A Question That Didn't Seem Like a Question
Imagine standing at one end of a tunnel, throwing a ball in, and timing it as it exits: 0.5 seconds inside. Nothing to discuss.
Now swap the ball for a photon and the tunnel for a cloud of rubidium atoms. You time it — and the photon exits *earlier* than it entered. It spent negative time in the cloud.
This is not a joke. It is an experimental result first observed in 1993 and finally 'confirmed by the atoms themselves' in a 2026 paper published in *Physical Review Letters*, vol. 136, no. 15, with first author Daniela Angulo (University of Toronto) and corresponding author Aephraim Steinberg.
Notably, Steinberg was a co-author of the 1993 paper. Then a PhD student at Berkeley, working with Raymond Chiao and Paul Kwiat, he used a two-photon interferometer to measure photon tunneling times and found a negative delay for photons crossing a 1.1-micrometer barrier. The paper has over 1,300 citations, but the physics consensus was: 'It's just a pulse-front effect. Don't take it seriously.'
Thirty years later, Steinberg went back to ask a question nobody thought to ask in 1993.
Why 'Negative Time' Sounds Like Nonsense
A photon enters an atomic cloud, interacts with atoms, gets absorbed and re-emitted — surely that takes time? How could it come out *earlier*? Doesn't that violate causality?
The key: a photon is not a billiard ball. A photon is a wavepacket — a diffuse cloud in space and time, with a leading edge, a trailing edge, and a thick middle.
What the atomic cloud does is not 'hold it and release it later' but something subtler: it reshapes the wavepacket. Near resonance, the cloud preferentially absorbs the trailing part of the wavepacket, letting the leading part pass more easily. The exiting wavepacket is thus 'advanced' — not because anything travels faster than light, but because the wavepacket's shape was trimmed so the front became the bulk.
This is the standard explanation of negative group delay: no information travels faster than light, so relativity is safe. Physicists breathed a sigh of relief: negative time isn't real, just a mathematical consequence of waveform reshaping. The 1993 Steinberg accepted this and moved on.
But one question was never asked: what do the atoms themselves have to say?
Asking the Atoms
This is the core idea of the 2026 paper.
If negative group delay is really just a wavepacket artifact, then asking the atoms 'how long was that photon in you?' should yield a positive number — the interaction and absorption/re-emission physically happened.
Conversely, if the atoms also answer with a negative number, then negative time has physically measurable consequences, beyond a mathematical game.
But 'asking atoms' is extremely tricky in quantum mechanics. Measurement disturbs the system: to know precisely when the photon is in the atom, you must constantly watch the atom's excited state — but each look disturbs it. Watch too often, and the photon never interacts with the atom at all. This is the quantum Zeno effect, named after Zeno's arrow paradox.
So you cannot ask the atom *precisely*. You must ask it *fuzzily*.
Weak Measurement: Peeking Without Being Caught
Enter weak measurement, proposed by Yakir Aharonov in 1988: measure without disturbing.
How? Lower the measurement's precision. A precise measurement collapses the quantum state and destroys the photon-atom entanglement. An imprecise one — a blurry signal — causes minimal disturbance. A single result is nearly pure noise, but average a million trials and the signal emerges.
Steinberg's team shone a weak laser, unrelated to the photons, sideways through the atomic cloud. This laser isn't absorbed or emitted; it only senses the atoms' refractive index, which shifts slightly depending on whether an atom is excited. This is the cross-Kerr effect, a nonlinear optical phenomenon.
In short: when the photon interacts with an atom, the atom is excited, the refractive index shifts slightly, and the weak laser's phase shifts a tiny bit. Measuring this phase shift indirectly tells you whether the photon was in the atom — without touching the photon directly.
It is a sideways, fuzzy, collective measurement. Each single shot means nothing, but a million averaged shots tell you: how long the photon spent in the atom.
Result: The Atoms Confirm Negative Time
The experiment ran across different pulse widths and optical depths. For the narrowest-band light pulses (longest in time, narrowest in frequency), the measured average atomic excitation time was (-0.82 ± 0.31) τ₀, where τ₀ is a baseline — the non-post-selected excitation time, equal to scattering probability times the atomic lifetime. The negative sign means: on average, atoms spent negative time in the excited state as photons passed through.
