奈米技術下考慮光學鄰近效應修正之連線最佳化

Abstract

As devices are packed in ever-closer proximity in the nanometer technology, the distortions of design shapes are getting more and more serious. Distortions introduced by the interference which is the diffraction caused by small design shapes can be regarded as a type of noise in lithographic problems. In this thesis, we present an efficient and effective Optical Proximity Correction (OPC) cost model, and this cost model can predict optimal positions to perturb wire segments for OPC performance optimization. We show that the effects of wire perturbation on interference slack have the property of unimodality or monotone within each perturbation range. Based on these properties, we can dramatically reduce the search space to locate the best position for each wire. Further, we can incrementally update the slacks of a segment or net by recomputing only the perturbed segments because our OPC cost model has the property of superposition. Experimental results show that our perturbation algorithm reduces the average Edge Placement Error (EPE) by 28% and the maximum EPE by 24%, and even has smaller total wirelength and less than 1 minute in the perturbation time performed on a 2.6 GHz AMD Opteron workstation.