三圖樣微影技術下之佈局分割方法

Abstract

對小於22 奈米的製程節點，雙圖樣微影技術(double patterning lithography)是目前用來增進可印刷性最受推崇的方法。然而，對於小於15 奈米的製程節點，有許多來自業界的文獻指出部分較複雜且密集的層，如閘層、接觸層和第一金屬層，將必須使用到三圖樣微影技術(triple patterning lithography)。不幸地，直至目前，專注在三圖樣微影技術之佈局分割的研究還相當少。最近才有學者對三圖影微影技術提出第一個系統性的方法。然而，我們觀察到，在三圖樣微影技術下，該演算法可能會遺漏相當多可能的縫合(stitch)位置，因此造成部分其實是可透過縫合解決的衝突(conflict)。在本論文中，我們先指出佈局分割在雙圖樣微影技術下和在三圖樣微影技術下主要的兩個不同之處。根據此兩個不同之處，我們提出一個包含四種削減技術的圖切割的方法，此方法可削減問題規模而不影響解的品質。接著，我們提出一個在分配光罩(mask)時考慮逢合的方法，此方法將會在分配光罩時讓衝突趨向產生在容易插入縫合的地方。由實驗結果可看出我們的圖切割的方法可有效的減少約93%的問題規模，且我們提出的光罩分配方法可減少約38%的衝突。

Double patterning lithography (DPL) has been widely recognized as one of the most promising solutions for the sub-22nm technology node to enhance pattern printability. However, much of the literature from industry states that, for the 15nm technology node and beyond, triple patterning lithography (TPL) will be required for the gate, contact, and metal one layers, which are too complex and dense to be split into two masks. Unfortunately, until now, there is very little research focusing on the layout decomposition for TPL. Recent work presented by Yu et al. proposed the first systematic study on the layout decomposition for TPL. However, we observe that their proposed algorithm may miss several possible stitch locations and result in some conflicts that can be resolved by inserting stitches. In this thesis, we first point out two main differences between DPL and TPL layout decompositions. Based on these two differences, we first present a graph division method consisting of four reduction techniques to reduce the problem size without degrading overall solution quality. Next, we propose a stitch-aware mask assignment algorithm, which is based on a heuristic that finds a mask assignment such that the conflicts among the features in the same mask are more likely to be resolved by inserting stitches. Experimental results show that our graph division method can significantly reduce the problem size by an average of 93%, and that the proposed stitch-aware mask assignment algorithm can achieve around 38% reduction of conflicts compared to a basic mask assignment method.