Simulation Method, System and Medium for Regulating Thermal Conductivity of Ferroelectric Domains in Bismuth Ferrite Films

Abstract

The present application discloses a simulation method, system and medium for regulating the thermal conductivity of ferroelectric domains in bismuth ferrite thin films, which mainly relates to the technical field of bismuth ferrite thin films, and is used to solve the problems that in the existing solutions, it is difficult to effectively analyze the structure-property relationship between the atomic configuration of ferroelectric domain walls and the thermal transport characteristics, and the thermal activation migration process of ferroelectric domain walls involves multi-physical field coupling effects, and it is difficult to achieve precision by using conventional simulation means. It includes: obtaining the NEP potential function; calculating the thermal switching ratio between single-domain and double-domain BFO molecular models at each preset temperature; obtaining the theoretical maximum value of the change of the thermal switching ratio with temperature, and obtaining the specific temperature corresponding to the theoretical maximum value; at the specific temperature, obtaining the change curve of the thermal switching ratio of the BFO molecular model with the change of the preset number of domain walls, and obtaining the specific number of domain walls corresponding to the theoretical maximum value; using the specific temperature and the specific number of domain walls corresponding to the theoretical maximum value as the final guiding conditions.

Description

Technical Field

[0001] This application relates to the technical field of bismuth ferrite thin films, and particularly to a simulation method, system, and medium for regulating the thermal conductivity of ferroelectric domains in bismuth ferrite thin films. Background Art

[0002] Bismuth ferrite thin film (chemical formula BiFeO3, abbreviated as BFO), as a typical perovskite-type room-temperature multiferroic material, has ferroelectricity and a thickness ranging from dozens of nanometers to several micrometers, and there are domain structures with different spontaneous polarizations inside it. As the interface dividing different polarized domains, ferroelectric domain walls can move, generate, and erase rapidly under the action of an external electric field. They can serve as excellent phonon scattering interfaces, affecting the phonon transport process and changing the thermal conductivity of the material. Therefore, the BFO thin film (BFO molecular model) can achieve dynamic reversible switching of thermal conductivity through domain wall density modulation. Through the design of ferroelectric domain density gradient and three-dimensional microstructure regulation, a new generation of intelligent thermal material system with wide-range continuous modulation ability and breakthrough thermal switching ratio can be constructed.

[0003] In the prior art, the performance optimization of bismuth ferrite thin film thermal switch materials is mainly limited by two major technical bottlenecks: First, traditional material design methods rely on empirical trial-and-error patterns, making it difficult to effectively analyze the structure-property relationship between the atomic configuration of ferroelectric domain walls and thermal transport characteristics; Second, existing molecular dynamics simulation methods are restricted by computational complexity. When simulating the heat conduction behavior of multi-domain structures at the nanoscale, there are defects such as limited time scale (usually less than 1 ns) and insufficient spatial resolution (unable to accurately characterize the 1-10 nm domain wall structure). Especially in multi-ferroic material systems such as bismuth ferrite, the thermally activated migration process of ferroelectric domain walls involves multi-physical field coupling effects, and it is difficult to achieve accuracy using conventional simulation means, severely restricting the optimization efficiency of material thermal switch performance. Summary of the Invention

[0004] This application provides a simulation method, system, and medium for regulating the thermal conductivity of ferroelectric domains in bismuth ferrite thin films to solve the problems that traditional material design methods rely on empirical trial-and-error patterns, making it difficult to effectively analyze the structure-property relationship between the atomic configuration of ferroelectric domain walls and thermal transport characteristics, and the thermally activated migration process of ferroelectric domain walls involves multi-physical field coupling effects, and it is difficult to achieve accuracy using conventional simulation means.

[0005] In a first aspect, this application provides a simulation method for regulating the thermal conductivity of ferroelectric domains in bismuth ferrite thin films, and the method includes:

[0006] Establish BFO molecular models with different supercell structures, and then use the AIMD technology to perform canonical ensemble and isothermal-isobaric ensemble simulations on the BFO molecular models respectively to obtain the phase space data of BFO at different preset temperatures as the initial training set;

[0007] Configure the hyperparameters of the GPUMD software, and use the NEP tool embedded in the GPUMD software with the initial training set to train the neural network evolved potential (NEP) until the initial training set can describe all the preset target MD scenarios, and then obtain the NEP potential function;

[0008] Obtain the single-domain and double-domain BFO molecular models. Based on the NEP potential function, simulate the thermal conductivity of the single-domain and double-domain BFO molecular models in the GPUMD software within the preset temperature range to obtain the thermal conductivity of the single-domain state and the thermal conductivity of the double-domain state at each preset temperature, and calculate the thermal switching ratio between the single-domain and double-domain BFO molecular models at each preset temperature; furthermore, obtain the variation curve of the thermal switching ratio with temperature change, obtain the theoretical maximum value of the thermal switching ratio with temperature change, and obtain the specific temperature corresponding to the theoretical maximum value;

[0009] Obtain the single-domain and BFO molecular models with a preset number of domain walls. At a specific temperature, perform isothermal-isobaric ensemble and canonical ensemble simulations on the BFO molecular models with the preset number of domain walls to obtain the thermal switching ratio between each preset number of domain walls and the single-domain BFO molecular model;

[0010] Obtain the variation curve of the thermal switching ratio of the BFO molecular model with the change of the preset number of domain walls, and obtain the specific number of domain walls corresponding to the theoretical maximum value;

[0011] Take the specific temperature and the specific number of domain walls corresponding to the theoretical maximum value as the final guiding conditions.

