Simulation method and related apparatus based on machine learning and molecular dynamics

Abstract

The application provides a simulation method based on machine learning and molecular dynamics and a related device, applied to the field of semiconductor manufacturing, after receiving a process menu; for each process condition in the process menu, inputting the process condition into a force field prediction model to output a machine learning force field under the process condition; finally, performing molecular dynamics simulation according to the machine learning force field under the process condition to obtain a simulation result under the process condition. Through the machine learning method, high-precision first-principle calculation data is learned, a machine learning force field is generated, the interaction of silicon carbide occurring in the ion implantation process is simulated, the ion implantation process is simulated, high-precision and high-efficiency simulation calculation is realized, the influence of different process conditions on device design process is comprehensively considered, a specific process menu is developed, and an optimization scheme is provided for silicon carbide ion implantation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of semiconductor manufacturing, and in particular, to a simulation method based on machine learning and molecular dynamics and a related device. BACKGROUND

[0002] At present, in the field of semiconductor device process simulation, traditional simulation methods such as SRIM use the Monte Carlo method, which mainly relies on existing experience or experimental data, and has poor accuracy and low efficiency in predicting experimental results and studying process mechanisms. SUMMARY

[0003] Therefore, the present application provides a simulation method based on machine learning and molecular dynamics and a related device, which realizes high-precision and high-efficiency simulation calculation.

[0004] The first aspect of the present application provides a simulation method based on machine learning and molecular dynamics, comprising:

[0005] receiving a process menu; wherein the process menu includes at least one process condition; the process condition is divided into implantation parameters and annealing parameters; the implantation parameters include the energy, dose, angle, temperature and crystal direction of implanted silicon carbide ions; the annealing parameters include annealing temperature, annealing time and temperature control rate;

[0006] For each process condition, input the process condition into a force field prediction model to output a machine learning force field under the process condition; wherein the force field prediction model is obtained by training a machine learning model with a training sample set; the training sample set includes at least one training sample;

[0007] According to the machine learning force field under the process condition, molecular dynamics simulation is performed to obtain the simulation result under the process condition.

[0008] Optionally, the construction method of the force field prediction model comprises:

[0009] constructing a training sample set; wherein the training sample set includes at least one training sample, and the training sample includes physical information of a training sample configuration;

[0010] For each training sample, the atomic structure of the training sample configuration is converted using the descriptor of the machine learning model to obtain training sample features;

[0011] Based on the training sample features, the machine learning model is trained to obtain the force field prediction model.

[0012] Optionally, the construction of the training sample set comprises:

[0013] generating an initial force field by a force field large model; wherein the initial force field comprises a plurality of initial configurations;

[0014] performing active learning by using the initial force field to obtain a plurality of target configurations;

[0015] performing first-principle calculation on each training sample configuration to obtain physical information of the training sample configuration; wherein the training sample configuration comprises the plurality of initial configurations and the plurality of target configurations.

[0016] Optionally, the physical information of the training sample configuration comprises interatomic forces and energy of the training sample configuration.

[0017] The second aspect of the present application provides a simulation device based on machine learning and molecular dynamics, comprising:

[0018] a receiving unit configured to receive a process menu; wherein the process menu comprises at least one process condition; the process condition is divided into implantation parameters and annealing parameters; the implantation parameters comprise energy, dose, angle, temperature and crystal direction of implanted silicon carbide ions; the annealing parameters comprise annealing temperature, annealing time and temperature control rate;

[0019] a machine learning unit configured to input the process condition into a force field prediction model for each process condition, and output a machine learning force field under the process condition; wherein the force field prediction model is obtained by training a machine learning model by a training sample set; the training sample set comprises at least one training sample;

[0020] a simulation unit configured to perform molecular dynamics simulation according to the machine learning force field under the process condition to obtain a simulation result under the process condition.

[0021] Optionally, the construction unit of the force field prediction model comprises:

[0022] a training sample set construction unit configured to construct a training sample set; wherein the training sample set comprises at least one training sample, and the training sample comprises physical information of a training sample configuration;

[0023] a conversion unit configured to convert an atomic structure of a training sample configuration by using a descriptor of a machine learning model to obtain a training sample feature for each training sample;

[0024] a training unit configured to train the machine learning model based on the training sample feature to obtain the force field prediction model.

