Basin jumping algorithm based on machine learning recommendation system and application thereof
The placeholder configuration of perovskite oxide Ba0.5Sr0.5Co0.8Fe0.2O3 is optimized through the basin jump algorithm based on the machine learning recommendation system, and the high cost and disorder problems in the optimization of perovskite catalyst structure are solved, and the problem of high cost and disorder is achieved. A more efficient understanding of the relationship between catalytic properties and material structure-effects.
Patent Information
- Application Number
- CN202410674029.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-28
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-05-28
AI Technical Summary
The prior art has high cost and structural disorder problems in the structural optimization of perovskite oxide catalysts, which makes it difficult to understand and improve catalytic activity.
Using a basin jump algorithm based on machine learning recommendation system, the placeholding configuration of perovskite oxide Ba0.5Sr0.5Co0.8Fe0.2O3 is optimized to avoid expensive DFT calculations , improve structural search efficiency.
It improves the probability of the basin jump algorithm searching for the lowest local area and finding a lower energy space-occupying configuration, which promotes the understanding of the structure-effect relationship of the material and reduces the calculation cost.
Smart Images

Figure CN119067205A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of artificial intelligence structure optimization, and in particular to a basin jumping algorithm based on a machine learning recommendation system and its application. Background Art
[0002] Hydrogen is a green and clean energy source. Hydrogen production by water electrolysis is a clean way to produce hydrogen, but the oxygen evolution reaction (OER) occurring at the anode is slow. An effective method is to use a suitable catalyst to reduce the OER reaction energy barrier and increase the reaction rate. At present, precious metal oxide catalysts are mainly used commercially, which have relatively good stability and catalytic activity, such as IrO2 and RuO2. However, the high cost of precious metals limits their widespread application, so it is necessary to find a low-cost catalyst with high catalytic performance.
[0003] Perovskite oxides have become a research hotspot recently due to their low cost, diverse chemical composition, huge configuration space and high catalytic activity. The chemical formula of perovskite is ABX3. A and B cover almost all metal cations in the periodic table, and X is usually a chalcogen or halogen. 0.5 Sr 0.5 Co 0.8 Fe 0.2 O3 is a recently discovered OER catalytic material with high catalytic activity. However, this multi-cation perovskite is characterized by multiple different cations on a sublattice, which leads to disordered atomic arrangement or site configuration in the structure. This makes it difficult to determine the site configuration and hinders the understanding of the structure-activity relationship of the material.
[0004] Computational studies of site-occupying configurations include special quasi-random structural methods and cluster expansion methods. Site-occupying configuration optimization is subordinate to structural optimization, so structural optimization algorithms can also be used for site-occupying configuration optimization, such as particle swarm optimization, evolutionary algorithm, simulated annealing algorithm, and basin hopping algorithm. These structural optimization algorithms often require expensive DFT calculations. The emergence of artificial intelligence has accelerated structural search, such as using machine learning potential functions to replace expensive DFT calculations. However, there is currently little research on introducing machine learning into basin hopping global optimization algorithms. Summary of the invention
[0005] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a basin jumping algorithm based on a machine learning recommendation system and its application, which at least solves some of the problems in the prior art.
[0006] To achieve the above purpose, the technical solution adopted by the present invention is as follows:
[0007] A basin jumping algorithm based on a machine learning recommendation system includes the following steps:
[0008] Step 1: obtaining an initial random placeholder configuration for placeholder configuration optimization;
[0009] Step 2: Determine the real-time machine learning force field calculation parameters;
[0010] Step 3: Determine the basin jump parameters, including temperature and jump rules;
[0011] Step 4: Use the real-time machine learning force field as a recommendation system and introduce the basin jumping algorithm to Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 The O3 occupancy configuration is globally optimized;
[0012] Step 5: Verify the optimized lowest energy occupied configuration by density functional theory (DFT).
[0013] In a further embodiment, in step 1, the initial random site-occupying configuration includes: structure model size, number of atoms, lattice parameters and atomic category information in the sub-lattice.
[0014] In a further embodiment, in step 1, the initial random occupancy configuration is a structural model of a 3×4×5 supercell; the initial random occupancy configuration contains 300 atoms, and its lattice parameters are Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 The A-site sublattice of O3 is occupied by Ba and Sr ions, the B-site sublattice is occupied by Co and Fe ions, and O acts as a "bystander" in the anion sublattice.
