Feedforward learning applied to RTP of semiconductor wafers

Our recent algorithmic results (1994) in feedforward learning control (LC) are applied to the control of semiconductor wafer temperature in a rapid thermal processing (RTP) reactor. Although the first attempt is experimental in nature, the results are very encouraging and deserve serious consideration. Applying this LC approach has resulted in a speed-up of a well-tuned feedback loop by a factor of 8 which amounts to more than 20 seconds saving in one processing step-quite significant for RTP. Additionally, the experiment has demonstrated the applicability of the LC theory in a real-world manufacturing setting. Many repetitive manufacturing tasks are potential applications of this LC procedure