Physical Property Data Correction for DFT and VQE Materials Screening
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing materials informatics methods struggle to accurately calculate physical property values for a wide range of compounds due to limitations in calculation methods, particularly in regions where density functional theory (DFT) fails to provide results, limiting the expansion of materials databases for new materials search.
Innovation Solution
A data processing method utilizing a combination of DFT and variational quantum eigensolver (VQE) calculations to generate correction models, enabling the correction and storage of calculated values as true values in a database, thereby expanding the range of compounds with known physical properties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If only DFT calculation method is used, then calculation speed is high, but calculation results cannot be obtained in certain regions (low reliability)
Solution Approach 1:
The patent segments the calculation space into multiple regions based on DFT calculation reliability. A determination unit divides the compound space into a first region where DFT calculations are reliable and a second region where they are unreliable. This segmentation allows the system to apply appropriate calculation methods to each region, maintaining high speed in the first region while ensuring reliability in the second region through alternative methods.
Solution Approach 2:
The patent employs a composite calculation approach by combining DFT calculation results with machine learning prediction results. The correction unit uses machine learning models trained on DFT data to correct and improve the reliability of DFT calculations in the second region. This composite method leverages the speed advantage of DFT while compensating for its reliability limitations through machine learning corrections.
2Measurement precision
If machine learning is used to correct calculated values, then accuracy is improved, but device complexity increases
Solution Approach 1:
The patent implements preliminary action by pre-training machine learning models using DFT calculation data before actual material screening. The training unit prepares correction models in advance that can quickly adjust DFT results during production use. This preliminary preparation reduces the computational complexity during actual operation, as the heavy lifting of model training is completed beforehand.
Solution Approach 2:
The patent introduces machine learning models as intermediary components between DFT calculations and final material evaluation. The correction unit acts as an intermediary that takes DFT calculated values and transforms them into more accurate predictions through learned correction functions. This intermediary layer simplifies the overall system by providing a standardized correction mechanism that can be applied uniformly across different materials.
Data Source
AI summary
A data processing method acquires a first calculated value by a first calculation method as a physical property value and a second calculated value by a second calculation method as the physical property value, of each of a plurality of first compounds whose true value of the physical property value is known. The method then generates a first correction model and a second correction model for correcting the first calculated value and the second calculated value, respectively, to the true value. The method stores, in a database, corrected values corrected by the first and second correction models of calculated values as the true values, where the calculated values are acquired by the first and second calculation methods as a physical property value of a second compound whose true value of the physical property value is unknown.