[0015] In a further embodiment, in step 2, the real-time machine learning force field parameters include: the training weight ratio of atomic forces and system energy, the atomic distribution of the broadened descriptor, and the number of basis functions of the extended descriptor.
[0016] In a further scheme, in step three, in the process of determining the basin jump parameters, the temperature K is set to 2500K, and the jump rule is set to perform only one exchange between different cations in the same sublattice for each jump.
[0017] In a further embodiment, in step 4, the Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 The O3 occupancy configuration is globally optimized, including the following steps:
[0018] (1) Implementation of basin jumping algorithm:
[0019] During each jump, two different cations in a certain sublattice in the occupied configuration are exchanged, and the new configuration is structurally relaxed or locally minimized;
[0020] If the Boltzmann factor exp(-ΔE / kT) is greater than a random number between (0, 1), the new occupancy configuration will be accepted;
[0021] If the conditions are not met, the new occupied configuration is rejected and returns to the previous local minimum state, repeating the jump process until the preset maximum number of steps is reached;
[0022] in,
[0023] k represents the Boltzmann constant; T represents temperature; ΔE is the difference between the current local minimum energy E and the previous local minimum energy;
[0024] (2) Relaxation / local minimum process:
[0025] During the energy minimization process for each configuration, the first two steps of structural relaxation are performed by DFT calculations, while the real-time machine learning force field (on-the-fly MLFF) is updated based on the obtained DFT data; then the remaining 15 relaxation steps are completed quickly using on-the-fly MLFF; this relaxation based on on-the-fly MLFF is effective because the local environment of the atoms changes little;
[0026] (3) The recommendation system determines the jump direction:
[0027] Before each “jump”, 100 random pre-“jumps” are performed to obtain a new perturbation configuration, and its energy is predicted using on-the-fly MLFF. The configuration with the lowest predicted energy is used as the next actual “jump” direction.
[0028] In a further scheme, in step 5, DFT calculation is performed on the optimized lowest energy occupied configuration, and the lowest energy occupied configuration is visualized. This helps to analyze the relationship between structure and material properties.
[0029] On the other hand, the above basin hopping algorithm based on machine learning recommendation system is effective in perovskite catalyst Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 Application in O3 site occupancy configuration optimization.
[0030] Compared with the prior art, the present invention has the following beneficial effects:
[0031] 1. The present invention introduces a machine learning recommendation system into the framework of a conventional basin-hopping algorithm, which improves the probability of the conventional basin-hopping algorithm to search for the lowest local minimum nearby and avoids exploring a large number of unnecessary high-energy occupied configurations through relatively expensive DFT calculations.
[0032] 2. The method of the present invention can more efficiently find lower energy site-occupying configurations and promote the understanding of the structure-activity relationship of materials. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1 This is a flow chart of the basin jumping global optimization algorithm based on the machine learning recommendation system in Example 1 of the present invention.
[0034] Figure 2 It is a comparison diagram of the new method of Example 1 of the present invention and the traditional basin jumping method, special quasi-random structure method and random search method.
[0035] Figure 3 This is the RBH operation result of Example 1 of the present invention when the size of the machine learning model data set is changed.
[0036] Figure 4 It is a comparison between the prediction of the relaxation part machine learning system of Example 1 of the present invention and the DFT calculation results.
[0037] Figure 5 This is a comparison between the predicted energy and the relaxation energy based on the machine learning recommendation system in Example 1 of the present invention.
[0038] Figure 6 It is a comparison diagram of the initial configuration of Example 1 of the present invention and the lowest energy occupied configuration obtained after optimization and confirmed by DFT calculation. DETAILED DESCRIPTION
[0039] The following non-limiting examples can enable those skilled in the art to more fully understand the present invention, but do not limit the present invention in any way. The following content is only an exemplary description of the scope of the present invention, and those skilled in the art can make various changes and modifications to the invention of the present invention based on the disclosed content, and they should also fall within the scope of the present invention.