[0012] In an implementation manner of this application, establish BFO molecular models with different supercell structures, and then use the AIMD technology to perform canonical ensemble and isothermal-isobaric ensemble simulations on the BFO molecular models respectively to obtain the phase space data of BFO at different preset temperatures as the initial training set, specifically including:

[0013] Construct a 1×1×1 BFO unit cell, use the phonopy software to obtain supercells with different numbers of atoms, and perform AIMD simulations at different preset temperatures in the first-principles calculation program respectively to obtain the initial training set; where the preset temperatures include at least: 100K, 200K, 300K, 400K, 500K, and 600K;

[0014] Use the read function of ASE to read the BFO model.xyz file of the 3×3×3 supercell, store the read atomic structure data in the result file, use the write function of ASE to save the atomic structure data in the result file in a format that can be called by the first-principles program, use a normal distribution to randomly generate the perturbation amount of the atomic positions, and call the command of the System object to generate a structure file for each perturbed system and save it as a new crystal structure, and add it to the initial training set.

[0015] In an implementation manner of the present application, using the initial training set and the NEP tool embedded in the GPUMD software, train the neural evolution potential NEP until the initial training set can describe all preset target MD scenarios, and then obtain the NEP potential function, specifically including:

[0016] S0. Perform single-point energy calculations on the structures in the initial training set in the first-principles program to obtain energy and virial information, and then obtain the converged first training data, and then obtain the first converged training set;

[0017] S1. Use the first converged training set to train the neural evolution potential NEP to obtain the 1-NEP function;

[0018] S2. Use the supercell atomic BFO molecular model, the 1-NEP function, and the preset condition information to obtain the phase space existing between 100K and 700K;

[0019] S3. Use farthest point sampling to select BFO molecular models that meet the preset limit range from the phase space;

[0020] S4. Perform single-point energy calculations on the BFO molecular models that meet the preset limit range in the first-principles program to obtain energy and virial information, and then obtain the converged second training data, and then add the second training data to the first converged training set to obtain the second converged training set;

[0021] S5. Use the second converged training set to train the neural evolution potential NEP to obtain the 2-NEP function;

[0022] S6. Repeat S2 - S5 until the preset stability condition is met to obtain the final NEP potential function.

[0023] In an implementation manner of the present application, obtain single-domain and double-domain BFO molecular models, and based on the NEP potential function, simulate the thermal conductivity of the single-domain and double-domain BFO molecular models in the GPUMD software within a preset temperature range to obtain the thermal conductivity of the single-domain state and the double-domain state at each preset temperature, and calculate the thermal switching ratio between the single-domain and double-domain BFO molecular models at each preset temperature, specifically including:

[0024] Use HNEMD for simulation, and obtain a number of preset temperatures at a preset temperature interval, and then obtain the thermal conductivity of the single-domain and double-domain BFO molecular models corresponding to each preset temperature;

[0025] Through the formula:

[0026] , calculate the thermal switching ratio between the single-domain and double-domain BFO molecular models corresponding to the current preset temperature ;

[0027] Among them, represents the thermal conductivity of the single-domain BFO molecular model, represents the thermal conductivity of the double-domain BFO molecular model.

[0028] In one implementation manner of the present application, a single-domain BFO molecular model with a preset number of domain walls is obtained. At a specific temperature, isothermal-isobaric ensemble and canonical ensemble simulations are performed on the BFO molecular model with the preset number of domain walls to obtain the thermal switching ratio between each preset number of domain walls and the single-domain BFO molecular model, specifically including:

[0029] Perform HNEMD simulation calculations at a specific temperature;

[0030] Release the stress in the BFO molecular model using the isothermal-isobaric ensemble;

[0031] Use the canonical ensemble to perform molecular dynamics simulations, with the driving factor along the y direction, and output the thermal conductivity of the single-domain BFO molecular model with the preset number of domain walls every preset step;

[0032] Calculate the thermal switching ratio between the single-domain BFO molecular model and the BFO molecular model with the preset number of domain walls corresponding to the specific temperature by dividing the thermal conductivity of the single-domain BFO molecular model by the thermal conductivity of the BFO molecular model with the preset number of domain walls.

[0033] In a second aspect, the present application provides a simulation system for regulating the thermal conductivity of ferroelectric domains in a bismuth ferrite thin film. The system includes:

[0034] A training set acquisition module for establishing BFO molecular models with different supercell structures, and then using the AIMD technology to perform canonical ensemble and isothermal-isobaric ensemble simulations on the BFO molecular models respectively to obtain the phase space data of BFO at different preset temperatures as the initial training set;

[0035] A function acquisition module for configuring the hyperparameters of the GPUMD software, using the initial training set to train the neural evolution potential NEP using the NEP tool embedded in the GPUMD software until the initial training set can describe all preset target MD scenarios, and then obtaining the NEP potential function;

[0036] A temperature acquisition module for obtaining single-domain and double-domain BFO molecular models, simulating the thermal conductivity properties of the single-domain and double-domain BFO molecular models in the GPUMD software based on the NEP potential function within a preset temperature range, obtaining the thermal conductivity of the single-domain state and the thermal conductivity of the double-domain state at each preset temperature, calculating the thermal switching ratio between the single-domain and double-domain BFO molecular models at each preset temperature; then obtaining the change curve of the thermal switching ratio with temperature change, obtaining the theoretical maximum value of the thermal switching ratio with temperature change, and obtaining the specific temperature corresponding to the theoretical maximum value;

[0037] The domain wall number acquisition module is used to obtain a single-domain BFO molecular model with a preset number of domain walls. At a specific temperature, perform isothermal-isobaric ensemble and canonical ensemble simulations on the BFO molecular model with the preset number of domain walls to obtain the thermal switching ratio between each preset number of domain walls and the single-domain BFO molecular model; obtain the change curve of the thermal switching ratio of the BFO molecular model with respect to the change in the preset number of domain walls, and obtain the specific number of domain walls corresponding to the theoretical maximum value.

