The Iron Ball at the Center of the Earth Keeps Changing Its Skin

A Solid Ball That Keeps Not Acting Like One

Start with the part that is not in dispute. About 5,150 kilometres beneath your feet, at a pressure of roughly 330 gigapascals and a temperature north of 5,000 kelvin, the molten iron-nickel alloy of the outer core freezes. It has been freezing, slowly, for well over half a billion years, growing a solid ball out from the centre of the planet one crystal at a time. That ball is the inner core. It is almost pure iron, laced with a little nickel and a scatter of lighter elements, and by every reasonable definition of the word it is solid rock — solid metal, rather. Solid things are supposed to be boring on human timescales. A boulder does not rearrange itself between one Tuesday and the next.

The inner core has spent thirty years failing to be boring. Since Xiaodong Song and Paul Richards first reported in 1996 that its seismic signature seemed to be drifting relative to the rest of the planet [3], a small, argumentative subfield has grown up around exactly how a two-thousand-kilometre-wide iron sphere floating in a sea of its own molten parent material actually behaves. Hrvoje Tkalčić and colleagues extended the same repeating-earthquake logic in 2013 to argue the rotation was not smooth but "shuffling" — speeding, slowing, occasionally reversing [4]. Does it rotate faster than the mantle above it, the way early results suggested? Does that rotation reverse over decades, the way more recent results suggest [5]? Or is some of what looks like rotation actually the ball itself changing shape at the edges, an idea that has been on the table, contested, since at least the early 2000s, and that a 2014 review by Arwen Deuss catalogues as one of the field's oldest open questions [10]?

The 2025 paper by Vidale and four co-authors does not resolve that fight by picking a winner. It resolves it by showing both sides were half right, using a dataset built specifically to catch the inner core in the act of doing whichever one it does [1]. "Later on," Vidale told reporters after publication, "I'd realize I was staring at evidence the inner core is not solid" — not liquid, he means, but not perfectly rigid either [14].

How You Take the Pulse of Something You Will Never Touch

You cannot drill to the inner core. The deepest hole anyone has ever bored, the Kola Superdeep Borehole, reached 12.3 kilometres, about a thousandth of the way down. Everything known about Earth's centre is inferred from how vibrations move through it, which means the entire field runs on one trick: send a wave through the thing you cannot see, and listen very carefully to what comes out the other side.

The specific wave that matters here has a name only a seismologist could love: PKIKP. Read it as a itinerary. A P-wave (compressional) leaves the earthquake, crosses the mantle, refracts into the liquid outer core (the K leg), continues into the solid inner core (the I leg), refracts back out through the outer core (K again), and crosses the mantle a second time to reach a distant seismometer (P again). Only earthquakes and receivers separated by more than about 110 degrees of arc — more than an eighth of the way around the planet — produce a PKIKP wave at all, because anything closer never gets steep enough to dive through the core. This is, not coincidentally, the reason almost every paper in this literature pairs an earthquake source near one pole with a receiver near the other: South Sandwich Islands earthquakes, near Antarctica, recorded in Alaska and northern Canada, sit close to opposite ends of a chord that threads directly past the planet's centre [1–3].

The second trick is the one that actually makes the measurement possible. An ordinary earthquake's PKIKP wave tells you almost nothing on its own, because you don't know how much of its shape came from the source, how much from the receiver, and how much from the inner core in between. But some faults keep rupturing in almost exactly the same way, in almost exactly the same place, years apart — repeating earthquakes, sometimes called doublets. Two members of a doublet have, to a very good approximation, an identical source and are recorded at an identical station. Whatever is different between their seismograms has to have come from whatever lies on the ray path between them and changed in the meantime. For a wave that dives through the inner core, that difference is a direct readout of the inner core changing.

Two Stations, Two Very Different Views Through the Same Ball

The paper leans on two seismic arrays in the far north: ILAR, at Eielson Air Force Base near Fairbanks, Alaska, and YKA, near Yellowknife in Canada's Northwest Territories [1]. Both are medium-aperture arrays — clusters of instruments a few kilometres across whose signals can be combined to boost a faint arrival out of the noise — and both have spent decades listening to the same South Sandwich Islands earthquake sequence.

What makes the pairing powerful is that the two arrays are not interchangeable. Because Fairbanks and Yellowknife sit at slightly different distances and azimuths from the South Sandwich Islands, their PKIKP rays cut through the inner core along different chords. The ILAR path bottoms out deep, passing relatively close to the planet's centre. The YKA path is shallower — in the paper's own words, "inner-core-grazing" — skimming through the outer part of the inner core rather than plunging toward its heart [1]. Figure 1 draws the geometry that this whole argument depends on.

