previous up next contents
Prev: 7.3.2 L-Shaped Trench Up: 7. High-Pressure Chemical Vapor Next: 8. Conclusion and Outlook


7.4 Reactor/Feature Scale Integration

The presented CVD simulation model works on feature scale and accounts for inhomogeneous gas concentrations. The origin of the inhomogeneities is the gas depletion resulting from the competition between gas consumption at the surface and supply of reactants by diffusion. Boundary conditions like the concentration of the gas species at the top of the simulation domain are considered as constant and homogeneous within the dimensions of the simulation domain whose characteristic length is up to a few microns.

It is obvious that the assumption of homogeneous boundary conditions is not correct when applying wafer scale to consider variations in heat distribution, gas flow and gas concentrations. In general, inhomogeneous thermal distributions and flow conditions within the reactor lead to strong variations in overall film thickness, film composition, profile evolution, and step coverage across the wafer.

Figure 7.8: CVD reactor scale simulation.
\begin{figure}\begin{center}
\ifthenelse{\boolean{nopics}}{\fbox{\texttt{eps-cvd...
...graphics[width=0.5\textwidth,clip]{eps-cvd/fluent.eps}}
\end{center}\end{figure}

It is beyond the scope of the presented approach to address also simulation and modeling on reactor scale. Reactor scale simulators dealing with different types of mass transport on non-moving grids such as FLUENT1are commercially available. As an example Fig. 7.8 shows the distribution of the ${\rm WF\hspace*{-0.2ex}_6}$ concentration across the wafer for a reactor scale simulation of tungsten CVD with ${\rm WF\hspace*{-0.2ex}_6}$ reduction. Nevertheless, the feature scale model allows the integration of the results from such equipment level simulations. Dirichlet boundaries at the top of the simulation domain can be set according to the concentration resulting from the equipment simulation. Together with the heat variations they account also for changes in the species effective diffusivities which influence the profile evolution by determining the balance between diffusion velocity and reaction rate. Variations in growth rate and overall film thickness varying across the wafer can also be covered by adjusting the local deposition rate. By these means of integrating results from reactor scale simulation our CVD model represents a link to the final prediction of the feature scale profile evolution in an integrated back-end process simulation.



Footnotes

... FLUENT1
See http://www.fluent.com/.
previous up next contents
Prev: 7.3.2 L-Shaped Trench Up: 7. High-Pressure Chemical Vapor Next: 8. Conclusion and Outlook

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