A simulation method for aging process of multi-component alloy system
By combining first principles, big data and artificial intelligence, we build a structure-energy database and machine learning potential function model, and use the Monte Carlo method to simulate the high-temperature aging process of multi-component alloys, solving the difficulty in simulating multi-component elements in the existing technology. The problem of alloy aging process realizes multi-scale evolution simulation and performance mechanism explanation.
Patent Information
- Application Number
- CN202210415181.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-04-18
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2042-04-18
AI Technical Summary
The existing technology is difficult to effectively simulate the aging process of multi-component alloy systems, especially at high temperatures. Traditional methods such as first-principle calculation and Monte Carlo simulation have high computational complexity, high time cost, and difficult to describe multi-component systems. and it is difficult to achieve long-term simulation.
Combining first principles, big data and artificial intelligence methods, by constructing structure-energy databases and machine learning potential function models, the Monte Carlo method is used to simulate high-temperature aging process to realize multi-scale evolution simulation of multi-component alloy systems.
Accurate multi-scale simulation of the aging process of the multi-component alloy system is realized, the mechanism that can explain the alloy performance from the atomic scale is reduced, the calculation complexity and time cost are reduced, and the high-temperature aging heat treatment process of the alloy can be simulated for a long time.
Smart Images

Figure CN114692508B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of simulation and prediction of microscopic properties of alloy materials, and in particular to a method for simulating the aging process of a multi-component alloy system. Background Art
[0002] Modern alloy materials are usually complex materials composed of multiple components. The coupling between multiple components makes the study of alloy processing and properties extremely difficult. When studying alloy materials, it is crucial to find an alloy composition ratio that meets specific properties in engineering. In addition, explaining the influence of composition on alloy properties from the microscopic scale of atoms and molecules and studying the microscopic physical and chemical mechanisms of alloy properties are still challenging research problems and have important scientific significance.
[0003] For precipitation-strengthened alloy materials such as nickel-based alloys, cobalt-based alloys and aluminum-based alloys, high-temperature aging treatment is usually involved in heat treatment. According to the different alloy compositions, precipitation-strengthening phases with different contents, morphologies and distributions are precipitated, which seriously affects the key macroscopic properties of the alloy such as strength and toughness. Traditional experimental research has the disadvantages of long research cycle and high cost, and it is difficult to conduct in-situ research on the dynamic precipitation process. Computational simulation methods can just make up for this defect. Computational simulation from a microscopic perspective can restore the real alloy precipitation process to a certain extent and give its inherent physical and chemical mechanism. However, classical material computational simulation methods usually only involve binary alloys and a small number of ternary alloys, and rarely target multi-component alloy systems. In addition, classical material computational simulation methods only involve a single scale, while the performance of material properties is multi-scale. Therefore, single-scale computational simulation methods are difficult to fully explain the mechanism and source of alloy properties. For example, the first-principles calculation of material systems contains at most a few hundred or a few thousand atoms, and it is difficult to study the effects of multiple alloying elements due to periodic boundary conditions; the classical Monte Carlo simulation relies heavily on empirical potentials and has many limitations: the development of empirical potentials is very time-consuming and labor-intensive, it is difficult to describe multi-component systems, and there is a lack of atomic-scale interpretation; although the first-principles molecular dynamics achieves the coupling of atomic and molecular scales, this method has very high computational complexity, the time cost of computational simulation is very high, and it is difficult to achieve long-term simulation of material systems. However, actual alloy materials have multiple components, and complex coupling effects will occur between the components. In addition, high-temperature aging heat treatment of alloys usually takes a time span of several hours or even days, involving processes such as alloy atomic diffusion and precipitation phase separation. The single use of the above methods is difficult to simulate the aging process of multi-component alloy systems. Therefore, it is necessary to develop a method for complex multi-component alloy systems that can accurately describe the physical properties at the atomic scale and have long-term molecular-scale simulation, so as to achieve multi-scale evolution simulation of the aging process of multi-component alloy systems. Summary of the invention
[0004] The purpose of the present invention is to provide a method for simulating the aging process of a multi-component alloy system, so as to realize the multi-scale evolution simulation of the aging process of the multi-component alloy system.
