The LHC made the early universe in a smaller cup this time. Read what "signs of" is doing in that sentence.

For about ten days in the summer of 2025, physicists at the Large Hadron Collider did something they had never done before: they collided nuclei of oxygen with oxygen, neon with neon, and oxygen with protons, at energies high enough to melt matter into its constituent parts. A year later, the four large experiments that ringed those collision points — ALICE, ATLAS, CMS and LHCb — have each reported that they saw, in the debris, the fingerprints of quark–gluon plasma. That is the state of matter the entire universe was made of for a few millionths of a second after the Big Bang, before it cooled enough for quarks to bind into the protons and neutrons that everything is now built from. Recreating it, even for an instant, even in a volume smaller than an atomic nucleus, is one of the genuinely astonishing things human beings have learned to do.

So let me be honest about the wonder first, because it is real, and because the skepticism I am about to spend the rest of this piece on only means anything if you understand what it is aimed at. Nobody disputes that the collisions happened. Nobody disputes that the detectors recorded exactly what they recorded. The measurement is solid. The question — the entire question — is what the measurement is a measurement of, and here the physicists have been far more careful than the coverage. The word that keeps appearing in their own summaries is "indications." Signs. Consistent with. Evidence for. Those are not synonyms for "we made quark–gluon plasma in oxygen," and the gap between them is the story.

Why you smash lead, and why this time they didn't

To make quark–gluon plasma on purpose, the standard recipe is to collide the heaviest nuclei you can accelerate. At the LHC that means lead: 208 protons and neutrons per nucleus, slammed together so violently that for a fleeting moment a region a few femtometres across reaches trillions of degrees, hot enough that quarks and gluons stop being confined inside individual protons and flow instead as a single, almost frictionless fluid. That fluid is the plasma. It exists for something like ten-yttoseconds — a ten-septillionth of a second — and then it cools and freezes back into ordinary particles that fan out into the detector. You never see the plasma. You see the spray it leaves, and you reason backwards to the fluid that must have produced it.

For two decades that reasoning worked cleanly because lead is big. A big nucleus makes a big, hot, long-lived droplet, and the signatures are loud. The trouble started around a decade ago, when experiments began looking at collisions that were supposed to be too small to make plasma at all — a single proton striking a lead nucleus, and then, more provocatively, one proton hitting another — and found some of the same signatures anyway, quieter but present. That was not supposed to happen. A proton is a single particle; there is no room in it for a fluid. Either the signatures did not mean what everyone thought they meant, or the smallest systems were making tiny droplets of plasma too, and our intuition about how small is too small was simply wrong.

Oxygen and neon are the experiment designed to sit exactly in that argument. An oxygen nucleus has sixteen protons and neutrons; neon has twenty. They are far heavier than a lone proton and far lighter than lead — a bridge across the very gap where the physics gets confusing. If quark–gluon plasma switches on somewhere between a proton and a lead nucleus, oxygen and neon are where you would go looking for the switch. That is why this run mattered enough to carve dedicated beam time out of a collider whose every hour is fought over.

Nobody disputes that the collisions happened. The question is what the measurement is a measurement of — and there the physicists were far more careful than the headlines.

What they actually measured

Two signatures are doing most of the work, and they are worth understanding, because once you see what each one is sensitive to, you can judge for yourself how strong the claim is.

The first is called elliptic flow. When two nuclei collide slightly off-centre, the region where they overlap is not round; it is shaped like an almond, elongated in one direction. If that region is filled with an ordinary gas of independent particles, the almond shape means nothing — the particles fly out every which way, blind to the geometry. But if it is filled with a fluid, the shape matters enormously: the fluid pushes harder along its short axis, where the pressure gradient is steepest, and the particles that stream out carry that asymmetry with them. Measure the debris flying out preferentially in one plane, and you have measured, indirectly, that a fluid was there to squeeze it. ALICE, ATLAS and CMS all report this elliptic flow in the oxygen and neon data, and — this is the elegant part — they see it ordered: strongest in neon, weaker in oxygen, weaker still than in lead. That ordering is not arbitrary. Neon-20 is a subtly deformed, clustered nucleus, less round than oxygen, so its head-on collisions start out more almond-shaped to begin with. The plasma interpretation predicts that the intrinsic shape of the nucleus should print itself onto the flow of the debris, and the intrinsic shapes line up with the measured order. When a prediction about nuclear structure shows up in the pattern of particles hours of analysis later, that is the kind of coincidence that is hard to fake.

