Implementation Of Hotspot Clustering Method Based On Improved Tangent Space Distance Metric

With the rapid development of the micro-electronics industry in recent decades, integrated circuit technology has been improving continuously, and the critical dimension keeps scaling down. The process variations also become increasingly severe and lithographic hotspot issues directly lead to a drop in chip yield, with higher expense of design and manufacturing. In order to meet the requirements of design for manufacturing (DFM) and design for yield (DFY), it would greatly improve the yield if the lithographic hotspots problems could be solved in the early stage of IC design. Hotspots classification, as well as hotspots detection and correction, plays a significant role in lithographic hotspots field.This thesis studies on lithographic hotspots classification field, introduces the related theory and research status at home and abroad, and elaborated on improved tangent space (ITS) based distance metric and incremental hotspots clustering method. This thesis describes the work of implementation of this hotspot clustering method using C language, according to the requirement of a commercial lithographic software platform. Then a series of lithographic hotspots cases have been tested with this hotspots classification program to determine some empirical parameter. By comparing the testing result with the original Matlab implementation using the same examples, we could tell the C language implementation is at most 9 times faster than the original Matlab version while with similar accuracy. The C language implementation improves the efficiency, and it’s ready for the integration into commercial lithographic EDA tools.