A compass follows the core's field, and the field sometimes turns around
A compass needle is a small magnet, and a magnet turns until it lines up with the magnetic field around it. Earth's field is made by liquid iron moving in the outer core, and nothing in the physics of that motion prefers one direction for the field over the other. Old rocks show the field turning around many times, in the recent geological past about four times per million years on average.
A needle that turns to line up
Lisa Tauxe's textbook Essentials of Paleomagnetism starts with the simplest compass there is: a magnetised needle on a cork in a bucket of water slowly turns until it lines up with the local field. A magnet sitting in a field has an energy that depends on its angle to the field, and that energy is lowest when the two are aligned. So the needle swings until its north-seeking end points along the field, and stays there.
Earth's field does not lie flat. Tauxe describes it pointing north and downward in the northern hemisphere, and north and upward in the southern. A needle free to tilt, called a dip needle, shows this: it lies nearly flat near the equator and stands upright near the magnetic poles. For a field made by a single magnet at Earth's centre, lined up with the spin axis, the tilt follows a rule Tauxe calls the dipole formula: the tangent of the tilt is twice the tangent of the latitude. At 45° latitude that gives a tilt of about 63°. A compass that can only turn flat follows the horizontal part of the field.
The north magnetic pole is a south pole
That is a clash of names, not a mistake in the physics. "North magnetic pole" names a place: where the needle's north end points. As a magnet, that place holds the south end of Earth's field, and it pulls the north end of every compass toward it.
The place also moves. The World Magnetic Model, which NOAA and the British Geological Survey produce as a standard model for navigation, is revised every five years. Its 2025 report, by Arnaud Chulliat and colleagues, puts the north dip pole, where the field points straight down, at 85.76° N, and finds the best single centred magnet tilted 9.21° from Earth's spin axis. The two agencies' December 2025 update measured the north dip pole moving 36 km a year. Near a dip pole the field points almost straight down, so the flat part that a compass can feel is small. The model marks "blackout zones" where that flat part is under 2,000 nT (nanotesla), and says compass accuracy there is degraded.
A dynamo, not a bar magnet
There is no giant bar magnet inside the Earth. Tauxe explains that the field comes from convection in the liquid outer core, which is iron, nickel and some lighter elements. The energy is thought to come partly from the core cooling and partly from buoyancy: as pure iron freezes onto the solid inner core, the liquid left behind is richer in light elements and rises. A conducting liquid moving through a field carries electric currents, and those currents make the field. That loop can sustain itself, which is what a dynamo means here. From outside, the result still looks much like one magnet: Tauxe's breakdown of the field at the surface puts the dipole part at some 90% of it.
Interactive A slice through the Earth from pole to pole, with the field lines outside it. Set how strong the axial dipole is, from today's strength through zero to fully reversed, and where you stand. On the right, a needle free to tilt shows the field's direction where you stand, and a level compass seen from above shows which way its red end points. Every control works with Tab, Enter and the arrow keys.
Why the field can turn around
François Pétrélis, Stéphan Fauve, Emmanuel Dormy and Jean-Pierre Valet start from a symmetry. The equations for magnetism in a moving, conducting fluid are unchanged if every field arrow is reversed. A dynamo that can hold a field one way can hold the mirror image just as well, so nothing in the core prefers today's direction.
In their simple model the field has two stable states, one for each direction, with a hump between them. The turbulent core keeps jostling the field. Climbing toward the hump is slow, because the dynamo pushes back; once over the top, falling into the opposite state is fast. A field that climbs partway and falls back where it started makes an excursion rather than a reversal. Because the push over the hump comes from random jostling, the model makes the time between reversals random too, with no regular schedule.
The rocks show the same lopsided shape. In 2024 Ahmed Nasser Mahgoub, Monika Korte and colleagues compared nine global models of the last reversal, the Matuyama–Brunhes. Most showed it driven mainly by the axial dipole fading away, and every model showed a slow decay followed by a fast recovery in the opposite direction. The main change came close to 778,000 years ago, and the whole process took 21,000 to 32,000 years depending on the model. Tauxe describes a reversal record in which the field drops to about 10% of its average strength while its direction swings from one pole to the other over several thousand years. She also notes that the field spends only 1–2% of its time between the two states.
How often it happens has changed a great deal. Tauxe's summary of the record shows a stretch of some 38 million years in the Cretaceous with no reversals or very few, after which the rate rose to the present average of about four per million years.
Is it flipping now?
The field is getting weaker. Maxwell Brown and colleagues, in 2018, put the decline at about 5% per century since direct measurements of its strength began in 1840. It is still going: in the 2025 World Magnetic Model the dipole term loses 12 nT a year out of 29,352, about 4% per century if that pace held. The December 2025 update also has the South Atlantic Anomaly, the region where the field is weakest, deepening by about 30 nT between the start of 2025 and the start of 2026.
Weakening is not the same as reversing. Brown and colleagues rebuilt the field from 50,000 to 30,000 years ago and found two spells, about 49,000 and 46,000 years ago, with a weak patch much like today's South Atlantic Anomaly. Both times the field recovered without an excursion or a reversal, and they concluded that today's field is probably not reversing. Mahgoub and colleagues found that the field at the start of the last reversal did not look like today's anomaly. Neither result dates the next reversal. Brown and colleagues note that methods to forecast the field cannot yet predict it over long times, so no one can say when the next one will come.
In short
A compass needle is a magnet that turns to line up with Earth's field, and its north end points to the place where, as a magnet, Earth has its south pole. The field is made by moving liquid iron in the core, and the physics works just as well with the field reversed. Every so often the dipole fades slowly and grows back the other way quickly. That has happened about four times per million years in the recent past. The field is weakening today, but the two past weak patches most like today's recovered on their own, and nobody can yet forecast the next flip.
Where this comes from
- A simple mechanism for the reversals of Earth's magnetic field linked only, not reproduced
arxiv.org/abs/0806.3756 - Essentials of Paleomagnetism, 5th web edition, ch. 1: The Physics of Magnetism linked only, not reproduced
pmagpy.github.io/Essentials-JupyterBook/chapter1/ - Essentials of Paleomagnetism, 5th web edition, ch. 2: The Geomagnetic Field linked only, not reproduced
pmagpy.github.io/Essentials-JupyterBook/chapter2/ - Essentials of Paleomagnetism, 5th web edition, ch. 14: The Ancient Geomagnetic Field linked only, not reproduced
pmagpy.github.io/Essentials-JupyterBook/chapter14/ - The US/UK World Magnetic Model for 2025-2030: Technical Report linked only, not reproduced
repository.library.noaa.gov/view/noaa/71569 - December 2025 State of the Geomagnetic Field linked only, not reproduced
www.ncei.noaa.gov/sites/default/files/2025-12/WMM%20SoGF%20Dec2025%20508.pdf - Earth's magnetic field is probably not reversing (PNAS) linked only, not reproduced
pmc.ncbi.nlm.nih.gov/articles/PMC5960311/ - Robustness of characteristics of the Matuyama-Brunhes geomagnetic field reversal found in global models (Frontiers in Earth Science) linked only, not reproduced
www.frontiersin.org/journals/earth-science/articles/10.3389/feart.2024.1443095/full