Detecting the health and proper operation of equipment during manufacturing is an important part of maintaining high yields. This paper describes a model-based approach to detecting changes in equipment during manufacturing. Physics-based mathematical models of the electrostatic chuck (ESC) temperature, RF plasma impedance, and chamber pressure were built for a commercial etch system. Each model consists of a small number of “tuning” parameters that relate to specific physical properties of the equipment. For example, for the ESC temperature model, a heat loss parameter describes the thermal coupling between the ESC and the cooled cathode of the ESC. For model-based fingerprinting, these parameters are adjusted for a given set of time series data until the model fits the data. The variations in the model parameters describe variations in the physical properties of the equipment. A “spider-plot” of these model parameters provides a convenient fingerprint of the condition of the equipment, either with time, wafer-to-wafer, or chamber-to-chamber.

To evaluate this approach, a database of 35 process variables, including ESC temperature and heater power, RF power and RF tuning capacitor values, and chamber pressure and valve position were collected at approximately 1-second intervals for eight different chambers over a period of several months – for a total of approximately 64,000 wafers. In addition, etch rate data, leak rate data, and a maintenance log were made available.

error: Content is protected !!