Paper PS2-WeM2
Experimentally Guided Development of a Dielectric Etch Plasma Model

Wednesday, November 12, 2014, 8:20 am, Room 308

Smaller technology nodes in the semiconductor industry place increased emphasis on etch productivity requirements, such as etch rate and critical dimension. Modeling and simulation play a central role in new developments (design of new hardware and exploration of novel processing options) to address the concurrent demand for improved performance and shorter development cycle. Validation against experimental data is a critical step in making these models a mature development tool. In this study, we have developed, refined and validated a dielectric etch process model based on blanket wafer etching results.

In an earlier study, we tested a 2-dimensional model for capacitively coupled plasmas (CCP) in combination with a surface mechanism model against experimental data for etching of blanket SiO₂ wafers in a dual-frequency CCP plasma etcher. The process parameters for this c-C₄F₈/O₂/Ar plasma were varied over a wide range of pressures (25-150 mTorr), bias powers (500-1500 W), and c-C₄F₈ and O₂ flows. The etch rate increased with bias power and c-C₄F₈ flow rate, weakly decreased with increasing O₂ flow rate, and moderately increased with pressure. The reactor simulations were performed using CRTRS, a 2/3-dimensional fluid plasma model. The plasma simulations provided fluxes of various fluorocarbon polymerizing species, atomic oxygen and atomic fluorine. We also calculated fluxes and energies of the ions impacting the wafer. Based on comparisons to the experimental data, we selected a coverage based etch mechanism. This mechanism described center-point etch rates well but indicated that the model needed some improvements to predict the radial etch rate profile and to capture the sensitivity to pressure.

Closer examination of the fluid plasma modeling results revealed that the electron density, and consequently the etch reactants, peaked near the wafer edge. The experimental profiles, on the other hand, showed a slight center-high profile. In the fluid plasma model, the electrons absorbed power at the wafer edge and increased reaction rates close to this power-absorption region. At lower pressures (with fewer collisions), this model was not capturing the non-local behavior of high-energy electrons. A Monte Carlo model provided better spatial representation of electron kinetics and this was coupled with the fluid plasma model. This hybrid plasma model significantly improved the experimental match. Both coverage and thickness based dielectric etching mechanisms were tested. In addition to these improvements, careful accounting for the power going into DC and RF modes gave greater model fidelity to the observed pressure sensitivity.