Cell-to-Cell Comparison for Reliable Semiconductor Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing defect inspection methods in semiconductor devices inaccurately estimate noise, leading to unreliable defect detection due to noise factors like pattern discoloration or lightening, which affects the semiconductor yield.
Innovation Solution
A cell-to-cell comparison method that compares the inspection area of an inspection target unit cell with corresponding areas of multiple unit cells using a two-dimensional vector arrangement, determining pixel gray levels and deviations from an average with a specified threshold, thereby improving defect detection reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If noise estimation is performed to distinguish defect gray levels from noise gray levels, then defect detection capability is improved, but measurement precision deteriorates due to inaccurate noise estimation
Solution Approach 1:
The patent segments the inspection area into multiple unit cells, each containing pixels at corresponding positions. By dividing the image into reusable unit cells and comparing pixels across multiple segments, the method eliminates the need for noise estimation while improving defect detection reliability through statistical comparison.
Solution Approach 2:
The patent creates multiple copies of unit cells from the same physical location on the wafer by reusing the captured image data. Instead of capturing multiple images (which would require alignment), the system generates virtual copies by dividing and recombining pixel data from unit cells, enabling statistical analysis without additional measurement errors.
2Reliability
If multiple images are captured to perform cell-to-cell comparison, then defect detection reliability is improved, but loss of time increases due to multiple image capture and alignment processes
Solution Approach 1:
The patent performs preliminary division of the captured image into unit cells and stores them for later reuse. By pre-processing the image data into reusable unit cell segments, the system eliminates the need for time-consuming alignment operations when multiple comparisons are needed, as all unit cells are already extracted and organized from the single captured image.
Solution Approach 2:
The patent makes the captured image data universally reusable by extracting unit cells that can be applied to multiple comparison operations. A single captured image serves multiple functions: it can be divided into multiple unit cells, each of which can be compared with others from the same image, eliminating the need for multiple separate image captures and their associated alignment procedures.
Data Source
AI summary
A cell-to-cell comparison method for inspecting a defect in a subject in which unit cells that may be expressed by a two-dimensional (2D) vector are periodically arranged according to an embodiment of the present invention includes: obtaining an image of the subject including an array of the unit cells; generating a tile that is a set of pixels corresponding to the same position in a plurality of unit cells within the array in the image; determining a gray level of each of a plurality of pixels within the tile; and determining whether the plurality of pixels within the tile are defective on the basis of the determined gray level of each of the plurality of pixels.