OPC Model Error Verification via ML Image Blending
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for verifying errors in optical proximity correction (OPC) models in semiconductor processes are inefficient, leading to inaccuracies and increased time due to the need for extensive pattern image comparison across the semiconductor substrate.
Innovation Solution
A method utilizing a machine learning model, specifically generative adversarial networks, to predict errors in OPC models by blending images from layout data, simulation data, and actual pattern images, allowing for swift verification and optimization of OPC models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If extensive pattern image comparison is performed across the semiconductor substrate to verify OPC model errors, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The verification process is segmented into multiple stages: (1) Extract training patterns from layout data and generate corresponding simulation patterns, (2) Train a machine learning model using paired training patterns and simulation patterns, (3) Use the trained model to predict errors on test patterns. This segmentation allows the system to avoid exhaustive comparison across the entire substrate while maintaining verification accuracy through targeted ML-based prediction.
Solution Approach 2:
A machine learning model is introduced as an intermediary between layout data and pattern verification. The ML model learns the mapping between layout patterns and expected pattern outcomes from training data, then uses this learned knowledge to predict verification results on test patterns without requiring direct extensive comparison across the entire semiconductor substrate.
2Manufacturing precision
If traditional OPC model verification methods are used, then manufacturing precision is maintained, but productivity decreases
Solution Approach 1:
The traditional mechanical/exhaustive comparison-based verification system is replaced with a machine learning-based prediction system. Instead of performing comprehensive pattern comparisons across the entire substrate, the ML model predicts verification results based on learned patterns from training data, significantly improving productivity while maintaining manufacturing precision through accurate error prediction.
Solution Approach 2:
The system performs preliminary training using training patterns and simulation patterns before actual verification. This preliminary action allows the ML model to learn the relationship between layout data and pattern outcomes in advance, enabling rapid and accurate verification during production without compromising manufacturing precision.
3Manufacturing precision
If comprehensive pattern comparison is performed to reduce difference between layout data and formed patterns, then manufacturing precision is improved, but device complexity increases
Solution Approach 1:
A machine learning model serves as an intermediary that simplifies the verification system architecture. Instead of implementing complex exhaustive comparison mechanisms, the ML model encapsulates the verification logic, reducing system complexity while maintaining the ability to achieve high layout-pattern conformity through accurate error prediction.
Solution Approach 2:
The verification approach changes from spatial parameter-based exhaustive comparison to parameter-based prediction using ML. By transforming the verification problem into a parameter prediction task (predicting pattern dimensions, shapes, and variations from layout parameters), the system achieves high manufacturing precision with reduced device complexity.
Data Source
AI summary
A method of fabricating a semiconductor device includes generating a mask based on second layout data obtained by applying an OPC model to first layout data and performing a semiconductor process using the mask on a substrate, obtaining a plurality of pattern images by selecting a plurality of sample patterns from the substrate, selecting sample images corresponding to the sample patterns from each of the first layout data, the second layout data, and simulation data obtained by performing a simulation based on the second layout data, generating a plurality of input images corresponding to the sample patterns by blending the sample images corresponding to the sample patterns, respectively, and generating an error prediction model for the OPC model by training a machine learning model using a data set including the input images and the pattern images.