Quantum-Computed Training Data for Material Property Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The prediction of material properties, particularly color, using density functional theory (DFT) is resource-intensive and lacks an efficient classical algorithm for solving the electronic Schrödinger Equation, leading to inaccurate approximations.
Innovation Solution
A method utilizing a quantum computer to numerically solve the electronic Schrödinger equation for hypothetical materials, generating a training dataset to train a machine learning model, specifically a neural network, to predict material properties such as band gap, which is then used to determine material colors for security features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If density functional theory (DFT) is used to predict material properties, then the prediction can be obtained using classical computational methods, but the calculation is resource-intensive and lacks efficiency
Solution Approach 1:
The patent replaces classical computational methods (DFT) with quantum computing to solve the electronic Schrödinger equation. This substitution transitions from classical mechanics-based computation to quantum mechanics-based computation, enabling more efficient and accurate material property predictions without excessive resource consumption
Solution Approach 2:
The patent changes the fundamental computational parameters by moving from classical approximation methods to quantum mechanical exact solutions. By using quantum computers to solve the Schrödinger equation with quantum states and operators, the system achieves both higher efficiency and accuracy in predicting material properties
2Reliability
If the electronic Schrödinger equation is solved using classical algorithms, then the computation can be performed, but no efficient classical algorithm exists to find its solution
Solution Approach 1:
The patent substitutes classical computational algorithms with quantum computing algorithms to solve the electronic Schrödinger equation. This replacement enables both exact solutions (improving reliability) and efficient computation (improving productivity) by leveraging quantum mechanical principles native to quantum computers
3Productivity
If approximate DFT implementation is used, then the electronic structure can be calculated, but the density functional is not exactly known and thus may be approximated
Solution Approach 1:
The patent replaces approximate DFT methods with exact quantum mechanical solutions using quantum computers. This substitution eliminates the need for approximating the density functional, providing both feasible computation and high precision in material property predictions
Solution Approach 2:
The patent uses quantum computers to pre-compute accurate material properties and generate training datasets before applying machine learning models. This preliminary exact computation ensures high precision reference data is available for training, improving overall prediction accuracy
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
Disclosed is a method of generating a training dataset for training a machine learning model for prediction of a specific property of a material. The method comprises: repeatedly performing the following: selecting a hypothetical material from a parameter space, the parameter space being defined by a set of material parameters descriptive of materials; using a quantum computer for numerically solving an electronic Schrödinger equation for a quantum system representing the selected material; using the solution of the electronic Schrödinger equation for obtaining the specific property of the selected material; adding an entry to the training dataset, the entry indicating the selected material and a label, wherein the label indicates the obtained specific property.