3D defect detection using optical wide-flield microscopy

We report a method to detect signed differences in two similar data sets representing 3-dimensional intensity profiles recorded by optical wide-field microscopes. The signed differences describe missing or unexpected intensity values, defined as defects. In technical applications like wafer and mask inspection, data sets often represent surfaces. The reported method is able to describe the size and position especially in relation to the neighboring surface and is called Three-Dimension-Aberration (TDA)-Technology. To increase the tool performance and to handle different sizes of defects a scaled bottom-up method is implemented and started with high reduced data sets for the search of large defects. Each analysis contains three steps. The first step is a correlation to calculate the displacement vector between the similar data sets. In the second step a new data set is created. The new data set consists of intensity differences. Extreme values in the data set represent the position of defects. By the use of linear and non-linear filters the stability of detection can be improved. If all differences are below a threshold the bottom-up method starts with the next larger scaled data set. In the other case it is assumed that the defect is detected and step three starts with the detection of the convex hull of the defect and the search of the neighboring surface. As a result the defect is described by a parameter set including the relative position. Because of the layered structure of the data set and the bottom-up technique the method is suitable for multicore processor architectures.