Optical 3-D inspection and defect detection in silicon wafer using phase measuring deflectometry and artificial neural network

Abstract:

The technology relates to a method for predicting surface slope and height profiles and detecting, localizing, and classifying 3-D surface defects on silicon wafers. The method includes defining architectures and constructing at least one of: slope-predicting neural network, height-predicting neural network, defect segmentation neural network, and defect classification neural network. The method further includes training a slope-predicting neural network from among the plurality of neural networks. The method further includes training a height-predicting neural network from among the plurality of neural networks using a plurality of x- and y-directional slope maps. The method further includes training a defect segmentation neural network from among the plurality of neural networks using a plurality of height maps. The method further includes training a defect classification neural network from among the plurality of neural networks using a plurality of defect segmentation maps.