What happens when you try to chop a photon in half?

Let’s start with the example of a partially reflective mirror. When a single photon hits that mirror, it will either go through the mirror or reflect from the mirror. The photon is considered to enter a superposition state of having both reflected and passed through (the probabilities of each path depending on how reflective the mirror is). If we place detectors in the path of the reflected and transmitted photons, when one clicks, it collapses the superposition, and the other potential path disappears. There are no circumstances in which both detectors will click at the same time. We do not record half a photon each way. Naively, we could make the same argument for a fully reflective mirror that is removed midway through reflection. In this argument, the photon enters a superposition state of transmitted and reflected, with the probability determined by when the mirror was removed compared to the “size” of the photon. Again, when we try to measure which way the photon went, we’d expect the superposition to collapse, and only one detector would click. This picture is universal and applies to all time-varying signals—and far more than those, too. When I was growing up on the farm, listening to AM radio on an old-fashioned (even then) tube radio, the music was always interrupted by a clicking noise. The clicking was from our electric fence, which was always zapping some errant grass, a misbehaving sheep, or a horny bull. That short-sharp current generated a short electromagnetic pulse (and an angry bull). The very short pulse (in time) was present across a very broad spectrum, including, to my annoyance, the AM broadcast spectrum. The shorter an event in time, the more frequency spectrum is required to support it. On the flipside, a single tone that does not change for a very long time requires very little spectrum (only the tone itself). This rule also applies to photons reflecting from mirrors. The photon is reflecting from the mirror, and the electromagnetic field is varying regularly and smoothly changing from the incoming to the reflected wave. The transmitted wave doesn’t exist, so the amplitude is a happy zero. Then the mirror is yanked away. The reflected wave’s amplitude abruptly drops to zero, and the transmitted wave suddenly jumps from zero. Those are two sharp transitions that require a lot more bandwidth than the original photon had. Our photon that has been cut off is still in a superposition of reflected and transmitted. But it also has a sharp edge, which requires a multitude of photons at different frequencies. Cutting a photon in half generates a rainbow. And as far as I can tell, the generated photons are in a superposition of both reflected and transmitted light. But since there are potentially many photons, both transmitted and reflected light could be measured simultaneously. This will be a complex experiment to perform. Researchers will need a source that generates single photons on demand with a very narrow spectral bandwidth. This will spread them out in time so that any additional photons that come from cutting it off are observable. They then need to be able to trigger the mirror at the right time. This won’t be done with something like a bathroom mirror. The authors calculate that the transition from reflective to transmission needs to take place in about 10 femtoseconds (a femtosecond is 10-15 s), which is insufficient time to move a physical mirror. Some materials, like semiconductors, can be driven from reflective to transmission quite rapidly (30-100 fs) using ultrafast laser pulses as the switch. Unfortunately, that laser pulse gets in the way. It will be quite difficult to filter out the big laser pulse used to remove the mirror so that the photons generated by cutting the long single photon are observable though. We already have some evidence that this works, though. The mirrors I described above are used to shorten ultrashort pulses, which means that the reflected pulses have more frequencies after reflection, and new photons must therefore be generated. We just haven’t observed it for single photons yet. But give it about a year, and I’d bet we will. Physical Review Letters, 2026, DOI: 10.1103/94pm-hp34