A layer only a few to a few dozen kilometers thick may be draped across the boundary between Earth’s core and mantle, and researchers say it likely consists of ancient ocean floor pushed deep underground over geologic time.

That is the picture emerging from a study led by The University of Alabama, published in Science Advances, which used seismic data from Antarctica to probe a vast stretch of the Southern Hemisphere nearly 2,000 miles below the surface. The team found evidence that ultralow velocity zones, or ULVZs, are not just isolated patches in a few places. Instead, they may be widespread along the core-mantle boundary.

These zones slow seismic waves and appear denser than the surrounding deep mantle. The researchers argue that the best explanation is old oceanic material that sank through subduction, then spread and accumulated along the bottom of the mantle.

“Seismic investigations, such as ours, provide the highest resolution imaging of the interior structure of our planet, and we are finding that this structure is vastly more complicated than once thought,” said Dr. Samantha Hansen, the George Lindahl III Endowed Professor in geological sciences at UA and lead author of the study. “Our research provides important connections between shallow and deep Earth structure and the overall processes driving our planet.”

The core-mantle boundary is one of the sharpest transitions inside the planet. According to the study, the change in physical properties there is greater than the shift between solid rock at Earth’s surface and the air above it. That boundary hosts ULVZs, thin regions about 5 to 50 kilometers thick with strongly reduced seismic wave speeds and increased density.

Scientists have debated what causes them.

One idea holds that ULVZs form from partial melting tied to thermal anomalies near the core. Another suggests they include a compositional ingredient, especially because some ULVZs sit far from the hottest parts of the lowermost mantle. The new paper leans toward a broad role for subducted materials, especially former oceanic crust and sediments carried downward by plate tectonics.

That conclusion rests on a rare southern view of Earth’s deep interior.

Hansen, her students and collaborators used data from the Transantarctic Mountains Northern Network, or TAMNNET, a 15-station broadband seismic array deployed in Antarctica from 2012 to 2015. The stations recorded earthquakes from across the Southern Hemisphere, giving the researchers 227 usable events to examine. Those events let the team track PcP waves, seismic waves that reflect off the core-mantle boundary, and search for subtle signals arriving just before or after the main reflection.

Those faint arrivals matter. When PcP waves encounter a ULVZ, reflections from the top of the layer and reverberations within it can create extra signals around the main wave. Detecting them is not easy, because PcP waves are often weak and can be masked by noise.

To improve the signal, the researchers used a method called historical interstation pattern referencing, or HIPR. The approach applies weighted stacking to seismic records, using historical measurement patterns to better sort high-quality data from low-quality data.

From the 227 events, the team generated HIPR-weighted, high signal-to-noise PcP waveform stacks. They then compared the real waveforms with more than 38,000 one-dimensional synthetic models for each epicentral distance in the dataset. Those models tested different ULVZ thicknesses, wave-speed reductions and density changes against a standard Earth model known as PREM, which does not include a ULVZ.

The researchers set strict screening rules. PcP stacks needed a signal-to-noise ratio of at least 1.5. The average comparison quality over the 100 best-fit synthetic models had to exceed 0.2. And to count as robust evidence for a ULVZ, the average fit from those best models had to beat PREM by at least 10%.

After those filters, 183 events remained. Of those, 152 showed robust ULVZ evidence. The other 31 remained uncertain.

That broad pattern stood out because the study area lies mostly away from the giant large low-velocity provinces, or LLVPs, beneath Africa and the Pacific, and also away from present-day subduction zones. Even so, the data pointed to widespread thin anomalous layers across the region.

“Analyzing 1000’s of seismic recordings from Antarctica, our high-definition imaging method found thin anomalous zones of material at the CMB everywhere we probed.” said Garnero. “The material’s thickness varies from a few kilometers to 10’s of kilometers. This suggests we are seeing mountains on the core, in some places up to 5 times taller than Mt. Everest.”

The study describes these features as “mountains” along the core-mantle boundary, with heights ranging from less than about 3 miles to more than 25 miles.

The seismic results alone do not pin down one exact structure in each place. The paper notes an important tradeoff: a thinner but more extreme ULVZ can fit the data almost as well as a thicker but less extreme one. That means exact local properties remain hard to specify.

Still, the regional pattern was consistent. Most geographic families in the study showed ULVZ thicknesses of about 14 to 20 kilometers, with some thinner zones between about 3 and 10 kilometers. The ratio between changes in S-wave and P-wave speeds ranged from about 2.5:1 to 4:1, values the authors say are generally consistent with some degree of partial melting.

To test where such material might come from, the team also ran three-dimensional mantle convection simulations using 200 million years of plate motion history. In both versions of the model, subducted materials eventually reached the lowermost mantle and spread widely along the core-mantle boundary. By the present day, the models showed subducted material scattered broadly across the boundary, with highly variable accumulations.

That matches the idea that old oceanic crust and sediments can sink with slabs, segregate because they are denser than surrounding mantle, and later migrate along the core-mantle boundary. Some may also become caught up in mantle plumes that rise back toward the surface.

The authors considered alternatives, including remnants of an ancient magma ocean and partial melting of pyrolite. They do not rule those out entirely. But they argue that subducted material, possibly partially molten near the core, better explains the widespread global distribution suggested by their results and previous studies.

The study also includes limits. The geodynamic models cannot resolve very thin oceanic crust under 10 kilometers and do not capture all the complex processes that may separate or alter subducted materials at the boundary. The seismic modeling used one-dimensional structures, which the authors note may miss complexities tied to topography or layering. And although the findings suggest a broadly distributed ULVZ layer, the paper says proving a truly global, ubiquitous layer would require finer seismic surveys capable of resolving structures thinner than 5 kilometers.

If ancient seafloor really forms a widespread coating above Earth’s core, it could shape more than deep-mantle maps. The study says such material may influence how heat escapes from the core, which helps power Earth’s magnetic field.

It may also affect the formation and composition of mantle plumes and help explain small-scale chemical and seismic heterogeneity deep inside the planet.

In that sense, the rocks once carried downward at subduction zones may still be steering processes at the deepest edge of the mantle.