微影疊對之Mix-and-Match控制

Mix-and-Match Control for Lithography Overlay

許志良(Chih-Liang Hsu)

指導教授 ： 張耀仁

中原大學/工學院/機械工程研究所/碩士(2005年)

https://doi.org/10.6840/cycu200500472

摘要

中文摘要

半導體製程中，大部分的IC電路圖形都是藉由光學微影從光罩轉移圖形到晶圓上。準確的圖形定義，將有助於後續各式結構層的形成，在結構層疊置的過程中，會先透過微影設備上的對準動作來確保各式結構層能完整的座落在期望的位置上。但在微影的過程中，常受到曝光機本身、晶圓狀況、量測設備及外在環境等誤差源的影響，而造成結構層在疊置上的誤差，因此在疊對誤差源的歸屬及補償上，便有其需要性。各微影設備本身具有誤差源獨立性的特性，所以在各微影設備間的疊對誤差來源及影響大小也都會有所不同，在Mix-and- Match的情況下，疊對誤差的影響將更為明顯。

徑向基底網路擁有快速學習的優點，其利用函數逼近的方式找出輸入/輸出之間的映射關係。在微影製程的應用上，為了分離晶圓的不可控誤差，因此將其調整為混合式類神經網路，如此可求得微影設備上可調整之誤差項的參數解，並且由此建立個別微影設備上的疊對誤差特徵。之後，再藉由田口式基因演算法選取微影設備上的最佳可控制參數值，讓需調整的微影設備能增加與標準圖形或標準設備之間的疊對誤差相似度，使其在Mix-and-Match的生產環境中，能降低因疊對誤差而產生的良率下降或排程問題。

關鍵字

最佳化 ； 基因演算法 ； 類神經網路 ； 微影疊對 ； 渴望函數 ； 控制

並列摘要

ABSTRACT

In terms of semiconductor manufacture, as most IC circuit patterns are printed from the mask onto the wafer by optical lithography, to define an accurate pattern will play a key role to the formation of structure layers. While stacking the structure layers, the alignment device on the photolithography equipment should be adopted to ensure that each layer is located on the expected location. However, the process of photolithography is often influenced by error sources such as the stepper, the condition of wafer, surveying instruments and other external factors. To avoid the error which may occur during the stacking of structure layers, it is necessary to find out the problem behind the overlay error source as well as the corresponding compensation. Since each photolithography equipment shares the feature of independent error source, the overlay error source and its impact found on the equipment will differ from one to another, especially when it comes to the occasion of Mix-and-Match, the impact of overlay error will be even more obvious.

Radial basis function network shows the benefit of fast learning and the ability to observe the mapping relation between input and output through curve fitting. On the application to photolithography process, the network will be adjusted into mix neural network in order to separate undercontrol item occurred on the wafer. By this adjustment, the network is able to find the parameter solution to the adjustable error terms on the photolithography equipment, and set up overlay error characters on individual equipment. On the following step, the Taguchi-Genetic Algorithm will be adopted to select the optima control value from the photolithography equipment, which will assist to increase the similarity between the overlay error of photolithography equipment and that of Golden pattern or Golden equipment. In this way, yield rate decrease or schedule problems due to overlay error will be reduced under Mix-and-Match process environment.

並列關鍵字

genetic algorithm ； optima ； Mix-and-Match ； neural network ； overlay ； control

參考文獻

17. 趙育漢,“以類神經網路分析微影覆蓋幾何誤差”，中原大學機械所碩士論文，2003

1.D.S Perloff, Four-point electrical measurement technique for characterizing “mask superposition error on semiconductor wafer,”IEEE Journal of Solid-State Circuits, VSC-13, n4, pp.436-444, Aug, 1978.

2.D. MacMillen, and W.D Ryden,“Analysis of image field placement deviations of a 5X microlithography reduction lens,”proceedings of SPIE, Optical Microlithography Technology, pp.78-89, 1982

4.I. Fink, S. Neal, and S. L. James,“Overlay Sample Plan Optimization for the Detection of Higher Order Contribution to Misalignment,”SPIE Integrated Circuit Metrology, Inspection, and process control VIII, V2196, pp. 389-399, 1994.

5.Z. C. Lin and W. J. Wu,“Multiple linear regression analysis of the overlay accuracy model,”IEEE Transactions on Semiconductor Manufacturing, Vol. 12, n2, pp. 229-237, 1999.