The broadest-band pulses gave (0.54 ± 0.28) τ₀ — positive, as intuition expects.
The key result: the atomic excitation time equals the group delay exactly. The negative time measured from the photon's side and from the atoms' side is the same negative time.
Not an artifact. The atoms confirmed it.
The Odyssey Analogy
The popular version of the paper (published in *The Conversation*, written by collaborator Howard Wiseman) uses a brilliant analogy. Odysseus spends five years on Calypso's island on his way home from Troy. When Penelope asks how long he stayed, Odysseus says: negative five years. She doesn't believe him — so she asks Calypso, who says: yes, he stayed negative five years.
The photon is Odysseus, the atomic cloud is Calypso's island, the group delay is Odysseus's account, and the atomic excitation time is Calypso's testimony. Both match. Both negative.
Does this mean time travel is coming? No. The paper states clearly: standard physics fully explains the result. Negative time violates neither causality nor relativity — you cannot go back and kill your grandfather. It only means that in quantum mechanics, 'time' is far stranger than our intuition suggests.
Why This Matters
1. How Dangerous the Word 'Artifact' Is
In 1993, the community saw negative group delay and called it an artifact. That judgment wasn't wrong — wavepacket reshaping does explain negative group delay — but it closed a question: does negative time have physical consequences? For 30 years, no one asked the atoms. Not because the question was too hard, but because everyone assumed it was already answered.
This is a recurring pattern in the history of science: a counterintuitive result gets absorbed by a 'reasonable explanation,' filed away as 'solved,' until someone reopens the drawer and finds a deeper question inside.
2. Weak Measurement as the Art of Observing Without Disturbing
The quantum Zeno effect teaches: look too carefully and you destroy what you want to see. Weak measurement teaches: look fuzzily and you can see more. A precise measurement gives certainty at the cost of disturbance; a fuzzy one gives noise but preserves the system's quantum nature. The average of a million fuzzy measurements can be more informative than one precise measurement.
This trade-off exists beyond quantum mechanics.
3. Asking the Right Question Matters More Than Answering Correctly
The 1993 experiment answered: is the group delay negative? Yes. But nobody asked: is the time also negative from the atoms' perspective? Steinberg's team didn't invent new techniques (weak measurement and cross-Kerr detection already existed); their contribution was asking a question nobody had asked.
More Distant Associations
An AI Evaluation Analogy
The author notes: a rising benchmark score doesn't mean rising real capability. A model may genuinely improve, or may learn to game the benchmark. Looking only at scores (group delay) can't tell you — you must 'ask the atoms': inspect internal states and real-task performance. The team's move — cross-validating the photon's self-reported arrival time against the atoms' records — is isomorphic to the 'white-box' approach in AI safety: don't just look at outputs; look at internal states.
The Observer Effect in Philosophy
In the social sciences, studying a group changes the group; an anthropologist living three years in a tribe observes a tribe altered by their observation. The quantum Zeno effect is the extreme version: observing a quantum system freezes it. Weak measurement sidesteps the dilemma — observe fuzzily enough and the system keeps evolving while you collect faint signals.
A kind of practical wisdom emerges: sometimes, don't look too clearly. Look fuzzily, and you may see the truth.
Back to the Graduate Student of 30 Years Ago
In 1993, Steinberg measured negative tunneling delays at Berkeley, published in PRL, gathered 1,300+ citations — and then the field said 'artifact,' and he set the problem aside.
Thirty years later, as a professor in Toronto, his student Daniela Angulo picked it back up and used a weak laser to ask the atoms. The atoms answered: yes, the photons spent negative time with us.
The paper doesn't dramatize this history. But comparing Steinberg's 1993 PRL with the 2026 PRL, you see a 30-year dialogue between a scientist and his own past work. In 1993 he lacked the tools (weak measurement was immature) and the approach (cross-Kerr detection hadn't been applied to this problem). Thirty years later he had the tools, the approach, and the student. He reopened the drawer and found the 'solved' problem had never been answered.
This is not a hero narrative; it's a story of patience. Some problems aren't solved — they're shelved and reopened. Sometimes you need 30 years, for the technology to catch up with the question, and for you to catch up with yourself.
The photon spent negative time in the atomic cloud. It's not an artifact — the atoms said so themselves. If you don't believe it, go ask the atoms.
— But remember: ask fuzzily.