Coupling thermal and electrostatic solver

[aminer2k]

Hello,
I am (again) working on coupling HeatSolve and StatCurrentSolve for a steady state, iterative solution. It looks like this can be accomplished by setting the current source term in StatCurrentSolve (drho/dt) and by setting the heat source term in HeatSolve (h) after each iteration. The problem is complex in that drho/dt and h are:

drho/dt = -divergence(A(T)*grad(T))
h= B(T)*divergence(-T*C(C)*grad(Potential)

A, B and C are basically material parameters that are functions of T.

How can I:
1. extract the gradients of temperature and potential as vector fields
2. extract T
3. calculat drho/dt and h, and apply as Body forces.

Based on what I have read in the forum and from examples, these gradients can be extracted with "temperature convection field" or similar, but I am not getting very far beyond that. Looks like updating the body force with calculations using matc is certainly possible, but it is not clear how to get these fields for calculation.

Thanks so much for any advice.

[raback]

Hi

Basically you could extract what you need using FluxSolver and then operate on the the resulting vector with DivergenceSolver. This would introduce to you four additional solver so for me the solution seems rather suboptimal.

What you would normally do in FEM is to take the operators use partial integration to operators such as \div ( a \grad T) and thereby get rid of the difficult term. Personally facing this kind of problem I would be tempted to write a monolithic solver for both fields.

-Peter

[aminer2k]

Thank you for your guidance on adding a fluxsolver and using the divergence solver. I will look into that, and repost progress.

Yes, I agree that it seems to make better sense to make a modified solver. I'd like to try that, but it seems beyond my programing skills right now. Basically, the governing equations for HeatSolve and StatCurrent Solve each need one more term, to include thermoelectric effects. My 'hack' is to use drho/dt and h to add these terms calculated externally, at the expense of running 4 solvers, as you mentioned.
Thanks so much,
Andy