AVS 49th International Symposium
    Plasma Science Wednesday Sessions
       Session PS+MM-WeA

Paper PS+MM-WeA2
Physically Based Modelling of High-Density-Plasma-CVD on the Feature Scale

Wednesday, November 6, 2002, 2:20 pm, Room C-105

Session: Feature Profile Evolution /Plasma Processing for MEMS
Presenter: G. Schulze-Icking, Infineon Technologies AG, Germany
Authors: G. Schulze-Icking, Infineon Technologies AG, Germany
A. Kersch, Infineon Technologies AG, Germany
A. Knorr, Infineon Technologies
A. Hausmann, Infineon Technologies Dresden GmbH & Co. OHG
J. Radecker, Infineon Technologies Dresden GmbH & Co. OHG
Correspondent: Click to Email
Due to its low thermal budget and its highly directional deposition HDP-CVD of SiO2 has become an important process in IC fabrication. In order to study (and ultimately improve) the HDP-CVD process we have developed a physically based model for feature scale simulations. This model has been implemented into our custom Topography Simulator "Topsi" and extensive studies of the HDP-CVD process have been performed. In this presentation we give a survey of our simulation results and compare them to experiments. The model we propose retains the characteristics of a complex reactor scale model published by Meeks et al.@footnote 1@ and extends it to the feature scale. A key aspect of both models is a "structural passivation" of the surface due to the chemisorption of gas phase precursors. In contrast to conventional CVD (at much higher temperatures) in HDP-CVD this passivation is removed by cations striking the surface. Accordingly ions not only sputter surface material but also are responsible for its directional deposition. The final topography therefore is the result of simultaneous neutral deposition/passivation, ion induced activation, and sputtering. It is well known@footnote 2@ that the sputtering yield crucially depends on the ion energy and the angle of incidence. This probably also applies to ion induced surface activation, but very little is known about its energy and angular dependence. We therefore have performed deposition experiments and compared the final topography to simulations performed using our new model. With the derived set of parameters we are now able to predict the surface evolution as a function of process conditions. This is a major improvement over a more empirical model proposed by Conti et al.@footnote 3@ @FootnoteText@ @footnote 1@ E.Meeks et al.; J.Vac.Sci.Techn. A, 16, pp 544 (1998) @footnote 2@ C.Abrams et al.; J.Vac.Sci.Techn. A, 16, pp 3006 (1998) @footnote 3@ R.Conti et al.; DUMIC Conference (1999).