(92a) Process Trend Monitoring Based on Key Sensitive Index: Applications Semiconductor Manufacturing

Authors: 
Yu, C., National Taiwan University
Jeng, J., National Taiwan University
Su, A., National Taiwan University
Huang, H., Department of Chemical Engineering, National Taiwan University


Batch processes play an important role in the production and processing of chemicals, pharmaceutical, and semiconductor devices. Generally, a batch process has finite-duration and, typically, it consists of the following steps: (1) charging the feed, (2) processing sequence, and (3) discharging the product. For chemical process industry (CPI), upon completion of a batch, a range of quality measurements is usually made at the quality control laboratory, often hours later. Nomikos and MacGregor [1] pioneer the use of the large number of trajectory measurements, collected on-line, for process analysis and statistical process control. The concept of projecting the trajectory information down into lower-dimensional latent variable space is realized using multiway principal component analysis (MPCA) [1-3]. Successful applications can be found mostly in chemical process industry and direct extension to fault detection in a metal etcher was explored in [4]. Updated summaries on the modeling, analysis, and statistical process control using MPCA are given in [5-6]. Despite recent progresses batch process analysis using data-based approaches in chemical industry, equation-based modeling remains as the major modeling tool in semiconductor manufacturing industry. The reason for that is, unlike chemical processes, the IC processing has: (1) much shorter (seconds-minutes for a wafer as opposed to minutes-hours for chemicals) and often variable (deliberately adjusted) time duration, and (2) less frequent quality measurements with longer metrology delay. Moreover, not the entire batch trajectory, but some particular processing sequence constitutes the quality determining steps, and this is especially true for crystallization processes in CPI and thermal processes in IC industries. These characteristics make the data-based approaches difficult. In this work, an alternative is sought. Instead of incorporating large number of trajectory data with variable batch time and possibly ?missing? data for some process variables using MPCA, a key sensitive index (KSI) based approach is proposed. From process insight, the key sensitive time-slot (KST) is identified. Next, possible key sensitive process variables (KSV) are chosen, and then index for these variables (key sensitive index, KSI) is computed. Once a KSI is computed for each batch (wafer-to-wafer), autocorrelation is computed as the batch process progresses. If significant autocorrelation is found, a time-series model is established for the selected KSI, if not different KSI is sought. With the time-series model, process trend can thus be forecasted and possible maintenance action can therefore be called for, whenever necessary. This provides dynamical capability for process trend monitoring while maintain the simplicity of single-variate analysis. An IC processing example is used to illustrate the KSI-based approach and results clearly indicate that process trend is well followed using KSI-based time-series model.

References [1] Nomikos, P; MacGregor, J. F., Monitoring of Batch Processes Using Multi-way Principle Component Analysis, AIChE J. 1994, 40, 1361. [2] Kourti, T.; Nomikos, P; MacGregor, J. F., Analysis, Monitoring and fault Diagnosis of Batch Processes Using Multiblock and Multiway PLS, J. Process Control 1995, 5, 277. [3] Louwerse, D. J.; Smilde, A. K., Multivariate Statistical Process Control of Batch Processes Based on Three-way Models, Chem. Eng. Sci. 2000, 55, 1255. [4] Wise, B. M.; Gallagher, N. B.; Butler, S. W.; White, D. D.; Barna, G. G., A Comparison of Principal Component Analysis, Multiway Principal Component Analysis, Trilinear Decomposition and Parallel Factor Analysis for Fault Detection in Semiconductor Etch Process, J. Chemometrics 1999, 13, 379. [5] Kourti, T., Multivariate Dynamic Data Modeling for Analysis and Statistical Process Control of Batch Processes, Start-up and Grade Transitions, J. Chemometrics 2003, 17, 93. [6] Qin, S. J., Statistical Process Monitoring: Basics and beyond. J. Chemomotrics 2003, 17, 480.