Double patterning has gained prominence as the most likely methodology to help keep Moore's law going towards 22nm ½ pitch lithography. However, most designs cannot be blindly shrunk to run using only two patterning layers and a variety of constraints must be imposed on designs to allow for correct decomposition. These constraints are more onerous for the contact layer than for line/space patterns because they more easily form odd cycles on the 2D plane, which cannot be broken using polygon cutting. As this can adversely limit packing density, especially in bit cells, a triple patterning decomposition capability could be attractive for the contact layer. Pattern decomposition for contacts can be likened to coloring a map where minimum spaces between contacts are replaced with borders. It is well known that 4 colors can color any map [5,6], but it is an NP-complete problem to compute the minimum number of colors needed to color any given map [7]. This should place an upper limit on the scalability of any algorithm able to color large networks. A variety of test patterns that are known 3-colorable are needed to compare suitable algorithms. It has been proved that a set of aperiodic tiling known as "Penrose Tiles" is 3-colorable [8]. This paper compares the scalability of different coloring algorithms using a variety of contact patterns based on Penrose Tiles.
Sub-resolution assist features (SRAFs) are an important tool for improving through-process robustness of advanced lithographic processes. Assist features have generally been placed and adjusted according to heuristic rules. The complexity of these rules increases rapidly with shrinking features size requiring more wafer data for calibration and more effort on the part of engineers. For advanced nodes, a model-based approach may better account for the variety of two-dimensional geometries and reduce substantially the amount of user effort required for effective SRAF placement.There are many ways in which model-based methods can be used to improve the effectiveness of assist features; we investigate several here. In the investigations described here, process window models may be employed to: 1) derive optimal rules for initial AF placement in a rule-based process, 2) resolve mask rule violations in optimal ways, and 3) make post-placement corrections of mask sites with poor behavior. In addition, we discuss a method for replacing an initial rule-based assist feature placement with a model-based placement which can consider the local two-dimensional geometry.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.