Hybrid Hotspot Detection in IC Layout Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current hotspot detection methods in integrated circuit (IC) design verification, such as pattern matching-based tools and machine learning techniques, are inadequate in detecting previously unseen lithographic hotspot patterns and suffer from high computational complexity, limiting their effectiveness in early design phases.
Innovation Solution
A hybrid hotspot detection technique that combines machine learning classification, pattern matching, and process simulation to identify hotspots by classifying layout patterns into preliminary hotspots, non-hotspots, and potential hotspots, using false positives and negatives from model calibration to refine detection, and employing process simulation for boundary hotspot identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If model-based lithography simulation is used to detect hotspots, then detection accuracy is improved, but computational complexity and runtime increase significantly
Solution Approach 1:
The patent segments the hotspot detection process into multiple stages: initial pattern matching to identify candidate hotspots, followed by targeted lithography simulation only on those candidates. This segmentation allows the system to maintain high detection accuracy through simulation while reducing overall computational complexity by limiting simulation to a small subset of critical patterns.
Solution Approach 2:
The patent performs preliminary pattern matching and classification before lithography simulation to pre-identify potential hotspots. This preliminary action filters out non-critical patterns, so that the computationally intensive simulation is applied only to patterns that are likely to be actual hotspots, thereby reducing total computational burden while maintaining accuracy.
2Productivity
If pattern matching-based tools are used for hotspot detection, then computational efficiency is improved, but detection coverage for unseen patterns deteriorates
Solution Approach 1:
The patent introduces machine learning classifiers as an intermediary between pattern matching and final hotspot identification. The ML model learns from training data to recognize patterns that traditional rule-based matching might miss, thereby improving detection coverage for unseen hotspot patterns while the system maintains computational efficiency through the speed of pattern matching for known patterns.
Solution Approach 2:
The patent creates a composite detection system that combines multiple approaches: traditional pattern matching, machine learning classification, and targeted lithography simulation. This composite approach leverages the strengths of each method—pattern matching for speed, ML for recognizing novel patterns, and simulation for verification—achieving both high efficiency and comprehensive detection coverage.
3Adaptability or versatility
If DRC rules are expanded to cover more two-dimensional patterns, then detection coverage is improved, but rules become overly conservative or too relaxed
Solution Approach 1:
The patent implements feedback mechanisms where lithography simulation results are used to validate and refine the DRC rules and pattern matching criteria. The system learns from simulation outcomes to adjust rule sensitivity, preventing both overly conservative and overly relaxed rule applications. This feedback loop maintains high detection coverage while preserving accuracy by continuously optimizing rule parameters based on actual lithography performance data.
Data Source
AI summary
Aspects of the invention relate to hybrid hotspot detection techniques. The hybrid hotspot detection techniques combine machine learning classification, pattern matching and process simulation. A machine learning model, along with false hotspots and false non-hotspots for pattern matching, is determined based on training patterns. The determined machine learning model is then used to classify patterns in a layout design into three categories: preliminary hotspots, preliminary non-hotspots and potential hotspots. Pattern matching is then employed to identify false positives and false negatives in the first two categories. Process simulation is employed to identify boundary hotspots in the last category.