Method for constructing interatomic interaction potential function of multi-element doped silicon carbide reinforced aluminum matrix composite
By constructing a deep potential energy model, the problem of insufficient simulation accuracy of traditional potential functions in multi-element doped silicon carbide reinforced aluminum matrix composites is solved, and a low-cost and efficient simulation method is provided, which is suitable for describing the interatomic interactions of multi-element doped SiC/Al composites.
Patent Information
- Application Number
- CN202511539606.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-27
- Publication Date
- 2026-01-20
AI Technical Summary
Existing technologies struggle to effectively simulate interatomic interactions in multi-element doped silicon carbide-reinforced aluminum matrix composites using traditional empirical potential functions, especially at material interfaces where computational accuracy is insufficient.
A deep potential energy model was constructed using DPGEN software and the DeePMD-kit package in conjunction with the TensorFlow framework. The interatomic interaction potential function of multi-element doped SiC/Al composite material was constructed through training and validation processes, and molecular dynamics simulations were performed using the DP model.
This method enables a low-cost and efficient description of the interatomic interactions in multi-element doped SiC/Al composite materials, providing a theoretical basis for the deformation and damage mechanisms of the materials and reducing the cost of experimental screening.
Smart Images
Figure SMS_1
Abstract
Description
TECHNICAL FIELD
[0001] The application relates to a method for constructing an interatomic potential function of a composite material. BACKGROUND
[0002] As a new generation of structural and functional integrated material, silicon carbide reinforced aluminum matrix (SiC / Al) composite has become the core material in the fields of aerospace, high-end equipment and electronic packaging due to its excellent specific strength, high wear resistance, controllable thermal expansion coefficient and excellent thermal stability. In order to optimize the performance of SiC / Al composite, researchers usually modify the aluminum matrix through multi-element alloying. The addition of copper can form Al2Cu strengthening phase, which can significantly improve the room temperature and high temperature strength of the material; the introduction of magnesium not only improves the wettability of the melt to SiC particles, but also forms Mg2Si phase with silicon to strengthen the matrix; and the regulation of silicon element can inhibit the formation of harmful Al4C3 interface reaction product, thereby ensuring the interface stability. This multi-element synergistic modification strategy enables the aluminum matrix composite to adapt to more service environments.
[0003] However, multi-element alloying not only improves the performance, but also makes the atomic-scale interaction of the material system extremely complex, which brings severe challenges to traditional material design methods. Experimental research to clarify these problems is very time-consuming and requires expensive equipment, while computer simulation has a significant advantage in saving time and cost. Researchers simulate the deformation behavior of materials under different loads through molecular dynamics method, and then analyze the strengthening mechanism and failure mechanism of the composite material. The key to molecular dynamics is the selection of potential function, and a common method is to mix the traditional empirical potential functions, but this method ignores the influence of the atomic surrounding environment, especially the calculation accuracy at the material interface. At present, an interatomic potential function suitable for multi-element doped SiC / Al composite material has not been developed. SUMMARY
[0004] In order to solve the problem that the interatomic interaction in the silicon carbide reinforced aluminum matrix composite is too complex and is difficult to simulate by using a traditional empirical potential function, the application provides a method for constructing an interatomic potential function of a multi-element doped SiC / Al composite material according to a DPGEN software, which is suitable for describing the deformation behavior of a silicon carbide reinforced aluminum matrix composite formed by C, Si, Al, Cu and Mg.
