Johann Wolfgang von Goethe,
Italienische Reise, 16. 3. 1787
THERE are two main directions that are recommended for future work. First, progression of the diffusion module towards a model development environment, and second, further research to make PROMIS a reliable tool for simulation of complete fabrication processes.
The first area is vastly addressed in the existing version. The class of
parabolic equations and boundary conditions which can be treated is fairly
general. Additionally required is the treatment of internal material
interfaces. We suggest to store three concentration values for each point at
an interface. One for each adjacent material and one for the interface
itself. Interface conditions of the form 
should be allowed, where 
denotes the interface
concentration, 
and 
are the fluxes in the adjacent
materials and 
is an arbitrary function of the three concentrations.
This would facilitate model development for e.g. silicidation and
polysilicon diffusion sources.
A symbolic input language for the mathematical expressions describing the physical model and additional automatic consistency and stability analysis would support easy model implementation.
The second area offers multifarious fields of activity. High concentration effects and the along going processes of activation and deactivation are one of the least understood areas in process modeling. In this context at least a model for phosphorus diffusion at high concentrations and a dynamic precipitation model for boron is needed. Models for diffusion from polysilicon and silicides are of considerable interest.
Process engineers are talking in junction depths and thermal budgets rather than in point-defect parameters. Therefore, depending on the user's skill level, it should be possible to hide the physical models. Models should be implemented as a generic class containing three components: model equations, model parameters, and conditions under which the model applies. A decision tree, or even better a knowledge-based expert system containing descriptive information about physical phenomena, should be applied to select the `best available' model.
The topographies allowed currently for diffusion and oxidation are not rigorous enough. Multilayer capabilities are indispensable for future realistic simulation. A module for oxidation which calculates the oxide growth concurrent with the diffusion of dopants is necessary for considering concentration dependent oxidation kinetics. The module has to account for mechanical stress, should predict oxidation induced defects in silicon, and should allow topographical changes like in the polysilicon buffered LOCOS process.
These new activities should face towards full three-dimensional simulation.