[0040] Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2O3 is a recently discovered OER catalytic material with high catalytic activity. However, this multi-cation perovskite is characterized by a variety of different cations on a sublattice, which leads to disordered atomic arrangement or site-occupancy configuration in the structure. This makes it difficult to determine the site-occupancy configuration and hinders the understanding of the structure-activity relationship of the material. The computational study of site-occupancy configuration includes special quasi-random structure methods and cluster expansion methods, and structural optimization algorithms such as particle swarm optimization, evolutionary algorithms, simulated annealing algorithms, and basin jumping algorithms can also be used for site-occupancy configuration optimization. These structural optimization algorithms often require expensive DFT calculations. The emergence of artificial intelligence has accelerated structural search, such as using machine learning potential functions to replace expensive DFT calculations.
[0041] The present invention introduces a machine learning recommendation system into the framework of a conventional basin hopping algorithm, referred to as RBH (recommendation-based basin-hopping). Different from the traditional basin hopping algorithm (BH), on the one hand, during the process of structural relaxation or local minima, the structural relaxation is accelerated by combining density functional theory calculations (DFT) and machine learning; on the other hand, the recommendation system determines the change in the "jump" direction. The recommendation system improves the probability of the BH algorithm searching for the lowest local minimum nearby, and avoids exploring a large number of unnecessary high-energy occupying configurations through relatively expensive DFT calculations. This method can more efficiently find lower energy occupying configurations and promote the understanding of the structure-activity relationship of materials.
[0042] To solve the above problems, the present invention provides a basin jumping algorithm based on a machine learning recommendation system. The flowchart is as follows: Figure 1 As shown in the figure. The black curve represents the original potential energy surface; the blue represents the transformed potential energy surface; the white circle represents the perturbed occupying configuration; the green dot represents the energy local minimum configuration in the accepting state; the red dot represents the energy local minimum configuration in the rejecting state; the black dotted line represents the process of combining DFT calculation and on the fly MLFF to accelerate the local minimization / relaxation process; the green arrow represents the machine learning recommendation system, which is used to specify the next "jump" direction.
[0043] A basin jumping algorithm based on a machine learning recommendation system specifically includes the following steps:
[0044] Step 1: obtaining an initial random placeholder configuration for placeholder configuration optimization;
[0045] Step 2: Determine the on-the-fly MLFF calculation parameters;
[0046] Step 3: Determine the basin jump parameters, including temperature and jump rules;
[0047] Step 4: Use the real-time machine learning force field as a recommendation system and introduce the basin jumping algorithm to Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 The O3 occupancy configuration is globally optimized;
[0048] Step 5: Perform DFT calculation verification on the optimized lowest energy occupied configuration.
[0049] In a further embodiment, in step 1, the initial random site-occupying configuration includes: structure model size, number of atoms, lattice parameters and atomic category information in the sub-lattice.
[0050] In a further embodiment, in step 1, the initial random occupancy configuration is a structural model of a 3×4×5 supercell; the initial random occupancy configuration contains 300 atoms, and its lattice parameters are Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 The A-site sublattice of O3 is occupied by Ba and Sr ions, the B-site sublattice is occupied by Co and Fe ions, and O acts as a "bystander" in the anion sublattice.
[0051] In a further embodiment, in step 2, the real-time machine learning force field parameters include: the training weight ratio of atomic forces and system energy, the atomic distribution of the broadened descriptor, and the number of basis functions of the extended descriptor.
[0052] In a further scheme, in step 2, the training data of on-the-fly MLFF comes from the DFT data of structural relaxation during RBH operation, that is, DFT calculation, construction, update and application of potential function are nested with each other, and the structure search and update of potential function are carried out synchronously.
[0053] In a further scheme, in step 2, the parameters of on-the-fly MLFF include: "ML_IWEIGHT = 3" for energy, force and stress normalization; "ML_WTOTEN = 100", "ML_WTIFOR = 4" and "ML_WTSIF = 1" correspond to the weights applied to energy, force and stress during on-the-fly MLFF training, respectively. Since this paper focuses on the energy of the configuration, the weight for energy fitting accuracy is deliberately increased; "ML_SION1 = 0.3" and "ML_SION2 = 0.3" are used to broaden the atomic distribution of the descriptor; "ML_MRB1 = 12" and "ML_MRB2 = 12" are used to expand the number of basis functions of the descriptor.