[0025] Optionally, the training sample set construction unit comprises:

[0026] An initial force field generation unit is configured to generate an initial force field by using a force field large model, wherein the initial force field comprises a plurality of initial configurations.

[0027] An active learning unit is configured to perform active learning by using the initial force field to obtain a plurality of target configurations.

[0028] A calculation unit is configured to perform first-principle calculation on each training sample configuration to obtain physical information of the training sample configuration, wherein the training sample configuration comprises the plurality of initial configurations and the plurality of target configurations.

[0029] Optionally, the physical information of the training sample configuration comprises interatomic forces and energy of the training sample configuration.

[0030] The third aspect of the present application provides an electronic device, comprising:

[0031] One or more processors;

[0032] A storage device having one or more programs stored thereon;

[0033] When the one or more programs are executed by the one or more processors, the one or more processors implement the simulation method based on machine learning and molecular dynamics according to any one of the first aspect.

[0034] The fourth aspect of the present application provides a computer storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the simulation method based on machine learning and molecular dynamics according to any one of the first aspect.

[0035] From the above solution, the present application provides a simulation method based on machine learning and molecular dynamics and related devices. After receiving a process menu, for each process condition in the process menu, the process condition is input into a force field prediction model, and a machine learning force field under the process condition is output. Finally, molecular dynamics simulation is performed according to the machine learning force field under the process condition to obtain a simulation result under the process condition. Through the machine learning method, high-precision first-principle calculation data is learned to generate a machine learning force field, simulate the interaction of silicon carbide in the ion implantation process, and thus simulate the ion implantation process, realize high-precision and high-efficiency simulation calculation, and further comprehensively consider the influence of different process conditions on device design process to develop a specific process menu and provide an optimization scheme for silicon carbide ion implantation. BRIEF DESCRIPTION OF DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required by the embodiments or the prior art description. Obviously, the drawings in the following description only represent some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of the provided drawings.

[0037] Figure 1 A specific flowchart of a simulation method based on machine learning and molecular dynamics provided for an embodiment of the present application;

[0038] Figure 2 A specific flowchart of a construction method of a force field prediction model provided for another embodiment of the present application;

[0039] Figure 3 A specific flowchart of a construction method of a training sample set provided for another embodiment of the present application;

[0040] Figure 4 A schematic diagram of a simulation device based on machine learning and molecular dynamics provided for another embodiment of the present application;

[0041] Figure 5 A schematic diagram of an electronic device for implementing a simulation method based on machine learning and molecular dynamics provided for another embodiment of the present application. DETAILED DESCRIPTION

[0042] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0043] The term "comprising" and its variants as used herein are open-ended, that is "including but not limited to". The term "based on" is "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Related definitions of other terms will be given in the following description.

[0044] It should be noted that the information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0045] It should be noted that the "first", "second", and the like mentioned in the present application are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0046] It should be noted that the "one", "multiple" modification mentioned in the present application is illustrative but not restrictive, and those skilled in the art should understand that unless the context clearly indicates otherwise, it should be understood as "one or more".

[0047] The embodiment of the present application provides a simulation method based on machine learning and molecular dynamics, as shown in Figure 1 The specific steps include:

[0048] S101, receiving a process menu.

[0049] The process menu includes at least one process condition. The process condition is divided into implantation parameters and annealing parameters; the implantation parameters include but are not limited to the energy, dose, angle, temperature, crystal direction, etc. of the implanted silicon carbide ions, which are not limited here; the annealing parameters include annealing temperature, annealing time, temperature control rate, etc. which are not limited here.

[0050] S102, for each process condition, input the process condition to the force field prediction model, and output the machine learning force field under the process condition.

[0051] The force field prediction model is obtained by training a machine learning model with a training sample set; the training sample set includes at least one training sample.

[0052] Specifically, the interaction of silicon carbide in the ion implantation process is simulated by the force field prediction model, that is, the microscopic atomic reaction process that will occur under the process condition, thereby assisting the process improvement of the process condition.

[0053] Optionally, in another embodiment of the present application, an embodiment of the method for constructing the force field prediction model, as shown in Figure 2 The specific steps include:

[0054] S201, constructing a training sample set.

[0055] The training sample set includes at least one training sample, and the training sample includes physical information of the training sample configuration.

