Semiconductor Process Window Discovery via Nuisance Filter
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional methods for determining the process window in semiconductor fabrication are lengthy and inefficient, relying on a two-step technique that combines optical inspection with SEM analysis, which limits the ability to perform the technique multiple times due to inefficient sampling.
Innovation Solution
A method that combines optical inspection with SEM analysis using a nuisance filter to classify defects, iteratively training the filter to distinguish between defects of interest and nuisance defects, allowing for the determination of a preliminary, revised, and further revised process window, ultimately generating a report specifying the final process window.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional two-step technique combining optical inspection and SEM analysis is used, then defect identification is performed, but the process is lengthy and uses inefficient sampling
Solution Approach 1:
The patent combines optical inspection and SEM analysis into a unified multi-step process that simultaneously performs defect detection, classification, and process window determination. The nuisance filter training integrates both optical and SEM data to create a comprehensive defect classification system, eliminating the need for separate two-step analysis and reducing overall process time while maintaining accuracy.
Solution Approach 2:
The patent performs preliminary defect classification using optical inspection data before committing to full SEM analysis. The nuisance filter is trained on optical images to pre-identify potential defects of interest, allowing selective and targeted SEM analysis only on regions likely to contain relevant defects. This preliminary sorting action reduces the sampling burden and accelerates the overall process.
2Reliability
If traditional two-step technique is used, then process window determination is achieved, but inefficient sampling limits the ability to perform the technique multiple times
Solution Approach 1:
The patent applies partial action by performing SEM analysis selectively only on regions pre-identified by the trained nuisance filter as potential defects of interest. Rather than analyzing all regions exhaustively, the system performs targeted partial analysis that achieves sufficient reliability for process window determination while consuming fewer resources, thereby enabling multiple iterations.
Solution Approach 2:
The patent implements feedback loops where the nuisance filter is iteratively trained and refined using results from previous analysis cycles. The system learns from accumulated defect data to improve classification accuracy over time, allowing multiple iterations with progressively better performance. Each iteration feeds results back into filter training, enhancing reliability while maintaining productivity.
3Measurement precision
If optical inspection is combined with massive SEM analysis, then defect classification is performed, but the process is lengthy
Solution Approach 1:
The patent applies local quality by directing full SEM analysis resources only to specific local regions identified as containing potential defects of interest. The nuisance filter enables spatially selective analysis where high-resolution SEM is applied locally to suspicious areas rather than uniformly across the entire wafer. This localized approach maintains classification accuracy while dramatically improving throughput by avoiding unnecessary analysis of defect-free regions.
Data Source
AI summary
To evaluate a semiconductor-fabrication process, a semiconductor wafer is obtained that includes die grouped into modulation sets. Each modulation set is fabricated using distinct process parameters. The wafer is optically inspected to identify defects. A nuisance filter is trained to classify the defects as DOI or nuisance defects. Based on results of the training, a first, preliminary process window for the wafer is determined and die structures having DOI are identified in a first group of modulation sets bordering the first process window. The trained nuisance filter is applied to the identified defects to determine a second, revised process window for the wafer. A third, further revised process window for the wafer is determined based on SEM images of specified care areas in one or more modulation sets within the second, revised process window. A report is generated that specifies the third process window.