[0054] In a further scheme, in step three, in the process of determining the basin jump parameters, the temperature K is set to 2500K, and the jump rule is set to perform only one exchange between different cations in the same sublattice for each jump.
[0055] In a further embodiment, in step 4, the Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 The O3 occupancy configuration is globally optimized, including the following steps:
[0056] (1) Implementation of basin jumping algorithm:
[0057] Each time a "jump" occurs, two different cations in a sublattice in the occupied configuration are exchanged, and the new configuration is structurally relaxed or locally minimized; if the Boltzmann factor exp(-ΔE / kT) is greater than a random number between (0, 1), the new occupied configuration will be accepted; if the condition is not met, the new occupied configuration will be rejected and return to the previous local minimum state, and the "jump" operation is iterated until the preset maximum number of steps is reached;
[0058] Among them, in the Boltzmann factor, k represents the Boltzmann constant; T represents temperature; ΔE is the difference between the current local minimum energy E and the previous local minimum energy.
[0059] (2) Relaxation / local minimum process:
[0060] During the energy minimization process of each configuration, the first two steps of structural relaxation are performed by DFT calculations, and the on-the-fly MLFF is updated based on the obtained DFT data; then the remaining 15 relaxation steps are quickly completed using on-the-fly MLFF; this relaxation based on on-the-fly MLFF is effective because the local environment of the atoms changes little.
[0061] (3) The recommendation system determines the jump direction:
[0062] Each time a "jump" is made, 100 random pre-"jumps" are performed to obtain a new perturbed configuration, and its energy is predicted using on-the-fly MLFF, and the configuration with the lowest predicted energy is used as the actual "jump" direction for the next time. This recommendation system improves the probability of the BH algorithm searching for the lowest local minimum nearby and avoids exploring a large number of unnecessary high-energy occupied configurations through relatively expensive DFT calculations.
[0063] In a further scheme, in step 5, DFT calculation is performed on the optimized lowest energy occupied configuration, and the lowest energy occupied configuration is visualized. This helps to analyze the relationship between structure and material properties.
[0064] The following embodiments are combined with the accompanying drawings and Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 O3 introduces the basin jumping algorithm based on machine learning recommendation system of the present invention.
[0065] Example 1
[0066] A basin jumping algorithm based on machine learning recommendation system, comprising the following steps:
[0067] Step 1: Obtain an initial random placeholder configuration for placeholder configuration optimization.
[0068] Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 O3 is a cubic structure, and the initial occupancy configuration is a 3×4×5 supercell, which contains 300 atoms. 30 Ba ions and Sr ions; 48 Co ions; 12 Fe ions; 180 oxygen ions. The lattice parameters of a single unit cell are There are two sublattices, where the A-site sublattice is occupied by Ba and Sr ions, and the B-site sublattice is occupied by Co and Fe ions. Oxygen ions act as "bystanders". The ions in the initial occupation configuration are randomly arranged.
[0069] Step 2: Determine the on-the-fly MLFF calculation parameters, including the training weight ratio of atomic forces and system energy, the atomic distribution of the broadened descriptor, and the number of basis functions of the extended descriptor.
[0070] The training data for on-the-fly MLFF comes from the DFT data of structural relaxation during RBH or BH operation, that is, DFT calculations, function construction, update and application are nested with each other, and the structure search is carried out synchronously with the update of the potential function.
[0071] The parameters of on-the-fly MLFF include: "ML_IWEIGHT = 3" for energy, force and stress normalization; "ML_WTOTEN = 100", "ML_WTIFOR = 4" and "ML_WTSIF = 1" correspond to the weights applied to energy, force and stress during on-the-fly MLFF training, respectively. Since this paper focuses on the energy of the configuration, the weight for energy fitting accuracy is deliberately increased; "ML_SION1 = 0.3" and "ML_SION2 = 0.3" are used to broaden the atomic distribution of the descriptor; "ML_MRB1 = 12" and "ML_MRB2 = 12" are used to expand the number of basis functions of the descriptor.
[0072] Step 3: Determine the basin hopping parameters, set the temperature K to 2500K, and set the hopping rule to perform only one exchange between different cations in the same sublattice in each hopping.
[0073] Step 4: Use the real-time machine learning force field as a recommendation system and introduce the basin jumping algorithm to Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 The O3 occupancy configuration is globally optimized.
