What STAR actually counted
The measurement is a ratio. In the isobar run — ruthenium-96 on ruthenium-96 and zirconium-96 on zirconium-96, at 200 GeV per nucleon pair — STAR took the baryon number left near mid-rapidity, the direction perpendicular to the beamline, and divided it by the difference in electric charge that arrived with it. Two nuclei of the same mass number and different proton count make that division clean: the charge changes, the baryon number does not. In the 10 percent most central collisions the ratio came out at 1.84, with a statistical uncertainty of 0.02, a systematic one of 0.09 and one of 0.16 from feed-down contributions. If every valence quark carried a third of a baryon and nothing else carried any, the ratio would sit at one.[1]
The second handle uses no ratio at all. In photonuclear events, where a photon from one gold nucleus strikes the other, at 54.4 GeV per nucleon pair, STAR followed how the net proton yield falls as it moves away from the photon's side. That fall is close to exponential, and the slope came out at 1.04, plus or minus 0.22. Regge theory puts the junction's slope between 0.42 and 1. Two measurements, two different observable quantities, and disagreement with the quark bookkeeping in the same direction.[1]
Where the junction enters
The junction is an old idea. Proposed in the seventies, it describes the three valence quarks of a proton as tied together at a Y-shaped meeting point of neutral gluons, and it places the baryon number at that meeting point rather than on the quarks. The picture predicts what STAR sees: because no single quark carries the junction, the collision strips the quarks away, the knot lags behind near the middle while they travel on, and the baryons gather where the electric charge did not arrive. Other readings are available too. The predictions in the comparison come from event generators built on the valence-quark assumption, and a generator that lets colour reconnect among the outgoing fragments moves its prediction up toward the data without invoking a junction at all. The force of the measurement therefore lies in how far the standard generators fall short of it.[1]
The paper is careful about this, and the carefulness is the part worth carrying away. It writes that neither account of what carries baryon number has been unequivocally verified by experiment to date, and it frames its own result as disfavouring the valence-quark picture. Between disfavouring one model and establishing another, a gap is left that only another measurement can close.[1]
What could close it?
RHIC stopped taking data in early 2026, so the isobar sample is closed; whatever else those collisions hold, no new ruthenium or zirconium event will be added to it. The discriminating step has to come from somewhere else: an apparatus that can see the junction's signature in a different initial state, or a fresh analysis that changes which assumptions the comparison generators carry. The signal to watch is narrow and specific: will an independent measurement of net-baryon transport again give a mid-rapidity ratio clearly above one, and will the feed-down share of 0.16, the largest single uncertainty in this result, shrink when it does?[1]