AI found 175,000 hidden signals inside Earth
Chinese Academy of Sciences researchers used deep learning to map hidden structures at the core-mantle boundary
Roughly 2,900 kilometres beneath your feet, where Earth's solid mantle gives way to its molten outer core, scientists have just found nearly 175,000 signals nobody had properly catalogued before, more than ten times what every previous study combined had managed.
The scientists from the Institute of Geology and Geophysics, Chinese Academy of Sciences, have developed a deep learning algorithm designed to detect PKP precursors: weak seismic waves that precede strong waves when scattered off the inhomogeneities around the core-mantle boundary.
Reporting their results in the Journal of Geophysical Research: Solid Earth in 2026, the researchers provided the algorithm with more than 2 million waveforms from about 5,000 earthquakes detected from 1990 to 2024 and fixed the algorithm's errors to train it again till it learnt to analyse data that no one could ever do manually.
Earlier maps of the lowermost mantle revealed these anomalies to be randomly distributed. Thanks to the greater amount of data by a factor of ten, the scientists discovered that a majority of the previously isolated patches were interconnected to form large bands along the core-mantle boundary.
These may include pieces of the oceanic crust and sediment that have been subducted for many millions of years. Some of these could even be associated with the enormous low shear-velocity zones known as “blobs” as well as the remains of the impact that led to the creation of the moon.
This material would melt or change to minerals that have different characteristics from the rock due to the great pressure existing at that depth.
The study flags six previously under-sampled regions, including zones beneath high-latitude Eurasia, Central Asia and the South Atlantic, as priorities for follow-up work using multiple types of seismic waves.
The study flags six previously under-sampled regions, including zones beneath high-latitude Eurasia, Central Asia and the South Atlantic, as priorities for follow-up work using multiple types of seismic waves.
