previous up next contents
Prev: 7.4 Reactor/Feature Scale Integration Up: Dissertation W. Pyka Next: The Last Word

8. Conclusion and Outlook

The goal of this thesis was a rigorous treatment of three-dimensional topography simulation, on the one side for the investigation of the topography related etching and deposition models itself, on the other side in order to generate input geometries which are accurate enough to assure reliable interconnect or device analysis.

Several milestones have been passed along the way towards this goal. The algorithmic core of the surface propagation has been optimized. The effect of the gained acceleration is, that the CPU time necessary for the surface propagation can be neglected with respect to the evaluation of the models needed for the calculation of local etching or deposition rates. Moreover the basic operations like surface normal calculation as well as handling of voids and unconnected parts in the geometry have been completed in three dimensions.

Concerning the three-dimensional geometry generation, the topography simulator has been expanded by means of solid modeling with mask operations including simple geometric primitives as well as layout information and aerial image simulation. Furthermore simple models for CMP and oxidation have been added. The control system for the geometry generation benefits from the automation obtained with the process-step-like setup which combines the IC design defined in the layout with the process recipe from manufacturing. With the application to several interconnect examples the combined solid modeling/topography simulation approach has proven its ability to improve the quality and significance of subsequent interconnect characterizations, especially for highly non-planar structures.

With respect to topography models, resist development not only turned out to be very helpful for predicting exposed and etched resist profiles within a rigorous lithography simulation but, by means of benchmark examples, also demonstrated the extraordinary stability of the proposed cellular surface movement algorithm.

Furthermore the foundations for extensive modeling of ballistic transport determined low-pressure etching and deposition processes have been laid and experimentally verified with a sputter deposition model for the deposition of TiN. Modeling of high-pressure CVD has been covered with a comprehensive and flexible model applicable to a wide range of process chemistries such as TiN CVD and tungsten CVD. For both low-pressure as well as high-pressure processes coupling to reactor scale simulations has been accomplished and enables comprehensive studies of across-wafer non-uniformities in topography simulation.

Yet, this thesis is not the finishing line for topography simulation. Although the algorithmic optimization and the program extensions for solid modeling and layout inclusion are treated and many geometry generation as well as etching and deposition modeling steps are considered, many topics are open to be developed further. As a direction to future work the most important ones are listed below.

The algorithms for the geometry generation lack of a direct definition of sidewall slope angles in order to replace the stepwise composition of tilted sidewalls and the generation of oxidation topographies has to be generalized.

For the low-pressure model, a more thorough investigation including a broader experimental verification of different processes is desirable. The most important aspects to be treated in addition are angular dependent surface interaction, simultaneous etching and deposition when passivation layers are formed, and implementation of multiple reflections, either by means of Monte Carlo simulations for stochastic particle tracing or by radiosity methods.

The high-pressure CVD model for sure is the most comprehensive part presented in this thesis. Nevertheless, a broader experimental verification with the simultaneous extension to process chemistries with gas phase reactions of precursors like in silicon-dioxide deposition from tetraethylorthosilicate (TEOS) would be an interesting challenge for the further confirmation of the presented concept.

Last but not least an analytic model for the transformation of the angular and radial target emission characteristics into incident off-center particle fluxes would make an adequate compromise between shifting of the origin of cosine or exponential distribution functions and expensive Monte Carlo simulations of the particle transport when considered on reactor scale.

The different topics addressed in this thesis have covered a wide range of simulations from single process optimization to integrated simulation flows including layout information. Summarizing, the presented cellular approach represents a way to fulfill the different requirements for the various types of simulations with one and the same program always working at an appropriate level of accuracy and within reasonable CPU times.

Subsections previous up next contents
Prev: 7.4 Reactor/Feature Scale Integration Up: Dissertation W. Pyka Next: The Last Word

W. Pyka: Feature Scale Modeling for Etching and Deposition Processes in Semiconductor Manufacturing