A day is not a fixed box of 24 hours. New work from the University of Alberta says a turning pull from the solid heart of the planet is one reason the box keeps changing size — by a few thousandths of a second, over decades. That is the measured claim. What sits around it is older, louder, and far less settled.
You wake up, look at a clock, and treat the day as a finished product. It is not. The planet spins a little faster, then a little slower. The change is too small to feel. It is not too small for satellites, bank stamps, radio telescopes, or the offices that decide when to add a leap second. On 23 September 2026 a paper in Nature put a name on one of the hidden gears: a gravitational tug between the solid inner core and uneven mass in the rocky shell above it.
The latest public write-up of that paper, carried on 29 September 2026 by ScienceDaily, repeats the same core finding. Mass media stops at “milliseconds” and “better GPS.” That is true as far as it goes. It is also the safest sentence. The harder sentence is this: we still cannot drill to the core, we still argue about how the inner core itself turns, and a small measured effect is being asked to stand in for a region we only hear through earthquake waves.
What the new study actually says
Huifeng Zhang, a physics doctoral student at the University of Alberta, and professor Mathieu Dumberry published Gravitational torque drives multidecadal variations in length of day on 23 September 2026. The university’s own note is here: Earth’s deep secret: why your day isn’t quite 24 hours.
Their claim is narrow. After you take out the known seasonal push of air and water, and after you take out the long, slow drag of the Moon, the remaining swing in day length over about 60 years still wobbles by a few milliseconds. That leftover wobble, they argue, is mainly a turning pull — a torque — from gravity. The solid inner core is not a perfect ball. Dense patches in it want to line up with dense patches in the mantle. When the inner core turns a little out of line, gravity tries to drag it back. That drag tugs the mantle. The mantle is the thick rock that carries the crust we live on. If the mantle speeds up or slows down even a little, the length of a day changes.
A second set of forces works the other way. At the boundary between the liquid outer core and the mantle there is electromagnetic drag and a kind of physical friction against the rough underside of the rock. Those forces resist the gravitational pull. They keep the swing from running wild. Zhang said, in the coverage that followed the paper, that before the model ran they did not know these forces were competing with each other. Dumberry put the inner-core habit in plain words: it “wants to be aligned” with the mantle.
That is the official result. It is a model fitted to a record, not a photograph of the centre of the Earth.
How anyone knows the inner core is turning at all
No camera goes there. The inner core is a solid metal ball, mostly iron, roughly the size of Mars, sitting inside a sea of liquid iron. What we have are earthquake waves that pass through it and come back slightly changed. From those travel times, teams have argued for years that the inner core does not keep exact pace with the surface.
A 2023 study that Zhang and Dumberry lean on — Yang and Song in Nature Geoscience — used seismic records back to the 1960s. It found the inner core running a little ahead of the rest of the planet until around 2010, then running a little behind. Other groups, including work led by John Vidale, have described a slowdown around the same years and even a backtrack along a similar path. The Alberta paper treats that reconstructed turn as an input. If that input is wrong, the torque history is wrong with it.
This is the first quiet gap. Headlines say “the inner core is changing the day.” The chain is: seismic model of inner-core motion → assumed shape and sticky flow of the inner-core boundary → calculated gravitational pull → match to a cleaned-up day-length curve from 1964 to 2019. Each step is reasoned. None of them is a direct look.
The numbers that do not need a theory
Day length is measured. Observatories track Earth’s facing against distant radio sources in the sky. The International Earth Rotation and Reference Systems Service keeps the books. After the weather and the oceans are removed, a multi-decade wave of a few milliseconds remains. That wave is not new. What is new is a model that says gravity from the inner core is the main driver of that particular wave, with the liquid-core boundary pushing back.
Separate from that wave, the planet has other clocks. The Moon steals spin through tides, stretching the average day over geologic time. Ice sheets melt and the crust rebounds; mass moves and the spin changes. Since the early 2020s some years have shown slightly shorter days, enough that researchers such as Duncan Carr Agnew warned that civil time might one day need a negative leap second — a second taken out of the clock, not added. Other analyses in 2024 said that burst of speed was already fading. As of July 2026, the IERS still had not called a leap second for the end of the year. The last added second was 1 January 2017. Twenty-seven leap seconds have been inserted since 1972. The pattern is not a law. It is a running correction.
So there are at least three stories stacked on one planet: a very long slowing from the Moon, shorter swings from ice, air and water, and a multi-decade core-mantle exchange. The Alberta paper is about the third story. Mixing the three in one scare sentence is how a millisecond becomes a prophecy.
Why GPS, banks and atomic clocks care
Atomic clocks do not care how the Earth feels. They count oscillations in atoms. Earth time is the turning of a lumpy, living ball. Civil time, UTC, is a compromise: atomic seconds, with occasional leap seconds so the clock does not drift too far from the Sun in the sky.
