Convergence

From Flooxs
Jump to navigation Jump to search

Device Convergence

DC Bias Increments

If your voltage steps are too large, the solution may not converge.

Continuity Equation Signs

The continuity equations should be of the form:

set eqnE "ddt(Elec) - ($Emob) * $Vt * sgrad(Elec, DevPsi/$Vt)"
set eqnH "ddt(Hole) - ($Hmob) * $Vt * sgrad(Hole, -DevPsi/$Vt)"

Grid Spacing

In areas with high e-fields, the grid spacing needs to be much smaller. Examples of this would be the depletion region of a diode or the SiO2/Si channel interface of a MOSFET.

Doping Profile

The doping profile you create needs to be defined for all regions. Be especially careful at boundaries.

If you are using some of the advanced device models, make sure that donor (ND) and acceptors (NA) level terms are defined everywhere in the device. For example, a diode that is 10 microns long in the x-direction starting at x=0.0, the term:

sel z=1.0e20*(x<3) name=ND

should be written so that the ND term exists over the entire structure (this applies for NA as well):

sel z=1.0e20*(x<3)+1.0 name=ND

Newton Step Number

The default number of Newton steps allowed is 20. In very rare cases the solution will be about to convergence but will run out of steps. To increase the number of Newton steps use:

pdbSetDouble Math iterLimit 50

where any number can be used in place of the last term.

Process Convergence