Taylor Robertson
Ansys Employee
Hello Yuzak,nnThis is limitation of lumopt, though not necessarily the adjoint method. The topology indices should be real dielectric values; although, I have seen examples where metal was present near the optimizable geometry. nIf you would like to optimize metals it may be possible to make the updates yourself, but there is a real challenge associated with the gradient calculation. At metal interfaces the fields will not be continuous and the FDTD mesh may not be sufficiently resolved to accurately solve for these changes. This is not something we can really support you with.nI would recommend this thesis http://optoelectronics.eecs.berkeley.edu/ThesisOwenMiller.pdf for a great discussion on the adjoint method. Additionally I had a similar discussion on this post. https://kx.lumerical.com/t/adjoint-optimisation-set-up-for-absorbing-geometry/75604.nnBestnn