Deep Learning Process Proximity Correction for Semiconductor Pattern Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor manufacturing processes face challenges in achieving reliable process proximity correction, particularly due to the limitations of traditional optical proximity correction methods which can result in shape deformation and etching skew issues during the photolithography process.
Innovation Solution
A deep learning-based process proximity correction method is introduced, which involves generating predictive models to correct layouts for After Development Inspection and performing deep learning-based and optical proximity corrections to improve pattern accuracy and reliability in semiconductor device manufacturing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional optical proximity correction (OPC) is used, then pattern correction can be achieved, but shape deformation and etching skew issues occur during photolithography
Solution Approach 1:
The patent changes the correction parameters by introducing process-specific parameters (etching conditions, deposition thickness, material properties) to the traditional OPC parameters. This allows the correction model to account for process variations that cause shape deformation and etching skew, thereby improving both pattern correction accuracy and shape consistency.
Solution Approach 2:
The patent performs process proximity correction in advance before the actual photolithography process. By predicting and correcting for process-induced deformations (etching skew, deposition effects) beforehand, the system prevents shape deformation issues from occurring during manufacturing, thus improving reliability while maintaining correction accuracy.
2Reliability
If deep learning-based process proximity correction is implemented, then reliability is improved, but computational complexity increases
Solution Approach 1:
The patent segments the correction process into distinct stages: optical proximity correction, process proximity correction, and predictive modeling. Each stage handles specific aspects of the correction, allowing the complex deep learning model to be broken down into manageable components that can be trained and applied separately, reducing overall computational complexity.
Solution Approach 2:
The patent uses predictive modeling to create virtual copies of the manufacturing process outcomes. By training deep learning models on simulated process data, the system can predict actual process results without repeatedly running complex physical simulations, thereby improving reliability while reducing computational burden during production.
3Manufacturing precision
If multiple correction steps (DL-PPC and OPC) are performed, then manufacturing precision is improved, but processing time increases
Solution Approach 1:
The patent implements a continuous correction workflow where deep learning-based process proximity correction and traditional optical proximity correction are performed in sequence without interruption. The output of one correction step directly feeds into the next, maintaining continuous useful action and minimizing idle time between corrections, thus improving precision while managing processing time efficiently.
Solution Approach 2:
The patent performs the computationally intensive deep learning-based process proximity correction as a preliminary step before the final optical proximity correction. By completing the more complex corrections first, the system establishes a corrected baseline that reduces the amount of additional correction needed in subsequent steps, thereby improving overall precision while reducing total processing time.
Data Source
AI summary
A deep learning-based process proximity correction method includes receiving a first layout associated with an After Cleaning Inspection (ACI), the first layout including a plurality of patterns associated with manufacturing a semiconductor device, generating a predictive model based on the plurality of patterns, through deep learning, generating a layout associated with an After Development Inspection (ADI) by correcting the first layout, and predicting an ACI using the layout of ADI, through the predictive model.