[0074] (1) Implementation of basin jumping algorithm:
[0075] Each time a "jump" occurs, two different cations in a sublattice in the occupied configuration are exchanged, and the new configuration is structurally relaxed or locally minimized; if the Boltzmann factor exp(-ΔE / kT) is greater than a random number between (0, 1), the new occupied configuration will be accepted; if the condition is not met, the new occupied configuration will be rejected and return to the previous local minimum state, and the "jump" operation is iterated until the preset maximum number of steps is reached;
[0076] Among them, in the Boltzmann factor, k represents the Boltzmann constant; T represents temperature; ΔE is the difference between the current local minimum energy E and the previous local minimum energy.
[0077] (2) Relaxation / local minimum process:
[0078] In the energy minimization process of each configuration, the first two steps of structural relaxation are performed by DFT calculation, and the on-the-fly MLFF is updated according to the obtained DFT data; then the remaining 15 steps of relaxation are quickly completed using on-the-fly MLFF; in this way, the energy minimization process can be carried out quickly with the assistance of machine learning. The maximum number of relaxation steps allowed for each energy minimization is set to 15. It has been verified that the convergence of most perturbed configurations is basically within 15 steps.
[0079] (3) The recommendation system determines the jump direction:
[0080] The direction of the "jump" is specified by a recommendation system. For each "jump", 100 random pre-"jumps" are performed to obtain a new perturbed configuration, and its energy is predicted using on-the-fly MLFF. The configuration with the lowest predicted energy is used as the actual "jump" direction for the next time. This recommendation system improves the probability of the BH algorithm searching for the lowest local minimum nearby and avoids exploring a large number of unnecessary high-energy occupied configurations through relatively expensive DFT calculations.
[0081] Step 5: Perform DFT calculation verification on the optimized lowest energy occupied configuration.
[0082] The results of RBH operation are compared with those of traditional BH, special quasi-random structure and cluster expansion methods, such as Figure 2 shown. Figure 2 (a) is the energy distribution of the configurations generated by random search and SQS, Figure 2 (b) is the traditional BH method for Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 Three runs of O3 occupancy configuration optimization, Figure 2 (c) is the three runs of the RBH method for optimizing the occupied configuration. It can be seen from the figure that the configuration energy distribution curve of the new method is significantly lower than that of other traditional methods. This shows the effectiveness of RBH.
[0083] In the previous RBH method, the data set for training on-the-fly MLFF contained at most 30 configuration data. The present invention expands this to 60 configuration data. The RBH running results are as follows: Figure 3 As shown in the figure, the results of RBH operation are similar to those of 30 configuration data. Whether it is 30 or 60 configuration data, RBH can effectively optimize the occupied configuration. This shows that in the case of a small data set, on-the-fly MLFF can still have a high degree of accuracy and effectively determine the jump direction.
[0084] The effectiveness of on-the-fly MLFF in RBH method is proved. Figure 4 As shown in the figure, in the relaxation section, the prediction of the machine learning system is compared with the DFT calculation results, and the prediction error is 0.31eV, that is, 1meV / atom, which is highly accurate. In the "jump" section, the prediction of the machine learning recommendation system is compared with the relaxation energy. Figure 5 As shown, more than one-third of the data are within the 0.3 eV error range, which is only a recommended system with acceptable accuracy.
[0085] The initial random site-occupied configuration is compared with the lowest energy site-occupied configuration confirmed by DFT. Figure 6 As shown, (a) is the initial random configuration, and (b) is the lowest energy configuration. Cyan represents Ba ions; pink represents Sr ions; green represents Co ions; yellow represents Fe ions; and gray represents O ions. 0.5 Sr 0.5 Co 0.8 Fe 0.2The Ba and Sr ions at the A site in O3 are arranged in layers, which can reduce the bond stress of Ba-O and Sr-O; while the Co and Fe ions at the B site are arranged disorderly, with configurational entropy as the main driving force.
[0086] The above description of the embodiments is to facilitate the understanding and use of the invention by those skilled in the art. It is obvious that those skilled in the art can easily make various modifications to these embodiments and apply the general principles described herein to other embodiments without creative work. Therefore, the present invention is not limited to the above embodiments, and improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the present invention should be within the scope of protection of the present invention.
