隨著半導體廠商正式邁入12吋晶圓量產，其投資成本也隨之高漲，在昂貴的設備成本及複雜的製造環境中，當晶圓於生產過程之製程物理特性產生偏移時，如能提高製程參數調整的準確性，便能減少機台的測機時間及測試晶圓的使用量，進而提高機台產能，增加整體的獲利。目前電漿化學氣相沉積（Plasma Enhanced Chemical Vapor Deposition, PECVD）薄膜的製程參數調整，主要是根據工程師現有的實驗數字及經驗加以調整。但由於影響電漿化學氣相沉積薄膜之物理特性有諸多互相關聯性的製程參數，單由實驗數字及經驗加以調整，往往造成預估值與實際值有某種程度的差距；且不同製程工程師的從業年資及經驗，常常會有參數調整次數的不同，造成機台使用率的降低。加以目前半導體產業於製程錯誤偵測及製程控制是採用統計製程管制（Statistical Process Control, SPC）為主要工具，而計量值的 SPC 基本上只允許檢測一個常態分佈的變數，但大部份半導體製程設備的運轉狀況通常具有非平穩性、自關聯性及交互關聯性的特性。因此發展一套製程參數調整的預測模型是有必要性的。本研究利用類神經網路（Artificial Neural Network）及複迴歸分析（Multiple Regression Analysis）發展一套電漿化學氣相沉積之製程參數對薄膜之物理特性的預測模式，並以某半導體代工廠之生產資料比較其預測績效之差異。
經由本研究實證結果，於電漿化學氣相沉積之 Silicon Nitirde 薄膜的物理特性之預測模型，類神經網路較複迴歸分析，有良好的預測效果。當機台參數的調整可能令薄膜的各項物理特性超出調整範圍，即可提出警示，並可供製程工程師執行參數調整的參考方案。

In today’s world, the business environment has become more globalized and the competition has become more dramatic. The semiconductor manufacturing has entered the stage of mass-production of 12-inch technology and thus the investment cost is also increasing. The high equipment cost and complicated manufacturing environment have made the industries to reduce the non-production wafers usage and tool unavailable time by increasing the accuracy of the parameter tuning. Thus, machinery capacity and overall profits are also increasing. Currently, the parameter turning of the Plasma Enhanced Chemical Vapor Deposition (PECVD) process is usually adjusted by the Design of Experimental (DOE) data and experiences of engineers. However, due to the multi-relationship of the physical characteristics of the thin film and the tunable process parameters, the experimental process tuning results in the bias between predict and actual values. The number of process tuning varies by the seniority of the engineers, and thus lower the machine usage. Besides, in semi-conductor industry, Statistical Process Control is the major tool for process detection and control. However, only the normal distribution can be used in the variable control charts. For most semi-conductor equipment, they possess the characteristics of non-stationary, auto-correlated and cross-correlated. Thus it is necessary to develop a predicting model of process parameter tuning to obtain better physical characteristics. This thesis will implement a predictive model of tunable process parameters in PECVD and the physical characteristics of thin-film by using Artificial Neural Network (ANN) and Multiple Regression Analysis (MRA). The real production data from one semiconductor manufacturing company are used to demonstrate predictive performance of the proposed model.
According to the experimental results, the ANN method has better prediction performance than MRA method. This model could provide the alerting message to the decision maker when tuning values could shift to out of specification. Moreover, through MRA results, the engineers can obtain the information on the parameters’ sensitivity and use these values as the reference for process parameter tuning.

中文部份
半導體科技，No.60，www.ssttpro.com.tw，2006年。
張勁燕，半導體製程設備，五南圖書出版股份有限公司，2005年。
許純君，預測的原則與應用，台灣西書出版社，1999年。
羅華強，類神經網路-MATLAB的應用，高立圖書有限公司，2005年。
葉怡成，類神經網路模式應用與實作，儒林圖書有限公司，2004年。
陳順宇，迴歸分析，華泰書局，2000年。

