Summary
A Nature study links broad, multidecadal changes in Earth’s day length to gravitational torque from the inner core. Its model-based reconstruction combines seismic estimates of inner-core rotation with core-flow models derived from changes in Earth’s magnetic field.
Earth’s inner core may be the main driver of the planet’s broad, multidecadal changes in day length, according to a model-based study by Huifeng Zhang and Mathieu Dumberry published in Nature. The researchers compared changes in Earth’s rotation from 1964 to 2019 with seismic estimates of inner-core rotation and models of flows in the fluid core, concluding that gravity-linked torque from the inner core drives the long-period changes while torques at the boundary between the core and mantle tend to oppose them.
The finding concerns variations lasting several decades, not the much more familiar 24-hour measure of a day. Earth’s rotation rate changes, so the length of day (LOD) varies by several milliseconds over timescales of roughly 10 to 70 years. The study focuses on the broad, approximately 60–70-year part of that variation.
How the study examined day length
The researchers analysed an LOD record for 1964–2019 after accounting for contributions from the atmosphere and oceans, lunar tidal friction and changes in mass at Earth’s surface. They then applied a low-pass filter with a 30-year cutoff to isolate the multidecadal signal. Changes in LOD reflect exchanges of angular momentum between the mantle and the core: a torque on the mantle changes its rotation rate.
To estimate the torques, the team combined two kinds of model. Seismic observations were used to reconstruct changes in inner-core rotation. Separately, changes in the geomagnetic field were used to infer flows near the top of the fluid core, which inform estimates of electromagnetic and topographic torques acting at the core–mantle boundary. The researchers tested predictions against the filtered LOD record using a Bayesian inversion, sampling different core-flow and inner-core-rotation histories.
Why the inner core matters
The inner core is not perfectly spherical: its boundary has a slight, uneven shape. In the study’s model, the gravitational pull associated with density variations in the mantle tends to align that shape in a preferred direction. As the inner core rotates relative to the mantle, the alignment shifts, producing a gravitational torque. Viscous deformation of the inner-core boundary gradually relaxes the misalignment, affecting the strength and timing of that torque.
The reconstructed inner-core rotation contains a broad oscillation of about 70 years, with an angular excursion of approximately 2.35° relative to the mantle. Using this history, the researchers found that the gravitational torque could reproduce both the timing and size of the multidecadal LOD changes. The best fits require the inner-core boundary to deform over a timescale of years, which shifts the peak in the predicted torque earlier than it would be for a rigid inner core.
Electromagnetic and topographic torques at the core–mantle boundary were modelled as opposing the gravitational torque in the multidecadal reconstruction. The study considered these two boundary mechanisms separately; either could produce a similarly good fit when combined with the gravitational torque. The results therefore support gravity as the main driver of the broad changes without identifying which boundary mechanism contributes most to resisting them.
The reconstruction also links rotation measurements to conditions deep inside Earth. The authors find that the inferred gravitational coupling is consistent with particular properties of the lowermost mantle, including near-neutrally buoyant thermochemical structures and a relatively low-viscosity post-perovskite phase. These are geophysical implications of the torque model, rather than direct observations of those structures. The analysis is centred on the multidecadal signal; shorter, decadal changes in day length involve additional questions about coupling between the core and mantle.