next up previous
Next: 3.4 Transformations of the Up: 3. Treatment of Interface Previous: 3.2 Discretization at the

3.3 Implementation of Interface Models in MINIMOS-NT

At each grid point equations for one or more unknowns, e.g., electrostatic potential, carrier concentrations, carrier temperatures, lattice temperature, have to be solved. Each of these equations is composed of several different physical quantities, e.g., carrier mobility, effective density of states, surface recombination rate, etc. Most of these quantities can be modeled in several ways with respect to which physical effects are taken into account or are neglected. This leads to different levels of accuracy of the result and to different a demand on computational resources.

MINIMOS-NT uses a sophisticated object-oriented model server [22] to manage and select the different physical models. For each physical model, e.g., carrier mobility model, carrier interface model or contact model, a model class is defined within the model server (see Fig. 3.5). Each model class is derived from a number of predefined model classes and implements one variation of the respective physical model. The derived model classes calculate the entries for the system of equations to be solved. For the carrier interface model class the derived model classes are a model with continuous quasi Fermi level (CQFL), a thermionic emission (TE) model or a thermionic field emission (TFE) model as described in [23]. The equations of the carrier interface models available in MINIMOS-NT are listed in Appendix A together with an example how to select a model.

For each interface of the simulated device the user can specify which of these models should be used. Each model class can have different additional parameters. For example the TFE model has an additional parameter for the effective tunnel length. The values of these parameters can be specified by the user in the input deck file for each interface [2].

To use a model function via the model server first all model parameters are handed to the model server. The model classes derived from a common parent model class all have the same number of input parameters, though not all of them might be used. The same is true for the output parameters. The input parameter types are the same for all derived model classes. This is not necessarily true for the output parameters. For example the carrier interface model which uses a continuous quasi Fermi level directly calculates the carrier concentrations and the carrier temperatures on both sides of the interface. On the other hand the TFE model determines a current flux and an energy flux across the interface. Because the resulting quantities are different the entries into the equation system have to be set up differently.

When entering the values calculated by a model class into the equation system certain transformations have to be performed as described in the next section. The entries into the equation system and the transformations depend on which output quantity is provided by the model.

Figure 3.5: Model server, model classes, and derived model classes.
\includegraphics[width=14cm]{eps/modelserver.eps}


next up previous
Next: 3.4 Transformations of the Up: 3. Treatment of Interface Previous: 3.2 Discretization at the
Martin Rottinger
1999-05-31