Region based virtual fourier filter

摘要

The present invention includes searching imagery data in order to identify one or more patterned regions on a semiconductor wafer, generating one or more virtual Fourier filter (VFF) working areas, acquiring an initial set of imagery data from the VFF working areas, defining VFF training blocks within the identified patterned regions of the VFF working areas utilizing the initial set of imagery data, wherein each VFF training block is defined to encompass a portion of the identified patterned region displaying a selected repeating pattern, calculating an initial spectrum for each VFF training block utilizing the initial set of imagery data from the VFF training blocks, and generating a VFF for each training block by identifying frequencies of the initial spectrum having maxima in the frequency domain, wherein the VFF is configured to null the magnitude of the initial spectrum at the frequencies identified to display spectral maxima.

主权项

1. A computer implemented method for detecting defects in imagery data acquired from a semiconductor surface, comprising:
performing a pattern search of one or more sets of imagery data in order to identify one or more patterned regions on a semiconductor wafer; generating one or more virtual Fourier filter (VFF) working areas, wherein each VFF working area includes at least one of the one or more identified patterned regions; acquiring an initial set of imagery data from each of the VFF working areas; defining two or more VFF training blocks within each of the one or more VFF working areas utilizing the initial set of imagery data so as to limit an intensity contribution from non-repeating pattern portions of the identified patterned region, wherein a first VFF training block within a selected working area is defined to encompass a first portion of the selected working area including a first identified patterned region of the one or more identified pattern regions including a selected repeating pattern, wherein an additional VFF training block within the selected working area is defined to encompass an additional portion of the selected working area including an additional identified patterned region of the one or more identified pattern regions including the selected repeating pattern; calculating an initial spectrum along at least one dimension for each VFF training block by transforming spatial domain intensity information associated with the two or more VFF training blocks to frequency domain spectral information utilizing the initial set of imagery data from each of the VFF training blocks; generating a VFF for each VFF training block by identifying frequencies of the initial spectrum having spectral maxima in the frequency domain, wherein the VFF is configured to null the magnitude of the initial spectrum at the frequencies of the initial spectrum identified to display spectral maxima; generating one or more filtered spectra for the one or more patterned regions by applying one or more generated VFFs to the initial spectrum of the one or more patterned regions; and creating a filtered set of imagery data for the one or more patterned regions by transforming the one or more generated filtered spectra of the one or more patterned regions to one or more spatial domain imagery data sets.