The Trick That Lets You Compare the Same Rock Twice

Here is the problem with just comparing any two repeating earthquakes years apart: the inner core does not sit still between them. If it is rotating relative to the mantle, even slowly, a wave that grazed one patch of the boundary in 1995 might graze a completely different patch in 2010. A mismatch would tell you nothing, because you would not know whether the rock changed or the ray simply moved to sample different rock.

The team's direct predecessor paper, published in Nature in 2024, solved exactly this problem for the question of rotation. Working from 121 repeating earthquakes between 1991 and 2023 — 143 distinct pairs, some belonging to sixteen larger "multiplet" families of events that all resemble each other — Wei Wang, Vidale, Guanning Pang, Keith Koper and Ruoyan Wang tracked how the inner core's PKIKP waveforms evolved and found something unexpectedly specific: the inner core super-rotated (drifted faster than the mantle) from roughly 2003 to 2008, then sub-rotated, drifting back the other way, from 2008 onward — two to three times more slowly than it had advanced [2]. Because the backward drift is slower, the inner core eventually swings back through a rotational position it already held years before. When that happens, a wave recorded today samples almost exactly the same patch of inner-core boundary that a wave recorded years earlier sampled. Rotation, in other words, occasionally undoes itself and hands you a free repeat of the experiment.

That is the engine behind the 2025 paper. Vidale and colleagues extended the earthquake catalogue with seven more recent events, adding twenty-five more pairs to the set, and then went looking specifically for pairs recorded at times when the backtracking history says the inner core had returned to a position it had occupied before [1]. If pure rotation is the whole story, those pairs should match at both stations — the ray is sampling the same patch of inner core it sampled last time, and nothing else should have changed. Any mismatch that survives this control is not about rotation any more. It has to be about the rock itself.

What Turned Up Between 2004 and 2008

Some pairs did not survive the control. Comparing repeaters recorded when the inner core had cycled back to a shared position, the paper found waveforms at YKA that did not match — specifically among pairs spanning parts of the 2004–2008 window — while the equivalent ILAR recordings, for the very same earthquake pairs, matched fine [1]. One example the paper documents at length is "pair 31," a repeating sequence tracked from 2004 to 2017, which the authors use to walk through exactly what a shallow-only mismatch looks like against a station that keeps agreeing with itself.

Put the two observations together and there is only one geometry that explains both. ILAR's ray, bottoming deep, kept sampling structure that had not changed. YKA's ray, grazing the boundary, sampled structure that had. The paper states the conclusion plainly: the changes "most likely" originate in the shallow inner core, not throughout its bulk, and they show up more strongly on ray paths that spend more of their length near the inner-core boundary [1].

"Annual-Scale" Does Not Mean "Ticks Like a Clock"

The paper's title promises "annual-scale variability," and it is worth being precise about what that phrase is and is not claiming. It does not mean the inner core changes on a fixed one-year cycle, the way a season does. It means the relevant process — whatever is happening near the boundary — unfolds over a span measured in single-digit years rather than in the thousands or millions of years that geological processes usually take. The window in which YKA and ILAR diverge is about four years wide (2004 to 2008); the example pair the paper walks through in detail spans thirteen years (2004 to 2017) [1]. That is "annual-scale" in the sense seismologists mean it: fast enough that a human research team, and not just a future geologist reading rock strata, can watch it happen within a single career.

This is the part of the finding that should stop you. Geological processes that reshape solid rock — recrystallization, slow creep, phase changes under pressure — are supposed to be the textbook example of something too slow to notice. Continents drift a few centimetres a year. Mountain belts rise over millions of years. The inner core, made of iron under pressure high enough to make steel look like putty, is not obviously a candidate for a process fast enough to leave a visible mark inside two decades of seismic records. And yet something near its boundary evidently is.

Why Solid Iron Is Allowed to Do This At All

The honest answer is that "solid" does the least useful amount of work here that a word can do. Solid materials under extreme pressure and heat, sitting only a few hundred degrees below their melting point, can flow — slowly, by viscous creep, the same physical mechanism that lets glaciers move and lets the solid mantle convect over millions of years, just potentially operating much faster near the inner core's edge. Bruce Buffett's 1997 geodynamic estimate put the inner core's effective viscosity somewhere between that of a very stiff solid and something that behaves, over the right timescale, almost like a very slow fluid [8]; more recent mineral-physics work on iron's strength at core pressures has pushed some estimates toward a mechanically "weaker" inner core than earlier models assumed [9]. None of this makes the inner core soft in any everyday sense. It means that "solid" and "immovable on human timescales" are not the same claim, and the inner core's edge, closest to the molten, turbulently convecting outer core, is exactly where you would expect any give to show up first.

