[Some like to argue that this is why light gets slowed down in glass – each photon had to “waste” some of its time sitting still in an atom – but for good reasons, physicists tend not to take that picture too seriously.]
You can ask how much time atoms spend in the excited state as light passes through, and this is measurable. If you do this experiment in a regime where there is never more than one photon around at a time, so you can neglect stimulated emission, the answer turns out to be: exactly the number of photons which got “scattered” or “lost” (absorbed and then re-emitted in a different direction, for instance), multiplied by the natural lifetime of the excited state.
One natural interpretation would be that once a photon excites an atom, it’s almost certain that in the process of spontaneous emission (which takes on average one lifetime), the photon will leave the original beam; such that when you see a transmitted photon, it’s simply because that photon was lucky enough never to get snatched up by an atom in the first place.
That’s what we expected when we did our experiment, but it turns out to be false. (Probably the first experiment I’ve done in my career where the answer really did not turn out to be more or less what we expected, even though [we believe] we did the experiment right.)
We learned how to measure how much time atoms spent in the excited state, conditioned on a photon showing up at a detector on the far side of the atom cloud. This means that we’re probing what we can learn about the “history” of a photon by observing where it is now – the kind of thing we’re usually taught we’re not allowed to do in quantum mechanics (viz. our tunneling-time experiments) !
We found that transmitted photons do cause atoms to spend some time in the excited state, although a bit less time than “scattered” photons do.
We have some semiclassical stories we can tell about this, but the kind of experiment we do can only be truly modelled by treating the light fully quantum mechanically. We are finalizing a theory paper with Howard Wiseman that we believe offers the proper answer to this question in general, and it has some (to me) very surprising results, which we will be testing in the laboratory soon (see Josiah Sinclair's poll, linked below).
For Further Reading
The technical paper can be found at PRX Quantum 3, 010314 (2022).
Edwin Cartlidge has written a column on our work for Physics World.
If you have intuitions about what this time should be (particularly in the "easy" limit of a narrowband pulse and a sample with very low absorption), Josiah Sinclair is running a poll on twitter! (We're setting up the follow-up experiments now, so place your bets quickly!)
Acknowledgments
This experiment builds on earlier work by Amir Feizpour, Matin Hallaji, and Greg Dmochowski -- we thank Amir, along with Howard Wiseman, Klaus Molmer, John Sipe. Alan Migdall, Amar Vutha, and Joseph Thywissen for helpful conversations.
This work was supported by NSERC and the Fetzer Franklin Fund of the John E. Fetzer Memorial Trust; AMS is a fellow of CIFAR.