[0005] To achieve the above object, the present invention provides the following solutions:
[0006] A method for simulating an aging process of a multi-component alloy system, comprising:
[0007] Acquiring atoms of a multi-component alloy system and constructing a structure file based on the atoms;
[0008] Calculating the static total energy of the multi-component alloy system using first principles according to the structure file;
[0009] Building a structure-energy database according to the static total energy and the structure file;
[0010] Using big data and artificial intelligence methods to construct a multi-component system machine learning potential function model based on the structure-energy database;
[0011] The high temperature aging process is simulated by using the Monte Carlo method according to the multi-component system machine learning potential function model to obtain the high temperature aging process simulation result.
[0012] Optionally, the obtaining of atoms of a multi-component alloy system and constructing a structure file according to the atoms specifically includes:
[0013] Obtain atoms of multi-component alloy systems;
[0014] constructing a plurality of different atomic structures according to the atoms and the crystal structure of the multi-component alloy system;
[0015] A structure file is constructed according to the plurality of atomic structures.
[0016] Optionally, the calculating the static total energy of the multi-component alloy system by using the first principles according to the structure file specifically includes:
[0017] Using the first principle to perform dynamic optimization according to the atomic structure of the structure file, to obtain an optimization result;
[0018] The static total energy is calculated based on the optimization results using the first principles to determine the static total energy.
[0019] Optionally, the method of using big data and artificial intelligence methods to construct a multi-component system machine learning potential function model according to the structure-energy database specifically includes:
[0020] The atomic structure in the structure-energy database is used as input, and the static total energy in the structure-energy database is used as output, and the machine learning potential function of the multi-component system is trained to obtain a machine learning potential function model of the multi-component system.
[0021] Optionally, the initial structure of the Monte Carlo simulation system is randomly generated according to the multi-component alloy system; the initial structure is used for simulating the high temperature aging process;
[0022] Setting simulation parameters; the simulation parameters include simulation temperature, number of iterations and random number seed;
[0023] exchanging the atomic positions of two different elements in the initial structure to obtain a final structure;
[0024] Calculating the system energy of the initial state structure and the system energy of the final state structure according to the multi-component system machine learning potential function model;
[0025] Determining an update result of the Monte Carlo simulation system using a Metropolis algorithm according to the system energy of the initial state structure and the system energy of the final state structure;
[0026] The high temperature aging process simulation result of the multi-component alloy system is determined according to the simulation parameters and the updated result of the Monte Carlo simulation system.
[0027] Optionally, determining the update result of the Monte Carlo simulation system by using the Metropolis algorithm according to the system energy of the initial state structure and the system energy of the final state structure specifically includes:
[0028] Determine whether the system energy of the initial structure is greater than the system energy of the modal structure to obtain a first determination result;
[0029] If the first judgment result is yes, the Monte Carlo simulation system accepts the exchange and determines that the final state structure is an updated result of the Monte Carlo simulation system;
[0030] If the first judgment result is no, a random number is generated and a setting acceptance probability is determined according to the system energy of the initial state structure and the system energy of the final state structure;
[0031] Determine whether the random number is greater than the set reception probability, and obtain a second determination result;
[0032] If the second judgment result is yes, the Monte Carlo simulation system accepts the exchange and determines that the final state structure is an updated result of the Monte Carlo simulation system;
[0033] If the second judgment result is no, the Monte Carlo simulation system does not accept the exchange and determines that the initial state structure is the updated result of the Monte Carlo simulation system.