[0005] The method for constructing the interatomic potential function of the multi-element doped silicon carbide reinforced aluminum matrix composite material is performed according to the following steps:
[0006] Step 1: Prepare initial data
[0007] ①, establish a crystal structure model, including:
[0008] establish a five-element SiC / Al interface model containing Si, Cu and Mg element doping, and establish a SiC / Al interface model containing point defects according to the five-element SiC / Al interface model;
[0009] establish a SiC / Al ternary interface model of Si, C and Al without alloy element doping, a SiC / Al interface model with one alloy element doping, and a SiC / Al interface model with two alloy elements doping;
[0010] establish an Al alloy model with one alloy element doping, an Al alloy model with two alloy elements doping, and an Al alloy model with three alloy elements doping;
[0011] establish a single-component crystal structure model;
[0012] establish a SiC crystal structure model;
[0013] The established five-element SiC / Al interface model containing Si, Cu and Mg element doping, the SiC / Al interface model containing point defects, the SiC / Al ternary interface model without alloy element doping, the SiC / Al interface model with one alloy element doping, the SiC / Al interface model with two alloy elements doping, the Al alloy model with one alloy element doping, the Al alloy model with two alloy elements doping, and the Al alloy model with three alloy elements doping are all crystal structure models;
[0014] The SiC crystal form in the five-element interface model containing Si, Cu and Mg element doping contains three crystal forms of 3C, 6H and 4H;
[0015] The SiC crystal form in the SiC / Al ternary interface model without alloy element doping, the SiC / Al interface model with one alloy element doping, and the SiC / Al interface model with two alloy elements doping contains three crystal forms of 3C, 6H and 4H;
[0016] The doping elements in the SiC / Al interface model with one alloy element doping are Cu, Mg and Si, and the doping elements in the SiC / Al interface model with two alloy elements doping are combinations of all two elements of Cu, Mg and Si;
[0017] The doping elements in the Al alloy model with one alloy element doping are Cu, Mg and Si, and the doping elements in the Al alloy model with two alloy elements doping are combinations of all two elements of Cu, Mg and Si;
[0018] The SiC crystal form in the SiC crystal structure model contains three crystal forms of 3C, 6H and 4H;
[0019] ②, structure optimization and ab initio molecular dynamics relaxation are performed on the five-element SiC / Al interface model doped with Si, Cu and Mg elements, the SiC / Al interface model containing point defects, the SiC / Al ternary interface model without doped alloy elements, the SiC / Al interface model doped with one alloy element, the SiC / Al interface model doped with two alloy elements, the Al alloy model doped with one alloy element, the Al alloy model doped with two alloy elements, the Al alloy model doped with three alloy elements, the single-component crystal structure model and the SiC crystal structure model established in step ① to obtain initial data as a training potential function;
[0020] Step two: training the potential function
[0021] The initial data obtained in step one is input into the DeePMD-kit package in the DPGEN software. After the data is converted in format, the smooth starting radius and the cutoff radius are set and input into the potential function network built by the TensorFlow framework in DPGEN for training to obtain an initialized deep potential model, i.e., a DP (Deep Potential) model.
[0022] Step three: verifying the accuracy of the potential function
[0023] The DP model trained in step two is provided to the LAMMPS software to simulate the multi-element doped SiC / Al interface model. After the simulation, if the DP model can correctly simulate the mechanical properties and deformation behavior of the material, a DP model that can accurately describe the interatomic potential in the multi-element doped SiC / Al composite material is obtained, the training is stopped, and a DP model that can accurately describe the interatomic potential in the multi-element doped SiC / Al composite material is obtained, and a potential function that can represent the mechanical properties and deformation behavior of the material is obtained. After the simulation, if a DP model that can accurately describe the interatomic potential in the multi-element doped SiC / Al composite material is not obtained, step four is performed.
[0024] Step four: exploration and labeling of the potential function model
[0025] The DP model of step two is called as an interatomic interaction potential, and one or more crystal structure models constructed in step one are subjected to molecular dynamics simulation to obtain a maximum force deviation value of atoms of the model; the temperature and pressure of the simulation are set, and the DP model prediction deviation value range is set as the upper limit and lower limit of the error criterion, the model with the maximum force deviation value within the upper limit and lower limit of the error criterion is selected as a hopeful improved model, and the model is marked, the marked model is subjected to DFT (Density Functional Theory) calculation, and then added to the training data set and subjected to step two and step three, until a potential function capable of representing the mechanical properties and deformation behavior of the material is obtained.
