A day on Earth isn’t exactly 24 hours. Physicists at the University of Alberta now say the solid inner core’s gravity explains a big part of why Earth’s length of day keeps wobbling. Huifeng Zhang and Mathieu Dumberry compared rotation data from 1964 to 2019 with seismic reconstructions of the inner core’s motion, and the pull fit swings of a few milliseconds.
Zhang is a PhD student in physics, and Dumberry is a physics professor. Canada’s Natural Sciences and Engineering Research Council funded the work, and Nature ran the paper on September 23, 2026.
Milliseconds traded with the core
Lots of things nudge the day, according to the paper. Tides act from day to day, and winds work over seasons and years. Over millions of years, friction from lunar tides makes the day longer, and some scientists even propose a billion years of 19-hour days in Earth’s past.
Zhang and Dumberry took those effects out, along with ice-age rebound and water shifting around the surface. A filter with a 30-year cutoff then left one broad swing of about 60 to 70 years.
According to the university, the liquid core speeds up over a few decades and then slows over a few more. The mantle, a rock layer about 1,860 miles (3,000 kilometers) thick, does the opposite because Earth’s angular momentum has to stay the same. It’s the mantle’s speed that sets how long our day is.
Scientists have known about this exchange for some 30 years. Clocks have to follow Earth’s spin, which is why leap seconds exist, at least until their planned end. Nobody had pinned down the force behind it.
“We still lack a successful ΔLOD prediction based on a modelling of the torque on the mantle,” the authors write. ΔLOD is their term for changes in the length of day.
Gravity against two boundary forces
The inner core is not perfectly round, and over time its high spots line up with dense regions at the bottom of the mantle.
In the study, flows in the liquid outer core use magnetic forces to pull the inner core slightly out of line. Gravity then twists the mantle.
The team tested this gravitational torque, as physicists call a twisting force, against two others at the core-mantle boundary. That boundary sits about 2,165 miles (3,485 kilometers) from Earth’s center.
One of them is electromagnetic drag on conducting rock at the base of the mantle. The other is pressure from liquid metal pushing on bumps in the boundary. Earlier predictions of that pressure effect came out about 100 times too large, the paper notes.
In their runs, gravity matched the observed curve, while the two boundary forces mostly worked against it. The authors conclude the swings “are driven by small differences between the (leading) gravitational and (resisting) CMB torques.” CMB stands for core-mantle boundary.
Earthquake waves from South Sandwich
For the inner core, the team used a 2023 reconstruction by Yi Yang and Xiaodong Song of Peking University. It rests partly on earthquake waves that traveled from the South Sandwich Islands to a station in College, Alaska.
That reconstruction shows the inner core rocking about 2.35 degrees back and forth against the mantle. It changed direction around 2010, much as liquid iron beneath the Pacific switched direction around 2010.
Outer core flows came from changes in the magnetic field. Each run picked one of 400 versions of that flow model.
An inner core that deforms
The model fit best when the inner core could slowly change shape. The best estimate for that adjustment is about 10 years, and the 95 percent confidence range runs from roughly four to 31 years.
The university says the core can “deform viscously” on that scale. That shape change moves the peak of the gravitational torque to about 2000 instead of 2010.
If electromagnetic drag supplies most of the resistance, the numbers fit an iron-rich layer about 1.2 miles (2 kilometers) thick at the base of the mantle. The study infers that layer but does not observe it.
One 70-year cycle on record
Every result depends on the inner core and flow reconstructions. The paper says its torque histories are “tied to the accuracy” of those models. With a different flow model, the best-fit values moved by up to 30 percent.
The 1964-2019 record also covers a single swing of the cycle. “Whether this flow structure is periodic, and repeats over time, or whether it only reflects the dynamics over the past seven decades, is unknown,” the authors write. Flows from 1915 to 1949 look partly similar, with notable differences.
Changes over 10 to 30 years are still unexplained. The paper says boundary forces probably cause them, yet current models of those forces show no large swings on that timescale.
The full study was published in Nature. The University of Alberta described it in a news article on September 23, 2026.
Image: Christopher C. Finlay, Julien Aubert, Nicolas Gillet / Nature Communications, CC BY 4.0
