Optical Proximity Correction Recipe Automation for Pattern Fidelity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The limitations of traditional optical lithography in semiconductor manufacturing, particularly in achieving precise and complex integrated circuit patterns, necessitate advanced computational techniques for optical proximity correction (OPC) to ensure high fidelity in photolithographic processes.
Innovation Solution
An automated system utilizing reinforcement learning (RL) agents and multi-modal large language models (LLMs) to generate OPC recipes, optimizing fragment points and edge placement error (EPE) measurement points, constructing decision trees for spatial reasoning, and generating photomasks for semiconductor wafers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional optical lithography is used for semiconductor manufacturing, then the manufacturing process is simpler, but the manufacturing precision deteriorates at nanometer scale geometries
Solution Approach 1:
The system enables automated OPC recipe generation through reinforcement learning agents that autonomously optimize correction parameters without requiring manual expert intervention. The AI model learns from training data and automatically generates optimized OPC recipes for different semiconductor patterns, making the system self-sufficient in recipe development.
Solution Approach 2:
The patent replaces manual expert-based OPC recipe development with an automated AI-driven system. The reinforcement learning model substitutes the mechanical process of trial-and-error optimization by engineers with intelligent algorithms that automatically learn and optimize correction parameters through training and inference.
2Productivity
If manual OPC recipe development is used, then the system is easier to operate, but the productivity deteriorates due to time-consuming trial and error
Solution Approach 1:
The system performs preliminary action by pre-training the reinforcement learning model on extensive training data before actual OPC recipe generation. This pre-training phase prepares the model in advance, enabling it to quickly generate optimized recipes during inference without requiring time-consuming trial and error during production use.
Solution Approach 2:
The reinforcement learning model incorporates feedback mechanisms where the agent receives rewards or penalties based on the quality of generated OPC recipes. This feedback loop allows the model to continuously learn and improve, adjusting its strategy to minimize edge placement errors and optimize correction effectiveness over time.
3Manufacturing precision
If advanced computational techniques are applied, then the manufacturing precision improves, but the device complexity increases
Solution Approach 1:
The patent introduces an intermediary layer between the lithography process and the OPC recipe generation. The reinforcement learning model acts as an intelligent intermediary that translates design requirements into optimized correction parameters, bridging the gap between simple input specifications and complex correction algorithms without exposing the full computational complexity to users.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Reduces key error metrics by over 10% without increasing runtime, streamlining the OPC recipe development process, and enhancing the efficiency and effectiveness of computational lithography in semiconductor manufacturing.
Implementation Method 1
OPE) that emerge from diffraction and interference during the photolithographic process
Implementation Method 2
OPE) that emerge from diffraction and interference during the photolithographic process
Data Source
AI summary
Systems and methods are provided for automated generation of optical proximity correction (OPC) recipes for producing photomasks for patterning semiconductor wafers, thereby improving the overall efficiency and effectiveness of computational lithography in modern semiconductor manufacturing. According to at least one embodiment, an OPC recipe is generated by a two-stage process that includes a reinforcement learning (RL) stage, for generating OPC actions for representative design patterns, and a large language model (LLM) stage, for generating an OPC recipe based on the OPC actions provided by the RL stage.