application significantly improves the convergence ratio of first-principles calculations compared to completely random methods, avoids the waste of time and resources caused by a large number of failed calculations, improves the generation efficiency of first-principles calculation data sets, and reduces the cost of producing training sets for machine learning force fields.

[0142] Currently, the production of high-quality training sets is one of the main bottlenecks in the application of machine learning force fields to the calculation of battery positive electrode materials. This application can generate high-quality first-principles calculation data sets, which can promote the application of machine learning force fields in positive electrode material calculations, expand the time and space scales of problems that can be handled by positive electrode material calculations, and accelerate the development of new positive electrode materials.

[0143] Afterwards, the machine learning force field is trained based on the obtained first-principles calculation data set to obtain the positive electrode material machine learning force field. The positive electrode material machine learning force field can improve the prediction accuracy in the prediction of properties in multiple aspects such as total energy, atomic force, atomic magnetic moment and lattice stress, and can be used for molecular dynamics (MD) simulation, Monte Carlo (MC) simulation and other calculations on the positive electrode material, thereby predicting the physical and chemical properties of the positive electrode material.

[0144] Different types of positive electrode materials correspond to different positive electrode material machine learning force fields.

[0145] Lithium iron manganese phosphate LiMn x Fe yThe data sets of the load ions not separated, the load ions completely separated and the load ions partially separated of the positive electrode material PO4 (x+y=1) were used as training sets, the learning rate was set to 0.01, and the CHGNet force field was iteratively trained for 60 rounds (epochs=60) to obtain the positive electrode material machine learning force field corresponding to the positive electrode material lithium manganese iron phosphate.

[0146] Lithium iron manganese phosphate LiMn x Fe y The positive electrode material PO4 (x+y=1) is used as an example to illustrate the predictive effect of the machine learning force field of positive electrode materials.

[0147] like Figure 3 As shown, Figure 3 This is a comparative schematic diagram of the atomic magnetic moments of transition metal atoms predicted by different methods provided in the embodiments of this application. The left side is a comparative schematic diagram of the atomic magnetic moment of Fe, and the right side is a comparative schematic diagram of the atomic magnetic moment of Mn. The force field before fine-tuning is the first structure optimization force field, and the force field after fine-tuning is the second structure optimization force field. The first principle calculation is the reference value given by the density functional theory method. Figure 3 It can be seen that compared with the first structure optimization force field, the second structure optimization force field has been fine-tuned on the two data sets of load ion non-detachment and load ion complete detachment, and its prediction results for the magnetic moment of transition metal atoms are closer to the first principle calculation and more accurate. Therefore, it is more reasonable to use the second structure optimization force field for the preparation of the load ion partial detachment data set.

[0148] The machine learning force field for cathode materials has the high prediction efficiency of the force field method while maintaining the prediction accuracy of the first principles calculation method.

[0149] The most complex structure of the partial detachment of the load ion data set (partial delithiation data set) was used for verification. Figure 4 , Figure 5 and Figure 6 . In the figure, the force field before fine-tuning is the pre-trained CHGNet force field without any fine-tuning training, that is, the first structure optimization force field; the force field after fine-tuning is the CHGNet force field trained based on the load ion non-detached data set (non-delithiation data set), the load ion completely detached data set (completely delithiation data set) and the load ion partially detached data set (partially delithiation data set), that is, the positive electrode material machine learning force field. In the figure, the data predicted by the first structure optimization force field are represented by squares, and the data predicted by the positive electrode material machine learning force field are represented by circles.

[0150] like Figure 4 As shown, Figure 4: is a schematic diagram of the total energy predicted by a fine-tuned and trained positive electrode material machine learning force field and an unfine-tuned and trained force field provided in an embodiment of the present application. Figure 4 It can be seen that the total energy predicted by the machine learning force field of the cathode material is closer to the oblique line in the figure, that is, closer to the total energy calculated by the first principles. The root mean square error of the total energy predicted by the first structure optimization force field is 16.5 meV / atom, and the root mean square error of the total energy predicted by the machine learning force field of the cathode material is 2.5 meV / atom.

