A Universe Without a Clock: What 24,000 Atoms Reveal About Time

Cool 24,000 rubidium atoms to a few billionths of a degree above absolute zero. Trap them in crossed laser beams. Slice the cloud in two with a barrier of light thin enough to let atoms seep across.

Then ask the cloud what time it is.

That is close to what Giovanni Barontini, a physicist at the University of Birmingham, actually did. His paper, "Testing the problem of time with cold atoms", appeared in Physical Review Research on 11 June 2026 as article L022047, and is published under a CC BY 4.0 licence. The university's news office publicised it the following day. Renewed popular coverage in late October 2026 revived the story; the research was published in June.

What Barontini built is a testbed for one of the oldest puzzles in physics. What he did not do is prove that time is an illusion, that time stopped in his laboratory, or that quantum gravity has been confirmed.

Why a universe is harder to time than a laboratory

In classical mechanics and ordinary quantum mechanics, time is a background parameter. You wind the clock, you watch the system evolve.

One approach to quantum gravity makes that familiar arrangement difficult. In the Wheeler–DeWitt framework, the quantum state of the universe satisfies a constraint written ĤΨ = 0, without an outside time parameter. That raises a question ordinary laboratory physics can avoid: if the system includes everything, what could possibly stand outside it and keep the time? This is a problem within a theoretical framework, not an established final description of our universe.

Physicists have explored time defined by relationships for decades. In the Page–Wootters approach, one subsystem serves as the clock and the rest is described relative to it. There have already been experimental illustrations: Moreva and colleagues reported one in 2014. Barontini adds a different laboratory platform: a many-atom system in which entropy exchange can be controlled and measured through repeated cycles.

A second question concerns direction. Many dynamical equations allow time-reversed descriptions, yet a recording of a breaking glass looks different when played backward. Thermodynamics connects that everyday arrow to increasing coarse-grained entropy. Explaining the universe’s unusually low-entropy past remains part of the larger puzzle.

What was actually built in Birmingham

The apparatus contained roughly 24,000 rubidium-87 atoms in a Bose–Einstein condensate. At sufficiently low temperatures, many atoms occupy the same quantum state, making collective quantum behavior accessible on a scale researchers can image. The cloud is still made of atoms; its usefulness comes from how coherently they behave together.

Crossed laser beams held the condensate in an optical trap. The optical setup let Barontini change the barrier separating its two halves and watch how that affected their interaction.

Across the middle, a digital micromirror device created an approximately 8-micrometre-wide barrier using light. One half was treated as the observed, or bright, subsystem; the other was excluded from that subsystem’s description and called dark. These are choices about what is included in the model, not a claim that half the atoms are intrinsically invisible. Over the roughly 100-millisecond experimental timescale, the whole cloud showed no measurable particle loss or dissipation.

The bright sector swelled and contracted, then did it again. The paper describes this as a "mini-universe" in the narrow sense of an effectively closed many-body system — a controlled analogy to Big Bang and Big Crunch cycles, not a cosmic event. The "dark sector" is merely the unobserved half of the trap. It is not dark matter.

The researchers followed the cloud for 120 milliseconds, collecting absorption images at 2-millisecond intervals. They extracted atom number, centre of mass, cloud width and an estimate of entropy per particle. Laboratory time remained available as a benchmark. The experiment asked whether a time variable built from internal changes could reproduce the ordering seen against that benchmark.

Entropy is more useful here than the word “disorder”

Calling entropy disorder is a convenient shortcut, but it obscures the measurement. Here, the construction uses a particular coarse-grained description of the cloud: how the measured distribution of atoms changes in the selected subsystem. That is different from tracking every detail of the complete quantum state.

Think of watching only one side of a doorway. People entering and leaving change what is happening in the room you observe, even when the building as a whole has lost nobody. The analogy is limited, but it captures why changes within one part can carry information that is absent from a single total for the whole.

How you tell time without a clock

Barontini's construction is disarmingly concrete. Measure the entropy in the bright sector. Measure a second internal variable — the centre-of-mass position, treated as a clock field — and track how entropy changes as that variable moves along its trajectory.

The entropic time is then an integral of the entropy gradient with respect to that field, scaled by an arbitrary unit. The absolute value in the integrand is the crucial detail: as long as entropy and the clock field change in the same direction, the arrow of time cannot flip.

That condition is the whole game, and the paper is explicit about it. Entropic time is monotonic only while dS and dφ share a sign. The ordering is not automatic; it is conditional.

