Allow nonlinearity with custom medium (FXC-3548)

[caseyflex]

This PR allows nonlinear_spec to be applied to CustomMedium and CustomDispersiveMedium. Anisotropic media are still not supported, and the nonlinearity cannot vary spatially.

Greptile Overview

Updated On: 2025-11-05 19:15:49 UTC

Greptile Summary

This PR extends nonlinearity support to CustomMedium and CustomDispersiveMedium classes. The implementation correctly validates that anisotropic media are rejected by checking the is_isotropic property before the type check.

Key changes:

• Reordered validation checks to check is_time_modulated first, then is_isotropic, and finally the type check
• Added CustomMedium and CustomDispersiveMedium to the list of allowed medium types in the isinstance check
• Updated imports to include the new custom medium types
• Updated error messages to be more specific about what is being rejected

Issues found:

• Missing CHANGELOG entry (per custom rule a109c62a-aa8c-4cc4-b515-0ad947c47929)

Note: The previous comment about missing anisotropic validation has been addressed - the code properly checks medium.is_isotropic before allowing custom media.

Confidence Score: 4/5

• This PR is safe to merge with minimal risk, pending CHANGELOG entry addition
• The implementation is correct and properly validates media constraints. The validation logic is clear: it checks time-modulation first, then isotropy, and finally the type. Both CustomMedium and CustomDispersiveMedium have the is_isotropic property, so the check will correctly reject anisotropic custom media. The only issue is a missing CHANGELOG entry, which is a documentation concern rather than a code quality issue.
• No files require special attention

Important Files Changed

File Analysis

Filename	Score	Overview
tidy3d/components/nonlinear.py	4/5	Enabled nonlinearity support for CustomMedium and CustomDispersiveMedium by reordering validation checks and updating allowed types. The validation properly checks is_isotropic before type checking.

Sequence Diagram

sequenceDiagram
    participant User
    participant NonlinearSpec
    participant NonlinearModel
    participant Medium
    
    User->>NonlinearSpec: Create NonlinearSpec with models
    User->>Medium: Create CustomMedium/CustomDispersiveMedium with nonlinear_spec
    NonlinearSpec->>NonlinearModel: Validate medium compatibility
    NonlinearModel->>Medium: Check is_time_modulated property
    alt is_time_modulated == True
        NonlinearModel-->>User: Raise ValidationError (time-modulated not supported)
    end
    NonlinearModel->>Medium: Check is_isotropic property
    alt is_isotropic == False
        NonlinearModel-->>User: Raise ValidationError (anisotropic not supported)
    end
    NonlinearModel->>Medium: Check medium type
    alt not in (Medium, DispersiveMedium, CustomMedium, CustomDispersiveMedium)
        NonlinearModel-->>User: Raise ValidationError (unsupported medium type)
    end
    NonlinearModel-->>NonlinearSpec: Validation passed
    NonlinearSpec-->>User: Medium with nonlinearity created successfully

Loading

[caseyflex]

@yaugenst-flex once we support anisotropic (not fully anisotropic, just diagonally anisotropic) nonlinear medium (a minor change perhaps), we can get rid of this _validate_medium_type function you don't like, in favor of the _has_incompatibilities in medium.py. But no need to change this here...

[greptile-apps]

_{1 file reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}

[caseyflex]

@greptile

[caseyflex]

@yaugenst-flex @momchil-flex I also got rid of the warning that evaluating the permittivity of a nonlinear medium ignores the nonlinearity. I think this is obvious (we aren't running the simulation so there is no source intensity anyway). And this warning comes up all the time when I don't need it, like in plotting, or computing source fields for simulations with nonlinear + plane wave for example. Let me know if you think this is ok; an alternative is suppressing this warning where we don't want it, but I personally would never want to see it.

[github-actions]

Diff Coverage

Diff: origin/develop...HEAD, staged and unstaged changes

• tidy3d/components/nonlinear.py (100%)

Summary

• Total: 4 lines
• Missing: 0 lines
• Coverage: 100%

[yaugenst-flex]

@yaugenst-flex @momchil-flex I also got rid of the warning that evaluating the permittivity of a nonlinear medium ignores the nonlinearity. I think this is obvious (we aren't running the simulation so there is no source intensity anyway). And this warning comes up all the time when I don't need it, like in plotting, or computing source fields for simulations with nonlinear + plane wave for example. Let me know if you think this is ok; an alternative is suppressing this warning where we don't want it, but I personally would never want to see it.

Hmm I'm not sure about this. I think the warning is good, even if it might be obvious considering the rest of the setup. I feel like it's an easy thing to miss and a reminder that you're getting the linear response is helpful. I think if warning noise is the issue, it might be better to perhaps demote it to log.info or logging only once per medium call, something along those lines. But dropping it entirely feels like a regression to me.

[momchil-flex]

Yeah not sure. I guess basically the definition of permittivity is 1 + chi(1), i.e. taking only the linear susceptibility. So indeed one shouldn't expect anything different when "getting epsilon". But of course in some cases we do think of a nonlinearity-induced change to the linear response...

I guess downgrading to INFO is fine, even though practically speaking it will never be seen by users. :)

[caseyflex]

Ok, I'll just restore the previous handling. If it gets annoying we can address it in a future PR.