[0056] Optionally, in another embodiment of the present application, an embodiment of step S201, as shown in Figure 3 The specific steps include:

[0057] S301, generating an initial force field by a force field large model.

[0058] The initial force field includes a plurality of initial configurations.

[0059] In the specific implementation of the present application, the initial force field can be generated using, but not limited to, the force field large model MPtrj and the force field large model GPTFF, which are not limited here.

[0060] S302, active learning is performed using the initial force field to obtain a plurality of target configurations.

[0061] Although the current force field large model has a wide range of applications and can provide high-quality initial force fields, it can also combine enhanced sampling schemes to more efficiently explore phase space, but the inference speed of the force field generated by the existing force field large model is slow, and the accuracy for specific vertical fields is poor. Therefore, the present application only uses the existing force field large model to participate in the construction of the initial training data set, and uses active learning technology to generate enough representative training data (target configurations).

[0062] S303, for each training sample configuration, performing first-principle calculation on the training sample configuration to obtain physical information of the training sample configuration.

[0063] The training sample configuration includes a plurality of initial configurations and a plurality of target configurations.

[0064] It should be noted that the first-principle calculation can use, but is not limited to, VASP, DASP, PWmat and other calculation software to calculate the interatomic force and energy of the training sample configuration, thereby obtaining the physical information of the training sample configuration.

[0065] It can be understood that due to many defect types caused by the injection of ions (such as Al ions) into the silicon carbide (SiC) matrix, these defects also need to be part of the training sample set, in addition, the configurations of different distances between ions (such as Al ions) and silicon (Si) ions, carbon (C) caused by ion (such as Al ion) injection also need to be considered, and the force and energy of these configurations are calculated as a data set.

[0066] Taking Al as an example, the training sample set can be composed of a large number of defect configurations related to Al-doped SiC. According to the 4H-SiC crystal structure, a supercell is constructed, and various defects and different Al doping concentrations are considered based on symmetry to generate a supercell containing defects and impurities.

[0067] S202, for each training sample, the atomic structure of the training sample configuration is converted using the descriptor of the machine learning model to obtain the training sample feature.

[0068] The descriptors are used to convert the atomic structure into mathematical features that can be processed by the neural network. The many-body interactions in the material are captured through the model training, enabling efficient prediction of the physical behavior of the material in large-scale molecular dynamics simulations.

[0069] In the practical application of the present application, the machine learning model can be a DP (Deep Potential) model, a NEP (Neuroevolution Potential) model, etc., which is not limited here. Moreover, multiple models can be trained in parallel, and after training, the best model is selected as the final force field prediction model. Thus, the accuracy and applicable range of the machine learning force field output by the final force field prediction model are ensured.

[0070] S203, training the machine learning model based on the training sample features to obtain a force field prediction model.

[0071] The machine learning model includes multiple iterations and verifications during training to ensure that it can generalize to ion (e.g., Al ion) implantation behavior under different conditions.

[0072] S103, performing molecular dynamics simulation based on the machine learning force field under the process conditions to obtain a simulation result under the process conditions.

[0073] In the specific implementation of the present application, when performing molecular dynamics simulation, Lammps can be used for simulation and simulation, but is not limited to this, which is not limited here. Since ion implantation can be ignored under low energy conditions, the simulation is simplified to ion (e.g., Al) atom bombardment of silicon carbide (SiC) substrate with certain kinetic energy, simulating the implantation path of ion (e.g., Al) in SiC, energy dissipation process, and defect (such as vacancy and interstitial atom) generation and migration mechanism. Through simulation, the effects of different implantation energy, dose, angle, temperature, etc. process parameters on implantation depth, distribution and uniformity, and defect distribution and evolution are analyzed, realizing high-precision and efficient simulation calculation.

[0074] In the practical application of the present application, a large number of simulation results are generated to establish a quantitative relationship between implantation conditions and defect formation. The simulation parameters include implantation parameters and annealing parameters after implantation.

[0075] The implantation parameters include implantation energy (e.g., 100-300 keV), dose (e.g., 10 16 -10 17 cm -3 ), angle (e.g., 0-20°), temperature (e.g., 20-600℃), and crystal orientation, etc., which are not limited here.

