[paul cotter (OP)]

Any electron following a circular path is accelerating and accelerating charges radiate. Why no em radiation from electromagnets? I only used the mri as an example of a rather extreme electromagnet. A comment by Hamdani triggered this query.

[alancalverd]

The drift velocity of electrons in a conductor or superconductor is minuscule - microns per second. So the acceleration of electrons in a 1 m diameter supercon magnet is negligible and any em radiation would be undetectable.

You could however argue that there is a minimum diameter of superconductor that cannot sustain its current because no electron could complete a loop. But since no energy is dissipated by resistance in a superconductor, the electrons do not require any energy to maintain their drift speed. However the magnetic field inside this sub-critical loop would be very large and probably exceed the critical field above which it cannot remain superconducting.

But I happily admit this is something of a "handwaving" argument and doesn't satisfy my pedantic alter ego.

[paul cotter (OP)]

As I am sure you realise by now that pedantic pursuits are my joie de vivre. The drift velocity is exceedingly slow but there is still acceleration occurring, albeit at a vanishingly small extent, and there will be energy leakage through radiation, even if we cannot detect it. I do, however, accept your analysis and thank you.

[Bored chemist]

I suspect that, for every electron traveling left to right, there's another one on the other side of the coil traveling right to left and (at least largely) cancelling out the radiation.

[paul cotter (OP)]

One cannot cancel radiation like that, that would be in conflict with the COE. Two sources of radiation from different spatial locations will produce a pattern of interference with areas of cancellation and enhancement. Only if they originate in the exact same space and 180 degrees phase difference will total cancellation occur.                                                                                                                                 Very late edit: as the frequency drops the wavelength increases and the requirement for coincident antiphase sources to effectively cancel relaxes. That is if the antiphase sources are a tiny fraction of lambda we can approximate coincidence. As any radiation from a magnet would be very low energy and hence low frequency I now believe BC to be correct and I am guilty of "jumping the gun". Mea culpa.