Deep earthquakes should not propagate far. At 125 kilometers below Chile's Atacama Desert, rock is too hot and too pressurized to fracture brittlely. The standard mechanism for intermediate-depth earthquakes — dehydration embrittlement, where minerals release water under pressure and the fluid lubricates a sudden slip — operates only within a narrow thermal window. Below a certain depth, the rock is too warm for the mechanism to work. The earthquake should stop.
Jia, Mao, Flores, and colleagues at UT Austin and Chilean institutions (Nature Communications, 2025) found that the 2024 Mw 7.4 Calama earthquake did not stop. It initiated at 125 km depth via dehydration embrittlement in a cold zone of the subducting Nazca slab, then propagated 50 km deeper into rock that was too hot for that mechanism. What carried it past the thermal boundary was a different process: thermal runaway. The friction generated by the initial rupture heated the surrounding rock, weakening it, enabling further slip, generating more heat, weakening more rock. The earthquake rewrote its own physics mid-event. It started as a fluid-assisted fracture and became a heat-driven melt.
This is the first observed instance of an earthquake transitioning between rupture mechanisms during a single event. The thermal boundary that should have stopped propagation was breached because the rupture itself generated the conditions for a mechanism that could operate on the other side.
The structural insight is about processes that generate their own enablers. Dehydration embrittlement cannot cross the thermal limit. Thermal runaway cannot initiate in cold rock — it requires an existing rupture to generate the seed heat. Neither mechanism alone can produce what happened. But the first creates the precondition for the second: the fracture generates friction, friction generates heat, heat activates a mechanism that friction alone could not. The earthquake is not one process. It is a relay — the output of the first mechanism is the input of the second, and the handoff happens at the exact point where the first mechanism would otherwise fail.
The thermal boundary was real. It did stop the first mechanism. It just couldn't stop the second one that the first one built.