In the previous section we discretized the transformed model equations which describe the physical problem. The result of this discretization is a large system of usually nonlinear algebraic equations (3.5-1).
The unknowns 
are the values of the concentrations at discrete
points. The rank of 
and is 
, the
number of quantities 
times the number of grid points 
.
For this set of nonlinear equations, in device simulation it is quite common to use decoupled iteration schemes [Moc83], [Sel84], e.g. Gummel's method or nonlinear Gauss-Seidel iteration. For a wide range of problems inexact Newton methods are very promising [Deu90]. With decoupled schemes convergence problems arises if the PDEs are strongly coupled, in that case one is forced to use a fully coupled Newton iteration.
