CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   Time dependent Maxwell equations (http://www.cfd-online.com/Forums/openfoam-solving/80236-time-dependent-maxwell-equations.html)

libe September 20, 2010 10:29

Time dependent Maxwell equations
 
Hello guys,
could someone give me a basic guidance on implementing in OpenFoam the system of time dependent Maxwell equations (e.g Microwave modeling,scattering etc ) . The system consists of time dependent electric field and magnetic field components. The mhdFoam solver I think is not able to perform such task.

Please help!

libe September 21, 2010 08:18

Hello again,
trying to solve the coupled system of Maxwell equation, for electric E and magnetic field H, i implemented the following solver, myMicro.C. My computational area is a cube ( 0.1m x 0.1m x 0.1m). On the left side of the cube (say inlet) I set a boundary condition oscillatingFixedValue for E and H in Oy and Oz direction respectively in order to take a wave propagation in Ox. The time step is very low 1e-13. After few iterations of the solver the wave stagnates, increases its value and the results are non physical at all...
Any ideaaa!!!


while (runTime.loop())
{
Info<< "Time = " << runTime.timeName() << nl << endl;

// #include "readPISOControls.H"
// #include "CourantNo.H"

tmp<fvVectorMatrix> EEqn
(
epsilon0*fvm::ddt(E)==fvc::curl(H)

);
// EEqn().relax();
EEqn().solve();


tmp<fvVectorMatrix> HEqn
(
mu*fvm::ddt(H)== -fvc::curl(E)


);
// HEqn().relax();
HEqn().solve();

Conny_T September 23, 2010 17:12

Quote:

After few iterations of the solver the wave stagnates, increases its value and the results are non physical at all...
Any ideaaa!!!
First of all, this is the expected behaviour. If you just translate some differential operators into the fvm::... and fvc:... function calls, you are very likely to have a stability problem. Most solvers require some sort of stabilisation.
If your approach were successful, you should have been very suspicious about the result.

Quote:

epsilon0*fvm::ddt(E)==fvc::curl(H)
This is the second problem. This system of equation couples for example the Ex component to the Hy and Hz components. Unfortunately Openfoam is not yet capable of solving this system with block matrices, but instead solves them component after component. If the Maxwell equation could be solved with Openfoam, a Maxwell solver would already be shipped with it.

Conny

libe September 24, 2010 02:55

Thanks conny,
I ve already realized that something important is missing from my approach with OpenFoam. Now I try to do my work with the classical FDTD adopting some AMR approaches. But I was wondering if openfoam could solve the Maxwell equation with the suggestion of Shankar, Hall et all for an Finite Volume Ttime domain CFD like solution.

mirko October 1, 2010 10:43

Quote:

Originally Posted by Conny_T (Post 276399)
First of all, this is the expected behaviour. If you just translate some differential operators into the fvm::... and fvc:... function calls, you are very likely to have a stability problem. Most solvers require some sort of stabilisation.
If your approach were successful, you should have been very suspicious about the result.

This is the second problem. This system of equation couples for example the Ex component to the Hy and Hz components. Unfortunately Openfoam is not yet capable of solving this system with block matrices, but instead solves them component after component. If the Maxwell equation could be solved with Openfoam, a Maxwell solver would already be shipped with it.

Conny

I am trying to understand this issue better. The `standard' FDTD (finite difference time domain) solution for Maxwell equations uses a leapfrog scheme: it first solves for all dE/dt (and updates E) from the curl of H for one half the time step, and then solves for dH/dt (and updates H) from curl of E and J for the other half the time step

Taking one step & one component as an example, in cartesian coordinates, dE_x/dt will depend on (curl H)_x, all of which are known. Integrating this over a volume, this can be viewed that the (curl H)_x is the source term for E_x in that cell.

I did not write the formulas in detail, but I don't see here a matrix solution problem here. Only evaluation of the source term (curl H), and updating the LHS (E).

I consider myself an OF and numerical arts novice, but I wonder if the OP used a time step that is too large, or if he set up the initial conditions correctly.

Mirko

vitoco September 21, 2015 12:47

Did you get a Solution???
 
Quote:

Originally Posted by Conny_T (Post 276399)
This is the second problem. This system of equation couples for example the Ex component to the Hy and Hz components. Unfortunately Openfoam is not yet capable of solving this system with block matrices, but instead solves them component after component.
Conny

Hello Conny:
Can you show me an example (link, code, pdf, etc...) to understand how to apply "solves them component after component" ???? Thanks in advance, I need it in this moment to simulate a electromagnetic wave for my thesis.
Best regards,


Victor


All times are GMT -4. The time now is 04:38.