[0038] The guidance module is used to use the specific temperature and the specific number of domain walls corresponding to the theoretical maximum value as the final guidance conditions.

[0039] In an implementation manner of the present application, the training set acquisition module includes a training set acquisition unit.

[0040] It is used to construct a 1×1×1 BFO unit cell, use the phonopy software to obtain supercells with different numbers of atoms, and perform AIMD simulations at different preset temperatures in the first-principles calculation program respectively to obtain an initial training set; where the preset temperatures at least include: 100K, 200K, 300K, 400K, 500K, and 600K.

[0041] Use the read function of ASE to read the BFO model.xyz file of the 3×3×3 supercell, store the read atomic structure data in the result file, use the write function of ASE to save the atomic structure data in the result file in a format that can be called by the first-principles program, use a normal distribution to randomly generate the perturbation amount of the atomic positions, and call the command to generate the structure file of the System object to save each perturbed system as a new crystal structure and add it to the initial training set.

[0042] In an implementation manner of the present application, the temperature acquisition module includes a temperature acquisition unit.

[0043] It is used to perform simulations using HNEMD, obtain a number of preset temperatures according to the preset temperature interval, and then obtain the thermal conductivities of the single-domain and double-domain BFO molecular models corresponding to each preset temperature.

[0044] Through the formula:

[0045] , calculate the thermal switching ratio between the single-domain and double-domain BFO molecular models corresponding to the current preset temperature ;

[0046] Among them, represents the thermal conductivity of the single-domain BFO molecular model, represents the thermal conductivity of the double-domain BFO molecular model.

[0047] In an implementation manner of the present application, the domain wall number acquisition module includes a domain wall number acquisition unit,

[0048] which is used to perform HNEMD simulation calculations at a specific temperature;

[0049] The stress in the BFO molecular model is released using the isothermal-isobaric ensemble;

[0050] Molecular dynamics simulations are performed using the canonical ensemble, with the driving factor along the y-direction, and the thermal conductivity of the BFO molecular model with a single domain and a preset number of domain walls is output every preset step.

[0051] The thermal switching ratio between the BFO molecular models with a single domain and a preset number of domain walls corresponding to the specific temperature is calculated by comparing the thermal conductivity of the single-domain BFO molecular model with that of the BFO molecular model with a preset number of domain walls.

[0052] In a third aspect, the present application provides a non-volatile computer storage medium, on which computer instructions are stored, and when the computer instructions are executed, a simulation method for regulating the thermal conductivity of ferroelectric domains in a bismuth ferrite thin film as described in any one of the above is realized.

[0053] From the above technical solutions, it can be seen that the present application has the following advantages:

[0054] 1. The structure-property relationship between the atomic configuration of ferroelectric domain walls and heat transport is accurately analyzed:

[0055] By combining AIMD (ab initio molecular dynamics) and the fourth-generation NEP (neural network potential), high-precision modeling of the atomic-level configuration of ferroelectric domain walls in BFO thin films (BFO molecular models) is achieved, overcoming the problem of insufficient accuracy of traditional empirical potential functions in describing complex polarization-lattice coupling effects, and for the first time quantitatively revealing the non-linear correlation law between the number and distribution of domain walls and thermal conductivity.

[0056] 2. The problem of simulating thermal activation migration of multi-physical field coupling is efficiently solved:

[0057] Using the NEP tool embedded in GPUMD, multi-field coupling dynamic simulations of ferroelectric domain polarization flipping (electric field response), lattice vibration (heat transport), and stress relaxation (mechanical deformation) are realized under a unified framework. Compared with the traditional finite element method, the calculation efficiency is improved, and the transient heat flux fluctuation characteristics at the sub-picosecond level during the domain wall migration process can be captured.

[0058] 3. The directional design of high-performance thermal switch materials is guided:

[0059] By establishing the theoretical variation curves of the thermal switch ratio with temperature and the number of domain walls, the optimal operating temperature range (such as the specific temperature corresponding to the theoretical maximum value) for regulating the thermal conductivity of the BFO molecular model and the critical domain wall density threshold (such as the specific number of domain walls) are clarified, providing a quantifiable design criterion for the material doping and domain engineering strategies of low-power thermal management devices.

[0060] 4. Break through the computational bottleneck of the traditional trial-and-error mode:

[0061] By adopting a method combining the adaptive generation of supercell structures and the iterative training of the NEP potential function, the computational resource consumption of the all-atom MD simulation is reduced, and at the same time, the prediction of cross-scale heat transport behaviors covering from single-domain homogeneous phases to multi-domain heterogeneous phases is achieved, laying a foundation for the high-throughput virtual screening of complex ferroelectric materials. Brief Description of the Drawings

[0062] To more clearly illustrate the technical solutions of the present invention, the drawings required for description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0063] Figure 1 It is a flowchart of a simulation method for regulating the thermal conductivity of ferroelectric domains in a bismuth ferrite thin film provided by an embodiment of the present application.

[0064] Figure 2 It is a schematic diagram of the internal structure of a simulation system for regulating the thermal conductivity of ferroelectric domains in a bismuth ferrite thin film provided by an embodiment of the present application.

[0065] Figure 3 It is a schematic diagram of the 180° domain structure of a BFO molecular model provided by an embodiment of the present application.

[0066] Figure 4 It is a schematic diagram showing the variation of the thermal switch ratio of the BFO molecular model with temperature provided by an embodiment of the present application.

[0067] Figure 5 It is a schematic diagram showing the variation of the thermal switch ratio of the BFO molecular model with the number of domain walls at 300K provided by an embodiment of the present application. Detailed Description of the Embodiments

[0068] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, rather than all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0069] Those skilled in the art should understand that the embodiments described below are only the preferred embodiments of the present disclosure, and do not mean that the present disclosure can only be implemented through these preferred embodiments. These preferred embodiments are only used to explain the technical principles of the present disclosure, rather than to limit the protection scope of the present disclosure. Based on the preferred embodiments provided by the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative efforts should still fall within the protection scope of the present disclosure.

