[6/n][Adjoint Module] Register a material grid's susceptibilities with the structure - #3286
Draft
smartalecH wants to merge 1 commit into
Draft
[6/n][Adjoint Module] Register a material grid's susceptibilities with the structure#3286smartalecH wants to merge 1 commit into
smartalecH wants to merge 1 commit into
Conversation
`add_susceptibilities` collects poles only from materials that satisfy `is_medium`, which is true only for `which_subclass == MEDIUM`. A `MaterialGrid` is `MATERIAL_GRID`, so its constituents' poles were never registered: nothing called `sigma_row` for them and the amplitudes that `epsilon_material_grid` interpolates went nowhere. A dispersive material grid was silently non-dispersive in the forward simulation. The adjoint side reads `medium.E_susceptibilities` directly through `get_material_pt`, so `get_chi1_tensor_disp` did see the poles. The result was a confident gradient for a structure that was never simulated -- a finite difference of the objective with respect to a design variable came back exactly zero while the adjoint reported a large value. Three changes: * Collect poles from `medium_1` and `medium_2` for a material grid, via a small `add_material_pols` helper used at all three enumeration sites. Electric only, since the grid does not interpolate a magnetic response. * Accumulate in `sigma_row` instead of stopping at the first match. The grid's list is the concatenation of its constituents' poles, so when both media carry the same pole -- same frequency and damping, different strength -- two entries match the one registered susceptibility and the mixture's amplitude is their sum. Taking the first match dropped `medium2`'s contribution. Identical poles superpose, so this is also the right answer for a plain medium that lists a pole twice. * Refuse constituent properties a material grid cannot interpolate. It mixes epsilon, the electric susceptibilities and the diagonal `D_conductivity`; `mu`, `H_susceptibilities`, the nonlinearities and `B_conductivity` are never written to the mixed medium and were accepted and then ignored, which is the same failure mode. Also check the invariant `epsilon_material_grid` relies on -- that the mixed medium's pole list is the concatenation of the two constituents', assembled in `python/typemap_utils.cpp` -- rather than indexing past the end of a list built through some other path. The new tests are endpoint and superposition identities that hold exactly, so each compares against a value known independently of the grid machinery: a grid at u=1 is indistinguishable from a block of its second constituent, at u=0 from its first, and a pole shared by both constituents has the summed amplitude.
smartalecH
changed the title
smartalecH
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.