I originally asked this question Does a single photon propagate with phase velocity or front velocity through a dispersive material? about the speed of a single photon in a dispersive medium. I received no answer but some good comments which made me re-thinking and concentrate on the core question.)


Consider the following hypothetical experiment: We have a small volume, completely evacuated except for one single atom that we can excite. The volume is surrounded by a giant shell made of glass, refractive index $n=1.5$. It's expensive glass, so no absorption and imaginary part of the refractive index for our wavelength of interest. The glass shell is one light day thick and is surrounded by photodetectors.

We excite our atom in the middle at $t_0 = 0$. After some lifetime it will emit a photon in some direction, which needs to transmit the huge glass shell an is then detected. The question is, when.. The glass shell is so thick that we can neglect Heisenberg uncertainty for the lifetime (traveling time will be much larger). I think we can also neglect photon shot noise at the receiving part of the experiment - although I'm not sure at this point, maybe that is the missing link...

Do we detect the photon:

  1. after roughly one day because it's a photon travelling with $c$?

  2. after 1.5 days because the photon was traveling with the phase velocity $v=\frac{c}{1.5}$?


From comments and this related question Phase and group velocities in QFT / Quantum Optics I meanwhile learned that even a single photon is composed of a frequency distribution and therefore has a group velocity (although the cited post is about the quantum wave function and I don't know how this relates to the electromagnetic function). So phase velocity and group velocity (which I think is the true speed of propagation if asking about when the photon is expected to arrive in this thought experiment) can differ even for a single photon.

In addition, in an answer to this question What really causes light/photons to appear slower in media? (fig. 2) it is stated that photons inside glass still travel at $c$, but as the wavelength is shorter in glass, the phase velocity is smaller. This makes sense, even if the propagation speed is $c$, planes of same phase will propagate slower if the wavelength is shorter. However, this answer was downvoted, no idea if this means something...

So, for now I think the photon is expected after one day in my thought experiment above, although phase velocity in the glass was smaller than the vacuum speed of light $c$ (the group velocity was $c$). Could someone confirm or correct?


2 Answers 2


Emission process is always finite in time. This means that the EM field associated with the photon will be something like a wave packet, with a reasonably* well defined beginning and end. The detection, since at the very least it must happen after the atom has been excited to subsequently emit the photon, will also happen after some time.

During the time between emission and detection the photon will propagate through the medium. How a wave packet travels through the medium is governed by medium's dispersion relations and wave packet's frequency spectrum (which is at the very least affected by natural broadening). The overall motion of the wave packet envelope is described by group velocity. By contrast, phase velocity will only make the ripples of the EM field move along the wave packet, but not affect the overall motion of the packet. In fact, it may even be backwards compared to wave packet motion.

Since when you detect a photon, detection probability is proportional to the square of the EM field, the peak of the wave packet arriving at the detector is the time when it's most probable for the detection event to happen. Thus, we can conclude that the time it takes between emission and detection is defined by group velocity.

*E.g. you could take the cut-off at the tail as 0.1% of peak amplitude, or whatever else that suits your needs.

  • $\begingroup$ Many thanks! But is the wave packet you mention a real electromagnetic wave or a quantum mechanical probability amplitude (the two must be related, but I don't know how...)? Second, I agree about the group velocity. But taking the refractive index of normal glass and reasonable assumptions for the photon... is the group velocity then closer to c or to the phase velocity, meaning does it take more like one day, or one and a half in that experiment? $\endgroup$ Commented Jul 30, 2022 at 22:37
  • 1
    $\begingroup$ @CharlesTucker3 since it's a single photon, it's just the wave function. You may be interested in this question. As for group velocity, this is given by group index, which for silica is a bit higher than the refractive index. So it takes more time to propagate than it would take in a vacuum. $\endgroup$
    – Ruslan
    Commented Jul 30, 2022 at 23:03
  • $\begingroup$ According to the cited webpage the group index for say a 500 nm photon is even larger than the phase index. So if doing my experiment, the outcome is that on average I detect the photon even a bit later than 1.5 days, correct? $\endgroup$ Commented Jul 31, 2022 at 6:53
  • 1
    $\begingroup$ @CharlesTucker3 well, given that you've chosen phase index as 1.5, then yes, most likely so. $\endgroup$
    – Ruslan
    Commented Jul 31, 2022 at 7:58
  • $\begingroup$ Many thanks, question is answered! $\endgroup$ Commented Jul 31, 2022 at 8:55

If you're talking about a real experiment, the answer is $c/v_g$, with $v_g$ the group velocity. In some cases $v_g$ may simply be equals to phase velocity, but it's still important to distinguish the two.

A single photon will always travel with velocity $c$, but inside matter light keeps being absorbed and reemitted, which creates an "effective" velocity inferior to $c$.

Edit: if the medium is a simple transparent medium, then you have the dispersion relation $k=n\frac{\omega}{c}=nk_0$ (with $k_0$ the wavenumber in vacuum and $n$ assumed constant), which implies that phase velocity and group velocity are equals.

  • $\begingroup$ So your response to the question is "neither 1. nor 2."? $\endgroup$
    – garyp
    Commented Jul 29, 2022 at 11:11
  • $\begingroup$ @garyp I edited my answer to give more details. $\endgroup$
    – Miyase
    Commented Jul 29, 2022 at 11:16
  • $\begingroup$ @Miyase The experiment is designed to treat the photon by its particle nature and since it's a single photon there is no group velocity. If one would treat the photon as a wave, the answer is phase velocity, I agree. The question is if a single photon is enough to treat it as a wave slowed down as interfering with secondary waves induced in the glass? $\endgroup$ Commented Jul 29, 2022 at 11:49
  • 1
    $\begingroup$ @CharlesTucker3: "since it's a single photon there is no group velocity" - no. The statement assumes that a single photon implies a single frequency, which is not in general valid. A single photon can (and in general would) represent a spectrum of frequencies. $\endgroup$ Commented Jul 29, 2022 at 12:30
  • 1
    $\begingroup$ If I could just think of a defined emission time and then a detection time, my bet would be on group velocity and not phase velocity. The problem that I am having is on how to define all terms in a more realistic thought experiment, because of the problem of lifetimes, linewidth of emission, etc. Nonetheless I almost always arrive at the conclusion that you would find a distribution that is consistent with group velocity. This would in principle be just a small correction to the 1.5 days (for green, silica, the group velocity is only 1.8% slower). $\endgroup$ Commented Jul 30, 2022 at 12:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.