[0026] The beneficial effects of the present application are:
[0027] 1. The present application provides a method for training SiC / Al-based composite material potential function, which solves the problems of high cost and high time consumption of first principle calculation method, and solves the problem that traditional potential function is difficult to describe multi-element system, especially the interatomic potential in metal matrix composite material doped with different alloy elements, provides a feasible method for metal matrix composite material potential function training, and is suitable for constructing interatomic potential function of metal matrix composite material with different matrix and different reinforcement.
[0028] 2. The present application can describe the dynamic deformation and damage evolution process of the system under external load from the atomic scale, including dislocation initiation and movement, grain boundary sliding, crack propagation and interface failure, and provides a theoretical basis for molecular dynamics simulation of damage mechanism research and other mechanical and thermal properties of metal matrix composite materials.
[0029] 3. The present application can describe the interatomic interaction force in the metal matrix composite material system of the same series through the trained potential function, simulate the material deformation and damage process, and can perform low-cost and high-efficiency preliminary screening without large-scale screening through experiments, and provide theoretical guidance for experiments. DETAILED DESCRIPTION
[0030] The technical scheme of the present application is not limited to the following specific embodiments, and any reasonable combination of the specific embodiments is also included.
[0031] Specific embodiment one: the construction method of the interatomic interaction potential function of the multi-element doped silicon carbide reinforced aluminum matrix composite material according to the following steps:
[0032] Step one: preparing initial data
[0033] ①, establishing a crystal structure model, including:
[0034] establish a five-element SiC / Al interface model containing Si, Cu and Mg element doping, and establish a SiC / Al interface model containing point defects according to the five-element SiC / Al interface model;
[0035] establish a SiC / Al ternary interface model without alloy element doping, a SiC / Al interface model with one alloy element doping, and a SiC / Al interface model with two alloy elements doping;
[0036] establish an Al alloy model with one alloy element doping, an Al alloy model with two alloy elements doping, and an Al alloy model with three alloy elements doping;
[0037] establish a single-component crystal structure model;
[0038] establish a SiC crystal structure model;
[0039] The established five-element SiC / Al interface model containing Si, Cu and Mg element doping, the SiC / Al interface model containing point defects, the SiC / Al ternary interface model without alloy element doping, the SiC / Al interface model with one alloy element doping, the SiC / Al interface model with two alloy elements doping, the Al alloy model with one alloy element doping, the Al alloy model with two alloy elements doping, and the Al alloy model with three alloy elements doping are all crystal structure models;
[0040] The SiC crystal form in the five-element interface model containing Si, Cu and Mg element doping contains three crystal forms of 3C, 6H and 4H;
[0041] The SiC crystal form in the SiC / Al ternary interface model without alloy element doping, the SiC / Al interface model with one alloy element doping, and the SiC / Al interface model with two alloy elements doping contains three crystal forms of 3C, 6H and 4H;
[0042] The doping elements in the SiC / Al interface model with one alloy element doping are Cu, Mg and Si, and the doping elements in the SiC / Al interface model with two alloy elements doping are combinations of all two elements of Cu, Mg and Si;
[0043] The doping elements in the Al alloy model with one alloy element doping are Cu, Mg and Si, and the doping elements in the Al alloy model with two alloy elements doping are combinations of all two elements of Cu, Mg and Si;
[0044] The SiC crystal form in the SiC crystal structure model contains three crystal forms of 3C, 6H and 4H;
[0045] ②, structure optimization and ab initio molecular dynamics relaxation are performed on the five-element SiC / Al interface model doped with Si, Cu and Mg elements, the SiC / Al interface model containing point defects, the SiC / Al ternary interface model without doped alloy elements, the SiC / Al interface model doped with one alloy element, the SiC / Al interface model doped with two alloy elements, the Al alloy model doped with one alloy element, the Al alloy model doped with two alloy elements, the Al alloy model doped with three alloy elements, the single-component crystal structure model and the SiC crystal structure model established in step ① to obtain initial data as a training potential function;
[0046] Step two: training the potential function
[0047] The initial data obtained in step one is input into the DeePMD-kit package in the DPGEN software. After the data is converted in format, the smooth starting radius and the cutoff radius are set and input into the potential function network built by the TensorFlow framework in DPGEN for training to obtain an initialized deep potential model, i.e., a DP (Deep Potential) model.