The paper offers two candidate mechanisms for what is physically doing the deforming, and is honest that it cannot yet distinguish between them [1]. One is that bumps and dips in the inner-core boundary's own topography are gravitationally coupled to density anomalies in the overlying mantle, and as the inner core's slow rotation carries a given patch of boundary past a fixed mantle anomaly, the coupling flexes that patch. The other is more direct: the outer core is a vigorously convecting liquid metal ocean, and its motion can simply drag — exert traction — on the solid surface it is in contact with, the way wind stresses the surface of open water. Either mechanism only needs to act on a thin near-surface layer to produce exactly the pattern the two arrays saw.

Separately, and using an entirely different seismic method, Dongdong Tian and Lianxing Wen reported in 2017 that the inner-core boundary is not seismically uniform everywhere. Beneath the southwest Okhotsk Sea, reflected compressional waves revealed a patch, 4 to 8 kilometres thick, whose seismic properties sit between those of the solid inner core and the liquid outer core — consistent with a genuinely "mushy," partially molten transition rather than a knife-edge boundary [6]. That is a different location, a different method, and almost certainly not the same patch of boundary that Vidale and colleagues' South Sandwich Islands rays sample. What it establishes is that the inner-core boundary is known, independently, to be capable of local structural softness at all — which is exactly the kind of material the 2025 result needs to exist somewhere for its viscous-deformation explanation to be physically reasonable.

Club Arithmetic: How Big Is the Ball, and How Thin Is the Skin

Two numbers are doing a lot of intuitive work in this story — how big the inner core actually is, and how small a fraction of it could plausibly be doing the moving — and neither number is stated outright in the paper. Both are things we can compute ourselves from published constants.

The Old Debate, and Why "Tentatively Both" Is the Right Answer

For most of the last three decades, the argument over inner-core temporal change has run along a single fault line: is the apparent variability in PKIKP waveforms caused by the inner core physically rotating, so that a fixed ray path samples new material over time, or is it caused by the inner-core boundary itself changing shape under a fixed or slowly rotating ray? Different research groups have published evidence for each side, sometimes about the very same earthquake catalogues, and the disagreement has been genuine rather than a matter of sloppy analysis — separating the two effects requires exactly the kind of controlled comparison this paper runs.

Vidale and colleagues' answer, stated directly in their abstract, is "tentatively both" [1]. The 2024 backtracking result established that a rotation signal is real and even lets you read its rate and its 2008 reversal off the data [2]. The 2025 result shows that once you control for rotation — by comparing pairs recorded specifically when the inner core had returned to a shared position — a genuine residual mismatch survives at the shallow-sampling station and not at the deep one. Rotation is real. Local structural change near the boundary is also real. Both operate on the same few-year timescale, and untangling exactly how much of the historical PKIKP literature's disagreements trace to one or the other is, per the authors, a project for more data and more repeating-earthquake pairs, not a settled question this single paper closes [1].

Why Any of This Matters Above Ground

It would be easy to file this away as an exotic fact about a piece of iron nobody will ever see. The inner core, however, is not an inert curiosity sitting quietly at the centre of the planet. Its growth releases latent heat and light elements into the base of the liquid outer core, and that buoyant release is one of the main engines believed to drive the convection that generates Earth's magnetic field — the same field that deflects charged solar particles and makes compasses point north. A boundary that flexes, sheds material unevenly, or couples mechanically to mantle structure overhead is a boundary condition on that convection, not a bystander to it. Papers going back to Cao, Masson and Romanowicz's 2007 mapping of short-wavelength inner-core boundary topography have treated the shape of this surface as geodynamically relevant for exactly this reason [7]. A boundary that changes shape on a timescale of years, rather than only over the hundreds of millions of years it took the inner core to grow to its current size, is a different kind of boundary condition than most geodynamo models have assumed.

What Would Make Us More Confident

The result rests on a genuinely small number of qualifying comparisons — the specific set of repeating pairs recorded during a window when the backtracking model says the inner core had cycled back to a shared position, at just two stations. More repeating earthquakes, especially any that happen to occur along different azimuths from the South Sandwich Islands sequence, would sample different chords through the inner core and either corroborate a genuinely localized shallow change or reveal that the YKA pattern was closer to a station-specific artefact than a global statement about the boundary. The 2025 paper's own extension — seven new events, twenty-five new pairs [1] — is exactly this kind of incremental strengthening, and the underlying earthquake sequence is still active, so the dataset should keep growing on its own.

