| For modern semiconductor manufacturing in this era of submicron technology, device size is continuously decreasing as component density increases. As a result of the nature of the inherent variability within sophisticated semiconductor equipment, stringent process and equipment control is required to maximize process yield. When unreliable equipment performance causes operating conditions to vary beyond an acceptable level, overall product quality can be jeopardized. Thus, timely and accurate malfunction diagnosis and prognosis are desirable for success in semiconductor manufacturing.; Malfunction diagnosis consists of two aspects. One is detecting the malfunction to avoid further faulty misprocessing, and the other is identifying the cause of the malfunction to prevent subsequent occurrences. Malfunction prognosis is the assessment of the current state and the prediction of the future state of a tool. Successful prognosis cannot be accomplished unless monitoring and diagnosis are encompassed. Each of the regimes is important for its purpose; however, prognosis becomes realistic and meaningful when they are considered all together.; This dissertation presents fault detection, real-time malfunction diagnosis, and malfunction prognosis using in-situ metrology data. The techniques recommend simultaneous usage of two in-situ metrology techniques for more precise plasma monitoring, provide numerical degree of belief on malfunction evidence, and suggest a methodology for malfunction prognosis, and they are applied to Plasma Therm 700 series reactive ion etching system located in Microelectronics Research Center at Georgia Institute of Technology. |
