Semiconductor Pattern Layout Correction with Random Bias and REEF
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for correcting layout errors in semiconductor manufacturing patterns, such as critical dimension (CD) variations and overlay mismatches, are inefficient and costly due to the need for multiple photomasks and extensive data collection.
Innovation Solution
Applying random biases to target patterns to simulate manufacturing distortions, calculating enhancement factors like Mask Error Enhancement Factor (MEEF) and Retarget Error Enhancement Factor (REEF) using machine learning models, which integrate these factors to predict and mitigate errors, reducing the need for multiple photomasks and extensive data collection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If multiple photomasks are used for error correction, then manufacturing precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent combines multiple correction functions (OPC and PPC) into a single integrated correction model that processes layout data through one photomask. The machine learning model integrates both optical proximity correction and process proximity correction into a unified framework, eliminating the need for separate correction steps and multiple photomasks while maintaining high manufacturing precision.
Solution Approach 2:
The correction model serves multiple functions simultaneously: it performs optical proximity correction, process proximity correction, and predictive error mitigation all through a single machine learning framework. This multi-functional approach allows one photomask to achieve what previously required multiple specialized photomasks.
2Measurement precision
If extensive data collection is performed for correction modeling, then measurement precision is improved, but loss of time and productivity decrease
Solution Approach 1:
The patent performs preliminary correction by predicting layout errors before photomask manufacturing using the trained machine learning model. The model has been pre-trained on historical data, allowing it to quickly predict and correct errors in new layouts without requiring extensive real-time data collection, thus reducing time loss while maintaining precision.
Solution Approach 2:
The machine learning model uses historical measurement data as training copies to learn enhancement factor patterns. Once trained, the model can predict errors for new layouts by copying learned patterns rather than requiring extensive new data collection, significantly reducing time while maintaining measurement precision.
Data Source
AI summary
A method of correcting a layout of a pattern includes: designing a layout of an original ADI target including target patterns; applying a plurality of biases to the target patterns to design a random biased ADI target including biased patterns; manufacturing a first photomask and a second photomask corresponding to the original ADI target and the random biased ADI target, respectively; performing an exposure process and a developing process on a photoresist layer by using the first and second photomasks to form first and second photoresist patterns, respectively; performing an etching process on an etching object layer by using the first and second photoresist patterns to form first and second patterns, respectively; measuring CDs of the first and second patterns to calculate a REEF; generating a PPC model by using the REEF; and performing a PPC by using the PPC model to correct the layout of the original ADI target.