Navigation satellites work in that gap. A position fix is a time fix. If the model of Earth’s facing is off, the ground point slides. The paper’s practical pitch — also in the source text you started from — is that a cleaner map of rotational wobble can tighten satellite tracks, phone links, money stamps and lab clocks. That part is ordinary engineering. It does not require a hidden hand. It does require that the people who run time and orbit systems keep feeding new Earth-orientation numbers into their machines.
Who those people are is not a secret society. It is a thin professional layer: IERS, national mapping offices, satellite operators, stock exchanges that timestamp trades to the millisecond. When a day stretches or shrinks, they adjust models. They do not hold a press conference about the inner core. That silence is often read as concealment. More often it is boredom. A three-millisecond drift over twenty years does not move a harvest. It does move a geodetic budget.
What the press leaves on the table
Mass outlets like a closed loop: scientists puzzled, scientists solved, GPS will get better, sleep well. Alternative outlets like an open pit: the core is reversing, the field is dying, the crust can slip, the calendar is a lie. Both habits waste part of the record.
The record says the inner core’s differential turn appears to have changed sense near 2010. The record says the magnetic field at the surface has weakened on the order of several percent since the nineteenth century, with a growing weak patch over the South Atlantic. The record says large low-shear regions sit at the base of the mantle and that outer-core flows can reorganize on decadal scales. The record does not say a 104-degree crustal flip is scheduled, and it does not say millisecond day-length changes are a countdown clock for one.
Planet Today already walked through the louder crustal-shift claim in July: Earth Preparing for 104° Crustal Shift? ECDO Theory and Hidden Forces. That piece laid official magnetic numbers next to the ECDO story, which treats core-mantle decoupling as a possible path to a rapid reorientation. Readers can hold both texts at once. The Alberta paper is about a small, repeating exchange of spin. ECDO is about a rare, violent unlock. Using one as proof of the other is a jump, not a measurement.
The same caution applies the other way. Calling every deep-Earth question “conspiracy” is also a dodge. For thirty years the coupling mechanism was admitted to be unknown. That is not a smear. It is the University of Alberta’s own starting point. “Understanding the mechanism of exchange between the core and mantle has so far eluded scientists,” the campus note says. A model that now fits one window of data is progress. It is not a tour of the basement.
A short history of a moving day
Ancient sky watchers treated the day as the Sun’s path. They were not wrong for farming. They were wrong for physics. Once clocks became sharper than the planet, the planet looked sloppy. In the twentieth century, quartz and then atoms made the sloppiness visible. Leap seconds were the political patch: keep the atomic second sacred, but jerk the civil clock when Earth drifts.
In 2022 the General Conference on Weights and Measures voted to widen the allowed gap between Earth time and UTC by 2035, which would make leap seconds rarer. Drafts in 2026 even named a possible start date for the new rule. That fight is not about the inner core. It is about software, power grids, and who wants a sudden extra second in a trading engine. Still, it sits in the same file: human systems hate a living planet.
Seismic work on inner-core spin is younger. In the 1990s some papers claimed a steady super-rotation of a degree or more per year. Later work cut that rate down, then replaced a one-way drift with a slow oscillation. A 2022 Science Advances paper spoke of an inner core rocking about a gravitational lock with the mantle, and tied that rock to a roughly six-year twitch in day length. Zhang has said she is looking at whether related coupling can explain that shorter cycle too. If both the 60–70 year wave and the 6-year twitch come from the same family of pulls, the inner Earth is less a rigid clock and more a set of coupled weights.
That image bothers people who want a finished machine. It also feeds older stories in which Earth is a body that breathes. Those stories are not evidence. They are the feeling people get when a “solid” planet starts to look viscous. Zhang and Dumberry say the inner-core boundary appears to “deform viscously” on a timescale of about ten years. In plain speech: the “solid” heart can sag and reshape, slowly, like cold pitch. That one line does more to unsettle the schoolbook picture than any slogan about elites.
Space, time, and the temptation to over-read
Space agencies need Earth orientation the way ships need charts. Hubble and later sky surveys keep finding that the Milky Way itself is a pile of swallowed parts, not a clean disk. One recent Planet Today science piece on that theme is Hubble Finds Lost Galaxy the Milky Way Swallowed 11.8 Billion Years Ago. The habit is the same above and below: we reconstruct hidden motion from leftover light or leftover shake, then argue about the story.
Geopolitics enters only at the clock. Time is infrastructure. Navigation, missiles, undersea cables, and cross-border payments all lean on a shared second. When Earth misbehaves, the fix is technical. When the fix is technical, the public sees nothing. Alternative media then fills the quiet with intent. Sometimes intent is real — states do hide military timing and gravity maps. Sometimes the quiet is just a small community doing unglamorous work. A Swiss reading keeps both files open and refuses to merge them without a document.