[0070] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device comprising a series of elements not only includes those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. Without further limitation, the element defined by the statement "comprising one..." does not exclude the existence of another identical element in the process, method, commodity or device comprising the element.

[0071] The technical solutions proposed in the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0072] The embodiment provides a simulation method for regulating the thermal conductivity of ferroelectric domains in bismuth ferrite thin films, as Figure 1 shown, the method provided by the embodiment of the present application mainly includes the following steps:

[0073] Step 110: Establish BFO molecular models with different supercell structures, and then use the AIMD technology to perform canonical ensemble and isothermal-isobaric ensemble simulations on the BFO molecular models respectively to obtain the phase space data of BFO at different preset temperatures as the initial training set.

[0074] This step can be specifically:

[0075] Construct a 1×1×1 BFO unit cell, use the phonopy software to obtain supercells with different numbers of atoms, and perform AIMD simulations at different preset temperatures in the first-principles calculation program to obtain the initial training set; where the preset temperatures include at least: 100K, 200K, 300K, 400K, 500K and 600K;

[0076] Use the read function of ASE to read the BFO model.xyz file of a 3×3×3 supercell, store the read atomic structure data in the result file, use the write function of ASE to save the atomic structure data in the result file in a format that can be called by the ab initio program. Use a normal distribution to randomly generate the perturbation amount of the atomic positions, call the command to generate the structure file of the System object to save each perturbed system as a new crystal structure, and add it to the initial training set.

[0077] Further exemplarily, construct the BFO unit cell, the lattice constants of the equivalent pseudo-cubic unit cell are a = b = c = 3.7893 (obtained by converting the rhombohedral structure to a pseudo-cubic approximation), use the phonopy software package to obtain supercells with different numbers of atoms, perform AIMD simulations at different temperatures in the ab initio calculation program package respectively to obtain the initial training set. Among them, the AIMD simulation process is as follows: to traverse the possible phase space configurations of the BFO model (supercell), run the canonical ensemble (NVT) at 100K, 200K, 300K, 400K, 500K, and 600K respectively, use the Andersen thermostat to run for 10 ps, take a structure every 0.2 ps, and a total of 300 structures are obtained for the initial training set. Use the read function of ASE (Atomic Simulation Environment) to read the BFO model.xyz file of a 3×3×3 supercell, store the read atomic structure data in the result file, use the write function of ASE to save the atomic structure data in the result file in a format that can be called by the ab initio program package, then call the perturb method of the System in the dpdata library to perturb the structure, specify the number of perturbations pert_num = 3, the perturbation amplitude of the lattice parameter cell_pert_fraction = 0.01 - 0.05, and the maximum distance of the atomic position perturbation is 0.1 - 0.5 , atom_pert_style = "normal", that is, use a normal distribution to randomly generate the perturbation amount of the atomic positions, then call the command to generate the structure file of the System object to save each perturbed system as a new crystal structure file, and a total of 100 structures are obtained and added to the initial training set.

[0078] Step 120: Configure the hyperparameters of the GPUMD software, use the NEP tool embedded in the GPUMD software with the initial training set to train the neural evolution potential NEP until the initial training set can describe all preset target MD scenarios, and then obtain the NEP potential function.

[0079] It should be noted that the process of configuring the hyperparameters of the GPUMD software can be specifically as follows:

[0080] Select appropriate hyperparameters in the GPUMD (Graphics Processing Units Molecular Dynamics) software. The cut-off radial and angular cut-off values are selected as 8 and 6, the number of neurons is selected as 50, and the others are default values. Use the NEP tool embedded in GPUMD to train the fourth-generation NEP (neuroevolution potential), and perform iteration and verification. That is, perform molecular dynamics simulation in GPUMD, use the farthest point sampling method to compare the GPUMD output results with the training set, select the structures with too large errors from the MD results compared with the training set data to perform density functional theory (DFT) single-point energy calculations and add them to the training set, and retrain NEP until the training set can effectively describe all target MD scenarios. That is, the training set can cover the thermodynamic / kinetic key regions of the target scenario, and evaluate the reliability of NEP through cross-validation, error analysis of energy and force, comparison of thermodynamic properties, etc. In addition, run long-time MD simulations to check its stability and physical rationality. Finally, obtain the NEP potential function that can accurately describe the interatomic interaction forces of BFO.

[0081] Among them, using the NEP tool embedded in the GPUMD software with the initial training set to train the neuroevolution potential NEP until the initial training set can describe all preset target MD scenarios, and then obtain the NEP potential function, which can be specifically as follows:

[0082] S0. Perform single-point energy calculations on the structures in the initial training set in the first-principles program to obtain energy and virial information, and then obtain the converged first training data, and then obtain the first converged training set;

[0083] S1. Use the first converged training set to train the neuroevolution potential NEP to obtain the 1-NEP function;

[0084] S2. Use the supercell atomic BFO molecular model, the 1-NEP function, and the preset condition information to obtain the phase space existing between 100K and 700K;

[0085] S3. Use the farthest point sampling to select the BFO molecular model that meets the preset limit range from the phase space;

[0086] S4. Perform single-point energy calculations on the BFO molecular model that meets the preset limit range in the first-principles program to obtain energy and virial information, and then obtain the converged second training data, and then add the second training data to the first converged training set to obtain the second converged training set;

[0087] S5. Train the neural evolution potential NEP using the second convergence training set to obtain the 2-NEP function;

[0088] S6. Replace the 2-NEP function with the 1-NEP function, repeat S2 - S5 until the preset stability condition is met to obtain the final NEP potential function.

