It's a bit dizzying if you think about it: We're hurtling through space on a rock that both spins on its axis and rotates around the Sun.
And about 2,900 kilometers (1,800 miles) beneath your feet, Earth's solid-but-viscous mantle gives way to the planet's liquid outer core.
As those layers slowly smoosh about, driven by heat, their movement slides, pushes, and deforms Earth's crust, the thin layer we live on above.
Incredibly enough, that dynamic movement is part of why this planet is a great place for a wide variety of life.
But that movement is also a danger, especially to us humans and our concrete cities. So understanding how Earth's innards move isn't just scientific curiosity; it also helps keep us safer from earthquakes and other natural disasters.
However, those same earthquakes are also a useful tool for geologists, who use data on Earth's inner rumblings to map our planet's insides far deeper than drills could ever go.
In a new study, published in the Journal of Geophysical Research: Solid Earth, researchers at the Chinese Academy of Sciences have used a machine learning program to sift through more than 2 million earthquake recordings collected over three and a half decades.
They uncovered nearly 175,000 faint seismic signals that reveal unexpected, small-scale structures near the boundary between Earth's mantle and core.
"We also discovered six areas that likely host significant heterogeneities that had never been documented before, providing clear priority targets for future exploration of Earth's deep interior," the researchers write in their paper.
The result is a comprehensive global map of these deep structures – which reveals that some of the strange features scientists have detected at the bottom of the mantle may be much more extensive than previously thought.
The study focused on a special type of seismic wave called PKP precursors. These are faint waves that arrive shortly before much stronger seismic waves (PKIKP).
When an earthquake occurs, both the stronger and weaker waves travel through Earth. If some of those waves encounter small variations in the material near the core-mantle boundary, though, they can scatter and take slightly different paths.
The scattered PKP waves, which pass through the outer liquid core but not the solid inner core, can then arrive ahead of the main PKIKP signal.
So, by analyzing wave times, these PKP precursors can 'show' unusual structures deep inside the planet.
The problem is that these waves are faint – and finding them in millions of earthquake recordings is an enormous job.

Traditionally, researchers have had to inspect seismic records manually, looking for these PKP clues, which is slow, painstaking work. And researchers might not always agree on whether a signal is a genuine precursor.
So the researchers trained a deep-learning system to do the initial screening, which enabled them to examine millions of seismic waveforms from almost 5,000 earthquakes between 1990 and 2024.
The algorithm first sorted recordings by quality, then determined whether they contained PKP precursors.
Importantly, the researchers manually checked and corrected the AI model's mistakes, then fed the corrected examples back into the algorithm during the training process.

Earthquakes with identified PKP precursors that traveled from the source (pink stars) to seismic array detectors (blue triangles). (Guan et al., J. Geophys. Res. Solid Earth, 2026)
Ultimately, the system identified 174,929 high-quality PKP precursor signals – more than ten times all previous studies combined.
That huge increase in observations matters because earlier maps of the bottom of the mantle tended to show relatively isolated, random-seeming structures.
With the expanded dataset, researchers could see that some of those patches actually connect into much larger, continuous belts.
These structures may have some surprisingly ancient origins.
The researchers suggest that some of the structures could be related to material dragged deep into Earth by subduction.

According to research from 2023, there are a variety of materials in the deep mantle, "such as continental crust, sediments, oceanic crust, and primitive mantle materials".
There may even be the remains of the object thought to have made the Moon when it crashed into Earth.
Under enormous pressures and temperatures at the core-mantle boundary, that material can chemically differentiate, undergo mineral transformations, or partially melt, creating pockets or structures with very different physical properties from the surrounding mantle.
Interpreting their findings cautiously, the researchers conclude that the small-scale structures they identified may be "thermochemical piles" shaped by some combination of multi‐episode subducted slab remnants, localized partial melting, mineral transitions, and interactions with known 'giant blobs' called large low-shear-velocity provinces.

Those structures can impact how the mantle moves and how hot magma gets beneath a volcano or along a deep-ocean ridge (since different rocks heat at different temperatures). And that can determine whether an eruption might occur or how severe it is, information that's useful to us.
They might even tell us more about scientific curiosities, such as how tectonic plates formed and imbalances in Earth's magnetic field.
The new map highlights six regions that had been poorly sampled by PKP precursor studies and now appear to be promising targets for closer investigation. They include areas beneath high-latitude Eurasia, Central Asia, and the South Atlantic, among others.
Related: A Giant Seismic Wave Bounced Off Earth's Core And May Have Shifted Japan
The researchers say these regions should now be priorities for future, higher-resolution studies using multiple types of seismic waves.
"As the catalog continues to expand, its high‐resolution spatiotemporal coverage will… advance the refinement of fine‐scale structural models of the lowermost mantle, and offer increasingly rich constraints for deepening our understanding of the geodynamic state of Earth's deep interior," the team concludes.
The study has been published in the Journal of Geophysical Research: Solid Earth.
This article was fact-checked by Clare Watson and edited by Clare Watson. While we pride ourselves on our process, we are only human. If you spot a mistake, please let us know.
