AVS 46th International Symposium
    Manufacturing Science and Technology Group Wednesday Sessions
       Session MS-WeM

Paper MS-WeM1
Low Open Area Endpoint Detection of Plasma Etching Processes

Wednesday, October 27, 1999, 8:20 am, Room 611

Session: Metrology I
Presenter: B.E. Goodlin, Massachusetts Institute of Technology
Authors: B.E. Goodlin, Massachusetts Institute of Technology
H.H. Sawin, Massachusetts Institute of Technology
Correspondent: Click to Email
Accurate determination of endpoint in plasma etching processes is essential to decrease defects due to both incomplete clearing of the etched material and excessive overetch of the underlying material, leading to a loss of dimension control. This is particularly challenging for low open area etches (<1%), where traditional sensors are at the limits of their sensitivities in determining endpoint. In previous work, we have investigated the use of multivariate analysis to improve the signal to noise of optical emission spectroscopy (OES) data. Improvements of signal to noise of well over 1000% was achieved versus the typical univariate endpoint detection mechanisms employed in industry. Nevertheless, in the lowest open area cases(<1%), difficulty was encountered due to the nonstationary time series behavior of the optical emission signals during main etch. This nonstationary behavior greatly obscured the ability to detect endpoint in these cases. In this work, we have explored several techniques by which to remove time series behavior in optical emission signals. After applying appropriate time series models, the revised data was then analyzed using a multivariate Hotelling's T2 method to see whether endpoint could adequately be detected in low open area etches. The results from historical data are very encouraging and further studies are underway to determine the robustness of this technique.