Scientists Discover Six Mysterious Structures 1,800 Miles Beneath Earth’s Surface
by David Lindfield
Scientists have discovered six previously unknown structures nearly 1,800 miles beneath Earth’s surface, revealing new details about one of the least understood regions of the planet.
In a new study published in the Journal of Geophysical Research: Solid Earth, researchers at the Chinese Academy of Sciences used earthquake recordings and artificial intelligence to map unusual features along the boundary between Earth’s solid mantle and its liquid outer core.
The structures, known as deep-seated scatterers, cannot be observed directly.
Instead, scientists detect them by analyzing subtle changes in seismic waves generated by large earthquakes as those waves travel through the planet.
The newly identified formations sit at the core-mantle boundary, an extreme environment where temperatures increase by roughly 1,000 degrees Celsius between layers, and slowly flowing solid rock meets molten nickel and iron.
Heat escaping from the core helps drive mantle circulation, which in turn influences volcanic and geological activity at Earth’s surface.
AI Analyzes Millions of Earthquake Recordings
Scientists cannot drill anywhere close to the depths involved, so researchers rely on seismic signals to study Earth’s interior.
One of the most useful signals is known as a PKP precursor.
These relatively weak seismic waves arrive shortly before the primary earthquake signal and scatter off irregular structures near the core-mantle boundary.
Identifying them has traditionally required substantial manual analysis.
“Manual identification of these precursors is inefficient, subjective, and insufficient for vast global seismic data sets,” the researchers wrote.
The team instead trained an artificial intelligence model using previously identified examples of PKP precursors.
Researchers then applied the system to more than two million seismic recordings generated by approximately 5,000 earthquakes.
The AI identified 174,929 high-quality precursor signals.
That represents more than ten times the number collected across all previous studies combined.