A Strategy for Optimizing Random Code Lithography Patterning in 0.18 um Generation Mask ROM

摘要

Among the non-volatize memory family, Mask ROM is the most cost competed product that wildly adopted in data storage application. The Mask ROM based on the 0.18 um technology generation is pushed to production in most of the advanced fabs that use KrF lithography. There are two critical steps in the process flow layers. Most of the existing studies deal with the small geometry poly patterning. The other challenge is to deal with random code implant layer. However, few studies have been done on the related problems. In a random code layer, the pattern merge algorithm is to fill random squares in a chessboard. The approach of merging X and Y will generate hole and post in a wafer. However, the corresponding two patterns (hole and post) are conflicted with each other either from the aerial theory or from the photo resist characteristics. It is difficult to achieve both resolution and PR loss requirements at the same time. However, experiments show that the strategy of applying the pattern merge algorithm with Y merge only is more flexible for lithography. The patterns change from hole and island to hole and trench. In this paper, we show an optimization methodology for Mask ROM random code implant layer (0.26um nominal CD in hole size and 0.5um nominal CD in post size) by using anal image and latent image such as Illumination conditions (numerical aperture and partial coherence), OPC (manual OPC and Automatically OPC), and interaction between antireflection layer and resist We have implemented the proposed strategy on real production wafers for validation. In particular, we will discuss the impacts on the basic process characteristics such as depth of focus, exposure latitude, PR loss, Kla defect inspection, cross-section profile, and electric test data.