[0048] Step three: verifying the accuracy of the potential function
[0049] The DP model trained in step two is provided to the LAMMPS software to simulate the multi-element doped SiC / Al interface model. After the simulation, if the DP model can correctly simulate the mechanical properties and deformation behavior of the material, a DP model that can accurately describe the interatomic potential in the multi-element doped SiC / Al composite material is obtained, the training is stopped, and a DP model that can accurately describe the interatomic potential in the multi-element doped SiC / Al composite material is obtained, and a potential function that can represent the mechanical properties and deformation behavior of the material is obtained. After the simulation, if a DP model that can accurately describe the interatomic potential in the multi-element doped SiC / Al composite material is not obtained, step four is performed.
[0050] Step four: exploration and labeling of the potential function model
[0051] The DP model of step two is called as an interatomic interaction potential, and one or more crystal structure models constructed in step one are subjected to molecular dynamics simulation to obtain a maximum force deviation value of atoms of the model; the temperature and pressure of the simulation are set, and the DP model prediction deviation value range is set as the upper limit and lower limit of the error criterion, the model with the maximum force deviation value within the upper limit and lower limit of the error criterion is selected as a hopeful improved model, and the model is marked, the marked model is subjected to DFT (Density Functional Theory) calculation, and then added to the training data set and subjected to step two and step three, until a potential function capable of representing the mechanical properties and deformation behavior of the material is obtained.
[0052] The embodiment has the following beneficial effects:
[0053] 1. The embodiment provides a method for training a SiC / Al-based composite material potential function, solves the problems of high cost and high time consumption of the first principle calculation method, solves the problem that a traditional potential function is difficult to describe a multi-element system, especially the interatomic interaction potential in a metal matrix composite material doped with different alloy elements, provides a feasible method for training a metal matrix composite material potential function, and is suitable for constructing an interatomic potential function of a metal matrix composite material with different matrices and different reinforcing bodies.
[0054] 2. The embodiment can describe the dynamic deformation and damage evolution process of a system under external load at the atomic scale, including dislocation initiation and movement, grain boundary sliding, crack propagation and interface failure, and provides a theoretical basis for molecular dynamics simulation of the damage mechanism and other mechanical and thermal properties of a metal matrix composite material.
[0055] 3. The embodiment can describe the interatomic interaction force in a metal matrix composite material system of the same series through the trained potential function, simulate the material deformation and damage process, and perform low-cost and high-efficiency preliminary screening without the need for large-scale screening through experiments and providing theoretical guidance for experiments.
[0056] Specific implementation method two: different from the specific implementation method one, the single-component crystal structure model of step one 1 includes Mg element crystal structure model, Cu element crystal structure model, Al element crystal structure model and Si element crystal structure model.
[0057] Specific embodiment three: the difference between this embodiment and specific embodiment one or two is that: step one ① the Si, Cu and Mg element doped five element SiC / Al interface model, the SiC / Al interface model containing point defects, the establishment of the SiC / Al ternary interface model of undoped alloy elements, the SiC / Al interface model doped with one alloy element, the SiC / Al interface model doped with two alloy elements, the Al alloy model doped with one alloy element, the Al alloy model doped with two alloy elements, the Al alloy model doped with three alloy elements, the establishment of single component crystal structure model, SiC crystal structure model is completed by VASP software.