[0151] like Figure 5 As shown, Figure 5 This is a schematic diagram of the atomic forces predicted by a fine-tuned positive electrode material machine learning force field and an unfine-tuned force field provided in an embodiment of the present application. The left side is a schematic diagram of the atomic forces predicted by the first structure optimization force field, and the right side is a schematic diagram of the atomic forces predicted by the positive electrode material machine learning force field. Figure 5 It can be seen that the atomic forces predicted by the machine learning force field of the positive electrode material are closer to the oblique line in the figure, that is, closer to the atomic forces calculated by the first principles. The root mean square error of the atomic forces predicted by the first structure optimization force field is The root mean square error of the atomic forces predicted by the machine learning force field of the cathode material is

[0152] like Figure 6 As shown, Figure 6 : This is a schematic diagram of the comparison between the atomic magnetic moment predicted by a fine-tuned machine learning force field of a positive electrode material and the atomic magnetic moment predicted by an unfine-tuned training force field provided in the embodiment of the present application. The left side is a schematic diagram of the atomic magnetic moment of Fe, and the right side is a schematic diagram of the atomic magnetic moment of Mn. Figure 6 It can be seen that the atomic magnetic moment predicted by the positive electrode material machine learning force field is closer to the atomic magnetic moment calculated by the first principles. The magnetic moment distributions of Fe and Mn predicted by the positive electrode material machine learning force field are both near the diagonal, eliminating the systematic deviation.

[0153] pass Figure 4 , Figure 5 and Figure 6 It can be seen that the machine learning force field for positive electrode materials trained based on the first-principles calculation data set generated in this application has a very high prediction accuracy when predicting complex structures such as structures in which the load ions are partially detached.

[0154] In the embodiment of the present application, based on the specific positive electrode unit cell structure, i.e., the unit cell structure from which the load ions are not separated, a supercell structure is constructed to expand the material scale and more comprehensively reflect the microstructural characteristics of the material. Subsequently, the atomic position and lattice vector are adjusted based on the constructed supercell structure, so that the stress state of all atoms in the structure reaches the preset ideal equilibrium value, promotes the convergence of the calculation, and ensures that the overall structure remains electrically neutral, thereby obtaining a stable and ideal structure to be calculated, which not only improves the physical rationality of the structure, but also lays a solid foundation for subsequent calculations. There are transition metal atoms in the structure to be calculated, and the interaction between their electrons is relatively complex. The magnetic moment parameters of all transition metal atoms in the structure to be calculated are determined as the input of the first principle calculation, which can not only promote the convergence of the calculation, reduce the number of iterations, and improve the calculation efficiency, but also significantly improve the accuracy of the calculation and the reliability of the calculation results. Finally, the first principle calculation is performed based on the magnetic moment parameters of the structure to be calculated and the structure to be calculated. After calculation iteration, the convergence of the first principle calculation data set that meets the set convergence standard is obtained. This application improves the convergence success rate of first-principles calculations of battery positive electrode materials, provides a first-principles calculation data set with good convergence for the training of machine learning force fields of positive electrode materials, and avoids the waste of time and resources caused by a large number of failed calculations, thereby improving the generation efficiency of first-principles calculation data sets.

[0155] See also Figure 7 , Figure 7 This is a structural diagram of a system for generating a first-principles computing data set provided in an embodiment of the present application. For ease of explanation, only the parts related to the embodiment of the present application are shown.

[0156] The system 700 for generating a data set for first principle calculations includes: a construction module 701 , an adjustment module 702 , a determination module 703 , and a calculation module 704 .

[0157] The construction module 701 is used to construct a supercell structure based on a specific positive electrode unit cell structure, where the specific positive electrode unit cell structure is a unit cell structure in which the load ions are not separated.