[0034] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:
[0035] The present invention obtains atoms of a multi-component alloy system and constructs a structure file according to the atoms; calculates the static total energy of the multi-component alloy system using the first principle according to the structure file; constructs a structure-energy database according to the static total energy and the structure file; constructs a multi-component system machine learning potential function model according to the structure-energy database using big data and artificial intelligence methods; and simulates a high-temperature aging process according to the multi-component system machine learning potential function model using the Monte Carlo method to obtain a high-temperature aging process simulation result. The present invention combines the first principle method, big data, artificial intelligence, and Monte Carlo simulation to achieve multi-scale simulation of the aging process of a multi-component alloy system. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0037] Figure 1 A flow chart of the method for simulating the aging process of a multi-component alloy system provided by the present invention;
[0038] Figure 2 A schematic diagram of a method for simulating the aging process of a multi-component alloy system provided by the present invention;
[0039] Figure 3 A flowchart of the multi-component alloy system aging process simulation method provided by the present invention in practical application;
[0040] Figure 4 This is a schematic diagram of data related to the energy prediction results of machine learning potential functions;
[0041] Figure 5 Schematic diagram of data related to the calculation results of the evolution of the precipitated phase fraction in the aging process simulation. DETAILED DESCRIPTION
[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0043] The purpose of the present invention is to provide a method for simulating the aging process of a multi-component alloy system, so as to realize the multi-scale evolution simulation of the aging process of the multi-component alloy system.
[0044] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0045] like Figure 1-Figure 3 As shown, the present invention provides a method for simulating the aging process of a multi-component alloy system, comprising:
[0046] Step 101: Acquire atoms of a multi-component alloy system and construct a structure file based on the atoms.
[0047] Step 101 specifically includes:
[0048] Acquire atoms of a multi-component alloy system; construct a plurality of different atomic structures according to the atoms and the crystal structure of the multi-component alloy system; and construct a structure file according to the plurality of atomic structures.
[0049] Obtain the atoms of a multi-component alloy system, construct symmetrically unequal atomic structures based on the atomic and alloy system crystal structures, and construct a structure file. Construct unequal atomic structures based on the components, and you can manually select symmetrically unequal structures or use enumeration methods to build them on a large scale. For a multi-component system, you need to construct multiple groups of structural systems, including single-component structural systems, binary structural systems, or multi-component structural systems, and even defective structural systems.
[0050] Step 102: Calculate the static total energy of the multi-component alloy system using the first principles according to the structure file.
[0051] Step 102 specifically includes:
[0052] The first principle is used to perform dynamic optimization according to the atomic structure of the structure file to obtain an optimization result; the first principle is used to perform static total energy calculation according to the optimization result to determine the static total energy.
[0053] Step 103: constructing a structure-energy database according to the static total energy and the structure file.
[0054] Step 104: Use big data and artificial intelligence methods to construct a multi-component system machine learning potential function model based on the structure-energy database; Use big data and artificial intelligence methods to construct a multi-component system machine learning potential function using the structure-energy database. Use the atomic coordinates and element types in the structure in the database as input, and the structural energy as output to train a machine learning model with sufficient energy accuracy to construct a machine learning potential function. The model can be constructed manually, or it can be trained using an existing machine learning potential model framework, such as Amp and DeePMD, etc.
[0055] Step 104 specifically includes:
[0056] Taking the atomic structure in the structure-energy database as input and the static total energy in the structure-energy database as output, the machine learning potential function of the multi-component system is trained to obtain a machine learning potential function model of the multi-component system, wherein the machine learning potential function model is a machine learning potential function model that can predict the structure energy of the multi-component system.
[0057] Step 105: Use the Monte Carlo method to simulate the high temperature aging process according to the machine learning potential function model of the multi-component system to obtain the simulation results of the high temperature aging process. Use the machine learning potential function as the system energy calculator and use Monte Carlo to simulate the evolution of the high temperature aging process of the multi-component system. Perform Monte Carlo simulation based on the Metropolis algorithm, read the machine learning potential function to calculate the system energy in real time, and simulate the high temperature aging process of the disordered supersaturated multi-component system.
[0058] Step 105 specifically includes:
[0059] Set simulation parameters; the simulation parameters include simulation temperature, number of iterations and random number seed.
[0060] The atomic positions of two different elements in the initial structure are exchanged to obtain a final structure.
