Three Mile Island Unit 2
Half the core of a US pressurised water reactor melted, and almost no radioactivity escaped. The containment worked. The accident nonetheless ended new US reactor orders for three decades.
| Date | 1979-03-28 |
|---|---|
| INES rating | 5 Accident with wider consequences |
| Facility | power reactor |
| Unit | Three Mile Island nuclear power plant 2 |
| Reactor type | Pressurised water reactor (PWR) |
| Country | United States |
| Evacuated | No evacuation ordered |
| Deaths at the time | 0 |
|---|---|
| Acute radiation deaths | 0 |
No deaths or injuries. Epidemiological follow-up has found no detectable health effects in the surrounding population; the average dose to a nearby resident was a small fraction of natural background.
A stuck-open relief valve let coolant escape, and the control room instrumentation did not show valve position. Operators, misreading a pressuriser level indication, throttled back the emergency cooling that was working correctly.
What happened
A minor fault stopped feedwater to the steam generators. The reactor scrammed correctly and a relief valve opened to relieve pressure, as designed — and then stuck open, which was not designed. Coolant escaped for over two hours.
The control room had no indicator for the valve's actual position, only for the signal sent to it. Meanwhile the pressuriser level read high, which operators were trained to treat as the reactor being full of water. It was not: steam voids were displacing the level while the core was uncovering.
Acting on that reading, they throttled back the emergency core cooling that had started automatically and was doing the right thing. About half the core melted.
Consequences
The containment building held. Off-site releases were small and no one was hurt.
The industrial and political consequences were enormous. No new US reactor was ordered for thirty years. The reactor at TMI-2 was defuelled over more than a decade; Unit 1 next door ran until 2019 and is now being restarted.
What changed afterwards
TMI reshaped how the industry thinks about operators. The failure was not a component but the interface between people and machine: instrumentation that showed a demand rather than a state, and training that taught a rule rather than a mental model of the plant.
It produced the Institute of Nuclear Power Operations, simulator training against realistic degraded scenarios, symptom-based rather than event-based emergency procedures, and a fundamental redesign of control room information displays.
Other accidents
| Accident | INES | Date |
|---|---|---|
| Chernobyl | 7 | 1986-04-26 |
| Fukushima Daiichi | 7 | 2011-03-11 |
| Kyshtym (Mayak) | 6 | 1957-09-29 |
| Windscale Pile 1 fire | 5 | 1957-10-10 |
| SL-1 | 4 | 1961-01-03 |
| Saint-Laurent A2 fuel melt | 4 | 1980-03-13 |
| Tokaimura criticality accident | 4 | 1999-09-30 |
| Vandellòs I turbine fire | 3 | 1989-10-19 |
| Davis-Besse vessel head corrosion | 3 | 2002-03-06 |
Sources
- World Nuclear Association, Three Mile Island Accident · retrieved 2026-09-07