[0158] The adjustment module 702 is used to adjust the atomic positions and lattice vectors based on the supercell structure to obtain a structure to be calculated in which the atomic force values ​​belong to a preset force range and the structure is electrically neutral.

[0159] The determination module 703 is used to determine the magnetic moment parameters of all transition metal atoms in the structure to be calculated.

[0160] The calculation module 704 is used to perform first-principles calculation based on the structure to be calculated and the magnetic moment parameters of each transition metal atom in the structure to be calculated, and obtain a first-principles calculation data set whose convergence meets a set convergence standard.

[0161] In some embodiments, the specific positive electrode unit cell structure comprises a plurality of transition metal atoms, and the building block is specifically used for:

[0162] Based on the specific positive electrode unit cell structure, a supercell structure is constructed in which a plurality of transition metal atoms are mixed and arranged and the load ions are not detached, a supercell structure in which a plurality of transition metal atoms are mixed and arranged and the load ions are completely detached, and a supercell structure in which a plurality of transition metal atoms are mixed and arranged and the load ions are partially detached.

[0163] In some embodiments, the specific positive electrode unit cell structure comprises one of the transition metal atoms, and the building block is specifically used for:

[0164] Based on the specific positive electrode unit cell structure, a supercell structure is constructed in which the structural size is a preset size and the load ions are not detached, a supercell structure in which the structural size is the preset size and the load ions are completely detached, and a supercell structure in which the structural size is the preset size and the load ions are partially detached.

[0165] In some embodiments, the adjustment module is specifically used to:

[0166] Adjusting the atomic positions and the lattice vectors of the supercell structure to obtain a balanced structure in which the atomic force values ​​belong to a first preset force range;

[0167] Selecting a structure to be disturbed whose structure is electrically neutral from the equilibrium structure;

[0168] For each of the structures to be disturbed, the steps of applying a first perturbation to the atomic positions of the structure to be disturbed and applying a second perturbation to the lattice vectors of the structure to be disturbed are repeated according to a set number of perturbations, so as to obtain the structure to be calculated in which the atomic force values ​​corresponding to the set number of perturbations of each structure to be disturbed belong to a second preset force range; wherein the disturbance direction and disturbance amount corresponding to the first perturbation and the second perturbation are random values ​​within a preset range.

[0169] In some embodiments, the first principle calculation data set includes a load ion non-detachment data set, a load ion complete detachment data set, and a load ion partial detachment data set, and the generation time of the load ion non-detachment data set and the load ion complete detachment data set is earlier than the generation time of the load ion partial detachment data set. The system further includes a training module, and the training module is used to:

[0170] Force field training is performed on the first structure optimization force field based on the load ion non-detachment data set and the load ion complete detachment data set to obtain a second structure optimization force field.

[0171] In some embodiments, the adjustment module is specifically used to:

[0172] Based on the first structure optimization force field, adjusting the atomic positions and the lattice vectors of the supercell structure from which the load ions have not been separated, to obtain the equilibrium structure from which the load ions have not been separated and whose atomic force values ​​belong to the first preset force range; and

[0173] Based on the first structure optimization force field, the atomic positions and the lattice vectors of the supercell structure from which the load ions are completely detached are adjusted to obtain the equilibrium structure from which the load ions are completely detached and whose atomic force values ​​fall within the first preset force range.

[0174] Based on the second structure optimization force field, the atomic positions and the lattice vectors of the supercell structure from which the load ions are partially detached are adjusted to obtain the equilibrium structure from which the load ions are partially detached and whose atomic force values ​​fall within the first preset force range.

[0175] In some embodiments, the system further comprises an oxidation state identification module, wherein the oxidation state identification module is configured to:

[0176] Predicting the predicted magnetic moment of each of the transition metal atoms in the equilibrium structure from which the load ions are partially separated based on the second structure optimization force field;

[0177] Based on the predicted magnetic moment, determining the oxidation state identification result of each of the transition metal atoms in the equilibrium structure from which the load ion is partially separated;

[0178] Based on the oxidation state assignment results of each of the transition metal atoms, the charge number of the equilibrium structure from which each of the charged ions is partially separated is determined.

