AVS 49th International Symposium
    Plasma Science Monday Sessions
       Session PS-MoM

Paper PS-MoM10
Endpoint Strategies for Recess Etch Processes in DRAM and eDRAM Applications

Monday, November 4, 2002, 11:20 am, Room C-105

Session: Conductor Etch I
Presenter: V.C. Venugopal, Lam Research Corporation
Authors: J.P. Merceron, Ecole Polytechnique, France
V.C. Venugopal, Lam Research Corporation
A.J. Perry, Lam Research Corporation
A.J. Miller, Lam Research Corporation
Correspondent: Click to Email
Developing a robust and reliable strategy to determine the end point of recess etch processes for DRAM and eDRAM applications presents some unique challenges. These processes involve etching the poly-Si back-filled into trenches in a Si substrate. The tight depth control required necessitates accounting for incoming material variations, mask erosion, and variations in the densities of incoming patterns. In addition, if an optical diagnostic tool such as an interferometer is used, the high aspect ratio structures and low open area lead to low signal levels, low fringe contrast and signal noise. Incoming nitride mask thickness could vary depending on the planarization process used (either CMP or etch) and the device being fabricated. Target depths are usually specified with respect to the bottom of the nitride layer. The starting recess depth (after planarization) is usually not known either. These uncertainties necessitate measurement of the starting nitride thickness as well as the initial recess depth to achieve the required accuracy. We have developed a robust endpoint strategy based on a broadband (UV-VIS-NIR) reflectometer, adapted to provide a high signal-to-noise ratio (SNR), to achieve the desired degree of control for recess etch processes. The broadband reflectance spectrum carries sufficient information to be able to determine the starting parameters of interest. The evolving recess structure causes a discernible modulation of the reflected light from the wafer, especially at short wavelengths (typically <300nm). A robust fringe counting method which accounts for mask erosion and etch rate variations but is insensitive to signal noise is then used to determine the change in depth of the recess relative to its initial value.