Simulation of dielectric etch in high density plasma reactors

Abstract

High density plasma (HDP) reactors applied to dielectric etch of fine-line, high aspect ratio features poses a challenge for future technology (below 0.25 jum). The complex nature of HDP reactors and the lack of basic chemical information has made it difficult to apply numerical simulation to enhance process development. This work focuses on a threestep process for modeling complex chemical HDP processes, where some of the necessary reaction rates, excitation energies and surface reactivities are not available. We will concentrate specifically on a C2F6 plasma for oxide etch, although the strategy can be applied to other plasma systems. The first step in modeling a HDP process is to develop gas phase and surface etch mechanisms from available data. Sensitivity analysis is applied to reduce the number of species and reactions ,this is required for performing 2D simulations. The second step is to validate the mechanism and the plasma models through data comparisons. A suite of three simulation tools was used during this process: Aurora, MPRES and Icarus. Experimental data used for mechanism refinement and validation included Langmuir probe data, laser diode data and oxide etch rate data taken in both an experimental and a commercial reactor. The final step is application of the developed model to investigate a commercial process. 2D plasma model results are presented to Currently resides at IBM Microelectronics Hopewell Junctions, NY 12533. This paper is declared a work of the U.S. Government and is not subject to copyright protection in the United States. Introduction The trend in semiconductor manufacturing towards higher aspect ratios and larger wafer diameters (300mm) continues to require higher uniformity and larger selectivity for etch processes. It is expected that dielectric, high density plasma (HDP) etch processes will face this challenge. Current commercial dielectric etch processes have been developed empirically. The process operates in a regime where a delicate balance exists between polymer deposition and ion assisted etch.. The challenge of improving this fragile etch process to meet new process requirements has generated incentive to improve process understanding. One approach to enhance process understanding and decrease process development time is through process simulation. Applying models to enhance understanding of these systems is nontrivial due to the complex chemistry (both gas and surface) and the lack of fundamental reaction rate information. To compensate for the lack of reaction rate data, simulations and experimental validation data can be combined to develop a mechanism that is valid for a defined operating regime. This work focuses on a three-step process of mechanism development for a HDP reactor. Specifically, a C2F6 plasma mechanism will be developed to predict oxide etch rates for a commercial reactor. The first step in modeling an HDP process the mechanism compilation. The initial C2F6 mechanism was built with data from the literature. When gas phase and surface reaction rates were not available, extrapolations from similar chemistries were used. Sensitivity analysis was applied to reduce the number of reactions and species in the mechanism while still capturing the salient features of the plasma. Species with low densities and reactions that did not significantly impact electron properties or ion fluxes were removed. For example, the initial C2p6 mechanism contained approximately 40 species while the reduced mechanism contains 14 species; fewer species are necessary for convergence and improved run times of the 2-D models. Surface chemistry mechanisms were required for the different wall materials and the oxide wafer. At this time a photoresist etch mechanism has not been developed. This work was funded by a Sandia-SEMATECH CRADA and was performed at Sandia National Laboratories which is operated for DOE under Contract DE-AC04-76DP00789. Copyright© 1998, American Institute of Aeronautics and Astronautics, Inc. The second step in HDP modeling is validating the mechanism and the models through comparison to experimental data. The reduced C2F6 mechanism was refined and validated by comparing 0-D and 2-D reactor simulations to data. Data was taken on a commercial HDP reactor and a Gaseous Electronics Conference (GEC) reference cell modified with an inductive coil and a quartz liner. The computational codes applied in this study include; Aurora, a OD model, MPRES, a 2D continuum finite element code and Icarus, a 2D Direct Simulation Monte Carlo (DSMC) code6. Data comparisons included Langmuir probe data, CF laser diode measurements, and oxide etch rate data. The third step in the process of HDP modeling is to apply the final mechanism to investigate process questions. The focus of this work was to investigate the changes in etch rates and selectivity (oxide to photoresist) as a function of gas injection location and flowrates for a commercial oxide etch reactor. Icarus simulations are presented to explain reported etch rate and selectivity variations.