A second, harder test would be finding an independent geometry — a different source region, different receiving stations — that also brackets a shallow-versus-deep ray-path contrast and checking whether it reproduces a shallow-only signal on a comparable timescale. Until then, the finding stands as a single, carefully controlled comparison rather than a global map of a "mushy," restless inner-core rim. What it establishes is narrower and, we think, more solid precisely because of that: somewhere along one specific pair of ray paths, near the inner core's edge, something changed in a few years that had no business changing that fast.

What We Would Tell Someone Who Only Read the Headline

Three things. First, nobody found a liquid layer, a hidden ocean, or a crack in the inner core; "mushy" here is a term of art borrowed from a different, unrelated study of a different patch of the same boundary [6], describing a seismic property that sits between solid-like and liquid-like, not an actual slush. Second, "annual-scale" describes how fast the process unfolds, not a strict yearly cycle — the window in question runs about four years, with a documented example stretching to thirteen. Third, and most important: this paper does not claim the inner core is secretly liquid or that solidity itself is in doubt. It claims something narrower and, to us, more interesting — that the outermost sliver of a two-thousand-kilometre solid iron sphere is soft enough, at the right pressure and temperature and given enough time, to visibly deform within a human research career, while the deep interior a few hundred kilometres further in shows no such thing. A planet apparently keeps its solid centre very still and its edges a little bit alive.

References

python inner-core-deforming.py > inner-core-deforming-output.txt

1. 1.

   Vidale, J. E., Wang, W., Wang, R., Pang, G. & Koper, K. (2025). Annual-scale variability in both the rotation rate and near surface of Earth's inner core. Nature Geoscience 18, 267–272. doi:10.1038/s41561-025-01642-2

2. 2.

   Wang, W., Vidale, J. E., Pang, G., Koper, K. D. & Wang, R. (2024). Inner core backtracking by seismic waveform change reversals. Nature 631, 340–343. doi:10.1038/s41586-024-07536-4

3. 3.

   Song, X. & Richards, P. G. (1996). Seismological evidence for differential rotation of the Earth's inner core. Nature 382, 221–224. doi:10.1038/382221a0

4. 4.

   Tkalčić, H., Young, M., Bodin, T., Ngo, S. & Sambridge, M. (2013). The shuffling rotation of the Earth's inner core revealed by earthquake doublets. Nature Geoscience 6, 497–502. doi:10.1038/ngeo1813

5. 5.

   Yang, Y. & Song, X. (2023). Multidecadal variation of the Earth's inner-core rotation. Nature Geoscience 16, 182–187. doi:10.1038/s41561-022-01112-z

6. 6.

   Tian, D. & Wen, L. (2017). Seismological evidence for a localized mushy zone at the Earth's inner core boundary. Nature Communications 8, 165. doi:10.1038/s41467-017-00229-9

7. 7.

   Cao, A., Masson, Y. & Romanowicz, B. (2007). Short wavelength topography on the inner-core boundary. Proceedings of the National Academy of Sciences 104, 31–35. doi:10.1073/pnas.0609810104

8. 8.

   Buffett, B. A. (1997). Geodynamic estimates of the viscosity of the Earth's inner core. Nature 388, 571–573. doi:10.1038/41534

9. 9.

   Gleason, A. E. & Mao, W. L. (2013). Strength of iron at core pressures and evidence for a weak Earth's inner core. Nature Geoscience 6, 571–574. doi:10.1038/ngeo1808

10. 10.

    Deuss, A. (2014). Heterogeneity and anisotropy of Earth's inner core. Annual Review of Earth and Planetary Sciences 42, 103–126. doi:10.1146/annurev-earth-060313-054658

11. 11.

    Kennett, B. L. N., Engdahl, E. R. & Buland, R. (1995). Constraints on seismic velocities in the Earth from traveltimes. Geophysical Journal International 122, 108–124. doi:10.1111/j.1365-246X.1995.tb03540.x

12. 12.

    Dziewonski, A. M. & Anderson, D. L. (1981). Preliminary reference Earth model. Physics of the Earth and Planetary Interiors 25, 297–356. doi:10.1016/0031-9201(81)90046-7

13. 13.

    NASA Goddard Space Flight Center, National Space Science Data Center (2024). Moon Fact Sheet. nssdc.gsfc.nasa.gov/planetary/factsheet/moonfact.html

14. 14.

    Phys.org (2025). Earth's inner core is less solid than previously thought: Study reveals structural transformation. News report on [1], including comment from J. E. Vidale.