[0058] Specific embodiment four: the difference between this embodiment and one of specific embodiments one to three is that: step one ② the structure optimization adopts init_bulk module in DPGEN software.
[0059] Specific embodiment five: the difference between this embodiment and one of specific embodiments one to four is that: step one ② the initial data includes atomic coordinates, atomic energies, forces and virial information.
[0060] Specific embodiment six: the difference between this embodiment and one of specific embodiments one to five is that: in step two, the embedded neural network in the deep potential model is 3 layers, and the node numbers of the 3 layers are 25, 50 and 100 in turn; the fitting neural network is 3 layers, and the node numbers of the 3 layers are 240, 240 and 240 in turn, the learning rate is exponentially decaying, and the starting learning rate, the stopping learning rate and the decay step are set to 0.001, 3.51e -8 and 50.
[0061] Specific embodiment seven: the difference between this embodiment and one of specific embodiments one to six is that: step two uses different random seeds to train 4 DP models at the same time, and trains for 800000 steps.
[0062] Specific embodiment eight: the difference between this embodiment and one of specific embodiments one to seven is that: step four uses lammps molecular dynamics software to perform molecular dynamics simulation under NVT ensemble or NPT ensemble.
[0063] Specific embodiment nine: the difference between this embodiment and one of specific embodiments one to eight is that: step four sets the simulation temperature to 100K, 200K, 300K, 400K, 500K, 600K, 700K, 800K, 900K and 1000K.
[0064] Specific embodiment ten: the difference between this embodiment and one of specific embodiments one to nine is that: step four sets the simulation pressure to 1 standard atmosphere.
[0065] Example 1
[0066] The method for constructing the interatomic interaction potential function of the multi-element doped silicon carbide reinforced aluminum matrix composite material of the embodiment is performed according to the following steps:
[0067] Step one: prepare initial data
[0068] ①, establish a crystal structure model, including:
[0069] establish a five-element SiC / Al interface model containing Si, Cu and Mg element doping, and establish a SiC / Al interface model containing point defects according to the five-element SiC / Al interface model;
[0070] establish a SiC / Al ternary interface model of Si, C and Al without alloying element doping, a SiC / Al interface model doped with one alloying element, and a SiC / Al interface model doped with two alloying elements;
[0071] establish an Al alloy model doped with one alloying element, an Al alloy model doped with two alloying elements, and an Al alloy model doped with three alloying elements;
[0072] establish a single-component crystal structure model;
[0073] establish a SiC crystal structure model;
[0074] The established five-element SiC / Al interface model containing Si, Cu and Mg element doping, the SiC / Al interface model containing point defects, the SiC / Al ternary interface model without alloying element doping, the SiC / Al interface model doped with one alloying element, the SiC / Al interface model doped with two alloying elements, the Al alloy model doped with one alloying element, the Al alloy model doped with two alloying elements, and the Al alloy model doped with three alloying elements are all crystal structure models;
[0075] The SiC crystal form in the five-element interface model doped with Si, Cu and Mg elements contains three crystal forms of 3C, 6H and 4H;
[0076] The SiC crystal form in the SiC / Al ternary interface model without alloying element doping, the SiC / Al interface model doped with one alloying element, and the SiC / Al interface model doped with two alloying elements contains three crystal forms of 3C, 6H and 4H;
[0077] The doping elements of the SiC / Al interface model doped with one alloying element are Cu, Mg and Si, and the doping elements of the SiC / Al interface model doped with two elements are all combinations of two elements from Cu, Mg and Si;
[0078] The doped elements of the Al alloy model doped with one alloying element are Cu, Mg and Si, and the doped elements of the Al alloy model doped with two elements are all combinations of two elements of Cu, Mg and Si;
[0079] The SiC crystal structure model includes three crystal types of 3C, 6H and 4H;
[0080] The single-component crystal structure model includes Mg element crystal structure model, Cu element crystal structure model, Al element crystal structure model and Si element crystal structure model;
