AI Mask Image Generation with Gradient-Based Optical Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optical proximity correction (OPC) methods for mask fabrication in semiconductor manufacturing rely heavily on user expertise, limiting performance and speed, and are not optimized for diffraction effects during photolithography.
Innovation Solution
Utilizing a first AI model to generate a mask image from a target pattern, modifying it with an activation function, and updating it based on the gradient of a loss function derived from optical simulation by a second AI model to improve accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If general OPC methods are used to create masks by repeating modifications according to user-defined rules, then masks meeting creation regulations can be produced, but performance and speed are limited by user ability to choose optimal rules
Solution Approach 1:
The patent replaces the manual mechanical process of rule selection and iterative modification with an automated AI-based system. The AI model automatically generates and optimizes mask patterns by learning from training data, eliminating the need for users to manually select and apply correction rules, thereby significantly improving both accuracy and fabrication speed
Solution Approach 2:
The patent transforms the discrete rule-based modification approach into a continuous optimization process using AI models. By adjusting model parameters and using gradient-based optimization, the system can continuously refine mask patterns to achieve higher precision without being constrained by predefined rule sets, while the automated process accelerates production
2Reliability
If multiple iterative modifications are performed according to user-defined rules to update mask shape, then mask creation regulations can be met, but the process requires significant user expertise and time
Solution Approach 1:
The patent performs preliminary training of AI models using extensive training data that encodes mask creation regulations and optimal correction patterns. This preliminary action allows the model to automatically comply with regulations during inference without requiring iterative user-guided modifications, significantly reducing the time needed for mask pattern optimization while ensuring regulatory compliance
Solution Approach 2:
The AI-based system enables self-service mask pattern generation by automatically learning from training data and performing optimizations without continuous user intervention. The model independently adjusts mask patterns to meet regulations, eliminating the time-consuming iterative process that requires user expertise and manual rule selection
Data Source
AI summary
A method for generating a mask image may include generating the mask image from a target pattern by using a first artificial intelligence (AI) model, modifying the mask image by using an activation function, calculating a gradient of the activation function by using a gradient of a loss function determined based on a difference between the target pattern and a pattern predicted through an optical simulation the modified mask image performed by a second AI model, and updating the modified mask image based on the gradient of the activation function.