Tokamak plasmas have a density ceiling. The Greenwald limit, established empirically in 1988, predicts the maximum electron density a tokamak can sustain before the plasma disrupts — a violent termination where confinement collapses in milliseconds, driving intense currents into the reactor walls. Decades of theoretical work have attempted to derive this limit from first principles, treating it as a property of magnetically confined plasma: something about radiation losses, or edge cooling, or transport physics imposes a maximum density that the plasma cannot exceed.
Researchers at EAST, China's fully superconducting tokamak, published results in Science Advances demonstrating that the density limit disappears when the startup protocol changes. By combining control of initial fuel gas pressure with electron cyclotron resonance heating during the ohmic startup phase, they optimized plasma-wall interactions from the first moments of the discharge. Under this modified protocol, the plasma operated stably at densities well beyond the Greenwald limit — in what the plasma-wall self-organization (PWSO) theory calls a density-free regime.
The mechanism is about impurities. Under conventional startup, the initial plasma-wall interaction regime is dominated by conditions that cause sputtered wall material to accumulate in the plasma, increasing radiation losses and driving the density limit. The PWSO theory, originally proposed by Escande and colleagues, predicts that when physical sputtering dominates — which it does when wall interactions are properly managed from the beginning — the plasma and wall reach a self-organized balance where the density-limiting impurity accumulation does not occur. The limit wasn't a property of magnetically confined plasma. It was a property of how the plasma was started.
This is structurally different from discovering that a limit is in the wrong place. The Greenwald limit is accurately located — it correctly predicts when plasmas disrupt under conventional startup. The scaling law works. The physics it describes is real. What the EAST result shows is that the physics is conditional on the interface protocol. Change the protocol and the same plasma, in the same machine, with the same magnetic geometry, operates in a regime where the limit has no purchase. The limit wasn't misidentified. It was never intrinsic.
Forty years of theoretical work asked why tokamak plasmas have a density ceiling. The question embedded the assumption that they do. The Greenwald limit, validated across dozens of machines over four decades, made the assumption empirically robust. But empirical robustness under uniform conditions is indistinguishable from universality until someone changes the conditions. Every tokamak in the world used similar startup procedures — because startup procedures are treated as engineering choices, not physics parameters. The limit appeared universal because the procedure was universal.
The general pattern: when every experiment shares a protocol, properties of the protocol are indistinguishable from properties of the phenomenon. The Greenwald limit is not wrong. It is a limit of the conventional startup procedure, accurately measured and correctly described. Calling it a limit of tokamak plasmas was the error — an error invisible as long as startup protocols varied less than the physics they were studying.