[0061] The system energy of the initial state structure and the system energy of the final state structure are calculated according to the machine learning potential function model of the multi-component system.
[0062] The update result of the Monte Carlo simulation system is determined by using the Metropolis algorithm according to the system energy of the initial state structure and the system energy of the final state structure.
[0063] The simulation result of the high temperature aging process of the multi-component alloy system is determined according to the simulation parameters and the updated result of the Monte Carlo simulation system; specifically, the simulation result of the high temperature aging process of the multi-component alloy system is determined according to the number of iterations and the updated result of the Monte Carlo simulation system. The simulation ends when the simulation reaches the specified number of iterations; and the simulation result of the high temperature aging process of the alloy system is analyzed.
[0064] Determining the update result of the Monte Carlo simulation system by using the Metropolis algorithm according to the system energy of the initial state structure and the system energy of the final state structure specifically includes:
[0065] It is determined whether the system energy of the initial structure is greater than the system energy of the modal structure to obtain a first determination result.
[0066] If the first judgment result is yes, the Monte Carlo simulation system accepts the exchange and determines that the final state structure is an updated result of the Monte Carlo simulation system.
[0067] If the first judgment result is no, a random number is generated and the acceptance probability is determined based on the system energy of the initial state structure and the system energy of the final state structure.
[0068] It is determined whether the random number is greater than the set reception probability to obtain a second determination result.
[0069] If the second judgment result is yes, the Monte Carlo simulation system accepts the exchange and determines that the final state structure is an updated result of the Monte Carlo simulation system.
[0070] If the second judgment result is no, the Monte Carlo simulation system does not accept the exchange and determines that the initial state structure is the updated result of the Monte Carlo simulation system.
[0071] Determining the update result of the Monte Carlo simulation system by using the Metropolis algorithm according to the system energy of the initial state structure and the system energy of the final state structure, in practical application, specifically includes:
[0072] When the system energy E of the experimental exchange initial structure ini Greater than the system energy E of the final structure fin When , the experimental exchange is accepted; when the system energy of the initial state structure is less than or equal to the system energy of the final state structure, a random number a is generated and the acceptance probability is determined according to the system energy of the initial state structure and the system energy of the final state structure. The expression of the acceptance probability is Where k Bis the Boltzmann constant, T is the absolute temperature; determine whether the random number is greater than the acceptance probability. If so, accept the trial exchange and update the configuration to the trial exchange structure; otherwise, reject the exchange and the configuration remains unchanged.
[0073] The present invention combines the first principles, big data with artificial intelligence and Monte Carlo simulation methods to achieve multi-component alloy aging simulation at a certain temperature and on a long time scale, and accurately describe the kinetic behaviors such as element diffusion during the aging process. This method is not limited to a specific alloy system and number of components. It can not only give the system evolution of the aging process from the molecular scale, but also give a physical explanation of the evolution from the atomic scale and the electronic scale. In addition, most of the steps of this method are automatically completed in an automatic process-based manner, realizing the automatic generation, calculation, collection, processing and storage of material data, providing a basis for further using computer technology to carry out material research.
[0074] The present invention also takes the Ni-Al-Re alloy system as an example to provide a specific processing flow of the aging process simulation method for a multi-component alloy system.
[0075] The Ni-Al-Re alloy system is a typical face-centered cubic structure. After high-temperature solution heat treatment, it is a disordered supersaturated γ solid solution. During the aging heat treatment, a γ'-Ni3Al precipitate phase will be supersaturated and precipitated from the matrix γ phase, accompanied by the redistribution of alloy elements in the two phases. Specifically, it includes the following steps:
[0076] 1. Use icet and ASE software to enumerate binary and multi-element structures with unequal symmetry, store them in the structure database in the form of ASE.db and generate structure files in batches. Use VASP-kit software to generate equivalent input files for each structure, perform VASP calculations to dynamically optimize each structure first, and then calculate the static total energy. After the calculation is completed, write the energy into the structure database to obtain the structure-energy database.
