The ocean floor beneath the Pacific has been falling through the planet for 250 million years, and new measurements have finally tracked where it lands.
Findings published in The Seismic Record in April 2026 quantify, for the first time at near-global scale, exactly where those dead plates have ended up: piled against the boundary between the rocky mantle and the molten iron outer core, 2,900 kilometres below the surface. The survey processed 16 million seismograms from 4,700 earthquakes recorded at 25,000 stations across the planet, generating 70,000 individual measurements that now cover 75 percent of Earth’s deepest rock layer. That coverage nearly triples what had previously been examined at this level of detail.
Earth’s surface is broken into a dozen or so rigid plates. Where two plates meet and one dives beneath the other, the descending plate sinks through the mantle like a freight train driving into wet concrete in extreme slow motion. The descent takes hundreds of millions of years. When the plate finally reaches the bottom, it hits the outer core, which is liquid iron and will not give way. The plate has nowhere to go. It buckles, folds, and spreads sideways across the boundary, deforming both itself and the surrounding rock for hundreds of kilometres in every direction. That deformation leaves a physical signature in the rock, one that seismic waves from earthquakes can read.
The reading method works through a property of rock under sustained pressure. When rock is compressed uniformly from all sides, seismic waves travel through it at the same speed in every direction. When rock is deformed by enormous one-directional stress, the mineral crystals inside it rotate and line up in a shared orientation, the way iron filings line up beneath a magnet. Once aligned, seismic waves moving along the alignment direction travel measurably faster than waves cutting across it. The survey measured that speed difference by comparing pairs of seismic waves from the same earthquake that followed slightly different paths through the deepest mantle before arriving at the same surface station. Any difference in their behaviour pointed directly to conditions in the deepest 200-to-300 kilometres of rock, not to anything above it.
In two-thirds of the area sampled, the deformation signal was clearly present. That is not a narrow band near known plate boundaries. That is the floor of the planet, deformed across most of its accessible surface area, on a scale that no previous survey had been able to confirm. The measurement threshold the survey used to separate genuine deep deformation from interference caused by shallower rock was a speed-difference value of 0.4 on the splitting intensity scale. Every zone that cleared that threshold counted as confirmed deformation. Roughly two-thirds of the 75 percent sampled did exactly that.
The geographic pattern sharpens the finding considerably. The research team cross-referenced the deformation map against a separate catalogue of 16 ancient slab remnants, dead plates identified through a combination of historical plate reconstruction and seismic imaging of the deep Earth. In 85 percent of the zones associated with those ancient plates, the deformation signal was strong. In zones with no known slab connection, that figure dropped to 63 percent. To test whether the overlap was coincidental, the team rotated the slab catalogue through 1,000 random orientations and recalculated the match each time. The actual observed alignment fell outside every one of those random results.
Some of that deformation is fresh, generated at the moment the slab hits the core-mantle boundary and spreads. Some of it is far older, preserved inside the slab material itself from deformation that happened long before the plate began its descent. Ancient ocean floor carries a structural memory. The crystal alignment built into the rock during its formation and early movement persists through hundreds of millions of years of sinking, through multiple high-pressure mineral transformations across hundreds of kilometres of depth, all the way to the moment of impact at the base of the mantle. In some of the zones the survey mapped, what the seismic waves are reading is a record of geological events that predate the dinosaurs by tens of millions of years.
The coldest parts of the deep mantle, the zones where ancient slabs have accumulated, carry a specific mineral that amplifies the deformation signal. At the pressures present in the final 200-to-300 kilometres above the core-mantle boundary, the dominant mantle mineral transforms into a denser, layered form called post-perovskite. Slabs are colder than the surrounding rock, and colder rock reaches that transformation at shallower depths, meaning slab zones carry a thicker layer of it. Post-perovskite crystal structure responds strongly to deformation: laboratory experiments established that a strain value above 0.5, a measure of how severely the rock has been worked, generates the detectable speed-difference signal. Flow models of the entire lower mantle predict that almost all of the deepest mantle zone globally exceeds that strain threshold, which is consistent with the near-universal deformation signal the survey recorded across the sampled area.
Two massive structures sitting in the deep mantle complicate the picture in ways the survey cannot yet fully resolve. Beneath the Pacific and beneath Africa, two continent-sized regions of anomalously slow rock extend roughly 1,000 kilometres upward from the core-mantle boundary. They are chemically different from the surrounding mantle and have almost certainly held their positions for hundreds of millions of years, possibly since before the current arrangement of continents existed. The survey’s seismic wave paths sample these structures only at their edges. Their interiors remain outside the reach of the current dataset because of the geometry of the wave paths used. Prior regional studies suggest the deformation structure inside them operates at finer spatial scales than this survey’s 4-by-4-degree measurement grid can resolve.
What the survey cannot yet determine is where all the deformed material goes after it spreads. Dead plates hit the bottom, buckle, and push outward. That displaced rock has to go somewhere. Some fraction of it likely feeds into the slow upward currents that eventually reach the surface as volcanic hotspots, places like Hawaii and Iceland where the magma source sits far deeper than a standard plate boundary. Whether material from a plate that subducted beneath ancient South America 200 million years ago is currently feeding a hotspot somewhere in the modern Pacific is a question this dataset cannot answer. Resolving flow direction requires seismic wave types with different path geometries, and the current global network does not generate those in sufficient numbers across the parts of the planet that matter most.
The 25 percent of the deepest mantle that remains unsampled sits beneath the southern oceans, the central Pacific, and parts of the Southern Hemisphere, where earthquakes and seismic stations are too sparse to generate usable measurement pairs. No currently operating global network covers those zones. The two continent-sized slow-rock structures beneath the Pacific and Africa sit partly within that gap, their interiors unresolved, each holding a position it has occupied for hundreds of millions of years while the plates above them have repeatedly assembled, broken apart, and reassembled into entirely new configurations.
Source
Wolf, J., Romanowicz, B., Garnero, E., Zhu, W., and West, J.D. (2026). Widespread Deformation at the Base of the Mantle Linked to Subducted Slabs. The Seismic Record, 6(2), 117–127. https://doi.org/10.1785/0320260001
