Medium perturbation

[dbochkov-flexcompute]

Heat and charge perturbation functionality. Currently, implemented for Medium and PoleResidue, but this is ready for an initial "big picture" review, to check if anything seems bad/needs a rework. Here's a notebook with demonstrations and explanations to help the review https://github.com/flexcompute-readthedocs/tidy3d-docs/blob/daniil/perturbation-medium/docs/source/notebooks/PerturbationMedium.ipynb. If everything looks more or less alright, I will go ahead and finish some remaining to-do's:

• do the rest of dispersive models, that is, add PerturbationDebye, PerturbationLorentz, etc
• add front-end tests
• check all code documentation is in place

tagging @momchil-flex just in case

[tylerflex]

Thanks Daniil, this is a ton of work and I think it is really well organized and covers everything I can think of very well. Most of my comments are just refining it for public consumption. The notebook is also extremely useful and will be invaluable. There were a few spelling issues, I recommend this https://pypi.org/project/jupyterlab-spellchecker/ to catch some of them before we do a more thorough proofread.

Thanks!

[weiliangjin2021]

Overall looks quite good! A few small comments.

[dbochkov-flexcompute]

big cleanup, added many docstrings, added tests, and fixed many loose ends. I am considering adding rest of dispersive models PerturbationDebye, PerturbationLorentz, etc, later, and merging this without them. What do you think?

[weiliangjin2021]

If allow_gain=True, conductivity can be < 0.

[dbochkov-flexcompute]

so, these validators only warn if conductivity/permittivity can potentially go out of bounds (within the provided/calculated temperature_range, electron_range, and hole_range). The strict validation is left to the corresponding CustomMedium. So perhaps it's still ok to warn the user about potentially negative conductivity even if allow_gain=True?

[weiliangjin2021]

I see, so it will warn and then error. In that regard, there seems to be missing allow_gain in the return of

    return CustomMedium(
            permittivity=permittivity_field,
            conductivity=conductivity_field,
            name=self.name,
            subpixel=True,
        )

and it seems better to add subpixel as a field of PerturbedMedium so that one can choose to apply subpixel or not. But warn that for a uniform medium, supixel parameter will be ignored.

[dbochkov-flexcompute]

Oh, you're right, thanks for catching that. Actually, frequency_range should also be passed. Fixed it.

[weiliangjin2021]

big cleanup, added many docstrings, added tests, and fixed many loose ends. I am considering adding rest of dispersive models PerturbationDebye, PerturbationLorentz, etc, later, and merging this without them. What do you think?

Agree, as long as we can have a good heat example in the current stage.

[dbochkov-flexcompute]

if there are no other comments at this point, is it ok to merge this into pre/2.4?

[tylerflex]

Yea I've given all of my comments. I think it's ok if everyone else does

[weiliangjin2021]

Green light from me.