Neural Model Semiconductor Pattern Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor manufacturing processes face challenges in accurately correcting patterns due to errors caused by diffraction and process variations, which existing methods like optical proximity correction (OPC) and process proximity correction (PPC) do not fully address, especially in terms of data efficiency and cost-effectiveness.
Innovation Solution
A processor-implemented method that identifies and transforms input and output components of semiconductor patterns using a neural model, generating augmented images through operations like removing, scaling, shifting, or rotating transformation targets, and trains the model based on differences between original and result images to improve pattern transformation estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If data augmentation techniques are used to reduce training data requirements, then training cost and time are reduced, but the complexity of the processing system increases
Solution Approach 1:
The system performs preliminary actions by pre-generating augmented training images through various transformations (rotation, scaling, shifting, removal of transformation targets) before the actual neural network training process. This prepares diverse training data in advance, reducing the time needed during model training while the complexity is managed through automated preprocessing pipelines
2Manufacturing precision
If existing OPC and PPC methods are used for pattern correction, then manufacturing process is simplified, but pattern correction accuracy is insufficient due to diffraction and process variations
Solution Approach 1:
The patent replaces traditional mechanical/optical correction methods (OPC, PPC) with a data-driven neural network approach. The neural network learns complex pattern transformations from augmented training images and applies learned corrections, achieving higher accuracy by substituting physics-based models with machine learning-based predictions that can capture non-linear effects
3Measurement precision
If traditional training methods are used with limited data, then processing speed is maintained, but model accuracy deteriorates due to insufficient training samples
Solution Approach 1:
The system creates multiple copies and variations of limited original training images through data augmentation techniques including rotation, scaling, shifting, and removal of transformation targets. This generates diverse training samples from limited source data, improving model accuracy without requiring additional physical manufacturing data collection time
Data Source
AI summary
A processor-implemented method includes: identifying input components of a semiconductor pattern of an original input image from the original input image corresponding to an application target of a process for manufacturing a semiconductor, generating an augmented input image by transforming a transformation target comprising one or more of the input components from the original input image; and executing a neural model for estimating pattern transformation according to the process based on the augmented input image.