[0081] The establishment of the SiC / Al interface model containing Si, Cu and Mg element doping, the SiC / Al interface model containing point defects, the SiC / Al ternary interface model without alloying element doping, the SiC / Al interface model doped with one alloying element, the SiC / Al interface model doped with two alloying elements, the Al alloy model doped with one alloying element, the Al alloy model doped with two alloying elements, the Al alloy model doped with three alloying elements, the single-component crystal structure model and the SiC crystal structure model is completed by the VASP software;
[0082] ②, the SiC / Al interface model containing Si, Cu and Mg element doping, the SiC / Al interface model containing point defects, the SiC / Al ternary interface model without alloying element doping, the SiC / Al interface model doped with one alloying element, the SiC / Al interface model doped with two alloying elements, the Al alloy model doped with one alloying element, the Al alloy model doped with two alloying elements, the Al alloy model doped with three alloying elements, the single-component crystal structure model and the SiC crystal structure model established in step ① are subjected to structure optimization and ab initio molecular dynamics relaxation to obtain initial data as the training potential function; the structure optimization adopts the init_bulk module in the DPGEN software; the initial data includes atomic coordinates, atomic energies, forces and virial information;
[0083] Step two: training potential function
[0084] The initial data obtained in step one is input into the DeePMD-kit package in the DPGEN software. The DeePMD-kit package converts the data format, sets the smooth starting radius to 0.5 Å, sets the cutoff radius to 6.0 Å, and inputs it into the potential function network built by the TensorFlow framework in DPGEN for training to obtain an initial deep potential model, i.e., a DP (DeepPotential) model. Four DP models are trained simultaneously using different random seeds, and the training is performed for 800,000 steps. The embedding neural network in the initial deep potential model is 3 layers, and the node numbers of the 3 layers are 25, 50, and 100, respectively. The fitting neural network is 3 layers, and the node numbers of the 3 layers are 240, 240, and 240, respectively. The learning rate is exponentially decaying, and the starting learning rate, stopping learning rate, and decay step are set to 0.001, 3.51e -8 and 50, respectively.
[0085] Step three: verifying the accuracy of the potential function
[0086] The DP model trained in step two is provided to the LAMMPS software to simulate the multi-element doped SiC / Al interface model. After simulation, if the DP model can correctly simulate the mechanical properties and deformation behavior of the material, a DP model that can accurately describe the interatomic potential in the multi-element doped SiC / Al composite material is obtained, and the training is stopped. The obtained DP model that can accurately describe the interatomic potential in the multi-element doped SiC / Al composite material is obtained, and a potential function that can represent the mechanical properties and deformation behavior of the material is obtained. If the DP model that can accurately describe the interatomic potential in the multi-element doped SiC / Al composite material is not obtained after simulation, step four is performed.
[0087] Step four: exploration and labeling of potential function model
[0088] The DP model of step two is called as an interatomic potential, and one or more crystal structure models constructed in step one are simultaneously subjected to molecular dynamics simulation using the lammps molecular dynamics software in an NVT ensemble to obtain a maximum force deviation value of the atomic force of the model; the temperature and pressure of the simulation are set, the temperature is 100K, 200K, 300K, 400K, 500K, 600K, 700K, 800K, 900K, 1000K, and the pressure is 1 standard atmosphere; at the same time, the upper limit of the DP model prediction deviation value is set to 0.15eV / Å, and the lower limit is set to 0.05eV / Å, which is used as an error criterion, the model with the maximum force deviation value within the upper limit and the lower limit of the error criterion is selected as a model that is expected to be improved, and the model is marked, the marked model is calculated by DFT (Density Functional Theory) and added to the training data set and subjected to steps two and three, until a potential function capable of representing the mechanical properties and deformation behavior of the material is obtained.
