Why does Earth's magnetic field flip?

The question

A compass needle points north because Earth is a giant magnet. Its field is generated deep inside the planet, in the liquid iron of the outer core, where churning metal acts like a dynamo.

In 1906 the French physicist Bernard Brunhes found volcanic rocks that were magnetised the "wrong" way, as if the compass had pointed south when they cooled. In 1929 Motonori Matuyama in Japan showed that such reversed rocks were systematically older. The conclusion was astonishing: the planet's magnetic field periodically turns upside down. It has done so hundreds of times. Why it happens, and why so irregularly, is still not understood.

What we know

Reversals are real and global. When lava cools, or sediment settles, its magnetic minerals record the direction of the field. Layers dated around the world show the same sequence of normal and reversed periods.
They proved plate tectonics. In 1963 Fred Vine and Drummond Matthews explained the striped magnetic pattern of the ocean floor: new crust forms at mid-ocean ridges and records the field as it spreads away, so reversals are written on both sides like a tape recording. This was one of the decisive arguments for moving plates.
Reversals are irregular. Some periods without a flip lasted only tens of thousands of years. One, the Cretaceous Normal Superchron, lasted about 38 million years, roughly 121 to 83 million years ago. The last full reversal happened about 773,000 years ago.
A reversal is not instant. During the transition the field weakens strongly and becomes complicated, with several poles. Estimates of its duration range from a few thousand years down to centuries, depending on the record.
There are also "excursions": the field wanders far from its usual direction and then returns without flipping. The best known, the Laschamp excursion, happened about 41,000 years ago.
In 1995 Gary Glatzmaier and Paul Roberts produced the first computer simulation of the core in which the field reversed by itself, without any outside push. Reversals turned out to be a natural behaviour of a dynamo.
The field has weakened by roughly 10% since systematic measurements began in the 19th century, and an area of especially weak field lies over the South Atlantic.

Main hypotheses

Chaotic internal dynamics

The flow in the core is turbulent. A reversal may be a random event, like a chaotic system jumping between two stable states. That would explain the irregular timing, but it would also mean that reversals can't be predicted precisely.

Control from the mantle

The rate of heat flow out of the core depends on the mantle above it, which changes slowly as plates sink and plumes rise. Changes in this heat flow could make reversals more or less frequent, explaining the long superchrons. This connects reversals to plate tectonics.

Triggers

Some researchers have proposed outside triggers, such as large impacts or sudden changes in the mantle. There is no convincing evidence for them, and simulations show that no trigger is needed.

Where understanding ends

We know that the geodynamo can reverse by itself, and simulations reproduce reversals. But simulations run at conditions far from those of the real core: the real liquid iron is much less viscous and the flow much more turbulent than any computer can model. Whether simulated reversals happen for the same reasons as real ones is not certain.

The records are also imperfect. Rocks capture the field only at certain moments, often smoothed over thousands of years, and details of the transition are hard to read. So we know little about what exactly happens during the flip, and how fast.

And there is no way to predict the next reversal. The current weakening is sometimes presented as a sign that one is beginning, but the field has weakened and recovered many times without flipping. Most researchers think a reversal in the coming centuries is unlikely but cannot rule it out. Its effects on satellites, power grids and life are also debated: the claimed link between the Laschamp excursion and environmental changes around its onset, 42,000 years ago, remains contested.

What would settle it

Simulations that come much closer to the real conditions of the core and show which processes control the timing of reversals.
More continuous and better-dated records of past reversals and excursions, especially high-resolution records of the transitions.
Better measurements of the present field and of the deep mantle, showing whether and how the boundary between core and mantle shapes the field.