Abstract
Accurate detection and classification of wafer defects
constitute an important component of the IC production process
because together they can immediately improve the yield and
also provide information needed for future process
improvements. One class of inspection procedures involves
analyzing surface images. Because of the characteristics of the
design patterns and the irregular size and shape of the defects,
linear processing methods, such as Fourier transform domain
filtering or Sobel edge detection, are not as well suited as
morphological methods for detecting these defects. In this paper,
a newly developed morphological gradient technique using
directional components is applied to the detection and isolation
of wafer defects. The new methods are computationally efficient
and do not rely on a priori knowledge of the specific design
pattern to detect particles, scratches, stains, or missing
pattern areas. The directional components of the morphological
gradient technique allow direction specific edge suppression and
reduce the noise sensitivity. Theoretical analysis and several
examples are used to demonstrate the performance of the
directional morphological gradient methods.