[0076] It should be noted that different implantation energies and doses will significantly affect the ion trajectory in the material, the doping depth and the defect generation behavior. By comparing the simulation results under different parameter combinations, the influence of different implantation conditions on the internal defect distribution and structural changes of the material is evaluated. The change of implantation angle and temperature will affect the path and residence position of ions in the material, and then affect the final doping distribution and defect density. By selecting different crystal orientations, the anisotropy of the material structure and the influence of channeling effect on ion implantation behavior can be further studied. According to the simulation results, an optimization scheme of implantation energy, dose and angle parameters can be proposed to ensure that the defect generation is controlled while the implantation depth and uniformity of the material are maximized.

[0077] It can be understood that the regulation of defects depends on the implantation conditions on the one hand and on the other hand on the annealing process after implantation. By comparing different annealing conditions (e.g. 1400-1900℃), the relationship between annealing conditions and lattice repair is established to ensure that impurities are effectively electrically activated and the performance of SJ devices is improved.

[0078] Since ion implantation and defects are closely related to lattice damage, including atomic displacement, cascade collision, dislocation formation, vacancy and interstitial atom generation, etc. On the basis of simulation results, these physical mechanisms are analyzed in depth, and the application can also determine which process parameters are most critical to doping and defect control. Physical mechanism analysis will reveal the interaction process of ions (e.g. Al) and SiC matrix, especially how to form initial lattice damage through collision, how to gradually expand the damage area and finally form stable defects through collision cascade and energy dissipation.

[0079] Therefore, the simulation will focus on the implantation path of ions (e.g. Al), the formation and evolution of defects, including the initial generation of defects, lattice damage during cascade collision, and the final defect aggregation behavior. Through these analyses, an optimized ion implantation process menu will be proposed. The core of defect regulation is to ensure that the doping concentration and implantation distribution requirements are met while improving the electrical activation rate of implanted atoms and repairing lattice defects.

[0080] Since the accumulation of lattice defects in the material will cause the electrical performance of the device to decrease, such as increasing the on-resistance and reducing the breakdown voltage. Therefore, the generation and evolution mechanism of defects can be analyzed in detail to determine which process parameters are the key factors affecting defect generation. After ion implantation, the annealing process is a key step in defect control. By adjusting the annealing temperature, annealing time and temperature control rate, point defects in the material can be effectively reduced, lattice repair can be promoted, and lattice distortion generated during doping can be reduced. By analyzing the defect conversion and annihilation mechanism in the annealing process through simulation results under different annealing conditions, the optimal annealing process parameters are proposed.

[0081] From the above scheme, the application provides a simulation method based on machine learning and molecular dynamics. After receiving a process menu, for each process condition in the process menu, the process condition is input into a force field prediction model, and a machine learning force field under the process condition is output. Finally, molecular dynamics simulation is performed according to the machine learning force field under the process condition, and a simulation result under the process condition is obtained. Through the machine learning method, high-precision first-principle calculation data is learned, a machine learning force field is generated, the interaction of silicon carbide in the ion implantation process is simulated, and the ion implantation process is simulated, so that high-precision and high-efficiency simulation calculation is realized, and the influence of different process conditions on device design process is comprehensively considered, and a specific process menu is developed to provide an optimization scheme for silicon carbide ion implantation.

[0082] The embodiment of the application provides a simulation device based on machine learning and molecular dynamics, as shown in the figure, and specifically comprises: Figure 4

[0083] The receiving unit 401 is configured to receive a process menu.

[0084] The process menu includes at least one process condition. The process condition includes implantation parameters and annealing parameters. The implantation parameters include, but are not limited to, the energy, dose, angle, temperature, and crystal direction of the implanted silicon carbide ions, and the like, which are not limited here. The annealing parameters include annealing temperature, annealing time, and temperature control rate, which are not limited here.

[0085] The machine learning unit 402 is configured to input the process condition into a force field prediction model for each process condition, and output a machine learning force field under the process condition.

[0086] The force field prediction model is obtained by training a machine learning model by a training sample set. The training sample set includes at least one training sample.

[0087] Optionally, in another embodiment of the application, an embodiment of the construction unit of the force field prediction model comprises:

[0088] The training sample set construction unit is configured to construct a training sample set.

[0089] The training sample set includes at least one training sample, and the training sample includes physical information of a training sample configuration.

[0090] The conversion unit is configured to convert the atomic structure of the training sample configuration by using the descriptor of the machine learning model for each training sample, and obtain a training sample feature.

