Golden Pattern Defect Inspection for Semiconductor Wafer Sensitivity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect inspection devices for semiconductor wafers struggle to accurately distinguish between electrical defects and other surface anomalies, such as foreign matter or noise, due to similar brightness differences in SEM images, leading to false defect detection.
Innovation Solution
The implementation of a Golden Pattern (GP) inspection method, where a SEM image of a memory mat is used as a reference to create a teaching image, and subsequent SEM images are compared for defect detection, allowing for the setting of specific defect discrimination criteria to differentiate between intended and unintended defects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If SEM imaging is used to detect electrical defects through brightness differences, then detection sensitivity for electrical defects is improved, but false detection increases due to similar brightness differences from foreign matter and noise
Solution Approach 1:
The patent applies preliminary action by creating a teaching image from a defect-free wafer before actual inspection. This teaching image serves as a reference that captures the normal brightness distribution pattern, enabling the inspection device to distinguish between normal variations (foreign matter, noise) and actual electrical defects by comparing subsequent images against this pre-established baseline
Solution Approach 2:
The patent changes the inspection parameter from absolute brightness threshold comparison to relative brightness pattern matching against a teaching image. By transforming the reference from a fixed threshold to a dynamic teaching image representing normal conditions, the system can adapt to normal variations while maintaining sensitivity to actual defects
2Productivity
If low resolution imaging is used to reduce image data amount for rapid identification, then inspection speed is improved, but defect discrimination capability deteriorates
Solution Approach 1:
The patent applies partial action by performing two-stage inspection: first using low-resolution imaging for rapid defect candidate identification, then applying high-resolution teaching image comparison only to suspected defect regions. This selective application of high resolution maintains inspection speed while improving defect discrimination capability where needed
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 high sensitivity in detecting only the intended defects while minimizing false positives, improving the accuracy of defect identification on semiconductor wafers.
Implementation Method 1
a scanning electron microscope (SEM), and after processing the obtained images, detecting defects present on the circuit patterns. During SEM imaging, a deflected electron beam is scanned across the semiconductor wafer and then detectors acquire secondary electrons and/or electron reflections, both emanated from the wafer, and convert these electrons into image form.
Data Source
AI summary
A conventional pattern inspection, which compares an image to be inspected with a reference image and subjects the resulting difference value to the defect detection using the threshold of defect determination, has difficulty in highly-sensitive inspection. Because defects occur only in specific circuit pattern sections, false reports occur in the conventional pattern inspections which are not based on the position. Disclosed are a defect inspection method and a device thereof which perform a pattern inspection by acquiring a GP image in advance, designating a place to be inspected and a threshold map to the GP image on the GUI, setting the identification reference of the defects, next acquiring the image to be inspected, applying the identification reference to the image to be inspected, and identifying the defects with the identification reference, thereby enabling the highly-sensitive inspection.