AVS 49th International Symposium
    Plasma Science Friday Sessions
       Session PS-FrM

Paper PS-FrM10
Evolution of Line Edge Roughness During Poly-Si Gate Stack Etching

Friday, November 8, 2002, 11:20 am, Room C-103

Session: Plasma Surface Interactions II
Presenter: S. Rauf, Motorola Semiconductor Products Sector
Authors: S. Rauf, Motorola Semiconductor Products Sector
P.J. Stout, Motorola Semiconductor Products Sector
J. Cobb, Motorola Semiconductor Products Sector
Correspondent: Click to Email
As the transistor dimensions in integrated circuits continue to shrink, the influence of gate surface roughness on transistor electrical characteristics is expected to grow. As a consequence, there is considerable need for gate fabrication processes that minimize or mitigate photoresist and poly-Si gate surface roughness. An integrated 2-dimensional (2D) plasma equipment - feature evolution model has been used to study roughness evolution on photoresist and poly-Si gate surfaces during gate stack etching. The model considers a sequence of typical gate etch processes including resist trimming, anti-reflective coating etching, and poly-Si etching. The plasma equipment simulations of the low-pressure inductively coupled plasmas are based on the Hybrid Plasma Equipment Model (developed at the University of Illinois) and include validated plasma chemical mechanisms for the relevant gas mixtures. The 2D string based feature evolution model (BabyBean) has been developed at Motorola, and it considers reactive ion etching, ion assisted physical sputtering, polymer deposition and sputtering, and isotropic etching by neutral radicals. The feature evolution model has been validated by comparing with process experiments. Simulations show that ions are able to remove "rough" protrusions on the photoresist surface through physical sputtering. This is accompanied by changes in etching or polymer deposition within "rough" cavities due to reduction in local neutral flux. Roughness evolution is sensitive to roughness amplitude and frequency.