revathijambunathan / warpx Goto Github PK

Current LLG inputs are read and stored in the WarpX class.
These can be moved to either an LLG class or macroscopic class.

The MXH_eff
and M X M x Heff implementation can be cleaned up a bit using a cross_product function
(cross_product_xcomp(M,H))
where cross_product_xcomp(M,H) would return the x-component of (M X H)

Currently, there are a number of places where face-to-face, edge-to-face, face-to-edge interpolations are performed
using hard-coded IntVects based on the respective src and dst multifabs.
These need to be replaced with built-in functions in CoarsenIO.H, which include generic interpolations.

Currently, in MacroscopicEvolveM.cpp, mag_normalization_error=0.1 is hard-coded. This needs to be added as a user-defined inputs with appropriate docs.

Initial E-B fields are read using the parser at initialization.

For WarpX with particles, the externally applied E and B functions (x,y,z,t) are applied directly on the particles as forces.

But, for material implementations, we need to call the parser functions of the externally applied E-B fields on the fields, when updating them.

We had to turn managed memory back on because B field arrays exist on the GPU, but we need to modify them on the CPU to add the perturbation. We need to copy them from device to host, update, and then copy back from host to device.

See the title.

Delete redundant lines in Source/FieldSolver/FiniteDifferenceSolver/MacroscopicEvolveM_2nd.cpp.

For example, the material properties are repeatedly loaded by Line 280-285, while this has been done by Line 74-80.

Note similar redundancy exists in other lines as well, such as Line 289-294. Please clean those.

The coupled Maxwell (with zero E-B field)+ LLG is currently not agreeing with stand-along LLG test-case.

We have a specialized diagnostic for testing the convergence of our implementation.
This is currently performed by default.
We could add a run-time user-flag to turn this feature on for diagnostics.

The function void WarpX::MacroscopicfieldOutput ( ) is used for this.

A proposed way to do this could be to add a diagnostic type = "convergence_test"
that can read a user-defined (i1,j1,k1) and (i2,j2,k2)

1. Hard source needs to be called immediately after each of the evolve functions. The right sequence should be: evolveHM - hard H source - evolveE - hard E source - evolveHM - hard H source
2. Soft source needs an extra factor with Delta_t included in it. Derivation needed to figure out the exact formula of this factor.
3. Soft source calling strategy needs to be updated in one of the following three ways:
   • split soft source calls in every time step, apply soft source twice just as the way we apply hard source. Use Delta_t/2 instead of Delta_t in each call.
   • apply soft source once in the beginning of each time step.
   • ?