Wafer Defect Detection Using Deep Learning 3D Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting defects on wafers are time-consuming, especially when using highly accurate three-dimensional (3D) inspection techniques like confocal chromatic sensing, and there is a need for a faster method to obtain 3D defect information.
Innovation Solution
A method and system that perform two-dimensional (2D) inspection to generate 2D defect information and estimate 3D defect information using a supervised deep learning process, mapping between 2D and 3D defect information, which reduces the need for time-consuming 3D inspection processes during inference phases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If highly accurate 3D inspection methods such as confocal chromatic sensing are used, then measurement precision is improved, but productivity deteriorates due to time-consuming processing
Solution Approach 1:
The patent creates a mapping model that copies 3D defect information from 2D inspection data. Instead of performing time-consuming 3D inspection for every defect, the system uses a pre-trained mapping model to copy 3D characteristics (height, depth, shape) from corresponding 2D defect features, achieving fast 3D defect information acquisition without actual 3D measurement
Solution Approach 2:
The patent performs preliminary 3D inspection and mapping model training on a subset of defects before actual production inspection. The mapping model is pre-trained using supervised deep learning with paired 2D and 3D defect data, so that during production, 3D defect information can be rapidly estimated without repeating the full 3D inspection process
2Measurement precision
If 3D inspection is performed for all defects, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent applies different inspection strategies to different defect types and locations. The mapping model is trained to provide accurate 3D information for specific defect types where 2D-to-3D mapping is most effective, while potentially using actual 3D inspection only for critical defects that require precise height measurement, optimizing the balance between time and precision locally
Data Source
AI summary
There may be provided a method for determining three dimensional (3D) defect information, the method may include performing a two-dimensional (2D) inspection of an area of a wafer to generate 2D defect information related to defects of the area of the wafer; estimating 3D defect information regarding the defects of the area of the wafer, wherein the estimating is based on the 2D defect information related to defects of the area of the wafer, and a mapping between 2D defect information and 3D defect information, wherein the mapping is generated using a supervised deep learning machine process.