Summary
A Nature study models changes in Earth’s day length from 1964 to 2019 and finds that gravitational torque associated with the inner core can drive the broad multidecadal variations. In the models, electromagnetic or topographic torques at the core–mantle boundary tend to oppose that motion.
Earth’s day length varies by several milliseconds over periods of decades. A study published in Nature on 23 September 2026 models those changes from 1964 to 2019 and finds that gravitational torque associated with the inner core can account for the broad multidecadal pattern. In the researchers’ model, torques at the boundary between the fluid core and mantle tend to resist, rather than drive, that pattern.
The analysis combines a seismic reconstruction of changes in inner-core rotation with estimates of fluid-core motion derived from changes in Earth’s magnetic field. It offers a model-based explanation for how exchanges between Earth’s deep interior and mantle can alter the planet’s rotation rate.
How the inner core can affect the length of a day
Earth’s length of day, or LOD, changes when its rotation rate changes. Over decades, these fluctuations are caused mainly by exchanges of angular momentum between the mantle and core. The study focuses on variations of several milliseconds, after accounting for estimated contributions from the atmosphere, oceans, lunar tidal friction and other long-term processes.
The inner core’s boundary is not perfectly spherical. In the researchers’ explanation, its long equatorial axis can become misaligned with the gravitational pattern created by mass variations in the mantle. Gravity then exerts a torque on the inner core, with an opposing torque on the mantle. That exchange can change the mantle’s rotation rate and, in turn, the length of the day.
The inner core’s boundary also deforms viscously, gradually moving back towards alignment. The researchers modelled this relaxation alongside changes in inner-core rotation. This matters for the timing of the torque: with viscous deformation included, the model places its maximum near 2000, rather than near 2010 as in a rigid-inner-core scenario.
Comparing the torque models
The team reconstructed the required torque from observed day-length changes and compared it with estimates of gravitational torque and two possible forms of coupling at the core–mantle boundary: electromagnetic coupling, involving magnetic fields and electrically conducting material, and topographic coupling, in which fluid pressure acts on uneven boundary topography.
Using a Bayesian inversion, the researchers tested combinations of torque parameters while sampling 400 realizations of a core-flow model and alternative inner-core rotation histories. They considered electromagnetic and topographic coupling as separate alternatives in the inversion. In both cases, the model could reproduce the filtered multidecadal LOD pattern with a misfit below 0.2 milliseconds. The gravitational torque generally led the changes, while the boundary torque opposed it.
For the electromagnetic-coupling case, the best-fit inner-core boundary relaxation time was 10.2 years, with a 95% confidence interval of 4.1 to 30.5 years. For the topographic-coupling case, the best fit was 7.8 years, with an interval of 1.7 to 21.5 years. These ranges reflect the model fits, not direct measurements of the inner core’s viscosity or shape.
What the result suggests about Earth’s deep interior
The inferred torque strength is consistent, the researchers argue, with a lowermost mantle containing large, nearly neutrally buoyant thermochemical structures and a relatively low-viscosity post-perovskite phase. The analysis also points to a low-viscosity inner core. These are geophysical implications drawn from the torque model, rather than direct observations of those materials at depth.
The study’s reconstruction is limited to broad multidecadal changes over 1964–2019. Its torque estimates depend on the seismic inner-core rotation history and the core-flow models used; the authors report that different core-flow models can shift best-fit torque parameters by up to 30%. The analysis also does not determine whether the roughly 70-year flow pattern repeats regularly. The researchers suggest that shorter-period changes in day length may be governed more by coupling at the core–mantle boundary.