[0089] It should be noted that in S0, in order to ensure accurate energy and virial information in the training set during NEP training, single-point energy calculations are performed on the data in the initial training set in the first-principles program. To ensure that the energies and forces of all structures converge to a unified value, the cut-off energy ENCUT is selected as 600 , the k-point is selected using KSPACING = 0.2, and the convergence criterion for the force is below for each atom , and the convergence criterion for the energy is . After the calculation, use a bash script to check whether each structure converges, that is, use the find command to search for all result files, and judge whether the structure converges according to NSW, NELM, the actual number of iteration steps, and whether "aborting loop because EDIFF is reached" appears. Use wc -l to count the number of the converged result files, and N_count records the serial number. Then for each converged result file, syst_numb_atom extracts the number of atoms, latt extracts the lattice vector information, ener extracts the total energy, viri_logi extracts the virial information, ion_numb_arra and ion_symb_arra respectively extract the number and symbol of each type of atom, and then use the paste command to combine all the information and write it into the output file to extract the converged structures to obtain the first training data, and discard the non-converged ones.

[0090] S1 can be: Select appropriate hyperparameters in the GPUMD software package to train the NEP-4 potential function with the first training data. To ensure the accuracy of the potential function, cut off = 8 6 and the number of neurons neuron = 50 can be selected, and the others are default values. After the training, the 1-NEP function is obtained. The sign that the training tends to converge is that the energy, force, and stress information of the system tend to be flat as the number of training steps increases.

[0091] The process of S2 - S5 can be: Use the 1-NEP function to perform molecular dynamics simulations in GPUMD to explore The possible phase space of BFO with 135 supercell atoms at 100K - 700K. First, perform the canonical ensemble (NVT) with a time step of 1 fs for a total of 500 ps, and output a structure every 1 ps. Then, perform the isothermal - isobaric ensemble (NPT) with a time step of 1 fs for a total of 0.5 ns, and output a structure every 10 ps. After obtaining the output file dump.xyz, use farthest - point sampling to select structures with a limit of min_distance = 0.02. Calculate the DFT single - point energy for the selected structures using the same parameter file, add the calculated structures (the second training data) to the first convergent training set to obtain the second convergent training set, and retrain NEP - 4 to obtain the 2 - NEP function.

[0092] The process of S6 can be: repeat the iterative operations of S2 - S5 until The BFO in the supercell runs stably for 20 ns in the isothermal - isobaric ensemble NPT, and the training set can completely describe the motion trajectory of BFO within 20 ns at the same temperature. At this time, the 2 - NEP function is the final version, denoted as the NEP potential function.

[0093] Step 130: Obtain the single - domain and double - domain BFO molecular models. Based on the NEP potential function, simulate the thermal conductivity of the single - domain and double - domain BFO molecular models in the GPUMD software within a preset temperature range to obtain the thermal conductivity of the single - domain state and the double - domain state at each preset temperature, and calculate the thermal switch ratio between the single - domain and double - domain BFO molecular models at each preset temperature; further obtain the change curve of the thermal switch ratio with temperature change, obtain the theoretical maximum value of the thermal switch ratio with temperature change, and obtain the specific temperature corresponding to the theoretical maximum value.

[0094] Among them, the process of calculating the thermal switch ratio between the single - domain and double - domain BFO molecular models at each preset temperature can be:

[0095] Use HNEMD for simulation, obtain several preset temperatures at a preset temperature interval, and then obtain the thermal conductivity of the single - domain and double - domain BFO molecular models corresponding to each preset temperature;

[0096] Through the formula: , calculate the thermal switch ratio between the single - domain and double - domain BFO molecular models corresponding to the current preset temperature ; where, represents the thermal conductivity of the single - domain BFO molecular model, represents the thermal conductivity of the double - domain BFO molecular model.

[0097] More specifically, this step can be specifically:

[0098] Construct BFO molecular models with different domain structures using Materials Studio: the 180° single-domain structure (0DW) of BFO and the 180° double-domain structure (1DW) containing a parallel domain wall. The model with one domain wall is shown as Figure 3 shown. Expand the unit cells of the two structures along the x, y, and z axes , with a total of 147,000 atoms.

[0099] Use HNEMD for simulation to calculate the thermal conductivities of the two domain structures between 200 K and 500 K, with an interval of 50 K. In the editing of the GPUMD input file run.in, first initialize the velocities of all atoms in the lattice. For example, at 200 K, set the velocity to 200. Then use the isothermal-isobaric ensemble NPT to release the stress in the structure and fully relax it until the structure converges. Then use the canonical ensemble NVT to perform a 20-ns molecular dynamics simulation. The driving factor is along the y direction, and the average result of the thermal conductivity is output every 1000 steps. To ensure the accuracy of the data and reduce errors, each HNEMD simulation is independently performed 5 times and averaged.

[0100] The formula for calculating the thermal conductivity k of the BFO molecular model using HNEMD is , where represents the heat flux density in the direction in the non-equilibrium state, is the ensemble average within the simulation time , represents the thermal conductivity tensor, represents the temperature of the system, represents the volume of the system, and the non-equilibrium heat flux is caused by the external driving factor . Use the data output by GPUMD in Python for fitting, and finally obtain the variation of the thermal conductivities of the two BFO molecular model structures with temperature. Use the obtained thermal conductivity data in Origin to fit the curve of the thermal switching ratio with temperature. As shown in Figure 4 is the variation of the BFO molecular model thermal switch with temperature. Finally, it is calculated that at 300 K, the maximum thermal switching ratio of the 180° domain wall BFO molecular model is .

