Abstract: A method for determining alignment and critical dimensions of regions formed on a semiconductor structure during one or more process steps includes the steps of defining a pattern A at a first location on the semiconductor device during a process step, defining a second independent pattern B at the first location on the semiconductor structure during another process step, acquiring an image of the combination A and B of both the first and second patterns, filtering that image to attenuate higher spatial frequencies while preserving lower spatial frequencies, and comparing the sign result of the filtered image with the sign result of a stored image of the individual patterns to determine alignment errors. In the preferred embodiment the step of filtering includes taking the Laplacian of Gaussian convolution of the image and saving the sign of the result. The comparison between the filtered image and the stored image uses the correlation function for the filtered images.

Abstract: A method and apparatus for measuring feature dimensions uses selective dark-field illumination to illuminate a target from a single direction at a low angle to the plane of the target. Opposing edges of the target elements are distinguished and captured in separate images. The images are filtered using a Gaussian convolution operator and a Laplacian operator. The signs of the filtered images are correlated at various offsets. The relative displacement of the images which produces the maxium correlation value is used to calculate the average dimension of the target elements.