微影線寬虛擬量測與控制

Virtual Metrology to the Control of Critical Dimension in Lithography Process

張貴安(Kuei-An Chang)

指導教授 ： 張耀仁

中原大學/工學院/機械工程研究所/碩士(2012年)

https://doi.org/10.6840/cycu201200157

摘要

摘要 近年來隨著LCD產業日益發展，面板業已成為台灣一項主要的產業，鑑於CD線寬的文獻上，大部分皆在探討半導體上之應用，關於CD線寬在TFT LCD 上的相關文獻相對較少，因此決定本論文之研究方向以PHOTO CD線寬於TFT LCD Array 製程之應用為主題。 本文研究主要在探討應用配合虛擬量測於TFT-LCD微影技術之CD線寬誤差控制，當TFT迴路元件逐漸將解析度提高與線寬縮小化時，微影技術就成為重要瓶頸。所謂微影技術是將光罩上已設計好的畫素圖案轉移到玻璃基板上，光罩圖案轉移的精度會影響後續製程的進行。尤其微影線寬之重要性，在一次又一次重複曝光中，前一層與後一層畫素圖案疊合時若因CD線寬產生誤差，點燈後可能會有所謂的亮點與暗點。於TFT畫素尺寸不斷放大，線寬誤差之製程容忍度也相對嚴格，當線寬超過誤差容忍度，則層間畫素電路因為線寬誤差導致電路發生短路或斷路而無法通過電性的測試，而造成產能上之損失。 本文研究方法是收集量測機台所得到的量測數據，利用類神經網路對非線性函數的逼近能力與小波分解來濾除干擾雜訊，對於線寬相關可控制參數進行類神經網路訓練，論文中透過廠內VDPP4製程配方補值系統搭配虛擬量測的技術概念來提高預測即時性，並精確的預測掃描式投影機線寬誤差量來進行探討，配合Run-to-run控制方式達到製程狀態監控，建立一套強健又穩定的線寬補值方式，故用虛擬量測方式預測出精確的最佳化參數配方來降低改善線寬誤差量是本文目的，以達到改善製程品質，提昇設備總體效能為目標，並幫助廠內工程師做即時判斷的依據指標。

關鍵字

虛擬量測

並列摘要

Abstract

With the widespread application of LCD, Flat-Panel-Display has become one of the most important industries in Taiwan. Because CD (Critical dimension) is primarily addressed in semiconductor related field, this research will concentrate on its application in TFT LCD Array. This research will focus on how to apply “Virtual Measurement” in Lithography of TFT-LCD to minimize CD-tolerance. With the enhancement in resolution and the narrowing in CD, Lithography has become the bottleneck of the circuit design in TFT. Lithography is the transference of the specific pattern, designed on masks, to the glass-panels. The transfer-accuracy has a big impact on the succeeding processes. CD always plays the most important role in the transfer-accuracy. Exposure will be repeated to achieve the required patterns, inaccurate overlay between layers will lead result in bright and dark points in the final products. More stringency is required as pixel-dimension becomes bigger and bigger. Capacity-loss can be caused by Circuit-Short or Circuit-Open, both of which are resulted from CD-inaccuracy between different pattern-layers.
Based on the data collected from the measurement system after eliminating the interference by using the small wave analysis, this research exploits “Neural Network” to train the controllable parameters of CD-tolerance by its non-linear approximation capability. With process-recipe compensation system (VDPP4) supplemented by Virtual Measurement Methodology, the real-time predictability and CD-tolerance- accuracy-estimation of Scan-type Projector are enhanced. The final purpose of this thesis is to adopt Virtual Measurement Methodology to build a stable and robust compensation method of CD-tolerance, enabling the precise prediction of “Optimal Recipe Parameters”, in order to improve CD-tolerance, process quality, productivity, and to offer a real-time index for engineering judgments.

並列關鍵字

Virtual Metrology

參考文獻

[2] Martinez, Victor M. ; Finn, Karen ; Edgar, Th- omas F, “Adaptive on-line estimation and controller of overlay tool bias”,Processdings of SPIE –The International Society forOpti-cal Engineering,

[3] X. A. Wang ; R. L. Mahajan, Xing A ; Mahajan, R.L, “Artificial neural

network model-based run-to run process controller”,IEEE Transa-ctions

on Components,Packaging,and Manuf-acturing Technology Part C:

Manufacturing, Vol.19,n1,p 19-26,Jan,1996.