Holographic Optical Processing For Submicron Defect Detection

Automated defect detection in highly repetitive subject formats such as integrated circuit photomasks has been the subject of numerous papers in the past.1-6 Techniques previously reported generally fall into two basic categories - namely digital and optical processing methods. Digital techniques involve the bit by bit or pixel by pixel comparison of information associated with two subjects. One subject is generally a highly magnified image of a portion of a single die. The other subject is either a similar image of a reference die, such as an adjacent die, or the digital data base used to generate the photomask. Optical processing methods differ fundamentally from digital techniques in that the subject information is manipulated not in the image plane, but in the Fourier Transform plane normally formed by a lens. The optical information processing discussed in this paper is limited to spatial frequency filtering.