英文部份
Aburto, L., Weber. R., “Improved supply chain management based on hybrid demand forecasts.” Applied Soft Computing, 2005.
Bianchi, L., Jarrett, J., and Hanumara, R. C., “Improving forecasting for telemarketing centers by ARIMA modeling with intervention.” Intermational Journal of Forecasting, 14, 497-504, 1998.
Chen, C. W., Huang, C.C., Lin, Y. Y., Chen, L. C., Chen, K. H., and Su, W. F., “Optical properties and photoconductivity of amorphous silicon carbon nitride thin film and its application for UV detection.” Diamond & Related Materials, 14, 1010-1013, 2005.
Chen, W. C., Lee, H. I., Deng, W.J., and Liu, K. Y., “The implementation of neural network for semiconductor PECVD process.” Expert Systems with Applications, 2006.
Chow, H. K., and Choy, K. M., “Forecasting the global electronics cycle with leading indicator: A Bayesian VAR approach.” International Journal of Forecasting, 22, 301-315, 2006.
Choy, K. L., Lee, W. B., and Lo, V., “An intelligent supplier management tool for benchmarking suppliers in outsource manufacturing.” Expert System with Applications, 22, 213-224, 2002.
Chu, C. W., and Zhang, G. P., “A comparative study of linear and nonlinear models for aggregate retail sales forecasting.” International Journal of Production Economics, 86(3), 217-231, 2003.
Cui, W. Z., Zhu, C. C., and Zhau, H. P., “Prediction of thin film thickness of field emission using wavelet neural networks.” Thin Solid Films, 473, 224-229, 2005.
Davazoglou, D., “Optical absorption threshold of low pressure chemically vapor deposited silicon oxynitride films from SiCl2H2-NH3-N2O mixtures.” Thin Solid Films, 437, 266-271, 2003.
Dia, H., “An object-oriented neural network approach to short-term traffic forecasting.” European Journal of Operational Research, 131, 253-261, 2001.
Favennec, L., Jousseaume, V., Rouessac, V., Fusalba, F., and Durand, J., “Porous extreme low k (Elk) dielectircs using a PECVD porogen approach.” Materials Science in Semiconductor Processing, 7, 277-282, 2004.
Galdikas, A., Mironas, A., Senuliene, D., Strazdiene, V., Setkus, A., and Zelenin, D., “Response time based output of metal oxide gas sensors applied to evaluation of meat freshness with neural signal analysis.” Sensors and Actuators B, 69, 258-265, 2000.
Hsieh, K. L., and Tong, L. I., “Optimization of multiple quality responses involving qualitative and quantitative characteristics in IC manufacturing using neural networks.” Computers in Industry, 46, 1-12, 2001.
Kakiuchi, H., Nakahama, Y., Ohmi, H., Yasutake, K., Yoshii, K., and Mori, Y., “Investigation of deposition characteristics and properties of high-rate deposited silicon nitride films prepared by atmospheric pressure plasma chemical vapor deposition.” Thin Solid Films, 479, 17-23, 2005.
Kang, B. S., Choe, D. H., and Park, S. C., “Intelligent process control in manufacturing industry with sequential processes.” Int. J. Production Economics, 60-61, 583-590, 1999.
Kenyou,G., Canel, C., and Neureuther, B. D., “The impact of lot-sizing on net profits and cycle times in the n-job, m-machine job shop with both discrete and batch processing.” Int. J. Production Economics, 97, 263-278, 2005.
Kim, B., Kim, D. W., and Han, S. S., “Refraction properties of PECVD of silicon nitride film.” Vacuum, 72, 385-392, 2004.
Kim, B., and Kim, S., “Diagnosis of plasma processing equipment using neural network recognition of wavelet-filtered impedance matching.” Microelectronic Engineering, 82, 44-52, 2005.
Ko, Y. D., Kang, H. S., Jeong, M. C., Lee, S. Y., Myoung, J. M., and Yun, I., “Neural network based modeling of PL intensity in PLD-grown ZnO thin films.” Journal of Materials Processing Technology, 159, 159-163, 2005.
Kulkarni, S. S., “The impact of uncertain yield on capacity acquisition in process plant networks.” Mathematical and Computer Modelling, 43, 704-717, 2006.
Langoni, D., Weatherspoon, M. H., Foo, S. Y., and Martinez, H. A., “A speed and accuracy test of backpropagation and RBF neural networks for small-signal models of active devices.” Engineering Applications of Artificial Intelligence, 19, 883-890, 2006.
Lee, C. Y., and Uzsoy, R., “A new dynamic programming algorithm for the parallel machines total weighted completion time problem.” Operations Research Letters, 11, 73-75, 1992
Li, D. C., Lina, Y. S., Tsai, T. I., and Wu, D. C., “Using mega-trend-diffusion and artificial samples in small data set learning for early flexible manufacturing system scheduling knowledge.” Computers & Operations Research, 34, 966-982, 2005.
Liau, L. C., Huang, C. J., Chen, C. C., Huang, C. S., Chen, C.T., Lin, S. C., and Kuo, L. C., “Process modeling and optimization of PECVD silicon nitride coated on silicon solar cell using neural networks.” Solar Energy Materials & Solar Cells, 71, 169-179, 2002.
Medjdoub, M., Courant, J. L., Maher, H., and Post, G., “Inductively coupled plasma -- plasma enhanced chemical vapor deposition silicon nitride for passivation of InP based high electron mobility transistors (HEMTs).” Materials Science and Engineering, B80, 252-256, 2001.
Mahabir, C., Hicks, F., and Fayek, A. R., “Neuro-fuzzy river ice breakup forecasting system.” Cold Regions Science and Technology, 46, 100-112, 2006.
Nazzal, D., Mollaghasemi, M., and Anderson, D., “A simulation-based evaluation of the cost of cycle time reduction in Agere Systems wafer fabrication facility – a case study.” Int. J. Production Economics, 100, 300-313, 2006.
Phillips, J. C., Lucovsky, G., “Minimization of dangling bond defects in hydrogenated silicon nitride dielectrics for thin film transistors ( TFTs ).” Journal of Non-Crystalline Solid, 227-230, 1221-1225, 1998.
Sher, P. J., and Yang, P. Y., “The effects of innovative capabilities and R&D clustering on firm performance: the evidence of Taiwan’s semiconductor industry.” Technovation, 25, 33-43, 2005.
Shetty, G. R., Malki, H., and Chellam, S., “Predicting contaminant removal during municipal drinking water nanofiltration using artificial neural networks.” Journal of Membrane Science, 212, 99-112, 2003.
Su, C. T., and Wong, J., T., “Designing MIMO controller by neuro-traveling particle swarm optimizer approach.” Expert Systems with Applications, 32, 848-855, 2006.
Trybula, W., “Cost of ownership – projecting the future.” Microelectronic Enginering, 83, 614-618, 2006.
Vlachos, D. S., Frngoulis, D. K., and Avaritsiotis, J. N., “An adaptive neural network topology for degradation compensation of thin film tin oxide gas sensors.” Sensors and Actuators, B45, 223-228, 1997.
Wu, M. C., and Chang, W. J., “A short-term capacity trading method for semiconductor fabs with partnership.” Expert Systems with Applications, 2006.
Yea, B., Osaki, T., Sugahara, K., and Konishi, R., “Improvement of concentration-estimation algorithm for inflammable gases utilizing rule-based neural networks.” Sensors and Actuators B, 56, 181-188, 1999.

相關網址
台灣半導體產業協會，http://www.tsia.org.tw/