Machine Learning Model for Rapid Density of State Prediction

Solution Overview

Problem

Conventional methods for predicting the electronic structure of materials, such as density functional theory (DFT), are inefficient as they require excessive time and resources to achieve high accuracy, making it impractical for rapid estimation of multiple materials.

Innovation Solution

A method using a trained model that converts input data on material elements into a grid image, determines principal component vectors, and generates a density of state graph for each energy level, allowing for rapid and accurate prediction of electronic structures by leveraging pre-input data from similar materials.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If DFT (density functional theory) is used to estimate the DOS of a material, then the accuracy of the electronic structure prediction is improved, but the time required for the prediction increases significantly (about 2-3 days for a single material)

Engineering Contradiction:

Improveaccuracy of electronic structure predictionVSAvoidprediction time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-calculating and storing DOS data for multiple reference materials in a database before actual prediction tasks. When a new material needs prediction, the system quickly retrieves and processes pre-computed data through machine learning models, avoiding the need to perform time-consuming DFT calculations from scratch for each material, thus reducing prediction time while maintaining accuracy

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent uses copying by creating machine learning models that replicate the behavior of expensive DFT calculations. The trained models copy the essential physics and chemistry relationships learned from DFT data, enabling rapid predictions that approximate DFT accuracy without requiring actual DFT computation for each new material

Inventive Principle:

Principle #26Copying

2Measurement precision

If DFT is used to estimate the DOS of multiple materials, then high accuracy is achieved, but the total time and computational cost become excessive

Engineering Contradiction:

Improveaccuracy of DOS estimationVSAvoidrate of material evaluation

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent implements universality by developing a machine learning framework that can predict DOS for multiple different materials using a single trained model. The model is trained on diverse reference materials and can then universally apply to predict DOS for new materials with similar compositions, enabling high-throughput screening of multiple materials at speeds much faster than individual DFT calculations

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent uses copying by creating machine learning models that replicate the behavior of expensive DFT calculations. The trained models copy the essential physics and chemistry relationships learned from DFT data, enabling rapid predictions that approximate DFT accuracy without requiring actual DFT computation for each new material

Inventive Principle:

Principle #26Copying

3Loss of time

If conventional methods are used for rapid prediction, then prediction time is reduced, but the accuracy of the electronic structure prediction becomes low

Engineering Contradiction:

Improveprediction timeVSAvoidaccuracy of electronic structure prediction

Core Design Contradiction:

Loss of timeVSMeasurement precision

Solution Approach 1:

The patent uses copying by creating machine learning models that replicate the behavior of expensive DFT calculations. The trained models copy the essential physics and chemistry relationships learned from DFT data, enabling rapid predictions that approximate DFT accuracy without requiring actual DFT computation for each new material

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent applies parameter changes by transitioning from first-principles DFT calculations with fixed high computational parameters to machine learning models with optimized inference parameters. The ML models use pre-trained parameters from DFT data but operate with much lower computational overhead during prediction, achieving a balance between speed and accuracy

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS11475340B2Method and electronic device for predicting electronic structure of material

Publication Date: 2022.10.18 KOREA INST OF SCI & TECH

AI summary

A method, performed by an electronic device, of predicting an electronic structure of a material includes: receiving input data of a user related to elements constituting the first material; applying the received input data to a trained model for estimating a density of state of the first material; and outputting a first graph indicating the density of state for each energy level of the first material output from the trained model, wherein the trained model is trained to generate the first graph based on pre-input data about a plurality of second materials composed of at least some of the elements constituting the first material and a plurality of second graphs representing the density of state for each energy level of the plurality of second materials.