| Summary: | 博士 === 國立臺灣大學 === 電子工程學研究所 === 105 === To further shrink layout features in advanced circuit designs, it is necessary to develop next-generation-lithography (NGL) technologies. According to the 2015 international technology roadmap for semiconductors (ITRS), directed self-assembly (DSA), electron beam lithography (EBL), extreme ultraviolet lithography (EUVL), and multiple patterning extensions may push the limits of lithography. Note that the main challenge of EUVL is its source power, which cannot be solved by methods of design for manufacturing. Therefore, this dissertation is focused on DSA, EBL, and multiple patterning. First, in the 2015 ITRS, DSA was recognized as a promising candidate for advanced circuit designs because DSA can have robust patterning capability in nanoscale. In DSA, specialized polymer molecules, called block copolymers (BCPs), can be directed by some guiding topographical features to form high-resolution self-assembled features. One common category of DSA involves using guiding templates to define the rough regions within which high-resolution features should be located. In particular, DSA with guiding templates has shown its capability for cut-mask fabrication in 1-D gridded designs (cut masks are responsible for creating wire patterns for 1-D layouts). However, the overlay accuracy of the self-assembled features and the printability of templates may vary with different topologies of templates. Moreover, two close templates might interfere each other and might not pattern desired cuts. Therefore, cut patterns of 1-D gridded designs might need to be redistributed such that the corresponding fabricated cut masks could pattern DSA guiding templates. Furthermore, traditional single 193i patterning may not be sufficient to create guiding templates for DSA without distortion in a nanoscale circuit design. Therefore, it is necessary to adopt hybrid lithography incorporating double patterning and DSA (using double patterning to create guiding templates for DSA) in cut-mask fabrication. In addition, the two-dimensional directed self-assembly (2D-DSA) technology with a square lattice of topographic features, denoted by posts, is another competitive option for advanced circuit designs. Directed by the posts, specialized BCPs can create two-dimensional metal wires, which can reduce the number of used metal layers. However, the orientations of posts must be determined before creating metal patterns, and some metal patterns have low yields in 2D-DSA. Therefore, the routing rules of 2D-DSA can be much different from those of traditional routing technologies, and thus the routability of 2D-DSA should be addressed in a new methodology. Then, to fabricate guiding topographical features, EBL can be used in high-resolution pattern fabrication. Unlike traditional optical technologies, where light diffracts when passing through the apertures on a photomask, maskless electron beams (e-beams) can be concentrated to the nanometer scale by using electromagnetic or electrostatic lenses. However, high-voltage beams often deposit a considerable amount of heat in a region, causing critical dimension (CD) distortion. To avoid the heating problem, subfield scheduling which reorders a sequence of subfields in the writing process is executed before layout fabrication. As a result, to utilize the benefits of these NGL technologies and address their difficulties, it is necessary to have solutions for a breakthrough. This dissertation proposes new methodologies to apply these NGL technologies. For 2D-DSA, we present the first detailed placement algorithm for 2D-DSA with a routability model based on the post-orientation probability. For DSA with guiding templates, we first propose a linear-time optimal dynamic-programming-based algorithm for a special case of the template guided cut redistribution problem, where there is at most one dummy wire segment on a track. We then extend our algorithm to general cases by applying a bipartite matching algorithm to decompose a general problem. Furthermore, for hybrid lithography, we first develop a linear-time optimal algorithm for a special case, consisting of a limited number of rows, of the template guided cut redistribution problem. Then, we develop a linear-time double-patterning aware partitioning method to decompose a general problem. Moreover, for EBL, to consider longer-range heat dissipation, we model a subfield scheduling problem with blocked region consideration as a constrained max-min m-neighbor traveling salesman problem (called constrained m-nTSP). To solve the problem, we decompose it into subproblems conforming to a special case with points on two parallel lines, solve each of them with a provably good linear-time approximation algorithm, and merge them into a complete scheduling solution. Experimental results show the effectiveness and efficiency of our proposed algorithms.
|
|---|