[0077] Specifically, icet software was used to enumerate symmetry-inequivalent architectures, including three systems: Ni-Al, Ni-Re, Al-Re and Ni-Al-Re. Among them, Ni-Al includes all structures with less than 10 atoms, totaling 2346; Ni-Re and Al-Re include all structures with less than 8 atoms, both 631; structures with less than 10 atoms in the Ni-Al-Re system were manually selected according to the Re content, totaling 978. A total of 4586 structures were created to create the ASE.db structural database. Combined with VASP-kit software, input files were generated for batch processing of all structures, and VASP structural optimization and static energy calculation were performed. After the calculation was completed, script post-processing was used to extract the static energy of each structure, and the energy was updated to the structure-energy database.
[0078] 2. Use an automated program to export the structure and energy information of the structure-energy database, set the cutoff radius and deep learning model parameters, and generate the input file of the DeePMD deep potential energy software, and train a multi-component machine learning potential function model that can achieve sufficient energy accuracy, so as to predict the energy of any structure of the system.
[0079] Specifically: extract structure and energy information from the structure-energy database as training data for the machine learning potential function model. Generate input files for the DeePMD software and set the cutoff radius (Common units in crystallography: Angstroms, 1e-10m), network parameters, learning rate, loss function and other parameters. Execute training and obtain the machine learning potential function model, and draw the first principle calculation energy and machine learning potential function prediction energy curves as shown in Figure 4 , specifically the DFT calculated energy and machine learning potential function predicted energy of the Ni-Al-Re system. From the figure, we can see that the energy prediction effect of the model is very good.
[0080] 3. The machine learning potential function model is input into the Monte Carlo simulation software based on Metropolis as the system energy calculator. At the same time, the initial structure of the simulation is read, and the simulation input parameters such as simulation temperature and maximum number of iterations are set to study the high-temperature aging process of the supersaturated alloy system.
[0081] Specifically, the energy calculator in the open source Monte Carlo simulation software source code was modified to the machine learning potential function model trained in step 2, a random initial structure was generated, and the simulation temperature was set to 1000K and the number of iterations was set to 5000,000. The Metropolis algorithm was executed to obtain the evolution of the system. Combined with the OVITO molecular simulation post-processing software, the order of the system was obtained. The results are as follows: Figure 5 , specifically the trajectory diagram and γ' phase fraction of MC simulation of 73Ni-25Al-2Re system at 1000K. Figure 5 Part (a) is the simulated initial structure, and the γ' phase fraction is 1.0%; Figure 5 Part (b) shows the structure after 1,000,000 simulation steps, and the γ' phase fraction is 50.6%. Figure 5 Part (c) shows the structure after 1,000,000 simulation steps, and the γ' phase fraction is 99.5%.
[0082] 4. Finally, restore the system simulation process according to the simulation trajectory file, use molecular simulation post-processing software such as OVITO to post-process the trajectory file, obtain the evolution of phase fraction, order degree, etc., and combine the first-principles calculation to analyze the microscopic electronic mechanism of the evolution.
[0083] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0084] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of the present invention. At the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.
