During process flow simulation, the PIF wafer model is subsequently undergoing a number of modifications of the boundaries of the material segments defining the wafer geometry and of the dopant distribution data in the bulk and along material interfaces of the wafer. To fulfill the requirements of the PIF wafer state convention adopted in VISTA/SFC for the PIF representation of a wafer model (cf. Section 4.3.1), grids have to conform to the boundaries of their respective segments to form a valid wafer state.
The reconciliation of existing grid and dopant data, modified grid and dopant data, and existing and modified geometry data is taken care of by external gridding tools that are called automatically by the framework. The selection of an appropriate gridding tool potentially poses problems similar to the choice of the right simulator. Various gridding tools have been reported, with new developments constantly challenging established solutions. Therefore, it is advisable to provide a mechanism for using the gridding tool of one's choice in a fashion similar to selecting the simulator best suited for a process step. At present, VISTA/SFC offers a choice of three grid generators, TRIANGLE [She96], TRIGEN [Ban90], and VORONOI [Hal94], for re-gridding purposes, with a clear interface provided for the integration of additional gridders. In the GUI, gridders are selectable by mouse click similar to other configuration parameters (Figure 4.10).
Figure 4.10: Selection of wafer state gridders: GUI
module.
At present, TRIANGLE has proven the most
reliable and robust
gridder of the three alternatives. It is based on a
highly flexible delaunay gridder and takes care of
enforcing the wafer state convention with respect to grid and dopant data,
re-gridding of the entire structure from scratch, using local grid
refinement with respect to dopant concentrations,
merging existing dopant data and additional data generated by an
implantation or diffusion step,
and for merging a geometry arising from a topography simulation step
with the dopant data of the original wafer.
As it makes use of existing grids, numerical inaccuracy is
reduced to a minimum and the creation of additional grid points is
restricted to areas where it is necessary.
TRIANGLE provides a mechanism for the generation of quality
meshes by means of various grid constraints including minimum angles
and maximum areas of triangles and dopant dose conservation
criteria.
An enhanced refinement technique delivers rapid variations
of grid density and thus minimizes the global number of grid points.
In general, the choice of gridding parameters and constraints exerts
a strong influence on the quality of the simulation result. Therefore,
all gridding-related resources are specific for a project and do not
need to be specified globally.
For example , the max-triangle-area
parameter specifies the maximum size in of all triangles
generated by TRIANGLE,
the min-angle parameter defines the minimum
angle in degrees. Typical values for these resources are 0.1 and 15,
respectively.
Wafer state gridding is mainly concerned with the
storage of distributed quantities, not with requirements arising
from numerical and computational considerations. Most simulation
tools employ their own strategy to construct an internal grid
representation that is determined by numerical aspects of the
equations to solve. In general, it is not possible to find a grid
that is optimally designed for all possible computational strategies.
Therefore, data representation on grids is supported by the
environment, whereas the generation of computational grids is
declared responsibility of the simulation and solver modules.