Off the top of your head, are you aware of any similar general-multiphysics NN work that's been applied to electromagnetics problems? In particular, some colleagues in my lab are investigating imaging via acoustic waves which are induced by microwave absorptive heating (in liquids, biological tissues, etc.); this approach is most commonly known as RF-induced thermoacoustic imaging [1]. It's very tricky to model this phenomenon in simulation, doubly so to measure it experimentally.
Most in my lab (myself included) are leery of throwing NNs at problems and seeing what sticks, but sometimes I wonder whether a model like yours might help us skip past the boring details to get at the novel technical stuff, or else extend simulations to more complicated boundary conditions.
"Effective" compared to the state of the art in the bands on either side: in RF/microwave we have very fast arbitrary waveform generators, very nice amplifiers, and well-characterized conductors, and on the optical side we have lasers, lights, fibers, and more. The terahertz gap is so named because it's too high in frequency for our usual RF devices to work well, and too low in frequency for our usual optical devices to work well; terahertz work ends up being a mix of both, taking from either column as needed for a specific application. (You might hear the word "quasi-optical" used in this sense, though I've never heard the dual word "quasi-microwave"!)
We do have terahertz devices - they're just very limited compared to devices in adjacent bands, usually stated in terms of power. But there are a lot of hardworking and talented people working on narrowing the THz gap from all sides. It's a very very hot research area at the moment.
With respect to [2] - I think this is partially our garden-variety, universal impostor syndrome, but not only that: even otherwise good papers can be written so poorly as to be nearly incomprehensible, and still get published! I've come across papers that seem to be genuinely valuable and interesting work, but the amount of mental manual labor required on the reader's part is horribly daunting - mishandling or absence of grammatical articles ("a", "the", etc.), inconsistent spelling, a feeling of constant ambiguity of meaning...
I have great sympathy for the many excellent scientists who have to overcome a language barrier to get published, since the lingua franca of virtually every major journal is English. It's not inherently bad that "the language of science is bad English"; these difficulties are a symptom of pulling together good science from everywhere in the world. I'm just deeply irritated with the publishers - IEEE in particular, though the fault is by no means theirs alone - who don't care to keep up a copyediting standard for their allegedly high-quality publications, since apparently their goal is not to communicate science well, but instead to make a profit.
(I distinctly remember one of my favorite math professors stating, in no uncertain terms, that the words "a" and "the" each have different connotations with respect to existence and uniqueness. Incorrect use of either would get points knocked off of your proof.)
If you can parlay it into an 1/8 wavelength, 25cm is probably not impossible to fit into some monstrously meandered beast that could fit in a phone's footprint. We've come a long way in terms of miniaturizing antennas, though I'm no expert.
But I think your second point hits the nail on the head - cell phones have enough antennas in them, and asking to add another one at such a low frequency is a great way to get your antenna and QA teams to look for a more sane employer, and to put your EMC compliance partner's kids through college :)
Fun fact: it's very difficult to get rid of built-up charge in space. And guess what incident EM radiation does to your electronics: that's right, it can build up charge! To my understanding, this is especially a problem when charge is built up in specific dielectrics/insulators or the very delicate structures of modern transistors and other semiconductor devices. This so-called "total ionizing dose" [1] can lead to "leakage currents, degrade the gain of a device, affect timing characteristics, and, in some cases, result in complete functional failure" [2].
Any RF or EMC engineer will tell you that ground is a dirty word, especially in extreme conditions. There's no one-size-fits-all approach, since so much depends on the physical context.
Depends on the definition of "simple", imo. The first thing that comes to mind is research into materials with good (tens of dB), wideband absorption in the mmWave bands. It's an area of active research [1] [2] (just a couple articles from a quick google, so caveat emptor).
My two cents, though I'm no expert: I'd bet it's what the computational EM community (and other fields) calls a "multiscale" problem. EM solvers - that is, simulators that numerically solve Maxwell over some geometry-and-source boundary conditions - find E- and H-fields at certain "mesh points". In other words, they discretize 3-space into a grid and calculate solutions to Maxwell at those points.
In general, you'll want your mesh to have subwavelength distance between points, and perhaps even less in regions with complicated geometry or parts of your geometry you're particularly interested in. In the microwave regime, this means mesh points will typically have tens of centimeters or less between them. However, given that the receiving antennas in satellite-based solar power are orders of magnitude larger than that, trying to simulate such a large structure and still keep your mesh points relatively dense is just asking for the curse of dimensionality to bite you.
In other words, it's certainly possible with enough compute time, but we have better things to do with our GPU cores, especially since the whole point of antenna simulation is to assist with design by allowing you to run a bunch of simulations to tune your design without having to fabricate a bunch of DUTs. Again, I'm not really an expert, but my understanding is that this kind of multiscale problem is a hot research topic right now, not only in computational EM but in many other areas of physics simulations, especially those governed by nasty PDEs (e.g. fluid dynamics) or those which involve complicated structures at multiple scales (e.g. VLSI design).
Off the top of your head, are you aware of any similar general-multiphysics NN work that's been applied to electromagnetics problems? In particular, some colleagues in my lab are investigating imaging via acoustic waves which are induced by microwave absorptive heating (in liquids, biological tissues, etc.); this approach is most commonly known as RF-induced thermoacoustic imaging [1]. It's very tricky to model this phenomenon in simulation, doubly so to measure it experimentally.
Most in my lab (myself included) are leery of throwing NNs at problems and seeing what sticks, but sometimes I wonder whether a model like yours might help us skip past the boring details to get at the novel technical stuff, or else extend simulations to more complicated boundary conditions.
[1] https://ieeexplore.ieee.org/abstract/document/6248685