Medicine and daily life barely touch this subject, except in one sideways way: human clocks are not planetary clocks. Sleep, heart rhythm, and mineral balance run on light, food, and nerves, not on a 2-millisecond drift in UT1. For that earthbound file, see a different Planet Today piece on what food tables leave out: Magnesium Beyond Pumpkin Seeds: What the Food Charts Leave Out. Mixing core torque with wellness claims is another way to lose the thread.
The two honest readings
Reading A, the lab reading. Angular momentum is conserved. When one deep layer speeds up, another must slow. Gravity between a slightly dented inner core and lumpy mantle can transfer that spin. Electromagnetic and topographic drag at the core-mantle boundary fight the transfer. A model using seismic inner-core history plus outer-core flow from the magnetic field can reproduce the multi-decade day-length curve from the mid-1960s through 2019. Implications: better constraints on how sticky and how bumpy the deep Earth is, and a slightly cleaner input for Earth-orientation services. Risk if oversold: treating a fit as a unique proof.
Reading B, the open-question reading. We have never seen the inner core. Seismic inversions disagree in detail. Magnetic field models of outer-core flow are themselves built on assumptions. The same decades that show a core-related day-length wave also show ice melt, groundwater shift, and a weakening dipole. Some independent researchers treat core-mantle coupling as a latch that can fail, not only as a gentle spring. Ancient flood stories and odd monument alignments are offered as cultural memory of past lurches. Risk if oversold: turning a millisecond into a catastrophe calendar.
Neither reading needs a villain. The first can become smug. The second can become a novel. The data pile is the same pile.
What still is not known
The Alberta window ends around 2019 for some of the flow models and 2021 for the inner-core reconstruction. The inner core’s turn after that is still being argued in the seismic literature. The six-year cycle is not put to bed. The exact height of bumps on the inner-core surface, the viscosity of that surface, and the strength of electromagnetic coupling are fitted parameters, not drilled samples. Large structures at the base of the mantle — the so-called large low-shear-velocity provinces — remain poorly explained. Whether they are old piles of oceanic rock, early metal-rich residue, or something else changes how one pictures the gravitational lock.
Also unknown: how much of the recent short-day spell belongs to the core and how much to the surface. If the core’s multi-decade wave is now in a phase that lengthens the day while melting ice tries to shorten it, the net you feel in leap-second debates is a sum, not a single cause. Papers that pick one cause and ignore the others are easier to read. They are worse maps.
How to sit with a three-millisecond planet
A millisecond is a thousandth of a second. A few milliseconds over decades will not make your afternoon late. It will, if ignored, make a satellite’s ground track drift and a time stamp argue with another time stamp. That is why the study matters even if it never trends on a phone.
Curiosity is the right tool here, not panic and not a shrug. The planet under the floor is not a dead ball. It has a solid heart that can change shape on a ten-year scale, a liquid shell that writes the magnetic field, and a rocky coat that we ride. Those layers pull on each other. Sometimes the pull shows up in the length of a day. Sometimes it shows up only in a seismogram. The people who write official briefs prefer the first kind of sentence because it has a use: clocks. The people who write unofficial briefs prefer the second kind because it has a mood: we are living on a machine we have not finished drawing.
Both sentences can be true at once. The drawing is unfinished. The clocks still have to run. The honest article is the one that keeps the millisecond in view and does not pretend the basement has been inspected.
Sources worth opening
- Primary paper: Zhang, H. & Dumberry, M., “Gravitational torque drives multidecadal variations in length of day,” Nature, 23 September 2026 — nature.com/articles/s41586-026-10999-2
- Journal news note: James Dinneen, “Gravitational tug-of-war inside Earth is changing the length of our days,” Nature, 23 September 2026 — nature.com/articles/d41586-026-03018-x
- University release: Geoff McMaster, University of Alberta, 23 September 2026 — ualberta.ca news spotlight
- Latest roundup: “Earth’s day isn’t exactly 24 hours…,” ScienceDaily, 29 September 2026
- Related seismic background on inner-core slowdown: coverage of Vidale and colleagues, 2024, e.g. Sci.News
Further reading on this site: Science · core-mantle shift claims · deep-time sky reconstruction.
Original source of the finding: University of Alberta / Nature, 23 September 2026. Full paper: https://www.nature.com/articles/s41586-026-10999-2. Campus summary dated 23 September 2026. This article prepared 29 September 2026.
Note for fact checkers: The millisecond-scale change in day length, the 1964–2019 window, the authors’ names, the journal, and the proposed lead role of gravitational torque are taken from the paper and the university note. Claims about crustal flips, hidden control of time, or medical effects are not in that paper and are marked here as separate arguments. Seismic models of inner-core rotation remain debated. A model that fits one cleaned data set is not a complete map of the core. Readers should open the primary Nature article before treating any secondary summary, including this one, as the last word.
Original article: Earth’s Inner Core Quietly Changes the Length of a Day on Planet Today 🚀
Automatically republished from the main blog.