[0101] Step 140: Obtain BFO molecular models with single domains and a preset number of domain walls. At a specific temperature, perform isothermal-isobaric ensemble and canonical ensemble simulations on the BFO molecular models with the preset number of domain walls to obtain the thermal switching ratios between each preset number of domain walls and the single-domain BFO molecular model; obtain the curve of the thermal switching ratio of the BFO molecular model varying with the preset number of domain walls, and obtain the specific number of domain walls corresponding to the theoretical maximum value. ​​​​​​​

[0102] In some embodiments, the calculation process of obtaining the thermal switching ratio between the number of each preset domain wall and the single-domain BFO molecular model may be as follows:

[0103] Perform HNEMD simulation calculations at a specific temperature;

[0104] Use the isothermal-isobaric ensemble to release the stress in the BFO molecular model;

[0105] Use the canonical ensemble to perform molecular dynamics simulations. The driving factor is along the y-direction, and the thermal conductivity of the BFO molecular model with a single domain and the number of preset domain walls is output once every preset number of steps.

[0106] Calculate the thermal switching ratio between the BFO molecular model with a single domain and the BFO molecular model with the number of preset domain walls corresponding to the specific temperature by dividing the thermal conductivity of the single-domain BFO molecular model by the thermal conductivity of the BFO molecular model with the number of preset domain walls.

[0107] More specifically, this step may be specifically as follows:

[0108] Construct BFO molecular model structures with different numbers of domain walls in Materials Studio, with an angle of 180°. The five numbers of domain walls are 1, 3, 5, 7, and 9, for a total of five structures.

[0109] Perform HNEMD simulation calculations at 300K (assuming 300K is the specific temperature). In the editing of the GPUMD input file run.in, initialize the velocity of all atoms in the lattice to 300. First, use the isothermal-isobaric ensemble NPT to release the stress in the structure and fully relax it. The structure will converge to a certain size. Then use the canonical ensemble NVT to perform 20 ns of molecular dynamics simulations, and the driving factor is along the y-direction, and the result of the averaged thermal conductivity is output once every 1000 steps. To ensure the accuracy of the data and reduce errors, each HNEMD is independently simulated 5 times and averaged.

[0110] In python, use the data output by GPUMD for fitting to obtain the variation of the thermal conductivity of the BFO molecular model domain structure with the number of domain walls. As Figure 5 shown in the variation of the BFO molecular model thermal switch with the domain wall, at 300K, when the number of 180° domain walls is 9, the BFO molecular model has the maximum thermal switching ratio .

[0111] Step 150: Use the specific temperature and the specific number of domain walls corresponding to the theoretical maximum value as the final guiding conditions.

[0112] More specifically, this step may be specifically as follows:

[0113] Obtain the theoretical conditions for the maximum thermal switching ratio of the BFO molecular model. For example, at 300 K (assuming 300 K is the specific temperature), when the number of 180° domain walls is 9, the thermal switching ratio of the BFO molecular model reaches the maximum, λ = 1.52.

[0114] In addition, this application Figure 2 is a simulation system for regulating the thermal conductivity of ferroelectric domains in a bismuth ferrite thin film provided by an embodiment of this application. As Figure 2 shown, the system provided by the embodiment of this application mainly includes:

[0115] A training set acquisition module 210, which is used to establish BFO molecular models with different supercell structures, and then use the AIMD technique to perform canonical ensemble and isothermal-isobaric ensemble simulations on the BFO molecular models respectively, so as to obtain the phase space data of BFO at different preset temperatures as the initial training set.

[0116] The training set acquisition module 210 includes a training set acquisition unit,

[0117] which is used to construct a 1×1×1 BFO unit cell, obtain supercells with different numbers of atoms by using the phonopy software, perform AIMD simulations at different preset temperatures in the first-principles calculation program respectively, and obtain the initial training set; where the preset temperatures at least include: 100 K, 200 K, 300 K, 400 K, 500 K, and 600 K;

[0118] Use the read function of ASE to read the BFO model.xyz file of the 3×3×3 supercell, store the read atomic structure data in the result file, use the write function of ASE to save the atomic structure data in the result file in a format that can be called by the first-principles program, use a normal distribution to randomly generate the perturbation amount of the atomic positions, and call the command for generating the structure file of the System object to save each perturbed system as a new crystal structure and add it to the initial training set.

[0119] A function acquisition module 220, which is used to configure the hyperparameters of the GPUMD software, use the initial training set to train the neural network evolved potential NEP by using the NEP tool embedded in the GPUMD software until the initial training set can describe all preset target MD scenarios, and then obtain the NEP potential function.

[0120] A temperature acquisition module 230 is configured to acquire single-domain and double-domain BFO molecular models, simulate the heat conduction properties of the single-domain and double-domain BFO molecular models in the GPUMD software based on the NEP potential function within a preset temperature range, obtain the thermal conductivities of the single-domain state and the double-domain state at each preset temperature, and calculate the heat switching ratio between the single-domain and double-domain BFO molecular models at each preset temperature; furthermore, obtain the variation curve of the heat switching ratio with temperature change, obtain the theoretical maximum value of the heat switching ratio with respect to temperature change, and obtain the specific temperature corresponding to the theoretical maximum value.

[0121] The temperature acquisition module 230 includes a temperature acquisition unit.

[0122] It is used to perform simulations using HNEMD, obtain a number of preset temperatures at a preset temperature interval, and further obtain the thermal conductivities of the single-domain and double-domain BFO molecular models corresponding to each preset temperature.

[0123] Through the formula:

[0124] , calculate the heat switching ratio between the single-domain and double-domain BFO molecular models corresponding to the current preset temperature. ;

[0125] Wherein, represents the thermal conductivity of the single-domain BFO molecular model, represents the thermal conductivity of the double-domain BFO molecular model.

