Linking TCAD and EDA through pattern matching

As the critical dimension in optical lithography shrinks to 90nm and below, determining where the layout is most affected by non-ideal process conditions is increasingly important. In many cases, combinations of local layout geometries that produce or are sensitive to residual effects can be found by locating theoretically problematic configurations of shapes. This dissertation explores the architectural, physical, and algorithmic feasibility of a prototype pattern matching approach as a novel Technology Computer Aided Design (TCAD) tool for linking to Electronic Design Automation (EDA).; The pattern matcher software architecture was created as a standalone Design for Manufacturability (DFM) tool that fits easily into the design flow and can be applied to many areas of lithography and integrated circuit processing. The development was motivated by the need to relate residual lens aberration effects back to the layout design. For this application, the pattern generator first reads a set of Zernike polynomials and takes the inverse Fourier transform (IFT) of the aberrated pupil function in order to generate the pattern bitmap. The pattern matcher then loads the pattern, a user input parameter file, and a multilayer mask layout in CIF or GDSII format. The match factor is computed along each edge and at each corner of interest, and a resulting sorted table of highest match factors for each pattern is output. The matching geometry is extracted for more rigorous process simulators such as SPLAT. The system also supports an interactive graphical display of the layout with pattern images drawn over the match locations.; The key contribution to this thesis is the collection of data structures and algorithms that implement pattern matching. The most efficient matching algorithm uses rectangle and triangle primitives and can efficiently process an entire chip in less than an hour on a standard desktop computer with near perfect scaling on parallel processor machines. This runtime is two orders of magnitude faster than Optical Proximity Correction (OPC). (Abstract shortened by UMI.)...