無損壓縮演算法應用於多電子束直寫系統

Lossless Compression Algorithm for Multiple E-Beam Direct Write Systems

白宇軒(Yu-Hsuan Pai)

指導教授 ： 陳中平

國立臺灣大學/電機資訊學院/電子工程學研究所/碩士(2015年)

https://doi.org/10.6342/NTU.2015.02781

摘要

電子束曝光是一種使用電子束在晶圓表面上製造圖樣的方式，是先進光刻(lithography)技術中極具潛力的一種技術，電子束曝光的精度可以達到次奈米(Sub-nm)級，電子束曝光如果能夠達到和光罩曝光技術能夠競爭的產率，則電子束曝光技術成為新一代主流的顯影技術將指日可待。
超大型積體電路中的電晶體數量及設計複雜度，根據摩爾定律呈指數成長，在生產產品時，將設計好的超大型積體電路資訊傳輸至電子束曝光機臺，交由機臺來進行曝光和生產，所需要的資料傳輸規格也越來越高，甚至是目前的傳輸技術無法達到的量級。
在這篇論文中，我們根據科磊公司(KLA-Tencor Corporation)設計的反射式電子束微影顯像系統(Reflective Electron-beam Lithography System)中使用的5位元灰階點陣圖規格，提出一個有效率運用記憶體的方式，將電路布局轉換成5位元灰階點陣圖；最重要的，運用辭典編碼的概念，改良目前已知灰階點陣圖的壓縮演算法，主要針對兩個目前技術上的瓶頸，第一是資料量的壓縮比，第二是解壓縮的速率，讓我們可以將資料經過壓縮後較容易傳輸至電子束曝光的機臺，並且透過簡易的解壓縮方式取得原本所需的資料，解決目前電子束曝光的困境。
實驗數據顯示出我們提出演算法在主要的兩大重要指標，資料量的壓縮比和解壓縮的速率，相較於2013年國立臺灣大學所發表的LineDiff Entropy演算法，分別提升了10%的壓縮比和7.5倍的解壓縮速率。

關鍵字

電子束 ； 曝光壓縮 ； 壓縮比 ； 解壓縮 ； 灰階點陣圖 ； 辭典編碼

並列摘要

Recently, electron-beam lithography is one of the candidates to draw custom shapes on the surface of wafer. The primary advantage of electron-beam lithography is that it can draw custom patterns with sub-nm resolution. Once the throughput of electron-beam lithography can be competitive with traditional optical lithography, the electron-beam lithography will be the main stream of the lithography.
As the VLSI circuit design getting larger and more complicated. If we want to apply the electron-beam lithography technology for a layout with 26 mm × 33 mm after rasterization with 7-nm pixel size, either we have to compress the bitmap of layout data with compression factor of 329.7 before fabrication and decompress the data inside electron-beam emitters or we need a transmission fiber with transfer bandwidth 105.5 Tbps at least in semi-conductor fabrication.
In this thesis, we proposed a more efficient memory-used algorithm to transform the layout data into a 5-bit gray-level bitmap, which is due to the specification of Reflective Electron Beam Lithography (REBL) system proposed by KLA-Tencor Corporation, and also a dictionary-based algorithm to compress the 5-bit gray-level bitmap. We focus on two of the bottlenecks of the electron-beam lithography. First one is the compression ratio and the other one is decompression rate.
According to the experimental results, our algorithm has achieved an at least overall 10% higher compression factor and at least 7.5x faster decompression rate in comparison with the LineDiff Entropy published in 2013.

並列關鍵字

electron-beam ； lithography ； compression ； compression ratio ； decompression ； 5-bit gray-level ； dictionary-based

參考文獻

[2] Chin-Khai Tang, Ming-Shing Su, and Yi-Chang Lu, “LineDiff Entropy: Lossless Layout Data Compression Scheme for Maskless Lithography Systems,” IEEE Signal Processing Letters, Vol. 20, No. 7, p. 645-648, July 2013.

[4] M. J.Wieland, H. Derks, H. Gupta, T. van de Peut, F.M. Postma, A. H. V. van Veen, and Y. Zhang, “Throughput enhancement technique for MAPPER maskless lithography,” in Proc. SPIE, vol. 7637, Alternative Lithographic Technologies II, doi:10.1117/12.849447, Apr. 2010.

[7] D. Salomon, “A Concise Introduction to Data Compression,” 1st Edition. London, U.K.: Springer-Verlag, 2008, ch. 1, sec. 2.

[9] Hsin-I Liu, Vito Dai, Avideh Zakhor, and Borivoje Nikolić, “Reduced Complexity Compression Algorithms for Direct-Write Maskless Lithography Systems,” Proc. SPIE, Vol. 6151, 61512B, doi:10.1117/12.656844, 2006.

[10] Chi-Hsiang Yeh, E. A. Varvarigos, B. Parhami, "Multilayer VLSI Layout for Interconnection Networks", International Conference on Parallel Processing 2000, pp. 33, 2000.