ML Calibration of Semiconductor Etching Profile Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing semiconductor device simulation models require perfect calibration of material parameters, which is time-consuming and resource-intensive, particularly in processes like etching, where conventional methods lack efficient optimization techniques.
Innovation Solution
A machine learning-based algorithm using Gaussian process regression (GPR) to calibrate simulation models for semiconductor devices, enabling rapid prediction of etching profiles and sidewall angles with high accuracy, utilizing minimal characterization data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional calibration methods are used for simulation models, then manufacturing precision is improved, but productivity deteriorates due to time-consuming and resource-intensive processes
Solution Approach 1:
The patent creates a digital twin (virtual copy) of the semiconductor device that replicates physical device behavior through machine learning models. This digital copy enables virtual prototyping and calibration without requiring physical device fabrication, thereby maintaining calibration accuracy while dramatically accelerating the prototyping process and reducing resource consumption.
Solution Approach 2:
The patent replaces traditional mechanical/physical calibration processes with machine learning-based computational methods. Instead of physically fabricating and measuring devices for calibration, the system uses ML algorithms to predict device behavior and optimize parameters computationally, substituting physical iteration with intelligent computation.
2Manufacturing precision
If extensive characterization data is collected for calibration, then manufacturing precision is improved, but loss of time and resources increases
Solution Approach 1:
The patent applies partial action by using minimal characterization data rather than exhaustive data collection. The machine learning models are designed to achieve effective calibration with a subset of critical parameters, avoiding the time-consuming process of collecting and processing all possible characterization data while maintaining sufficient model accuracy.
Solution Approach 2:
The patent performs preliminary action by using machine learning to predict device behavior and identify critical calibration parameters before actual device fabrication. This preliminary computational analysis guides the calibration process, reducing the need for extensive experimental characterization and accelerating the overall calibration timeline.
3Reliability
If multiple simulation models are calibrated for multiple performance metrics, then reliability is improved, but device complexity increases
Solution Approach 1:
The patent merges multiple simulation models and performance metrics into a unified machine learning framework. Instead of calibrating separate models for different metrics (etching rate, sidewall angle, etc.), the system uses integrated ML models that can predict multiple outcomes simultaneously, reducing calibration complexity while maintaining or improving overall reliability through consistent parameter optimization.
Solution Approach 2:
The patent creates universal machine learning models that serve multiple functions - predicting various performance metrics (etching rate, sidewall angle, profile shape) from a single calibration process. These multi-functional models eliminate the need for separate calibration procedures for each metric, simplifying the overall system while enhancing reliability through comprehensive performance prediction.
Data Source
AI summary
An machine learning (ML)-based algorithm used for prediction to calibrate simulation models for semiconductor devices. More than one simulation model for a semiconductor device for predicting more than one performance metric is contemplated. According to one embodiment, the predictive model using ML can be used specifically to estimate an etching profile and vertical sidewall angle.