Anisotropic MaterialGrid

[kdotom]

The documentation mentions that for MaterialGrid, with frequency-independent isotropic ε,
epsilon_diag and epsilon_offdiag are interpolated. Is this still the case or are anisotropic materials now handled as fully anisotropic?

[stevengj]

It sounds like you are confusing interpolation with anisotropy?

The material grid linearly interpolates between two materials, regardless of what kind of materials they are. This doesn't mean that it discards anisotropy.

If these two materials are anisotropic, e.g. 3x3 matrices $\varepsilon_1$ and $\varepsilon_2$, then the material grid results in anisotropic materials. For example, a grid value of $u \in [0,1]$ is converted to an interpolated material $\varepsilon = (1-u) \varepsilon_1 + u \varepsilon_2$, which is fully anisotropic. This is done in the code here:

meep/src/meepgeom.cpp

Lines 578 to 580 in 176eae6

	// Linearly interpolate dc epsilon values
	cinterp_tensors(m1->epsilon_diag, m1->epsilon_offdiag, m2->epsilon_diag, m2->epsilon_offdiag,
	&mm->epsilon_diag, &mm->epsilon_offdiag, u);