The Three Mile Island accident, step by step

Three Mile Island 2 was a Babcock & Wilcox pressurised water reactor of 906 electrical megawatts, on an island in the Susquehanna river. It had been in commercial operation for three months. A PWR has two separate circuits: the primary, at about 155 bar, stays liquid and carries heat from the core to the steam generators; the secondary boils on the other side of those tubes and drives the turbine.

Primary pressure is set by the pressurizer: a vertical vessel connected to the circuit, water at the bottom and steam at the top, with heaters and sprays to raise or lower pressure. At the very top sit relief valves, among them a pilot-operated one — the PORV — that opens by itself if pressure exceeds the limit and discharges to a drain tank inside containment.

Babcock & Wilcox steam generators are once-through: water goes in and comes out as steam with no recirculation, with very little water inventory on the secondary side. That makes them fast and efficient, and leaves very little margin: cut the feedwater and they dry out in a couple of minutes.

At four in the morning, a problem in the secondary water polishing system left the feedwater pumps with no flow. The turbine tripped at once. With no feedwater, the steam generators stopped taking heat out of the primary within seconds.

Heat from the core built up in the primary water, which expanded and raised the pressure. Seconds later the PORV opened, as designed, and eight seconds in the reactor tripped: the rods fell and fission stopped. Up to that point every piece of equipment did what it was supposed to do.

Auxiliary feedwater started automatically, but its block valves had been left closed after maintenance and nobody noticed for eight minutes. That was the first hint of what kind of accident this was: the systems were there, and the problem was knowing what state they were in.

With the reactor tripped, primary pressure fell and the PORV should have shut. It did not. It stayed open for two hours and twenty minutes, draining primary water into the drain tank until its rupture disc burst and radioactive water spilled onto the containment floor and from there, pumped, into the auxiliary building.

In the control room, the PORV lamp said it was closed. It did not measure stem position: it showed that the electrical signal to the solenoid had been removed. The valve had been told to close and the lamp reported that faithfully. The valve, mechanically stuck, was still open.

A loss-of-coolant accident was under way, and none of the indicators the operators were watching said so. The discharge line temperature was high, but that also happens after a normal opening, and there was no reading in the room that told one case from the other.

High-pressure injection started on its own, as it should on a pressure drop, and began pushing water into the primary. And then the operators throttled it back.

The reason is the crux of Three Mile Island. The only inventory indication a PWR gives its operators is pressurizer level, and that level was rising. It was rising because the primary was losing pressure and the water was starting to boil inside the circuit: steam forming in the high points pushed water up into the pressurizer. The level rose precisely because the reactor was running out of water.

Training said that filling the pressurizer solid — losing the steam bubble that controls pressure — was one of the worst things that could happen in a PWR. The operators did what they had been taught to do with what they were seeing: they cut back injection so as not to fill it. For hours, the system that could have saved the core was throttled by hand while a reliable instrument showed, quite accurately, a number that meant the opposite of what it looked like.

With the PORV open and injection throttled, the water level in the vessel fell below the top of the fuel. The primary coolant pumps were also stopped, because they were vibrating on the steam they were swallowing, and with them went forced circulation.

The uncovered fuel heated up, the zirconium cladding reacted with steam to make hydrogen and the core degraded. The later reconstruction showed that at least 45% of the core, some 62 tonnes, melted, and that about 19 tonnes of molten material reached the lower plenum of the vessel. The vessel did not fail.

The hydrogen that collected in containment deflagrated on the afternoon of the 28th and produced a pressure spike the building absorbed without damage. Over the following days, public alarm centred on a gas bubble at the top of the vessel, feared to be explosive; it was vented between 30 March and 1 April. The containment — the concrete-and-steel structure the RBMK did not have — held back essentially all the radioactive material that had left the core.

The release was about 370 PBq, almost entirely noble gases — krypton and xenon — which neither deposit nor enter the food chain, plus a very small amount of iodine. The average dose to the population within ten miles was 0.08 mSv and the highest estimated individual dose about 1 mSv: the order of a chest X-ray, against a natural background of some 3 mSv a year.

More than a dozen independent studies over decades found no health effects in the population attributable to the accident. The damage was of another kind: the cleanup ran from August 1979 to December 1993 and cost about 973 million dollars; the damaged fuel left the island for Idaho National Laboratory, with the last shipment completed in April 1990. The accident was rated level 5 on the INES scale.

Three Mile Island did not change reactors, it changed the people who run them and what those people are shown. Training moved from event-based procedures to symptom-based diagnosis, on simulators that replicate each control room. Critical valves got real position indication. Control-room design was redone with human-factors criteria, and INPO was created so the US industry would share operating experience instead of keeping it.

Unit 2 never operated again and sits in monitored storage. Unit 1, next door, kept generating until 2019, and in 2024 its owner announced a restart, renamed the Crane Clean Energy Center, with a return to service planned for 2027.