Signal Threshold Modeling for Semiconductor Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for identifying defects in semiconductor IC chip manufacturing using charged particle beam inspection tools face challenges in distinguishing genuine defects from nuisance signals, often requiring trial-and-error to set a signal strength threshold, which can lead to missed defects or inefficient analysis due to overlap in signal strength magnitudes between nuisance and defect signals.
Innovation Solution
A method involving fitting distribution models to create nuisance and defect distribution models, determining a signal strength threshold, and correcting for overlap to accurately select defect candidates, thereby optimizing the capture rate and reducing unnecessary processing of nuisance signals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a high signal strength threshold is set to reduce nuisance signals, then the number of nuisance signals for analysis is reduced, but genuine defects may be missed due to low signal strength
Solution Approach 1:
The patent transforms the fixed threshold approach into a dynamic parameter selection process. By establishing a relationship between capture rate and signal strength threshold, the system can adjust the threshold parameter based on the desired capture rate, converting a static parameter into a controllable variable that balances efficiency and reliability.
Solution Approach 2:
The patent implements feedback through the capture rate metric. By monitoring the capture rate and adjusting the signal strength threshold accordingly, the system creates a closed-loop control mechanism that continuously optimizes the balance between reducing nuisance signals and maintaining defect detection accuracy.
2Reliability
If a low signal strength threshold is set to capture all potential defects, then the capture rate of genuine defects is improved, but a large number of nuisance signals require further analysis
Solution Approach 1:
The patent enables dynamic adjustment of the signal strength threshold parameter based on the desired capture rate. Instead of using a fixed low threshold that processes all potential defects, the system calculates an optimized threshold that achieves the target capture rate while minimizing nuisance signal processing.
Solution Approach 2:
The patent applies partial action by selecting only the necessary portion of signals for detailed analysis. By setting the threshold to achieve a specific capture rate (e.g., 90%), the system processes only the essential subset of signals that meet the criteria, avoiding unnecessary analysis of excessive nuisance signals.
3Ease of manufacture
If trial and error is used to set the signal strength threshold, then the process is simple to implement, but it is time consuming and difficult to verify suitability
Solution Approach 1:
The patent performs preliminary action by pre-establishing the relationship between capture rate and signal strength threshold through distribution modeling. This preparatory work creates a lookup table or formula that allows rapid threshold selection without time-consuming trial and error during actual defect identification processes.
Solution Approach 2:
The patent introduces an intermediary computational model that mediates between the desired capture rate and the actual threshold value. This intermediary layer (the capture rate-threshold relationship model) eliminates the need for direct trial-and-error experimentation, providing a systematic and verifiable method for threshold selection.
4Productivity
If the threshold is set too high, then nuisance signal processing is minimized, but genuine defects with low signal strength are missed
Solution Approach 1:
The patent dynamically adjusts the threshold parameter based on the desired capture rate, ensuring that the threshold is optimized for each specific application. This prevents both excessively high thresholds that miss defects and excessively low thresholds that process unnecessary nuisance signals.
Solution Approach 2:
The patent uses capture rate as feedback to verify that the threshold setting achieves the desired balance. By monitoring whether the actual capture rate meets the target, the system can adjust the threshold to ensure no genuine defects are missed while maintaining processing efficiency.
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 enhances the efficiency of defect identification by setting a suitable signal strength threshold, reducing the risk of missing genuine defects and minimizing unnecessary analysis of nuisance signals, thereby improving the overall yield and throughput in semiconductor manufacturing.
Implementation Method 1
fitting a distribution model to the initial data set to create a nuisance distribution model; fitting a distribution model to the set of defect candidates to create a defect distribution model of detection signals
Implementation Method 2
determining a signal strength threshold dependent on at least the defect distribution model, wherein the determining comprises correcting the defect distribution model, desirably the correcting being suitable for correcting for overlap in magnitude between elements representative of nuisance signals and detection signals
Data Source
AI summary
The embodiments of the present disclosure provide a method of processing data derived from a sample, comprising processing an initial data set of elements derived from a detection by a detector for calibration, the data set comprising elements representing nuisance signals and detection signals. The processing of the initial data set comprising: fitting a distribution model to the initial data set to create a nuisance distribution model; setting a signal strength value, and selecting elements in the initial data set having a magnitude greater than the signal strength value as a set of defect candidates; fitting a distribution model to the set of defect candidates to create a defect distribution model of detection signals; and determining a signal strength threshold dependent on at least the defect distribution model. The determining comprising correcting the defect distribution model.