Claims
1. A basin jumping algorithm based on machine learning recommendation system, characterized in that: The following steps are involved: Step 1: obtaining an initial random placeholder configuration for placeholder configuration optimization; Step 2: Determine the real-time machine learning force field calculation parameters; Step 3: Determine the basin jump parameters, including temperature and jump rules; Step 4: Use the real-time machine learning force field as a recommendation system and introduce the basin jumping algorithm to globally optimize the system's occupancy configuration; Step 5: Perform density functional theory calculations to verify the optimized lowest energy occupied configuration.
2. The basin jumping algorithm based on machine learning recommendation system according to claim 1 is characterized in that: In step 1, the initial random site-occupying configuration includes: structural model size, number of atoms, lattice parameters and atomic category information in the sub-lattice.
3. The basin jumping algorithm based on machine learning recommendation system according to claim 1 is characterized in that: In step 2, the real-time machine learning force field calculation parameters include: the training weight ratio of atomic force and system energy, the atomic distribution of the broadened descriptor, and the number of basis functions of the extended descriptor.
4. The basin jumping algorithm based on machine learning recommendation system according to claim 1 is characterized in that: In step 3, in the process of determining the basin jump parameters, the temperature K is set to 2500K, and the jump rule is set to perform only one position exchange between different cations in the same sublattice for each jump.
5. The basin jumping algorithm based on machine learning recommendation system according to claim 1 is characterized in that: In step 4, the system in which the basin jumping algorithm is introduced to perform global optimization on the system occupancy configuration is Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 O3.
6. The basin jumping algorithm based on machine learning recommendation system according to claim 5 is characterized in that: The Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 The initial random site-occupying configuration of O3 is a 3×4×5 supercell structural model; the initial random site-occupying configuration contains 300 atoms, and its lattice parameters are Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 The A-site sublattice of O3 is occupied by Ba and Sr ions, the B-site sublattice is occupied by Co and Fe ions, and O acts as a "bystander" in the anion sublattice.
7. The basin jumping algorithm based on machine learning recommendation system according to claim 5 or 6, characterized in that: In step 4, the step of globally optimizing the system's placeholder configuration includes: (1) Implementation of basin jumping algorithm: During each jump, two different cations in a sublattice in the occupied configuration are exchanged, and the new configuration is structurally relaxed or locally minimized; If the Boltzmann factor exp(-ΔE / kT) is greater than a random number between (0, 1), the new occupancy configuration will be accepted; If the conditions are not met, the new occupied configuration is rejected and returns to the previous local minimum state, repeating the jump process until the preset maximum number of steps is reached; in, k represents the Boltzmann constant; T represents temperature; ΔE is the difference between the current local minimum energy E and the previous local minimum energy; (2) Relaxation / local minimum process: During the energy minimization process for each configuration, the first two steps of structural relaxation are performed by DFT calculations, while the real-time machine learning force field (on-the-fly MLFF) is updated based on the obtained DFT data; then the remaining 15 relaxation steps are completed quickly using on-the-fly MLFF; this relaxation based on on-the-fly MLFF is effective because the local environment of the atoms changes little; (3) The recommendation system determines the jump direction: Each time a "jump" occurs, 100 random pre-"jumps" are performed to obtain a new perturbation configuration, and its energy is predicted using on-the-flyMLFF. The occupied configuration with the lowest predicted energy is used as the next actual "jump" direction.
8. The basin jumping algorithm based on machine learning recommendation system according to claim 7 is characterized in that: Step five includes: 0.5 Sr 0.5 Co 0.8 Fe 0.2 Density functional theory calculations were performed on the lowest energy occupied configuration of O3, and the lowest energy occupied configuration was visualized.
9. A basin hopping algorithm based on a machine learning recommendation system according to any one of claims 1 to 8 in the perovskite catalyst Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 Application in O3 site occupancy configuration optimization.
Citation Information
Patent Citations
Method for calculating configuration entropy of high-entropy alloy based on atomic placeholder ordering behavior
CN112635000A
Glass hardness prediction method based on squirrel optimization algorithm and machine learning algorithm
CN114373523A
Crystal structure energy searching method, device and equipment and readable medium
CN117352089A
Orbital mixer machine learning method for predicting an electronic structure of an atomic system
US20230359929A1
