Endpoint prediction for polysilicon plasma etch via optical emission interferometry

In semiconductor manufacturing, the completion or “endpoint” of a plasma etch is typically controlled as a timed process, or monitored by use of optical emission spectroscopy. As etch process requirements and complexity increase (sequences of discrete recipe changes during the etch), consistency in timing changes is becoming an important issue. Newly developed full wafer interferometry sensors enable real-time monitoring of the entire wafer surface during an etch. In this article, endpoint prediction (or film thickness estimation) for etching polysilicon is considered. For the etch of a specific film structure, the interferometric signal is cyclical with a known number of cycles. Similar to the phase angle of a cosine function, data points of an interferometric curve can be associated with a linear phase function. If the phase at endpoint is known, in situ film thickness can be obtained by determining the phase of the incoming signal in real-time. Simulations show that the proposed film thickness estimation algorithm has good accuracy in the face of etch rate drift and variation in film structure. We also find that estimation is more robust if shorter wavelength signals are used. Finally, experimental data verifies that remaining film thickness can be estimated within reasonable accuracy, and endpoint effectively predicted for blanket and patterned films. These methods provide the critical information needed to make control decisions (e.g., when to switch to a more selective chemistry) based on reaching a desired known film thickness.