Earth’s Core Gravity Is Changing How Long Days Last

A Nature paper published on 23 September 2026 finds that Earth’s core is changing the length of a day by several milliseconds over decades. The driver is gravitational torque between the solid inner core and the rocky mantle, with two other deep forces pushing back.

Those milliseconds will not move a wristwatch, but they belong on the same ledger timekeepers use when they decide whether clocks must skip a second.

Gravity Wins a Three-Way Fight Underground

Huifeng Zhang, a University of Alberta doctoral candidate, and Mathieu Dumberry, a geophysics professor at the same university, are listed as equal contributors. They compared torque models with length-of-day records from 1964 to 2019, after winds, ocean currents, lunar tidal friction, glacial rebound, and water-mass shifts had been stripped out.

A low-pass filter with a 30-year cutoff left the decade-scale curve the torques had to match. The paper, titled gravitational torque drives multidecadal day changes, finds that gravity supplies the swing in the right direction. Electromagnetic and topographic torques at the core-mantle boundary generally oppose it.

“Before we obtained the result, we didn’t know they are competing with each other,” Zhang said. The best-fit gravitational strength is 1.4 × 10^19 N m. Dumberry said the inner core wants to be aligned with the mantle because dense regions attract each other, then westward flow in the liquid outer core drags it out of line, and gravity pulls it back east.

What I find particularly exciting is that these different pieces of information can come together to provide a more coherent picture of Earth’s deep interior, a region that is extremely difficult to observe directly.

Huifeng Zhang, doctoral candidate, University of Alberta

The reconstruction is only as good as the seismic inner-core track and the core-flow models taken from the changing magnetic field, a limit the paper states in plain language. Treat “the core” as one smooth residual, and that limit gets buried.

Three Torques, Only One With the Right Sign

Geophysicists have known since 1988, when Dominique Jault and colleagues published the angular-momentum budget, that decade-scale day-length swings come from exchange between the core and the mantle. The open question was which physical torque does the exchanging.

Zhang and Dumberry rebuilt three candidates. Gravitational coupling is the twist from density anomalies in the inner core pulling on mass anomalies in the mantle. Electromagnetic coupling is magnetic drag on a conducting layer at the base of the mantle. Topographic coupling is fluid pressure on bumps along the core-mantle boundary.

THE THREE CORE-MANTLE TORQUES

For plausible amplitudes, the electromagnetic and topographic predictions land in the right size range and still miss the phase. Their day-length curves run broadly opposite to the observations. Gravity matches both amplitude and phase once the inner core is allowed to change shape as it rotates.

Zonal flow in the outer core lines up with that picture. Near the core-mantle boundary, polar flow runs the other way from mid-latitude and equatorial flow, which is what a gravity-led torque budget expects.

The Inner Core Slowed Around 2010

The gravitational torque is built from a seismic history of how the inner core turns relative to the mantle. That history, used for 1964 to 2021, is a broad oscillation with an amplitude of about 2.35 degrees.

Around 2010 the differential rotation switched from prograde to retrograde, a turn other seismic studies have also picked up. A rigid inner core would have put the peak gravitational torque near that date. The fit instead needs the inner core to deform, which shifts the torque maximum to around 2000.

On a long average, westward flow at the core-mantle boundary still applies a westward torque on the mantle. A slow, permanent eastward inner-core drift balances it. The best-fit steady drift is 0.12 degrees per year, in line with dynamo models run at Earth-like conditions. Decade-scale day-length changes are the fluctuations around that mean balance, not a reversal of the planet’s spin.

Public discussion still folds this result into the 2023 headline that the inner core “changed direction.” Zhang and Dumberry are doing different work. They take that relative slowdown as an input, then ask which torque actually turns the mantle.

Older Models Overstated How Strong That Gravity Is

University of Alberta physics describes the same geometry in simpler terms. Small changes in inner-core spin exert gravitational torque on mantle mass anomalies, while friction and magnetic drag at the core-mantle boundary resist the tug. The observed milliseconds are the leftover imbalance. The rocky shell in that picture is about 3,000 kilometres thick.

The authors write that gravitational coupling is weaker than earlier estimates. Weak coupling fits a lower mantle in which the large piles are hot and compositionally denser than their surroundings, close to neutral buoyancy, with low-viscosity post-perovskite lenses between them.

WHAT THE FIT IMPLIES UNDERGROUND

• Mantle piles: Near-neutral thermochemical piles that are hot and compositionally dense, which lowers the gravitational coupling strength.
• Post-perovskite lenses: Low-viscosity material between those piles, consistent with a weaker gravity torque than older models used.
• Basal iron layer: A highly conducting, iron-enriched sheet a few kilometres thick at the bottom of the mantle, the layer electromagnetic coupling needs.
• Inner-core viscosity: A relaxation time of 10.2 years, so the inner core changes shape far faster than the mantle.