[0179] In some embodiments, the determining module is specifically used to:

[0180] Based on the oxidation state identification result, the magnetic moment parameter of each of the transition metal atoms in the structure to be calculated from which the load ion is partially detached is determined.

[0181] The system for generating a first principles computing data set provided in the embodiment of the present application can implement each process of the embodiment of the above-mentioned method for generating a first principles computing data set and can achieve the same technical effect. To avoid repetition, it will not be described here.

[0182] Figure 8 8 is a structural diagram of a terminal provided in an embodiment of the present application. As shown in the figure, the terminal 8 of this embodiment includes: at least one processor 80 ( Figure 8Only one is shown in the figure), a memory 81, and a computer program 82 stored in the memory 81 and executable on the at least one processor 80, wherein the processor 80 implements the steps of any of the above-mentioned method embodiments when executing the computer program 82.

[0183] The terminal 8 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The terminal 8 may include, but is not limited to, a processor 80 and a memory 81. Those skilled in the art will appreciate that Figure 8 It is only an example of terminal 8 and does not constitute a limitation on terminal 8. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the terminal may also include input and output devices, network access devices, buses, etc.

[0184] The processor 80 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0185] The memory 81 may be an internal storage unit of the terminal 8, such as a hard disk or memory of the terminal 8. The memory 81 may also be an external storage device of the terminal 8, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal 8. Further, the memory 81 may also include both an internal storage unit of the terminal 8 and an external storage device. The memory 81 is used to store the computer program and other programs and data required by the terminal. The memory 81 may also be used to temporarily store data that has been output or is to be output.

[0186] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0187] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0188] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0189] In the embodiments provided in the present application, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0190] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0191] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0192] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0193] The present application implements all or part of the processes in the above-mentioned embodiment method, and may also be implemented through a computer program product. When the computer program product runs on a terminal, the terminal can implement the steps in the above-mentioned method embodiments when executing the computer program product.

[0194] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for generating a first-principles calculation data set, characterized in that: include: Based on a specific positive electrode unit cell structure, a supercell structure is constructed, wherein the specific positive electrode unit cell structure is a unit cell structure in which the load ions are not separated; Based on the supercell structure, the atomic positions and lattice vectors are adjusted to obtain a structure to be calculated in which the atomic force values ​​belong to a preset force range and the structure is electrically neutral; Determining the magnetic moment parameters of all transition metal atoms in the structure to be calculated; A first-principles calculation is performed based on the structure to be calculated and the magnetic moment parameters of each transition metal atom in the structure to be calculated to obtain a first-principles calculation data set whose convergence meets a set convergence standard.

2. The method according to claim 1, characterized in that The specific positive electrode unit cell structure contains a plurality of transition metal atoms, and the supercell structure is constructed based on the specific positive electrode unit cell structure, including: Based on the specific positive electrode unit cell structure, a supercell structure is constructed in which a plurality of transition metal atoms are mixed and arranged and the load ions are not detached, a supercell structure in which a plurality of transition metal atoms are mixed and arranged and the load ions are completely detached, and a supercell structure in which a plurality of transition metal atoms are mixed and arranged and the load ions are partially detached.

3. The method according to claim 1, characterized in that The specific positive electrode unit cell structure contains one of the transition metal atoms, and the supercell structure is constructed based on the specific positive electrode unit cell structure, including: Based on the specific positive electrode unit cell structure, a supercell structure is constructed in which the structural size is a preset size and the load ions are not detached, a supercell structure in which the structural size is the preset size and the load ions are completely detached, and a supercell structure in which the structural size is the preset size and the load ions are partially detached.

