Machine Learning Model for OPC Strategy Determination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional optical proximity correction (OPC) methods rely on operator judgment, leading to inaccurate and inefficient defect correction in semiconductor photolithography, as different defects and patterns require tailored strategies, making experience accumulation difficult.
Innovation Solution
An OPC operation method and device utilizing a machine learning model to analyze local area patterns and determine corresponding OPC strategies, eliminating the need for operator judgment and enabling accurate and experience-based correction of mask layouts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If operator judgment is used to determine OPC strategies, then flexibility in handling different defects is maintained, but accuracy and consistency of defect correction deteriorate
Solution Approach 1:
The patent replaces the mechanical system of human operator judgment with an automated machine learning model. The model extracts local area patterns from mask layouts and automatically determines appropriate OPC strategies, eliminating reliance on human expertise while improving accuracy and consistency in defect correction.
Solution Approach 2:
The patent introduces a machine learning model as an intermediary between the mask layout and OPC strategy determination. This intermediary automatically analyzes local area patterns and maps them to appropriate correction strategies, serving as a bridge that transforms manual judgment processes into automated decision-making.
2Adaptability or versatility
If operator judgment is used to determine OPC strategies, then adaptability to different patterns is maintained, but experience accumulation and knowledge transfer deteriorate
Solution Approach 1:
The patent creates a digital copy of operator expertise by training a machine learning model on historical OPC cases and outcomes. This digital copy captures and stores experience knowledge in a structured format that can be consistently applied across different patterns and easily transferred to new situations without relying on individual operator memory or judgment.
Solution Approach 2:
The patent transforms qualitative operator judgment into quantitative parameters by extracting numerical features from local area patterns. This parameterization enables the system to adapt to different patterns through mathematical analysis while systematically accumulating experience data that can be used to refine and improve correction strategies over time.
3Productivity
If manual OPC strategy determination is used, then complexity of the process is reduced, but productivity and efficiency deteriorate
Solution Approach 1:
The patent segments the mask layout into multiple local area patterns around defect hotspots, allowing the machine learning model to focus on analyzing only relevant regions rather than the entire layout. This segmentation improves processing efficiency by reducing the computational scope while maintaining comprehensive defect correction coverage.
Solution Approach 2:
The patent performs preliminary extraction and analysis of local area patterns before applying OPC corrections. By pre-processing and identifying critical regions in advance, the system prepares optimized correction strategies that can be quickly applied, thereby improving overall productivity without requiring complex real-time processing during the actual correction phase.
Data Source
AI summary
An optical proximity correction (OPC) operation method and an OPC operation device are provided. The OPC operation method includes the following steps. A mask layout is obtained. If the mask layout contains at least one defect hotspot, at least one partial area pattern is extracted from the mask layout according to the at least defect hotspot. A machine learning model is used to analyze the local area pattern to obtain at least one OPC strategy. The OPC strategy is implemented to correct the mask layout.