6.1.2 Taxonomy of Problems

Problem 1.

Tool produces a result on a non-geometry-conforming grid and tool expects a geometry-conforming grid as input.

Problem 2.

Tool produces a result on a single grid that spans multiple segments (with different materials) and tool expects one grid for each segment (a wafer state compliant input, see also Section 2.5.5).

Problem 3.

Tool creates a result attribute that must be merged with an existing attribute which describes the previous wafer state (e.g. additional Boron doping by ion implantation) to produce a new, valid wafer state. Should the old or the new grid be used to represent the superposition of the attributes? This choice of the target grid is non-trivial. Especially when different spatial regions are affected, a grid-merge is desirable.

Problem 4.

(A variation of problem 3) Tool creates a result attribute that must be merged with the wafer state on a grid type that deviates from the wafer state grid.

Problem 5.

Tool alters the geometry of the wafer state, but does not care for the attributes and grid defined on the altered geometry.

Problem 6.

The wafer state is defined for a much larger area (and so are grids and attributes) than the sub-domain that shall be simulated by tool . A sub-geometry (a single segment or a rectangular sub-domain) is fairly easily constructed, but the grid and attributes on this sub-domain may be required as input for the tool.

There are many more grid-related problems and conflicts that do arise when multiple state-of-the-art simulation tools are used to simulate practical device fabrication steps (see also Chapter 7). Some of these problems can be solved by interpolation services. The problems 1-6 listed above, however, can not be solved satisfactory by interpolation alone. Moreover, the continued interpolation before and after each simulation step is a dangerous sink of accuracy and should be avoided when feasible alternatives exist.

Everything said so far is applicable for two-dimensional as well as three-dimensional simulation. Nevertheless, the conflicts are much more relevant for two-dimensional simulation, simply because a larger variety of two-dimensional simulators exist already and because other, more severe grid and geometry related problems dominate the field of research in the three-dimensional case.