4. The method according to claim 1, characterized in that: The adjusting of the atomic positions and lattice vectors based on the supercell structure to obtain a structure to be calculated in which the atomic force values ​​belong to a preset force range and the structure is electrically neutral, includes: Adjusting the atomic positions and the lattice vectors of the supercell structure to obtain a balanced structure in which the atomic force values ​​belong to a first preset force range; Selecting a structure to be disturbed whose structure is electrically neutral from the equilibrium structure; For each of the structures to be disturbed, the steps of applying a first perturbation to the atomic positions of the structure to be disturbed and applying a second perturbation to the lattice vectors of the structure to be disturbed are repeated according to a set number of perturbations, so as to obtain the structure to be calculated in which the atomic force values ​​corresponding to the set number of perturbations of each structure to be disturbed belong to a second preset force range; wherein the disturbance direction and disturbance amount corresponding to the first perturbation and the second perturbation are random values ​​within a preset range.

5. The method according to claim 4, characterized in that The step of adjusting the atomic positions and the lattice vectors of the supercell structure to obtain a balanced structure in which the atomic force values ​​belong to a first preset force range includes: Based on the first structure optimization force field, adjusting the atomic positions and the lattice vectors of the supercell structure from which the load ions have not been separated, to obtain the equilibrium structure from which the load ions have not been separated and whose atomic force values ​​belong to the first preset force range; and Based on the first structure optimization force field, the atomic positions and the lattice vectors of the supercell structure from which the load ions are completely detached are adjusted to obtain the equilibrium structure from which the load ions are completely detached and whose atomic force values ​​fall within the first preset force range.

6. The method according to claim 4, characterized in that The first principle calculation data set includes a data set of non-detached load ions, a data set of completely detached load ions, and a data set of partially detached load ions, and the generation time of the data set of non-detached load ions and the data set of completely detached load ions is earlier than the generation time of the data set of partially detached load ions. The method further includes: Based on the data set of the undetached load ions and the data set of the completely detached load ions, force field training is performed on the first structure optimization force field to obtain a second structure optimization force field; The step of adjusting the atomic positions and the lattice vectors of the supercell structure to obtain a balanced structure in which the atomic force values ​​belong to a first preset force range includes: Based on the second structure optimization force field, the atomic positions and the lattice vectors of the supercell structure from which the load ions are partially detached are adjusted to obtain the equilibrium structure from which the load ions are partially detached and whose atomic force values ​​fall within the first preset force range.

7. The method according to claim 6, characterized in that Before selecting the electrically neutral structure to be disturbed from the equilibrium structure, the method further comprises: Predicting the predicted magnetic moment of each of the transition metal atoms in the equilibrium structure from which the load ions are partially separated based on the second structure optimization force field; Based on the predicted magnetic moment, determining the oxidation state identification result of each of the transition metal atoms in the equilibrium structure from which the load ion is partially separated; Based on the oxidation state assignment results of each of the transition metal atoms, the charge number of the equilibrium structure from which each of the charged ions is partially separated is determined.

8. The method according to claim 7, characterized in that Determining the magnetic moment parameters of all transition metal atoms in the structure to be calculated includes: Based on the oxidation state identification result, the magnetic moment parameter of each of the transition metal atoms in the structure to be calculated from which the load ion is partially detached is determined.

9. A system for generating a first-principles computing data set, characterized in that: include: A construction module, for constructing a supercell structure based on a specific positive electrode unit cell structure, wherein the specific positive electrode unit cell structure is a unit cell structure in which the load ions are not separated; An adjustment module, used for adjusting the atomic positions and lattice vectors based on the supercell structure, to obtain a structure to be calculated in which the atomic force value belongs to a preset force range and the structure is electrically neutral; A determination module, used to determine the magnetic moment parameters of all transition metal atoms in the structure to be calculated; A calculation module is used to perform first-principles calculation based on the structure to be calculated and the magnetic moment parameters of each transition metal atom in the structure to be calculated, so as to obtain a first-principles calculation data set whose convergence meets a set convergence standard.

10. A terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

11. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images