Scientists measured a photon leaving an atom excited for negative time, and it's exactly as strange as it sounds, but not in the way headlines suggest

[KaiHeck]

University of Toronto physicists led by Professor Aephraim Steinberg have experimentally measured a negative weak value for how long atoms remained excited because of a photon passing through them, a result that finally passed formal peer review in Physical Review Letters this past April after first appearing as an eye catching preprint back in 2024. To be clear about what this doesn't mean first, light did not travel backward through time, and nothing violated causality, the leading edge of any pulse still respects the ordinary speed limit set by relativity

The underlying physics involves something called group delay. Light traveling through matter near an atomic resonance experiences interference between its different frequency components in a way that can reshape a pulse so its peak appears to exit earlier than a comparable pulse traveling through empty space. Physicists have understood negative group delay as a mathematical description of that reshaping for a long time, the debate was always whether that negative number corresponded to anything physically real happening inside the material, or whether it was purely a description of the outgoing waveform's shape

Steinberg's team tested this directly using a cold cloud of rubidium-85 atoms. A weak signal pulse passed through the cloud while a separate, off-resonant probe beam crossed the same region, and the atomic excitation caused by the signal photon subtly shifted the phase of that probe beam through what's called the cross-Kerr effect. By selecting only trials where a signal photon successfully made it through the cloud without scattering away, and using a technique called weak measurement that extracts a small amount of information per trial without significantly disturbing the underlying quantum system, the researchers found that the resulting excitation time tracked the photon's group delay, and under some conditions, both became genuinely negative, measured at as much as minus 0.82 relative to the experiment's own reference time

A follow up theoretical paper published in September 2025 explained the mechanism, a transmitted photon can be described as taking multiple possible quantum histories simultaneously, in one it barely excites the atoms at all, in another its energy briefly gets stored collectively among them before returning to the outgoing beam. Quantum mechanics combines the probability amplitudes for these different histories, and once the experiment keeps only the trials that end with a successfully transmitted photon, destructive interference between those histories can make the resulting weakly measured value come out negative. Notably, the peer reviewed paper's actual title swapped out the preprint's provocative language about photons spending negative time inside atoms for the more careful and precise phrase negative weak values, and physicists are still debating exactly how weak values should be philosophically interpreted even as this experiment strengthens the case that the negative number predicts a real, measurable laboratory effect rather than serving as pure mathematical abstraction

Scientists observe 'negative time' for the first time in a quantum experiment - The Brighter Side of Newswww.thebrighterside.news

[Quanta]

The peer reviewed title deliberately swapping negative time for negative weak values is such an important and honestly pretty responsible bit of scientific self correction, the original framing was always going to get misread as time travel

[Jordan89]

Multiple quantum histories interfering destructively once you postselect only the successfully transmitted photons is a genuinely elegant explanation, doesn't need anything to literally happen backward, just careful accounting of which outcomes you kept

[Lantern76]

Using a separate probe beam's phase shift as an indirect clock instead of trying to directly stop and inspect the photon itself is a clever experimental workaround for a question that seems impossible to measure any other way