Machine Learning Semiconductor Property Prediction Using Unrelaxed Charge Density
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current ab initio simulations are limited in size and accuracy for predicting semiconductor material properties, especially for large-scale systems, and require significant computational resources, making them inefficient for complex semiconductor device testing.
Innovation Solution
A method using machine learning models trained with ab initio simulation data, specifically employing unrelaxed charge density as a descriptor to predict semiconductor material properties, allowing for faster and more accurate predictions of material properties in larger systems and device testing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ab initio simulation is used to achieve accurate electronic structures, then measurement precision is improved, but device complexity increases and productivity decreases
Solution Approach 1:
The method pre-calculates and stores charge density data from ab initio simulations in a training system before actual device testing. This preliminary action creates a reusable database that eliminates the need to perform computationally intensive ab initio simulations during the actual testing phase, thereby maintaining accuracy while improving productivity.
Solution Approach 2:
The method creates a machine learning model that copies the essential relationships between charge density and material properties from the training system. This virtual copy allows predictions to be made without repeatedly executing the full ab initio simulation, significantly reducing computational complexity while preserving accuracy.
2Measurement precision
If ab initio simulation is used to study material properties, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The method performs ab initio simulations in advance to build a training system with pre-calculated material properties and charge density data. This preliminary computation stores the time-consuming results in a format that can be quickly queried and applied during actual device testing, thereby reducing the time loss in practical applications.
Solution Approach 2:
The method introduces a machine learning model as an intermediary between the ab initio simulation results and the actual device testing. This intermediary translates the complex simulation data into a form that can be rapidly processed, reducing the time required to obtain accurate material property predictions.
3Productivity
If machine learning model is used to predict material properties, then productivity is improved, but measurement precision may deteriorate
Solution Approach 1:
The method performs comprehensive ab initio simulations in advance to build a robust training system that captures accurate relationships between charge density and material properties. This preliminary action ensures that the machine learning model is trained on high-precision data, maintaining measurement precision while enabling fast predictions during actual use.
Solution Approach 2:
The machine learning model creates a virtual copy of the physical relationships governing material properties, learned from the training system. This copy preserves the accuracy of the original ab initio simulations while allowing rapid prediction without the computational overhead, thus maintaining precision while improving productivity.
Data Source
AI summary
Disclosed are methods of predicting semiconductor material properties and methods of testing semiconductor devices using the same. The prediction method comprises preparing a machine learning model that is trained with a training system and using the machine learning model to predict material properties of a target system. The machine learning model is represented as a function of material properties with respect to a descriptor. The descriptor is calculated from unrelaxed charge density (UCD) that is represented by summation of atomic charge density (ACD) of single atoms.