[0091] The training unit is configured to train the machine learning model based on the training sample feature, and obtain the force field prediction model. ​

[0092] The specific working processes of the units disclosed in the above embodiments of the application can be seen from the corresponding method embodiment contents, such as Figure 2 as shown, which will not be repeated here.

[0093] Optionally, in another embodiment of the application, an implementation of the training sample set construction unit comprises:

[0094] An initial force field generation unit is configured to generate an initial force field by using a force field large model. The initial force field comprises a plurality of initial configurations.

[0095] An active learning unit is configured to perform active learning by using the initial force field to obtain a plurality of target configurations.

[0096] A calculation unit is configured to perform first-principle calculation on each training sample configuration to obtain physical information of the training sample configuration.

[0097] The training sample configuration comprises the plurality of initial configurations and the plurality of target configurations. The physical information of the training sample configuration comprises interatomic forces and energy of the training sample configuration.

[0098] The specific working processes of the units disclosed in the above embodiments of the application can be seen from the corresponding method embodiment contents, such as Figure 3 as shown, which will not be repeated here.

[0099] The simulation unit 403 is configured to perform molecular dynamics simulation according to the machine learning force field under the process condition to obtain a simulation result under the process condition.

[0100] The specific working processes of the units disclosed in the above embodiments of the application can be seen from the corresponding method embodiment contents, such as Figure 1 as shown, which will not be repeated here.

[0101] As can be seen from the above solutions, the application provides a simulation device based on machine learning and molecular dynamics. After the receiving unit 401 receives a process menu, the machine learning unit 402 inputs each process condition in the process menu into a force field prediction model to output a machine learning force field under the process condition. Finally, the simulation unit 403 performs molecular dynamics simulation according to the machine learning force field under the process condition to obtain a simulation result under the process condition. Through the machine learning method, high-precision first-principle calculation data is learned to generate a machine learning force field, simulate the interaction of silicon carbide in the ion implantation process, and thus simulate the ion implantation process, realize high-precision and high-efficiency simulation calculation, and further comprehensively consider the influence of different process conditions on device design processes to develop a specific process menu and provide an optimization scheme for silicon carbide ion implantation.

[0102] The functionality described herein above can be performed, at least in part, by one or more hardware logic components. For example, and without limitation, illustrative types of hardware logic components that can be used include Field-programmable Gate Arrays (FPGAs), Application-specific Integrated Circuits (ASICs), Application-specific Standard Products (ASSPs), System-on-a-chip systems (SOCs), Complex Programmable Logic Devices (CPLDs), etc.

[0103] Another embodiment of the present application provides an electronic device, comprising Figure 5 as shown, comprising:

[0104] one or more processors 501.

[0105] a storage 502 having stored thereon one or more programs.

[0106] The one or more programs, when executed by the one or more processors 501, enable the one or more processors 501 to implement the machine learning and molecular dynamics based simulation method according to any one of the above embodiments.

[0107] Another embodiment of the present application provides a computer storage medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements the machine learning and molecular dynamics based simulation method according to any one of the above embodiments.

[0108] In the context of the present application, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0109] It should be noted that the computer readable medium in the present application can be a computer readable signal medium or a computer readable storage medium or any combination of the two. The computer readable storage medium may, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or apparatus, or any combination of the above. More specific examples of the computer readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, the computer readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device or apparatus. In the present application, the computer readable signal medium can include a data signal carried in a baseband or as a part of a carrier wave, which carries computer readable program code. Such a propagated data signal can take various forms, including but not limited to an electromagnetic signal, an optical signal or any suitable combination of the above. The computer readable signal medium can also be any computer readable medium other than the computer readable storage medium, which can send, propagate or transmit a program for use by or in conjunction with an instruction execution system, device or apparatus. The program code contained in the computer readable medium can be transmitted by any suitable medium, including but not limited to a wire, a cable, a RF (radio frequency) or the like, or any suitable combination of the above.

[0110] The computer readable medium described above can be contained in the electronic device described above; or can exist separately and not be assembled into the electronic device.

[0111] Another embodiment of the present application provides a computer program product, which, when executed, performs the simulation method based on machine learning and molecular dynamics described above.

[0112] In particular, according to embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, embodiments of the present application include a computer program product comprising a computer program carried on a non-transitory computer readable medium, the computer program comprising program code for performing the methods illustrated by the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network by a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processing device, the above-mentioned functions defined in the methods of the embodiments of the present application are performed.

