Algorithm Selector for Optical Mode and Detection Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor wafer inspection methods are inefficient in selecting optimal optical modes and algorithms due to reliance on signal-to-noise values, leading to prolonged processing times and increased human error, without effectively evaluating DOI capture versus nuisance rates during initial selection stages.
Innovation Solution
A system and method that utilize a processor to compare combinations of optical modes and detection algorithms based on capture rate versus nuisance rate, using difference images and attributes to identify optimal combinations for improved defect and nuisance event separation, eliminating the need for sensitivity tuner nuisance event filters and reducing processing time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If signal-to-noise ratio based selection is used for optical mode and algorithm selection, then the selection process is simple, but the selection accuracy is insufficient and does not effectively evaluate DOI capture versus nuisance rate
Solution Approach 1:
The patent changes the selection parameter from signal-to-noise ratio to capture rate versus nuisance rate metrics. The system calculates capture rate as the ratio of DOI events to total events, and nuisance rate as the ratio of nuisance events to total events, providing a more accurate basis for selecting optimal optical modes and algorithms
Solution Approach 2:
The system performs preliminary hot scans and frame-level analysis before final algorithm selection. By evaluating multiple algorithms on hot scan data and comparing their capture and nuisance rates in advance, the system identifies the best algorithm-optical mode combination before actual production inspection
2Measurement precision
If hot scans and STNEF tuning are performed to compare nuisance rates, then the selection accuracy improves, but the processing time increases significantly consuming many hours or even days
Solution Approach 1:
Instead of running complete hot scans with full STNEF tuning, the system performs partial analysis on frame-level data from hot scans. It evaluates capture rate and nuisance rate metrics on subsets of data without requiring complete recipe configuration and full wafer scans, achieving sufficient accuracy with reduced time investment
Solution Approach 2:
The system performs preliminary evaluation of multiple algorithms on hot scan frame data before final selection. By comparing capture and nuisance rates at the frame level using automated metrics, the system identifies promising algorithms without requiring time-consuming full STNEF tuning and complete wafer scans
3Reliability
If multiple algorithms and optical modes are evaluated with complete recipe configuration and STNEF creation, then the selection reliability improves, but the productivity decreases due to the time-consuming process
Solution Approach 1:
The patent replaces manual STNEF tuning and subjective algorithm evaluation with automated computational metrics. The system automatically calculates capture rate, nuisance rate, and separation metrics for each algorithm-optical mode combination, eliminating the need for time-consuming manual tuning while maintaining objective comparison criteria
Solution Approach 2:
The system performs self-evaluation of algorithms by automatically computing performance metrics on hot scan data. The automated metric calculation and comparison process eliminates dependency on manual intervention for algorithm selection, enabling rapid objective assessment without requiring expert tuning time
4Measurement precision
If wafer scans are run to compare algorithm performance, then the measurement accuracy improves, but the tool time is occupied that could otherwise be used for inspection or production
Solution Approach 1:
The system performs preliminary algorithm evaluation using hot scan frame data before actual production inspection. By comparing capture and nuisance rates on preliminary hot scan data, the system selects the optimal algorithm, ensuring that full production wafer scans use the best available algorithm from the start, avoiding the need to occupy tool time for comparative scanning
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables faster and more accurate selection of optimal optical modes and algorithms, reducing processing time and minimizing human error, allowing for more efficient semiconductor manufacturing by using capture versus nuisance rate analysis.
Implementation Method 1
a light source configured to direct a beam of light at a wafer, a detector that collects the beam of light reflected from the wafer
Data Source
AI summary
Based on job dumps for defects of interest and nuisance events for multiple optical modes, detection algorithms, and attributes, the best combination of the aforementioned is identified. Combinations of each of the modes with each of the detection algorithms can be compared for all the defects of interest detected at an offset of zero. Capture rate versus nuisance rate can be determined for one of the attributes in each of the combinations.