ML-Optimized TCAD Simulation Settings for Semiconductor Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
TCAD simulators face challenges in optimizing simulation settings, leading to slow and inconsistent results due to the need for manual intervention, especially when simulating new technologies without prior settings, resulting in increased development time and cost for semiconductor manufacturers.
Innovation Solution
A machine learning-based system that predicts optimized TCAD simulator settings by training on a repository of prior simulations, using a machine learning model to generate predicted simulation settings with associated confidence values, thereby automating the process and improving simulation efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If manual provision of simulation settings modifications is used, then simulation accuracy can be improved, but simulation speed and productivity deteriorate due to slow and inconsistent manual intervention processes
Solution Approach 1:
The system enables self-service by training a machine learning model on historical simulation data and prior settings, allowing the simulator to automatically select and apply optimal settings for new simulations without manual intervention. The model learns from past simulations and autonomously configures meshing, physics models, and numerical parameters based on the input structure and simulation type.
Solution Approach 2:
The system performs preliminary action by pre-training the machine learning model on a repository of prior simulations and their optimal settings before actual simulations are run. This pre-computed knowledge is then rapidly applied to new simulations, eliminating the need for manual settings adjustment during the simulation process itself.
2Reliability
If manual intervention is used to optimize simulation settings, then simulation reliability can be improved, but development time increases due to the iterative manual adjustment process
Solution Approach 1:
The machine learning model autonomously selects simulation settings based on learned patterns from historical data, eliminating reliance on manual expert intervention. This self-service approach ensures consistent, reliable results while dramatically reducing the time required to configure and execute simulations.
Solution Approach 2:
The system incorporates feedback by training the machine learning model on historical simulation results and settings data. The model learns from past successes and failures, continuously improving its ability to predict optimal settings, which enhances both reliability and speed of future simulations.
3Adaptability or versatility
If general purpose TCAD simulators are used to handle variety of structures and conditions, then adaptability is improved, but simulation convergence and accuracy deteriorate due to lack of optimized settings for specific cases
Solution Approach 1:
The system dynamically changes parameters by using the machine learning model to select different simulation settings based on the specific structure and simulation conditions. The model adjusts meshing parameters, physics models, and numerical settings according to the input features, ensuring optimal convergence and accuracy for each unique simulation case while maintaining general-purpose capability.
Solution Approach 2:
The simulator performs self-service by automatically adapting its settings based on the input structure characteristics. The machine learning model analyzes the simulation input and autonomously configures optimal parameters, enabling the general-purpose simulator to achieve specialized-level performance without manual intervention.
Data Source
AI summary
Embodiments provide efficient, robust, and accurate programmatic prediction of optimized TCAD simulator system settings for future simulation executions to be performed by a TCAD simulation system.