The second signature is subtler and, to my eye, the more interesting one. It is called parton energy loss, and the idea is this: when a very energetic quark or gluon is produced inside the collision and then has to travel out through whatever medium the collision created, it loses energy along the way, the way a bullet slows in water. If there is nothing there — no plasma, just vacuum — it sails out at full energy. If there is a hot, dense medium, it comes out degraded. CMS looked at this by counting high-energy charged particles coming out of oxygen–oxygen and neon–neon collisions and comparing that count to what you would expect from simply adding up a lot of ordinary proton–proton collisions. They found fewer than expected — a suppression — which is what you would see if the energetic particles had to plough through something on the way out. ALICE reports a matching suppression in neutral pions. LHCb, looking at particles built from charm quarks, finds the suppression growing as you move from oxygen to the heavier neon, exactly the direction you would expect if a larger nucleus makes a larger volume of medium to traverse.

The honest version, which is also the more interesting one

Here is where a good science reporter has to slow down. Every one of those measurements is real, and every one of them is consistent with a small droplet of quark–gluon plasma. "Consistent with" is not "caused by." Elliptic flow, in particular, has a well-known rival explanation that has nothing to do with plasma at all: the initial state of the colliding nuclei, before any fluid forms, may already carry correlations — a picture that goes by the name colour-glass condensate — that can mimic the flow pattern a fluid would produce. Physicists have argued about whether small-system flow is a final-state effect (a fluid squeezing) or an initial-state effect (correlations that were there from the start) for the better part of ten years, and the oxygen and neon data do not end that argument. They constrain it. They make the fluid interpretation more comfortable and the alternatives more strained. That is progress, and it is not the same thing as a verdict.

This is why the collaborations wrote "indications" and not "discovery," and it is the most credible thing about the whole announcement. The temptation, when you have four independent experiments all pointing the same way, is to round up. Four signatures, four detectors, one conclusion — surely that is a discovery. But the signatures are not fully independent of one another, and they share the same interpretive ambiguity: each is consistent with plasma and each has a non-plasma story that has not been fully killed. Reporting them as strong, mutually reinforcing evidence, while declining to call the case closed, is not hedging. It is the correct description of the state of knowledge.

• Demonstrated: the LHC collided oxygen and neon nuclei in 2025, and four experiments recorded elliptic flow and particle suppression in the debris.
• Strongly supported: those patterns are what a tiny, short-lived droplet of quark–gluon plasma would produce, and the flow even tracks the nuclei's intrinsic shapes.
• Not yet settled: whether a fluid actually formed, or whether initial-state correlations mimic part of the signal — the decade-old small-system debate the data narrows but does not close.
• Announced elsewhere: "the smallest drop of the early universe," a lovely phrase that runs slightly ahead of what "indications of" will support.

On what timescale we will actually know

The thing to understand about the 2025 oxygen and neon run is how short it was. Ten days. A collider that spends most of the year colliding protons was handed a sliver of time to do something genuinely new, and the experiments made the most of it, but a sliver is a sliver. The statistics are what they are. The decisive tests — the ones that could actually distinguish a fluid from an initial-state mimic — need more collisions, more collision species to fill in the gap between oxygen and lead, and the patient cross-checking between theory and data that turns a suggestive pattern into an established fact. None of that happens on the timescale of a press release. It happens over years, in the unglamorous back-and-forth of papers that refine the last one's error bars.

I have a soft spot for this kind of result precisely because it resists the hype cycle. There is no product here, no funding round riding on the announcement, no date on a slide that will quietly slip next year. There is a hard measurement, an honest interpretation of what it does and does not prove, and a community that has been arguing in good faith about the same question for a decade and just gathered better evidence without pretending the argument is over. The early universe did not fit in this cup, exactly. But we learned something real about the smallest volume in which its physics might switch on — and the people who learned it were careful enough to tell you which part they are still not sure of. In a field that too often sells the dream, that restraint is the finding I would frame.