[0126] A domain wall number acquisition module 240 is configured to acquire single-domain and BFO molecular models with a preset number of domain walls, perform isothermal-isobaric ensemble and canonical ensemble simulations on the BFO molecular models with the preset number of domain walls at a specific temperature, obtain the heat switching ratios between each preset number of domain walls and the single-domain BFO molecular model; obtain the variation curve of the heat switching ratio of the BFO molecular model with respect to the preset number of domain walls, and obtain the specific number of domain walls corresponding to the theoretical maximum value.

[0127] The domain wall number acquisition module 240 includes a domain wall number acquisition unit.

[0128] It is used to perform HNEMD simulation calculations at a specific temperature;

[0129] Use the isothermal-isobaric ensemble to release the stress in the BFO molecular model;

[0130] Use the canonical ensemble to perform molecular dynamics simulations, with the driving factor along the y direction, and output the thermal conductivities of the single-domain and BFO molecular models with the preset number of domain walls once every preset step.

[0131] Calculate the thermal switching ratio between the single-domain BFO molecular model and the BFO molecular model with a preset number of domain walls at a specific temperature by comparing the thermal conductivities of the two models.

[0132] The guidance module 250 is configured to use the specific temperature and the specific number of domain walls corresponding to the theoretical maximum value as the final guidance conditions.

[0133] In addition, an embodiment of the present application further provides a non-volatile computer storage medium, on which executable instructions are stored. When the executable instructions are executed, a simulation method for regulating the thermal conductivity of ferroelectric domains in a bismuth ferrite thin film as described above is implemented.

[0134] The foregoing description of the disclosed embodiments enables those skilled in the art to practice or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Thus, the invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A simulation method for regulating the thermal conductivity of ferroelectric domains in bismuth ferrite thin films, characterized in that, The method includes: Establishing BFO molecular models with different supercell structures, and then using the AIMD technique to perform canonical ensemble and isothermal-isobaric ensemble simulations on the BFO molecular models respectively to obtain the phase space data of BFO at different preset temperatures as the initial training set; specifically including: Constructing a 1×1×1 BFO unit cell, using the phonopy software to obtain supercells with different numbers of atoms, performing AIMD simulations at different preset temperatures in the first-principles calculation program respectively to obtain the initial training set; where the preset temperatures at least include: 100K, 200K, 300K, 400K, 500K and 600K; using the read function of ASE to read the BFOmodel.xyz file of the 3×3×3 supercell, storing the read atomic structure data in the result file, using the write function of ASE to save the atomic structure data in the result file in a format that can be called by the first-principles program, using a normal distribution to randomly generate the perturbation amount of the atomic positions, calling the command of the System object to generate a structure file to save each perturbed system as a new crystal structure and adding it to the initial training set; Configuring the hyperparameters of the GPUMD software, and using the initial training set to train the neural evolution potential NEP with the NEP tool embedded in the GPUMD software until the initial training set can describe all preset target MD scenarios, and then obtaining the NEP potential function; Obtaining single-domain and double-domain BFO molecular models, based on the NEP potential function, simulating the thermal conductivity properties of the single-domain and double-domain BFO molecular models in the GPUMD software within a preset temperature range, obtaining the thermal conductivity of the single-domain state and the thermal conductivity of the double-domain state at each preset temperature, and calculating the thermal switching ratio between the single-domain and double-domain BFO molecular models at each preset temperature; then obtaining the change curve of the thermal switching ratio with temperature change, obtaining the theoretical maximum value of the thermal switching ratio with temperature change, and obtaining the specific temperature corresponding to the theoretical maximum value; Obtaining single-domain and BFO molecular models with a preset number of domain walls, performing isothermal-isobaric ensemble and canonical ensemble simulations on the BFO molecular models with the preset number of domain walls at a specific temperature, obtaining the thermal switching ratio between each preset number of domain walls and the single-domain BFO molecular model; obtaining the change curve of the thermal switching ratio of the BFO molecular model with the preset number of domain walls, and obtaining the specific number of domain walls corresponding to the theoretical maximum value; Taking the specific temperature and the specific number of domain walls corresponding to the theoretical maximum value as the final guiding conditions.

2. The method for simulating the regulation of the thermal conductivity of ferroelectric domains in a bismuth ferrite thin film according to claim 1, wherein Using the initial training set to train the neural evolution potential NEP with the NEP tool embedded in the GPUMD software until the initial training set can describe all preset target MD scenarios, and then obtaining the NEP potential function, specifically including: S0. Performing single-point energy calculations on the structures in the initial training set in the first-principles program to obtain energy and virial information, and then obtaining the converged first training data and further obtaining the first converged training set; S1. Training the neural evolution potential NEP with the first converged training set to obtain the 1-NEP function; S2. Use the supercell atomic BFO molecular model, the 1-NEP function, and the preset condition information to obtain the phase space existing between 100K and 700K; S3. Using farthest point sampling to select BFO molecular models that meet the preset limit range from the phase space; S4. Perform single-point energy calculations on the BFO molecular models that meet the preset limit range in the first-principles program, obtain energy and virial information, and then obtain the converged second training data. Furthermore, add the second training data to the first converged training set to obtain the second converged training set; S5. Use the second converged training set to train the neural evolution potential NEP to obtain the 2-NEP function; S6. Replace the 2-NEP function with the 1-NEP function, and repeat S2 - S5 until the preset stability condition is met to obtain the final NEP potential function.

3. The method for simulating the regulation of the thermal conductivity of ferroelectric domains in a bismuth ferrite thin film according to claim 1, wherein Obtain single-domain and double-domain BFO molecular models. Based on the NEP potential function, simulate the thermal conductivity properties of the single-domain and double-domain BFO molecular models in the GPUMD software within a preset temperature range to obtain the thermal conductivities of the single-domain state and the double-domain state at each preset temperature, and calculate the thermal switching ratio between the single-domain and double-domain BFO molecular models at each preset temperature. Specifically include: Use HNEMD for simulation, obtain a number of preset temperatures at a preset temperature interval, and then obtain the thermal conductivities of the single-domain and double-domain BFO molecular models corresponding to each preset temperature; Through the formula: , calculate the thermal switching ratio between the single-domain and double-domain BFO molecular models corresponding to the current preset temperature ; Among them, represents the thermal conductivity of the single-domain BFO molecular model, represents the thermal conductivity of the double-domain BFO molecular model.