[0113] Although the subject matter has been described in language specific to structural features and / or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.

[0114] Although specific implementation details have been included in the above discussion, these should not be construed as limiting the scope of the application. Some of the features described in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment can also be implemented separately or in any appropriate subcombination. It is, therefore, contemplated that the application can include any combination of the features described in the foregoing disclosure.

[0115] The above description is merely exemplary of the application and the application principles that have been employed. It is to be understood that the application is not limited to the specific devices, methods, or the like, described above, and that numerous modifications can be made by one skilled in the art without departing from the application. For example, one of ordinary skill in the art will recognize that the application is not limited to the specific implementations described above, but will also include any implementation that is within the scope of the claims and that utilizes the principles of the application.

Claims

1. A simulation method based on machine learning and molecular dynamics, characterized in that, include: Receive process menu; wherein, the process menu includes at least one process condition; the process condition is divided into implantation parameters and annealing parameters; the implantation parameters include the energy, dose, angle, temperature and crystal orientation of the ions implanted into silicon carbide; the annealing parameters include annealing temperature, annealing time and temperature control rate; For each process condition, the process condition is input into the force field prediction model, and the machine learning force field under the process condition is output; wherein, the force field prediction model is obtained by training the machine learning model with a training sample set; the training sample set includes at least one training sample; the training sample includes physical information of the training sample configuration; Molecular dynamics simulations were performed based on the machine learning force field under the aforementioned process conditions to obtain simulation results under those conditions. The methods for constructing the force field prediction model include: An initial force field is generated using a large force field model; wherein the initial force field includes multiple initial configurations; Multiple target configurations were obtained by actively learning using the initial force field. For each training sample configuration, first-principles calculations are performed on the training sample configuration to obtain the physical information of the training sample configuration, i.e., the training sample; wherein, the training sample configuration includes multiple initial configurations and multiple target configurations; For each training sample, the atomic structure of the training sample configuration is transformed using the descriptor of the machine learning model to obtain the training sample features. Based on the features of the training samples, the machine learning model is trained to obtain the force field prediction model.

2. The simulation method based on machine learning and molecular dynamics according to claim 1, characterized in that, The physical information of the training sample configuration includes the interatomic forces and energy of the training sample configuration.

3. A simulation device based on machine learning and molecular dynamics, characterized in that, include: A receiving unit is used to receive a process menu; wherein the process menu includes at least one process condition; the process condition is divided into implantation parameters and annealing parameters; the implantation parameters include the energy, dose, angle, temperature, and crystal orientation of the ions implanted into silicon carbide; the annealing parameters include annealing temperature, annealing time, and temperature control rate; A machine learning unit is used to input the process conditions into a force field prediction model for each process condition and output the machine learning force field under the process conditions; wherein, the force field prediction model is obtained by training the machine learning model with a training sample set; the training sample set includes at least one training sample; the training sample includes physical information of the training sample configuration; The simulation unit is used to perform molecular dynamics simulation based on the machine learning force field under the process conditions, and obtain the simulation results under the process conditions. The building blocks of the force field prediction model include: An initial force field generation unit is used to generate an initial force field from a large force field model; wherein the initial force field includes multiple initial configurations; The active learning unit is used to actively learn from the initial force field to obtain multiple target configurations. The computing unit is used to perform first-principles calculations on each training sample configuration to obtain the physical information of the training sample configuration, i.e., the training sample; wherein, the training sample configuration includes multiple initial configurations and multiple target configurations; The transformation unit is used to transform the atomic structure of the training sample configuration for each training sample using the descriptor of the machine learning model, so as to obtain the training sample features. The training unit is used to train the machine learning model based on the features of the training samples to obtain the force field prediction model.

4. The simulation device based on machine learning and molecular dynamics according to claim 3, characterized in that, The physical information of the training sample configuration includes the interatomic forces and energy of the training sample configuration.

5. An electronic device, characterized in that, include: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the simulation method based on machine learning and molecular dynamics as described in any one of claims 1 to 2.

6. A computer storage medium, characterized in that, It stores a computer program, wherein the computer program, when executed by a processor, implements the simulation method based on machine learning and molecular dynamics as described in any one of claims 1 to 2.

Images