Neural Network for Material Electronic Structure Prediction

Solution Overview

Problem

Conventional methods for predicting electronic structures of materials, such as the Density Functional Theory (DFT) method, are time-consuming and expensive, making it challenging to estimate state densities of multiple materials quickly and accurately.

Innovation Solution

A trained model is used to predict electronic structures by converting user-input data into energy level-by-level state densities, determining principal components, and generating graphs based on pre-calculated data, allowing for fast and accurate estimation of state densities.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the DFT method is used to estimate state density, then the accuracy of the electronic structure prediction is improved, but the time required for estimation increases significantly

Engineering Contradiction:

Improveaccuracy of state density estimationVSAvoidtime required for estimation

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-calculating state densities for numerous materials using the accurate DFT method and storing them in a database before actual use. This allows the system to quickly retrieve and compare pre-computed data during actual predictions, avoiding the need to perform time-consuming DFT calculations for every new material query while maintaining high accuracy through the pre-computed reference data

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent uses copying by creating a comprehensive database of pre-calculated state density data from DFT calculations. Instead of performing original DFT calculations for each query, the system copies and retrieves relevant data from this pre-computed database, significantly reducing calculation time while preserving the accuracy benefits of DFT methodology

Inventive Principle:

Principle #26Copying

2Measurement precision

If the DFT method is used to estimate state density, then the accuracy of the electronic structure prediction is improved, but the cost of estimation increases

Engineering Contradiction:

Improveaccuracy of state density estimationVSAvoidcost of estimation

Core Design Contradiction:

Measurement precisionVSEase of manufacture

Solution Approach 1:

The patent applies preliminary action by pre-calculating state densities for numerous materials using the accurate DFT method and storing them in a database before actual use. This allows the system to quickly retrieve and compare pre-computed data during actual predictions, avoiding the need to perform time-consuming DFT calculations for every new material query while maintaining high accuracy through the pre-computed reference data

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent uses copying by creating a comprehensive database of pre-calculated state density data from DFT calculations. Instead of performing original DFT calculations for each query, the system copies and retrieves relevant data from this pre-computed database, significantly reducing calculation time while preserving the accuracy benefits of DFT methodology

Inventive Principle:

Principle #26Copying

3Productivity

If faster estimation methods are used, then the time required for prediction is reduced, but the accuracy of state density estimation deteriorates

Engineering Contradiction:

Improvespeed of estimationVSAvoidaccuracy of state density estimation

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent introduces an intermediary approach by implementing a neural network model that is trained on accurate DFT calculation data. This neural network serves as an intermediary that translates simple material composition inputs into accurate state density predictions by learning from the pre-computed DFT data, thereby achieving both fast prediction speed and high accuracy without performing direct DFT calculations

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent uses copying by creating a comprehensive database of pre-calculated state density data from DFT calculations. Instead of performing original DFT calculations for each query, the system copies and retrieves relevant data from this pre-computed database, significantly reducing calculation time while preserving the accuracy benefits of DFT methodology

Inventive Principle:

Principle #26Copying

Data Source

PatentUS11113611B2Method and electronic apparatus for predicting electronic structure of material

Publication Date: 2021.09.07 KOREA INST OF SCI & TECH

AI summary

A method of predicting an electronic structure of a material by an electronic apparatus includes receiving user's input data about elements constituting the material; applying the received user's input data to a trained model for estimating a state density of the material; and outputting a first graph representing energy level-by-level state densities of the material output from the trained model, wherein the trained model is trained to generate the first graph based on a plurality of second graphs representing pre-calculated energy level-by-level state densities respectively corresponding to a plurality of pre-input data about elements of various materials and the plurality of pre-input data.