Files in this item



application/pdfTIAN-DISSERTATION-2016.pdf (6MB)
(no description provided)PDF


Title:Layout decomposition for triple patterning lithography
Author(s):Tian, Haitong
Director of Research:Wong, Martin
Doctoral Committee Chair(s):Wong, Martin
Doctoral Committee Member(s):Hwu, Wen-mei; Chen, Deming; Rutenbar, Rob A.
Department / Program:Electrical & Computer Eng
Discipline:Electrical & Computer Engr
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):triple patterning
layout decomposition
graph coloring
standard cell designs
Abstract:Nowadays the semiconductor industry is continuing to advance the limits of physics as the feature size of the chip keeps shrinking. Products of the 22 nm technology node are already available on the market, and there are many ongoing research studies for the 14/10 nm technology nodes and beyond. Due to the physical limitations, the traditional 193 nm immersion lithography is facing huge challenges in fabricating such tiny features. Several types of next-generation lithography techniques have been discussed for years, such as {\em extreme ultra-violet} (EUV) lithography, {\em E-beam direct write}, and {\em block copolymer directed self-assembly} (DSA). However, the source power for EUV is still an unresolved issue. The low throughput of E-beam makes it impractical for massive productions. DSA is still under calibration in research labs and is not ready for massive industrial deployment. Traditionally features are fabricated under single litho exposure. As feature size becomes smaller and smaller, single exposure is no longer adequate in satisfying the quality requirements. {\em Double patterning lithography} (DPL) utilizes two litho exposures to manufacture features on the same layer. Features are assigned to two masks, with each mask going through a separate litho exposure. With one more mask, the effective pitch is doubled, thus greatly enhancing the printing resolution. Therefore, DPL has been widely recognized as a feasible lithography solution in the sub-22 nm technology node. However, as the technology continues to scale down to 14/10 nm and beyond, DPL begins to show its limitations as it introduces a high number of stitches, which increases the manufacturing cost and potentially leads to functional errors of the circuits. {\em Triple pattering lithography} (TPL) uses three masks to print the features on the same layer, which further enhances the printing resolution. It is a natural extension for DPL with three masks available, and it is one of the most promising solutions for the 14/10 nm technology node and beyond. In this thesis, TPL decomposition for standard-cell-based designs is extensively studied. We proposed a polynomial time triple patterning decomposition algorithm which guarantees finding a TPL decomposition if one exists. For complex designs with stitch candidates, our algorithm is able to find a solution with the optimal number of stitches. For standard-cell-based designs, there are additional coloring constraints where the same type of cell should be fabricated following the same pattern. We proposed an algorithm that is guaranteed to find a solution when one exists. The framework of the algorithm is also extended to pattern-based TPL decompositions, where the cost of a decomposition can be minimized given a library of different patterns. The polynomial time TPL algorithm is further optimized in terms of runtime and memory while keeping the solution quality unaffected. We also studied the TPL aware detailed placement problem, where our approach is guaranteed to find a legal detailed placement satisfying TPL coloring constraints as well as minimizing the {\em half-perimeter wire length} (HPWL). Finally, we studied the problem of performance variations due to mask misalignment in {\em multiple patterning decompositions} (MPL). For advanced technology nodes, process variations (mainly mask misalignment) have significant influences on the quality of fabricated circuits, and often lead to unexpected power/timing degenerations. Mask misalignment would complicate the way of simulating timing closure if engineers do not understand the underlying effects of mask misalignment, which only exists in multiple patterning decompositions. We mathematically proved the worst-case scenarios of coupling capacitance incurred by mask misalignment in MPL decompositions. A graph model is proposed which is guaranteed to compute the tight upper bound on the worst-case coupling capacitance of any MPL decompositions for a given layout.
Issue Date:2016-09-19
Rights Information:Copyright 2016 Haitong Tian
Date Available in IDEALS:2017-03-01
Date Deposited:2016-12

This item appears in the following Collection(s)

Item Statistics