[0089] After multiple rounds of training, the LAMMPS software and the first-principles calculation results are compared in this embodiment, and Table 1 shows the DFT (first-principles) and MD (LAMMPS molecular dynamics software) calculation values of the lattice constant, elastic constant and other parameters of SiC; the calculated lattice constant results are highly consistent, and the elastic constant is also relatively close, with a maximum relative error of less than 10%, which can correctly describe the interatomic interaction of the silicon carbide reinforced aluminum matrix composite material.
[0090] Using the same computer system and the same number of cores, the same model is subjected to first-principles molecular dynamics relaxation and MD relaxation using the machine learning potential function of this embodiment at the same temperature. It is found that the calculation time required for first-principles molecular dynamics relaxation for 1fs is 128 seconds, and the calculation time required for MD relaxation using the machine learning potential function of this embodiment for 1fs is only 0.0243 seconds, which is about 4 orders of magnitude faster than first-principles calculation, which fully embodies the advantage of high efficiency of this embodiment relative to first-principles calculation.
[0091] Table 1
[0092]
Claims
1. A method for constructing an interatomic potential function for a multi-element doped silicon carbide reinforced aluminum matrix composite material, comprising: The method for constructing an interatomic interaction potential function of a multi-element doped silicon carbide reinforced aluminum matrix composite is performed according to the following steps: Step one: preparing initial data ①, establishing a crystal structure model, including: establishing a five-element SiC / Al interface model containing Si, Cu and Mg element doping, and establishing a SiC / Al interface model containing point defects according to the five-element SiC / Al interface model; establishing a Si, C, Al ternary interface model of SiC / Al without alloy element doping, a SiC / Al interface model doped with one alloy element, and a SiC / Al interface model doped with two alloy elements; establishing an Al alloy model doped with one alloy element, an Al alloy model doped with two alloy elements, and an Al alloy model doped with three alloy elements; establishing a single-component crystal structure model; establishing a SiC crystal structure model; the established five-element SiC / Al interface model containing Si, Cu and Mg element doping, the SiC / Al interface model containing point defects, the SiC / Al ternary interface model without alloy element doping, the SiC / Al interface model doped with one alloy element, the SiC / Al interface model doped with two alloy elements, the Al alloy model doped with one alloy element, the Al alloy model doped with two alloy elements, and the Al alloy model doped with three alloy elements are all crystal structure models; the SiC crystal form in the five-element interface model doped with Si, Cu and Mg elements contains three crystal forms of 3C, 6H and 4H; the SiC crystal form in the SiC / Al ternary interface model without alloy element doping, the SiC / Al interface model doped with one alloy element, and the SiC / Al interface model doped with two alloy elements contains three crystal forms of 3C, 6H and 4H; the doping elements of the SiC / Al interface model doped with one alloy element are Cu, Mg and Si, and the doping elements of the SiC / Al interface model doped with two alloy elements are all combinations of two elements from Cu, Mg and Si; the doping elements of the Al alloy model doped with one alloy element are Cu, Mg and Si, and the doping elements of the Al alloy model doped with two alloy elements are all combinations of two elements from Cu, Mg and Si; the SiC crystal form in the SiC crystal structure model contains three crystal forms of 3C, 6H and 4H; ②, performing structure optimization and ab initio molecular dynamics relaxation on the established five-element SiC / Al interface model containing Si, Cu and Mg element doping, the SiC / Al interface model containing point defects, the SiC / Al ternary interface model without alloy element doping, the SiC / Al interface model doped with one alloy element, the SiC / Al interface model doped with two alloy elements, the Al alloy model doped with one alloy element, the Al alloy model doped with two alloy elements, the Al alloy model doped with three alloy elements, the single-component crystal structure model, and the SiC crystal structure model established in step ① to obtain initial data as a training potential function; Step two: training the potential function The initial data obtained in step one is input into the DeePMD-kit package in the DPGEN software, the DeePMD-kit package converts the data into a format, sets a smooth starting radius and a