Whip a mirror out of the path while a lone photon is midway through bouncing off it, and the result isn’t half a photon heading one way with the other half going the opposite direction. What you get instead is a rainbow.
That’s the argument put forward by three Norwegian physicists in a new paper (there’s an arxiv.org link), and I’ll confess the outcome sat much further from my gut instinct than I would have guessed.
Begin with the nature of a photon, since the entire result rests on it. A photon is one particle of light, and ordinarily it can’t be split. Yet it isn’t a particle in any neat sense either. It has no single location. Rather, it’s a spread-out object, smeared across both space and time.
Why the world isn’t a permanent acid trip
Photons splitting and merging isn’t something you meet in everyday life, and you ought to be thankful for that. Were shining a single color through glass to routinely divide or fuse photons, every surface would spit out colors that were never present in the source to begin with. Reality would resemble the most saturated legal LSD trip you can dream up. It doesn’t. So, no LSD.
Photons really do split and merge, but only in particular circumstances. The material the light passes through has to react to the light itself. Arrange that, and a single color can spread out into many.
Physicists label the ordinary behavior linear. Splitting and merging photons counts as nonlinear, and coaxing out nonlinearity generally calls for either a highly sensitive medium or a very intense source, such as a laser.
Snatching a mirror away partway through a reflection doesn’t seem nonlinear on first inspection. Dwell on it a moment, though, and it plainly is. What obscures that is the way we’ve been taught to imagine single photons striking mirrors.
The half-silvered mirror trap
Consider a partially reflective mirror. Aim one photon at it and the photon either goes through or bounces back. We describe it as entering a superposition of both routes, the probabilities determined by how reflective the glass is. Place a detector on each route and, the instant one fires, the superposition collapses and the other route disappears.
The two detectors never fire together. You’ll never log half a photon down each path.
Here comes the seductive step. Take a fully reflective mirror and yank it away midway through the reflection. By that same reasoning, the photon ought to rest in a superposition of reflected and transmitted, the odds depending on when you pulled the mirror relative to the photon’s “size.” Attempt to measure it, the superposition collapses, and one detector fires.
That isn’t what occurs. Once I paused and genuinely reasoned through it, the naive picture was clearly wrong. Grasping why demands a bit more equipment.
Why a sharp edge costs you a rainbow
Time and frequency are two faces of a single coin. Strike a note on a piano and there’s a time-domain view, a steady wobble of pressure that lasts a while. That note corresponds to one frequency at one amplitude. Messier sounds, like chords or staccato, hold messier structure in time and demand more frequencies layered together, each carrying its own amplitude.
This is true of every time-varying signal, and more besides. Raised on the farm, I used to catch AM radio on a tube set that was already ancient back then, and the music kept getting interrupted by a clicking noise. That was our electric fence zapping stray grass, a misbehaving sheep, or a horny bull.
Every brief, sharp burst of current threw off a brief electromagnetic pulse, and an angry bull. A very short pulse in time spreads across a very wide swath of spectrum, including, much to my annoyance, the AM band. The briefer the event, the more frequency it takes to assemble. Turn it around and a lone, unchanging tone needs almost no spectrum whatsoever.
Photons at mirrors follow the very same rule. As the photon reflects, the field changes smoothly, converting the incoming wave into the reflected one. The transmitted wave doesn’t exist, so its amplitude rests at a contented zero.
Then you yank the mirror. The reflected wave’s amplitude falls to zero while the transmitted wave leaps up from zero. Two hard edges, and hard edges call for far more bandwidth than the original photon ever carried. The severed photon remains in a superposition of reflected and transmitted, but it now also carries a sharp edge, and that edge requires a spread of photons at differing frequencies.
Cut a photon in half and you produce a rainbow. As best I can tell, those new photons are themselves in a superposition of reflected and transmitted. But since there can be a great many of them, you could register transmitted and reflected light simultaneously. That’s precisely the thing the half-silvered mirror never permitted.
The experiment nobody’s run yet
Managing this in a laboratory will be tough. You’d require a source that emits single photons on demand with a very narrow spectral bandwidth, which stretches each one out in time so the extra photons from the cut become visible. Then you’d have to flip the mirror at precisely the right instant.
A bathroom mirror won’t cut it. The authors work out that the shift from reflective to transmitting must occur in roughly 10 femtoseconds. A femtosecond is 10-15 s, which is far too fast to move anything physically. Certain materials, semiconductors among them, can be switched from reflective to transmitting in 30 to 100 fs using ultrafast laser pulses as the trigger. The snag is the trigger itself. That laser pulse is loud, and screening it out so you can spot the faint photons generated by cutting the long single photon will be a genuine headache.
We already have a clue that this works. The same sort of mirrors are used to shorten ultrashort pulses, which means the reflected pulses emerge carrying more frequencies than they arrived with, so new photons must be getting created. We simply haven’t caught it happening for single photons yet.
Give it about a year. I’d wager we will.
Physical Review Letters, 2026, DOI: 10.1103/94pm-hp34