A novel methodology for decomposition and stitching optimization

Multiple patterning plays an important role in advanced technology node development. Although the requirements of multiple patterning had been reduced after introducing Extreme Ultraviolet Lithography (EUV) into advanced logic technology, various decomposition techniques still show great flexibility and advantages in many occasions. Along with technology scaling down, variation caused by pattern diversity become more and more critical in extremely compact standard cell design. Patterns requiring decomposition still need to be optimized to meet manufacturing target. Stitching is an efficient methodology while solving color conflicts in decomposition and helps with variation and cost control. By following the recent standard cell design trend, a much stronger correlation is observed between patterns requiring decomposition and technology features, such as CPP, cell height, etc. Instead of extracting the stitch locations from same color spacing errors or hotspots, a technology feature based stitching method is proposed. However, pattern variation still cannot be effectively controlled by applying single step stitching. In this paper, a multiple-step stitching based decomposition and optimization method is proposed to reduce pattern variation impact. The candidates for stitch locations are determined by referring to the key architecture features of the standard cells. To deal with pattern complexity, we offer a novel decomposition solution by considering supplemental candidates for stitch locations without introducing new color conflicts. This method was used to decompose a typical 2D layer in standard cell of advanced technology node. A single step technology feature based stitching was made to solve the min spacing and Edge Placement Error (EPE) issues as the first attempt, the hotspots of certain patterns were observed and the second attempt was made through updating the stitching reference to remove the hotspots. To reduce the risk which might be caused by pattern variety, additional stitching steps were applied to optimize decomposition output. A proposed index was used to describe the pattern variety by considering pattern similarity and the pattern variety had been significantly reduced through decomposition optimization.