cutoff radius, and inputs them into a potential function network built by the TensorFlow framework in DPGEN for training, to obtain an initial deep potential energy model, namely a DP model; Step three: verifying the accuracy of the potential function The DP model trained in step two is provided to the LAMMPS software to simulate the multi-element doped SiC / Al interface model; if the DP model can correctly simulate the mechanical properties and deformation behavior of the material after simulation, a DP model that can accurately describe the interatomic interaction potential in the multi-element doped SiC / Al composite material is obtained, the training is stopped, and the DP model that can accurately describe the interatomic interaction potential in the multi-element doped SiC / Al composite material is obtained, and the potential function that can represent the mechanical properties and deformation behavior of the material is obtained; if a DP model that can accurately describe the interatomic interaction potential in the multi-element doped SiC / Al composite material is not obtained after simulation, step four is performed; Step four: exploration and labeling of potential function model The DP model in step two is called as the interatomic interaction potential, and one or more crystal structure models constructed in step one are subjected to molecular dynamics simulation to obtain the maximum force deviation value of the atomic force of the model; The temperature and pressure of the simulation are set, and the DP model prediction deviation value range is set as the upper and lower limits of the error criterion, the model with the maximum force deviation value within the upper and lower limits of the error criterion is selected as the model that is expected to be improved, and the model is labeled, the labeled model is subjected to DFT calculation, and then added to the training data set and subjected to steps two and three, until the potential function that can represent the mechanical properties and deformation behavior of the material is obtained.
2. The method of claim 1, wherein the method is characterized by: The single-component crystal structure model in step one ① includes Mg element crystal structure model, Cu element crystal structure model, Al element crystal structure model and Si element crystal structure model.
3. The method of claim 1, wherein the method is characterized by: The five-element SiC / Al interface model containing Si, Cu and Mg elements, the SiC / Al interface model containing point defects, the SiC / Al ternary interface model established without alloying elements, the SiC / Al interface model doped with one alloying element, the SiC / Al interface model doped with two alloying elements, the Al alloy model doped with one alloying element, the Al alloy model doped with two alloying elements, the Al alloy model doped with three alloying elements, the single-component crystal structure model, and the establishment of the SiC crystal structure model are completed by the VASP software.
4. The method of claim 1, wherein the method is characterized by: The structure optimization in step one ② uses the init_bulk module in the DPGEN software.
5. The method of constructing a multi-element doped silicon carbide reinforced aluminum matrix composite interatomic potential function according to claim 1, wherein: The initial data in step one ② includes the coordinates of atoms, the energy of atoms, force and virial information.
6. The method of constructing a multi-element doped silicon carbide reinforced aluminum matrix composite interatomic potential function according to claim 1, wherein: The embedded neural network in the deep potential model in step two is 3 layers, and the node numbers of the 3 layers are 25, 50, and 100 in turn; the fitting neural network is 3 layers, and the node numbers of the 3 layers are 240, 240, and 240 in turn, the learning rate is exponentially decaying, and the starting learning rate, the stopping learning rate, and the decay step are set to 0.001, 3.51e -8 -4, and 50 respectively. -8 7. The method of claim 1, wherein the method further comprises: determining the interatomic potential function of the multi-element doped silicon carbide reinforced aluminum matrix composite material. Step two uses different random seeds to train four DP models simultaneously for 800,000 steps.
8. The method of claim 1, wherein the method is characterized by: Step four uses the lammps molecular dynamics software to perform molecular dynamics simulation under NVT ensemble or NPT ensemble.
9. The method of claim 1, wherein the method is characterized by: The temperature of the simulation is set to 100 K, 200 K, 300 K, 400 K, 500 K, 600 K, 700 K, 800 K, 900 K, 1000 K, according to step four.
10. The method of claim 1, wherein the method is characterized by: The pressure of the simulation is set to 1 atmosphere, according to step four.