The oscillating-clock trap

Why not simply use the centre-of-mass position as the clock? Because in a recollapsing subsystem it does not evolve monotonically. It swings out and swings back.

In recollapsing minisuperspace models, the arrow aligns with an increasing scale factor and reverses at the classical turning point. There is no single arrow spanning bang to crunch. A naive clock would therefore run backwards halfway through each cycle.

Building the time out of entropy transport sidesteps that, because what is being integrated is not the clock's position but the entropy exchanged along its path. Across repeated expansion and recollapse, the paper reports that entropic time grows monotonically almost everywhere, with small wiggles attributable to sampling.

What the measurements showed

Three results carry the argument.

• The ordering works. Entropic time robustly sequenced events in the bright sector across repeated cycles, even as the cloud expanded and collapsed.
• The rate is physical. The slope of entropic time against laboratory time was set by the flow of entropy into and out of the bright sector. Slow the exchange and internal time slows with it.
• With no exchange, nothing to order. Raise the barrier and the sectors stop trading atoms; entropy stops changing; the internal clock has nothing left to measure. The paper notes a near-stationary, heat-death-like analogue.

Barrier height was the control knob: changing it changed the exchange between the two sectors. The paper reports that the entropy assigned to the whole system remained constant within experimental uncertainty. That matters because a subsystem can exchange entropy even when the combined system does not show a net increase. The approximately 5 per cent statistical uncertainty reported for plotted measurements should not be confused with a precision measurement of the entropy of the cosmos.

An equation that fits, and the limits of that fit

Barontini then did the harder thing. He derived an effective Schrödinger equation parameterised by entropic time and showed it could reproduce the measured evolution of the bright sector.

The machinery behind it is standard and approximate. The bright sector is modelled with an effective non-Hermitian Hamiltonian, the imaginary term accounting for atoms gained and lost across the barrier. In the mean-field regime the real part is the Gross–Pitaevskii Hamiltonian. The treatment averages the interaction term over density and uses spherical rather than axial coordinates, simplifications the paper states do not affect the physical content.

Now the caveat a careful reader should hold onto. When the simulated width was compared with measurement in the near-zero-barrier case, the agreement was good — but the parameters were inferred from the data themselves. That is model calibration, not independent confirmation of a cosmological model.

The paper is candid about this. It presents an experimentally grounded testbed for internal-time constructions, not a verdict on quantum gravity.

What “time stopped” really means

The result gives researchers a controllable test of an internal clock. It does not establish that time is an illusion, or select the correct theory of quantum gravity. The atoms obey ordinary laboratory physics; the cosmological connection is an analogy in the reduced mathematical model.

“Time stopped” therefore means the chosen entropic parameter stopped advancing. A laboratory clock could continue recording the interval. The interesting question is what counts as an observable sequence for the subsystem, rather than whether the researchers froze the world around their apparatus.

A small laboratory can ask a very large question

There is precedent for using ultracold atoms to investigate cosmological ideas. In work reported by NIST in 2018, an expanding atomic condensate stretched sound-like disturbances in a way analogous to cosmological redshift. That older experiment studied expansion, not Barontini’s entropy-based clock.

The common strategy is powerful: identify a mathematical relationship shared by a difficult-to-access system and one you can control, then test the relationship in the laboratory. The strength of that strategy depends on knowing which features carry over and which do not. A tabletop model earns its value through those boundaries.

Where the next tests point

The paper's own forward-looking list is the most honest guide to what remains open. It proposes trying alternative internal clocks beyond the centre of mass, examining a quantum bounce scenario, using a Loschmidt echo to probe reversibility, and adapting the platform toward analogue horizons.

Those proposals address different open questions: whether another internal clock gives a compatible description, how a collapse might turn into a bounce, how reversible the dynamics are, and whether related analogues can be engineered. They are possibilities described for future work, not accomplishments of this experiment.

Why a general reader should care

Three reasons, in ascending order of importance.

First, this is a rare case of a question usually confined to cosmology being placed on a laboratory bench. Ideas about the universe as a whole can now be probed with lasers and a camera.

Second, the method is instructive. Build an internal parameter, validate it against an external one you trust, then use it to order events the external one cannot reach. That pattern recurs across science.

Third, and most interesting, the result is conditional. Entropic time orders events because of a sign condition, and stops when entropy stops moving. The fascinating part is not that time emerged from a cloud of atoms. It is that the emergence came with an explicit instruction manual, a stated failure mode, and a list of experiments that could break it.

That is what a real advance in fundamental physics tends to look like: not a slogan, but a device that can be wrong in specific, testable ways.