Estimation and Closed-Loop Control for Rapid Thermal Processing Systems

  • Friedland, Bernard (PI)
  • Kosonocky, W. F. (CoPI)

Project: Research project

Project Details


9312451 Friedland A technique of rapid thermal processing (RTP) has recently been developed which promises to break a bottleneck in semiconductor manafacturing. This new technique involves the use of high power tungsten-halogen lamps to raise the temperature of a semiconductor wafer rapidly to the desired curing temperature. By use of multivariable feedback control it is possible to maintain the wafer temperature within an acceptable tolerance band. Until recently, the only accurate method of performing the required temperature mearsurments was to imbed thermocouples in the wafer. This technique, however, is not practical for use in production. A potential breakthrough in performing the required measurement is based on the use of a multi-wavelength infrared imaging pyrometer recently developed by NJIT's Electronic Imaging Laboratory in conjunction with the David Sarnoff Research Center. This sensor, in principle, is capable of measuring the temperature of the wafer through a sapphire window with the required accuracy. Extraction of temperature data will entail sophisticated data processing, owing to the requirement of estimating both emissivity and temperature at the different wavelengths. Extended Kalman filtering (EKF) is the preferred algorithm. Use of the EKF algorithm, however, impacts on how the control problem is formulated. In particular, it is necessary to use a nonlinear dynamic model of the process in order to capture the correct temperature dependence for estimation of emissivity. Hence, designing a closed loop controller necessitates dealing with a nonlinear control and estimation problem. The proposed investigation will address this problem and the theoretical issues it raises. ***

Effective start/end date9/1/932/28/98


  • National Science Foundation: $234,746.00


Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.