Conductivity of the lowermost mantle and possible stratification at the top of the fluid core are less tightly pinned. The paper treats those as open, and says better magnetic maps and better flow models will have to close them.

Ice Melt Bought Timekeepers Three Extra Years

Length of day is also a climate record, and that is where the torque fight stops being an interior curiosity. Duncan Agnew, a geophysicist at the Scripps Institution of Oceanography at the University of California San Diego, showed in 2024 that melt from Greenland and Antarctica has been slowing Earth’s spin by moving mass toward the equator. Remove that surface brake, and the core’s long contribution has been speeding the solid Earth up.

Agnew’s extrapolation said UTC would need a negative leap second by 2029. Without the recent surge in polar melt, that problem would have arrived 3 years earlier, in 2026. “This is an unprecedented situation and a big deal,” Agnew said. A negative leap second has never been inserted, and clock hardware is widely untested for a second that disappears.

His core term is a long acceleration, not the filtered 30-year-plus wobble Zhang and Dumberry now budget. They are different pieces of the same observable. Incomplete torque accounting is why a climate fingerprint in day length has been so hard to pull out, and why subtracting ice melt from one smooth core residual still leaves the clock partly unread.

THE CLOCK, THE CORE AND THE CALENDAR

1. 1988: Dominique Jault and colleagues show that core-mantle angular-momentum exchange can change Earth’s rotation.
2. Around 2010: Inner-core rotation relative to the mantle switches from prograde to retrograde.
3. 1 January 2017: The last leap second takes effect; none has been negative, and the running count of insertions reaches 27.
4. March 2024: Agnew’s ice-melt budget delays a first negative leap second from 2026 to 2029.
5. 6 July 2026: IERS Bulletin C 72, issued in Paris, announces no leap second at the end of December 2026, with UTC trailing TAI by 37 seconds since 2017.
6. 23 September 2026: Zhang and Dumberry publish the gravity-led torque budget for 1964 to 2019.
7. 2026: Draft Resolution C before the 28th CGPM would start continuous UTC on 20 May 2027 and raise the allowed |UT1-UTC| gap to 3,600 seconds.

Christian Bizouard, director of the IERS Earth Orientation Center at Observatoire de Paris, signed Bulletin C 72. A March 2025 workshop of the Consultative Committee for Time and Frequency and the IERS put the chance of a negative leap second at 30 percent by 2035, which is the pressure behind the draft. Until the 28th CGPM votes, 20 May 2027 is a proposed date, not a switch that has been thrown.

Can Gravity Also Drive the Six-Year Pulse?

Day length also wobbles on a cycle of about six years, a signal Zhang is now testing with the same coupling tools. In the authors’ own write-up they ask whether gravitational coupling excites that pulse too, and whether better core-flow maps could eventually forecast inner-core rotation and day length together.

“Even tiny changes in Earth’s rotation can provide valuable information about processes occurring thousands of kilometers beneath our feet,” Zhang said. The six-year feature is the leftover that dynamo models, gravity coupling, and magnetic waves in the outer core still contest. This paper does not close it.

Gravity now has a decade-scale budget with a named winner and two named resistors. Ice melt still pulls the other way on the same clocks. Leap-second planners who treat the core as one leftover trend are reading an average where the new work has started to itemize the fight.

Frequently Asked Questions

How Do Scientists Measure Changes in Earth’s Spin?

Geophysicists time Earth’s orientation against distant cosmic objects, then the IERS publishes the daily C04 series used after 1962 (with an older C02 series covering 1949 to 1961). Zhang and Dumberry took that record, removed seasonal atmosphere and ocean effects, and filtered for periods longer than 30 years before fitting torques.

What Would a Negative Leap Second Do to Clocks?

On a civil clock, 23:59:58 would be followed directly by 00:00:00, so 23:59:59 would never occur. Every leap second since 1972 has inserted an extra second instead of removing one, which is why network software is widely untested for a missing tick.

How Many Leap Seconds Separate UTC From Atomic Time?

UTC trails TAI by 37 seconds. The scale opened on 1 January 1972 with a 10-second offset, and 27 leap seconds have been inserted since, which is 10 plus 27. IERS Bulletin C 72 holds that 37-second offset until further notice.

Why Did Early Topographic Models Blow Up?

Older calculations that pushed reconstructed core-flow pressure against seismically mapped bumps on the core-mantle boundary produced day-length swings typically 100 times too large, and the answers jumped when either the topography map or the flow was tweaked. Zhang and Dumberry therefore treat topographic coupling as a scaled opposing term rather than a stand-alone prediction.

How Large Is Earth’s Core in This Picture?

Earth’s core is a Mars-sized ball made mostly of iron, with a solid inner core that can rotate separately inside a molten outer core. The gravitational torque in the new fit is the twist between density patterns in that solid inner core and mass anomalies in the mantle above the liquid layer.