4. The method for simulating the regulation of thermal conductivity by ferroelectric domains in a bismuth ferrite thin film according to claim 1, characterized in that, Obtain single-domain BFO molecular models with a preset number of domain walls. At a specific temperature, perform isothermal-isobaric ensemble and canonical ensemble simulations on the BFO molecular models with the preset number of domain walls to obtain the thermal switching ratio between each preset number of domain walls and the single-domain BFO molecular model. Specifically include: Perform HNEMD simulation calculations at a specific temperature; Release the stress in the BFO molecular model using the isothermal-isobaric ensemble; Use the canonical ensemble for molecular dynamics simulation, with the driving factor along the y-direction, and output the thermal conductivities of the single-domain and BFO molecular models with the preset number of domain walls once every preset step; Calculate the thermal switching ratio between the single-domain and BFO molecular models with the preset number of domain walls corresponding to the specific temperature by dividing the thermal conductivity of the single-domain BFO molecular model by the thermal conductivity of the BFO molecular model with the preset number of domain walls.

5. A simulation system for regulating the thermal conductivity of ferroelectric domains in a bismuth ferrite thin film, characterized in that, The system includes: A training set acquisition module for establishing BFO molecular models with different supercell structures, and then using the AIMD technology to perform canonical ensemble and isothermal-isobaric ensemble simulations on the BFO molecular models respectively to obtain the phase space data of BFO at different preset temperatures as the initial training set; The training set acquisition module includes a training set acquisition unit, To construct a 1×1×1 BFO unit cell, use the phonopy software to obtain supercells with different numbers of atoms, and perform AIMD simulations at different preset temperatures in a first-principles calculation program to obtain an initial training set. Among them, the preset temperatures include at least: 100K, 200K, 300K, 400K, 500K, and 600K. Use the read function of ASE to read the BFOmodel.xyz file of the 3×3×3 supercell, store the read atomic structure data in the result file, use the write function of ASE to save the atomic structure data in the result file in a format that can be called by the first-principles program, use a normal distribution to randomly generate the perturbation amount of the atomic positions, and call the command to generate the structure file of the System object to save each perturbed system as a new crystal structure and add it to the initial training set. The function acquisition module is used to configure the hyperparameters of the GPUMD software, and use the NEP tool embedded in the GPUMD software with the initial training set to train the neural evolution potential NEP until the initial training set can describe all preset target MD scenarios, and then obtain the NEP potential function. The temperature acquisition module is used to obtain the single-domain and double-domain BFO molecular models, and based on the NEP potential function, simulate the thermal conductivity of the single-domain and double-domain BFO molecular models in the GPUMD software within the preset temperature range to obtain the thermal conductivity of the single-domain state and the double-domain state at each preset temperature, and calculate the thermal switching ratio between the single-domain and double-domain BFO molecular models at each preset temperature. Furthermore, obtain the change curve of the thermal switching ratio with temperature change, obtain the theoretical maximum value of the thermal switching ratio with temperature change, and obtain the specific temperature corresponding to the theoretical maximum value. The domain wall number acquisition module is used to obtain the single-domain and BFO molecular models with a preset number of domain walls, and perform isothermal-isobaric ensemble and canonical ensemble simulations on the BFO molecular models with a preset number of domain walls at a specific temperature to obtain the thermal switching ratio between each preset number of domain walls and the single-domain BFO molecular model. Obtain the change curve of the thermal switching ratio of the BFO molecular model with the preset number of domain walls, and obtain the specific number of domain walls corresponding to the theoretical maximum value. The guidance module is used to use the specific temperature and the specific number of domain walls corresponding to the theoretical maximum value as the final guidance conditions.

6. The simulation system for regulating the thermal conductivity of ferroelectric domains in the bismuth ferrite thin film according to claim 5, characterized in that, The temperature acquisition module includes a temperature acquisition unit. It is used to perform simulations using HNEMD, obtain a number of preset temperatures at a preset temperature interval, and then obtain the thermal conductivity of the single-domain and double-domain BFO molecular models corresponding to each preset temperature. Through the formula: , calculate the thermal switching ratio between the single-domain and double-domain BFO molecular models corresponding to the current preset temperature ; Among them, represents the thermal conductivity of the single-domain BFO molecular model, represents the thermal conductivity of the double-domain BFO molecular model.

7. The simulation system for regulating the thermal conductivity of ferroelectric domains in the bismuth ferrite thin film according to claim 5, characterized in that, The domain wall number acquisition module includes a domain wall number acquisition unit. It is used to perform HNEMD simulation calculations at a specific temperature. Use the isothermal-isobaric ensemble to release the stress in the BFO molecular model. Use the canonical ensemble to perform molecular dynamics simulations, with the driving factor along the y direction, and output the thermal conductivity of the single-domain and BFO molecular models with a preset number of domain walls once every preset step. Calculate the thermal switching ratio between the single-domain BFO molecular model and the BFO molecular model with a preset number of domain walls at a specific temperature by comparing the thermal conductivities of the single-domain BFO molecular model and the BFO molecular model with a preset number of domain walls.

8. A non-volatile computer storage medium, characterized in that, It stores computer instructions, and when the computer instructions are executed, they implement a simulation method for regulating the thermal conductivity of ferroelectric domains in a bismuth ferrite thin film as described in any one of claims 1-4.

Images