[Pavithra]

Hello Everyone,

I have a solver for melting in the presence of electric field. In the solver, I calculate the gradient of electric potential to obtain the electric field, as follows:

E = -fvc::grad(Ue)

Here,
E -> Electric field (volVectorField)
Ue -> Electric potential (VolScalarField).

When I run the solver using a single processor. It gives accurate results and I can perfectly validate both electric field and electric potential distribution.

But, when I run in parallel mode, the gradient calculation goes wrong and I observe nonphysical values in electric field.

The electric potential distribution is physical and has accurate values (in both serial and parallel modes). I am able to validate with the literature. Thus, I am sure that the governing equations of the solver have no problem.

I observe errors only in the electric field. That too, only while running in parallel mode.

E is a final value calculated for post-processing. It does not affect any other part of the governing equations. Also, the most strange thing is that, the gradient of electric potential produced nonphysical values even while calculated using paraview after running reconstructPar.

Kindly please help me to debug this problem. Is there any specific way to calculate gradient while running in parallel mode?

(I have tried different numbers of cores, as well as, different methods of domain decomposition.)

Thank You.

With Thanks,
Pavithra.