I suspect you are going to be writing your own code.  To handle metals, you may want to consider the method of lines based on real-space discretization like the finite-difference method.  RCWA is essentially the method of lines where a Fourier transform is used to discretize the field and materials.  In real-space there is an equivalent thing to FFF involving dielectric smoothing, but I don't think you will need it.  Also, when you formulate a semi-analytical code to handle full tensors, you will have solve a 4x4 matrix equation instead of a 2x2.  This size isn't the problem.  You will have to sort the modes you calculate in forward and backward propagating to make it stable.

Also, anisotropy really bogs down the computation.  Your sims will be much slower than ordinary materials.

One suggestion I have...Is there any way to study your concept in two-dimensions first?  If so, that will be a lot faster than jumping straight into three dimensions.

Hi Raymond

Thanks for the answer. I feared that I would have to write my own code, but I think that it would take too much time and effort, and I think it wouldn't be worth it. The fact is that S4, at least in principle, supports dielectric tensors with e_xy and e_yx different from zero, but when I perform a calculation it really seems to struggle with it. I also have confirmation that commercial FDTD software, while capable of such a calculation, struggles when the off diagonal components are very small. I saw comsol perform successfully these calculations (http://onlinelibrary.wiley.com/doi/10.1002/adom.201400125/abstract), but the authors confirmed me that the calculation is extremely complex. So I am a bit at a loss here, at this point any code conceptually simplex and faster than comsol would do...

best

Giovanni

Hi Giovanni,

Indeed performing such calculations is far from trivial, specially when dealing with metals and/or MO. In our theoretical treatments we use a combination of our own codes (S-Matrix codes with FFF and MO for arbitrary magnetization directions -see  http://journals.aps.org/prb/abstract/10.1103/PhysRevB.71.205116 for Polar conf (no FFF in that one) and http://link.aps.org/doi/10.1103/PhysRevB.85.245103 for a full FFF and general formulation) that are used to obtain far field quantities, FDTD to obtain the field profiles (careful with this, while FDTD codes seem to deal more or less well with MO, (e.g. Lumerical) the parametrization of the Eps_tensor makes it really troublesome). Normally the near-field patters that we extract using FDTD are those where the MO is not the relevant information. Finally, we use COMSOL to get the MO modification of the field profiles (such as those in the reference you mention or others).

Normally we verify the convergence degree by comparing all three of them.....

At the moment our codes are not ready to be shared, but we could try to help you. Please contact me at [email protected] and we could see how we could  deal with your structures.

Best,

Antonio

http://journals.aps.org/prb/abstract/10.1103/PhysRevB.71.205116

http://link.aps.org/doi/10.1103/PhysRevB.85.245103

Hi Antonio

Nice to meet you here.  As you know I have read with attention all of your works...nevertheless implementing a code from scratch along those guidelines is going to be a very long work... furthermore if I remember correctly, the second reference contains FFF rules only for TMOKE case, while in this moment I am interested in PMOKE configurations. Anyway, correct me if I am wrong, it seems to me that in the first reference you were able to compute MO signals without FFF, and this for me is very good news! I will not esitate to contact you once I can get a bit more time to work on this problem.

Thanks

Giovanni