Machine Learning Process Proximity Correction for Semiconductor Mask Layouts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor manufacturing processes face challenges in maximizing the process window during photolithography, particularly in achieving precise pattern transfer and correcting for etching-induced deformations, which affects the quality and reliability of semiconductor devices.
Innovation Solution
A machine learning-based process proximity correction (PPC) method that utilizes pattern gauge-based critical dimension measurements, feature extraction, and clustering to generate a prediction model for maximizing the process window, enabling the generation of optimized mask layouts that account for etching-induced deformations and variations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional OPC methods are used for pattern correction, then manufacturing process is simple, but process window is not maximized and etching-induced deformations are not adequately corrected
Solution Approach 1:
The patent introduces an intermediary ML model that acts as a mediator between the layout data and the pattern correction process. The model is trained on measurement data from wafers and serves as a bridge to predict and correct etching-induced deformations, transforming complex correction tasks into predictable adjustments based on learned patterns from training data.
Solution Approach 2:
The patent performs preliminary actions by training the ML model in advance using measurement data from actual wafers. The model learns etching-induced deformation patterns beforehand, and then applies this pre-acquired knowledge to correct layouts before manufacturing, rather than attempting corrections during the actual production process.
2Measurement precision
If measurement data from actual wafers is used for training, then prediction accuracy is improved, but data acquisition time and measurement complexity increase
Solution Approach 1:
The patent creates a virtual copy of the measurement process by training the ML model on collected measurement data. Instead of continuously measuring actual wafers during production, the model learns from a training set and then predicts corrections for new layouts, effectively copying the learning process without requiring continuous physical measurements.
Solution Approach 2:
The system performs self-service by using its own measurement data to train and improve its prediction capabilities. The ML model learns from measurement data collected during normal operation and then uses this self-acquired knowledge to automatically correct future patterns without requiring external intervention or additional measurement time.
3Productivity
If gauge clustering is performed to group similar patterns, then model training efficiency is improved, but computational complexity increases
Solution Approach 1:
The patent segments the diverse set of measurement data into grouped clusters based on similarity. By dividing the data into meaningful groups (clusters of similar patterns), the system can train more efficiently on representative samples from each cluster rather than processing every individual measurement, reducing overall training complexity while maintaining accuracy.
Data Source
AI summary
The present disclosure relates to process proximity correction (PPC) methods based on machine learning (ML), optical proximity correction (OPC) methods, and mask manufacturing methods including the PPC methods. One example PPC method based on ML includes obtaining a pattern gauge-based bottom critical dimension (CD) and obtaining pattern gauge-based features from a first layout, performing a gauge clustering operation of grouping and classifying pattern gauges including similar features, calculating distribution parameters in a skew-normal distribution of the pattern gauge-based bottom CD in each cluster, performing ML between the distribution parameters and a feature in each cluster to generate a prediction ML model, predicting a distribution, a maximum limit, and a minimum limit of the pattern gauge-based bottom CD by using the prediction ML model, generating an after cleaning inspection (ACI) target including a maximum process window, and generating a second layout by performing an development inspection (ADI) retarget operation.