Genetic Algorithm Controller for Crystal Structure Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for searching new crystal structures using genetic algorithms face high calculating costs due to the large number of candidate solutions generated, making it impractical to evaluate all candidates for material properties like magnetism through first-principle calculations.
Innovation Solution
A material generation apparatus and method that employs a genetic algorithm with a neighborhood set generator for mutation and crossing-over operations, and a neural network for predictive value calculation, reducing the number of candidate structures by using structural relaxation and regression models to focus on promising solutions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a genetic algorithm is used to search for new crystal structures, then the search coverage is improved, but the calculating cost excessively increases
Solution Approach 1:
The patent applies partial action by performing first-principle calculations only on selected crystal structures from the genetic algorithm search, rather than evaluating all generated candidates. The regression model predicts properties for most candidates, and only those with promising predicted values undergo expensive first-principle verification, thus achieving thorough search coverage with controlled computational cost
Solution Approach 2:
The patent introduces a regression model as an intermediary between the genetic algorithm and first-principle calculations. This intermediary quickly evaluates crystal structure candidates and filters out unpromising ones, allowing the expensive first-principle calculations to be applied only to selected candidates, thereby resolving the contradiction between comprehensive search and computational cost
2Device complexity
If the number of atoms in crystal structure is fixed, then the search space is reduced, but the ability to discover new materials is limited
Solution Approach 1:
The patent implements dynamic atom number control where the number of atoms in crystal structures is not fixed but can change during the genetic algorithm evolution. The mutation operations include adding or removing atoms, allowing the search to explore crystal structures with varying atom numbers and compositions, thus maintaining both computational tractability and material discovery capability
3Measurement precision
If all candidate solutions are evaluated by first-principle calculation, then the accuracy of property prediction is improved, but the productivity of material search is reduced
Solution Approach 1:
The patent applies partial action by performing accurate first-principle calculations only on a small subset of selected crystal structures rather than all candidates. The regression model performs rapid screening of all candidates, and only those with promising predicted values undergo expensive first-principle verification, achieving both accuracy where needed and efficiency overall
Solution Approach 2:
The regression model serves as an intermediary that provides quick property predictions for all candidates, filtering out unpromising ones before they undergo expensive first-principle calculations. This two-stage evaluation process maintains high accuracy for final selections while dramatically improving overall search productivity
Data Source
AI summary
A genetic algorithm controller that controls respective processes using a genetic algorithm is configured. The processes include generation of a crystal structure of an inorganic material, a mutation operation of a crystal structure, a crossing-over operation of a crystal structure, structural relaxation calculation of a crystal structure, calculation of a predictive value of an objective function, selection and weeding out of a crystal structure based on a predictive value of an objective function, observation of an objective function value of a crystal structure by first-principle calculation, update of a regression model based on a result of observing the objective function value, and end determination for a material generation process.