Earth's magnetic field is generated by convection in the outer core — liquid iron flowing in complex patterns, driven by the planet's rotation and heat escaping from the inner core. The process is called the geodynamo, and for decades, models of it focused on the dynamics of the iron itself: how fast it flows, what patterns it forms, how those patterns generate and sustain the dipole field. The boundary conditions — the surfaces the iron presses against — were treated as approximately uniform. The bottom of the mantle was assumed to be a smooth, roughly isothermal ceiling above the churning core.
Biggin and colleagues (Nature Geoscience, 2026) combined 265 million years of palaeomagnetic data with supercomputer simulations of core dynamics and found that the ceiling is neither smooth nor isothermal. Two massive, intensely hot rock formations sit at the base of the mantle, approximately 2,900 kilometers beneath Africa and the Pacific Ocean, separated by a pole-to-pole ring of cooler material. The hot regions are not participants in the geodynamo — they are part of the mantle, essentially static on the timescale of core convection. But they determine the pattern of the field by controlling where the iron below them flows and where it stagnates. Beneath the hotter regions, liquid iron may idle rather than convecting vigorously. Beneath the cooler regions, convection runs hard.
The result: some features of Earth's magnetic field have remained stable for hundreds of millions of years, while others changed dramatically. The stable features correspond to the geometry of the mantle structures. The changing features correspond to the dynamics of the core. The field is a composite — partly determined by the engine's internal dynamics, partly determined by the geography the engine operates within. And the geography, because it is set by the mantle rather than the core, persists on a much longer timescale than the dynamics it controls.
This inverts the usual attribution. The magnetic field is produced by the core, so the core appears to be the explanatory object — understand the core, understand the field. But the pattern of the field — where it's strong, where it's weak, which features persist and which fluctuate — is set by a structure that doesn't produce the field at all. The mantle's thermal geometry is not a participant in the geodynamo. It is a boundary condition, and boundary conditions are the variables that the system cannot modify. The core can change its flow patterns, reverse its polarity, strengthen or weaken its output. It cannot change the thermal structure of the mantle above it. The thing the system cannot change is the thing that determines its large-scale behavior.
The general pattern: when a dynamic system operates within a fixed boundary, the boundary contributes nothing to the system's mechanism but determines the system's persistent structure. The mechanism explains how the output is produced. The boundary explains why the output has the shape it does. These are different explanations, and the boundary explanation is often invisible because boundaries are not doing anything — they are simply there, and the system works around them. The most enduring features of a dynamic system may not be products of its dynamics at all. They may be imprints of the container it cannot escape.