The Neural Network Implementation in Pattern Recognition of Semiconductor Etching Process

類神經網路應用於半導體蝕刻製程終點圖形之辨識

10.29977/JCIIE.200607.0001

陳文欽(Wen-Chin Chen)；陳振臺(Chen-Tai Chen)；蔡志弘(Chih-Hung Tsai)；何宗軒(Tsung-Hsuan Ho)

倒傳遞類神經網路 ； 半導體 ； 蝕刻 ； 製程終點圖形 ； 圖形辨識 ； back-propagation neural network ； semiconductor manufacturing ； etching process ； endpoint curve ； curve recognition

工業工程學刊

23卷4期（2006 / 07 / 01）

269 - 279

英文

在半導體製造過程中，蝕刻製程的產品終點圖形(End-Point Curve)可以判定產品在蝕刻製程是否有異常，在生產線上監控此一圖形變化，以避免異常損失的產生及擴大。本研究就目前現況研究一可行之方法，以取代生產線上，需要人工監測及判斷圖形異常的問題。所採用的方法是利用類神經網路圖形分類的特性，其原因為，它具有自我學習的特點，只要圖形採取較佳的z本，透過辨識訓練就可將圖形樣本分類結果學習下來。隨著新加樣本的學習，會更新分類結果加強使系統更具有判斷力。這套系統可改善製程圖形分類時，分類圖形的不確定性，並且可提供生產線，即時有效的停機建議。本文選擇類神經網路中，倒傳遞類神經網路(BPNN; Back-Propagation Neural Network)演算模式來進行研究。並採用晶圓廠蝕刻終點圖形數位化資料來做分析。經隨機挑選3個製程配方(Recipe)每個配方200個樣本訓練學習，再用剩餘的100個樣本對訓練成果做測試，結果發現分類效果良好，其正確率達96%以上。相信運用類神經網路可以發展成為一套頗為理想的蝕刻終點圖形製程圖形失效辨識系統。

In the semiconductor manufacturing, product endpoint curve in the etching process can be used to determine if any abnormality exists in the process. Real-time monitoring of changes in the product endpoint curve in situ can allow early detection of abnormalities and prevent losses from these abnormalities. This research has focused on discovering a feasible way to replace manual monitoring and the curve determination currently used on production lines. In this study, we take advantage of a unique neural network curve classification feature- the self-learning feature. The Back-Propagation Neural Network (BPNN) was used as the backbone of this study. Etching endpoint curve digital data collected from wafer fab was used as analysis database. This study selected 200 samples from each of the 3 randomly chosen process recipes to train the learning process. This study also used the remaining 100 samples as a comparison to our trained results. The classification results in an accuracy of over 96%. Therefore, this system can reduce the uncertainty in the process curve classification and provide in-time and effective suggestions for shutting down machines on the production line.

1. Aghaeizadeh Zoroofi, R.,H. Taketani,S. Tamura, Y. Sato,K. Sekiya(2001).Automated inspection of IC wafer contamination.Pattern Recognition,34(6),1307-1317.
2. Allen, R. L.,R. Moore,M. Whelan.(1996).Application of neural networks to plasma etch end point detection.Journal of vacuum Science & Technology B (Microelectronics and Nanometer Structures),14,498-503.
3. Bhatikar, R.,L. Mahajan(2002).Artificial Neural-Network-Based Diagnosis of CVD Barrel Reactor.IEEE Transactions on semiconductor,15,71-78.
4. Cheng, C. S.,C. A. Tseng(1995).Neural network in detecting the change of process mean value and variance.Journal of the Chinese Institute of Industrial Engineers,12(3),215-223.
5. Douglas, H.(2001).300mm and Overall Factory Efficiency (OFE).Software Solutions for 300mm Fabs, Hsinchu,,13(14)
6. Fogel, B.(1991).An Information Criterion for Optimal Neural Network Selection.IEEE Transaction on Neural Network,2,490-497.
7. Hush, D. R.,B. G. Horne(1993).Progress in supervised neural networks.IEEE Signal Processing Magazine January,8-39.
8. Khaw, J. F. C.,B. S. Lim,Lennie E.N. Lim(1995).Optimal design of neural network using the Taguchi method.Neurocomputing
9. Kim, B.,S. Gary(1997).Real-Time Diagnosis of Semiconductor Manufacturing Equipment: Using a Hybrid Neural Network Expert System.IEEE Transactions on components, packaging, and manufacturing technology-PARTC,20,39-47.
10. Kirschner, H.,R. Hillebrand(2000).Neural networks for HREM image analysis.Information Sciences,129,31-44.
11. Lippmann, R. P.(1987).An introduction to computing with neural nets.IEEE ASSP Magazine,4-22.
12. Looney, C. G.(1997).Pattern Recognition Using Neural Networks, Theory and Algorithms for Engineer and Scientists.
13. Marcoux, P. J.,P. D. Foo(1981).Methods of End Point Detection for Plasma Etching.Solid State Technol,115.
14. Maynard, H.,E. Rietman,J. T. C. Lee,D. Ibbotson. M. Meyyapan,D. J. Economou,S. W. Butler(1995).Plasma etching endpointing by monitoring RF power systems with an artificial neural network.Proceedings of the symposium on process control, Diagnostics, and Modeling in Semiconductor Manufacturing Proceedings of process Control Diagnostics, and Modeling in Semiconductor Manufacturing,189-207.
15. Mundt, R.,M. Mayyapan,D. J. Economou,S. W. Butler(1995).Model based training of a neural network endpoint detector for plasma etch applications.Proceedings of the symposium on process control, Diagnostics, and Modeling in Semiconductor Manufacturing Proceedings of process Control Diagnostics, and Modeling in Semiconductor Manufacturing,178-188.
16. Murata, N.,S. Yoshizawa(1994).Network Information Criterion-Determining the Number of Hidden Units for an Artificial Neural Network Model.IEEE transaction on neural network,5,865-872.
17. Ng, G. S.,H. C. Sim(1998).Recognition of Partially Occluded Objects with Back-Propagation Neural Network.International Journal of Pattern Recognition and Artificial Intelligence,12(5),645-660.
18. Onoda, T.(1995).Neural Network Information Criterion for Optimal Number of Hidden Units.Neural Networks Proceedings., IEEE International Conference,1,275-280.
19. White, D. A.,B. E. Goodlin,A. E. Gower,D. S. Boning,H. Chen,H. H. Sawin,T. J. Dalton.(2000).Low open-area endpoint detection using a PCA-Based T2 statistic and Q statistic on optical emission spectroscopy measurements.IEEE Transactions on Semiconductor Manufacturing,13(2),193-207.