Claims
1. A method for simulating the aging process of a multi-component alloy system. It is characterized in that include: Acquiring atoms of a multi-component alloy system and constructing a structure file based on the atoms; Calculating the static total energy of the multi-component alloy system using first principles according to the structure file; Building a structure-energy database according to the static total energy and the structure file; Using big data and artificial intelligence methods to construct a multi-component system machine learning potential function model based on the structure-energy database; Using the Monte Carlo method to simulate the high temperature aging process according to the multi-component system machine learning potential function model, to obtain a high temperature aging process simulation result; The high temperature aging process simulation is performed according to the multi-component system machine learning potential function model using the Monte Carlo method to obtain the high temperature aging process simulation result, specifically including: Randomly generating an initial structure of a Monte Carlo simulation system according to the multi-component alloy system; the initial structure is used for simulating a high temperature aging process; Setting simulation parameters; the simulation parameters include simulation temperature, number of iterations and random number seed; exchanging the atomic positions of two different elements in the initial structure to obtain a final structure; Calculating the system energy of the initial state structure and the system energy of the final state structure according to the multi-component system machine learning potential function model; Determining an update result of the Monte Carlo simulation system using a Metropolis algorithm according to the system energy of the initial state structure and the system energy of the final state structure; Determine the simulation result of the high temperature aging process of the multi-component alloy system according to the simulation parameters and the updated result of the Monte Carlo simulation system; specifically: Determining the simulation result of the high temperature aging process of the multi-component alloy system according to the number of iterations and the update result of the Monte Carlo simulation system; terminating the simulation when the simulation reaches the specified number of iterations; analyzing the simulation result of the high temperature aging process of the alloy system; Determining the update result of the Monte Carlo simulation system by using the Metropolis algorithm according to the system energy of the initial state structure and the system energy of the final state structure specifically includes: Determine whether the system energy of the initial state structure is greater than the system energy of the final state structure to obtain a first determination result; If the first judgment result is yes, the Monte Carlo simulation system accepts the exchange and determines that the final state structure is an updated result of the Monte Carlo simulation system; If the first judgment result is no, a random number is generated and a setting acceptance probability is determined according to the system energy of the initial state structure and the system energy of the final state structure; Determine whether the random number is greater than the set acceptance probability, and obtain a second determination result; If the second judgment result is yes, the Monte Carlo simulation system accepts the exchange and determines that the final state structure is an updated result of the Monte Carlo simulation system; If the second judgment result is no, the Monte Carlo simulation system does not accept the exchange and determines that the initial state structure is an updated result of the Monte Carlo simulation system; Determining the update result of the Monte Carlo simulation system by using the Metropolis algorithm according to the system energy of the initial state structure and the system energy of the final state structure, in practical application, specifically includes: When the system energy E of the experimental exchange initial structure ini Greater than the system energy E of the final structure fin When , the experimental exchange is accepted; when the system energy of the initial state structure is less than or equal to the system energy of the final state structure, a random number a is generated and the acceptance probability is determined according to the system energy of the initial state structure and the system energy of the final state structure; The expression of acceptance probability p is set as In the formula, K B is the Boltzmann constant, T is the absolute temperature; determine whether the random number is greater than the acceptance probability. If so, accept the trial exchange and update the configuration to the trial exchange structure; otherwise, reject the exchange and the configuration remains unchanged.
2. The method for simulating the aging process of a multi-component alloy system according to claim 1, It is characterized in that The obtaining of atoms of the multi-component alloy system and constructing a structure file according to the atoms specifically includes: Obtain atoms of multi-component alloy systems; constructing a plurality of different atomic structures according to the atoms and the crystal structure of the multi-component alloy system; A structure file is constructed according to the plurality of atomic structures.
3. The method for simulating the aging process of a multi-component alloy system according to claim 1, It is characterized in that The calculating the static total energy of the multi-component alloy system by using the first principles according to the structure file specifically includes: Using the first principle to perform dynamic optimization according to the atomic structure of the structure file, to obtain an optimization result; The static total energy is calculated based on the optimization results using the first principles to determine the static total energy.
4. The method for simulating the aging process of a multi-component alloy system according to claim 2, It is characterized in that The method of using big data and artificial intelligence to construct a multi-component system machine learning potential function model based on the structure-energy database specifically includes: The atomic structure in the structure-energy database is used as input, and the static total energy in the structure-energy database is used as output, and the machine learning potential function of the multi-component system is trained to obtain a machine learning potential function model of the multi-component system.
Citation Information
Patent Citations
Computer analogue simulation method and device for substances and materials
CN112071371A
High-entropy alloy mechanical property calculation method based on atom occupying behavior in sublattice
CN112395762A
Method and equipment for obtaining material phase diagram, and computer readable and writable storage medium
CN112800609A
Systems and methods for predicting structure and properties of atomic elements and alloy materials
US20200066376A1
Application of optimized quantum monte carlo simulation method in studying complex magnetic systems
US20210342505A1
