Wafer Inspection Defect Sensitivity Using Context Codes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect detection techniques in semiconductor wafer inspection are inefficient, as they rely on a priori information about critical regions, struggle with noise estimation, and fail to effectively group regions for outlier detection, leading to missed defects and excessive nuisance detections.
Innovation Solution
A system that separates design files into regions, generates context codes to encode geometry, aligns wafer images with design files, groups regions by noise levels, calculates thresholds, and determines defect criticality based on context codes, enabling improved sensitivity and coverage without relying on pre-defined patterns of interest.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If design rule based scripting language is used to define inspection regions, then inspection regions can be generated according to rules, but not all critical regions are known a priori and new regions cannot be easily discovered
Solution Approach 1:
The system automatically generates inspection regions by analyzing wafer images and design data without requiring manual definition or a priori knowledge of critical regions. The algorithm self-identifies potential defect locations by comparing actual wafer structures with design specifications, enabling automatic adaptation to new region types.
Solution Approach 2:
The system dynamically adjusts inspection parameters and region definitions based on analyzed wafer data. By changing the approach from fixed rule-based regions to dynamically generated regions based on actual wafer characteristics, the system adapts to different critical regions automatically.
2Measurement precision
If design rule based scripting language measures noise of each region, then noise characteristics can be analyzed, but the noise covers a large variety of contexts and mixes populations with distinct and different noise characteristics
Solution Approach 1:
The system segments the wafer into distinct inspection regions based on design data and image analysis. Each region is analyzed separately for its specific noise characteristics, preventing mixing of different noise populations. This segmentation allows precise noise measurement tailored to each region's unique context.
Solution Approach 2:
The system applies different noise analysis parameters and methods to different regions based on their local characteristics. Each region receives customized noise analysis appropriate to its specific context, rather than applying a uniform noise measurement approach across the entire wafer.
3Reliability
If hot spot care areas are increased to several thousand for better coverage, then more patterns can be covered, but every CA group would need to be tuned independently which is not practical
Solution Approach 1:
The system merges the functions of individual hot spot care area tuning into a unified automated process. By combining region generation, noise analysis, and threshold determination into an integrated algorithm that operates on all regions simultaneously, the system achieves comprehensive coverage without requiring independent tuning of each region.
Solution Approach 2:
The system automatically determines appropriate sensitivity thresholds for all inspection regions through algorithmic analysis of wafer images and design data. This self-service approach eliminates the need for manual independent tuning of each care area group, making the process practical even with thousands of regions.
4Reliability
If a uniformly hot inspection is run to discover weak areas, then a large number of defects can be found, but the resulting number of DBG bins can be excessive and reviewing each bin is too cumbersome
Solution Approach 1:
The system applies different sensitivity levels and inspection strategies to different regions based on their local characteristics and criticality. Rather than uniformly high sensitivity across all regions, the system optimizes sensitivity locally, reducing the number of false positives that require review while maintaining high detection capability where needed.
Solution Approach 2:
The system dynamically adjusts inspection parameters and sensitivity thresholds based on region-specific characteristics analyzed from design data and wafer images. This parameter optimization reduces the number of detected defects to a manageable level while maintaining high reliability for critical regions.
Data Source
AI summary
Criticality of a detected defect can be determined based on context codes. The context codes can be generated for a region, each of which may be part of a die. Noise levels can be used to group context codes. The context codes can be used to automatically classify a range of design contexts present on a die without needing certain information a priori.