AI identifies six unknown structures at Earth's core-mantle boundary

Researchers from the Institute of Geophysics and Geodynamics of the Chinese Academy of Sciences have used artificial intelligence to identify six previously unknown structural irregularities deep within the Earth. These anomalies are located at the core-mantle boundary, approximately 2,900 kilometers below the surface.

By utilizing deep learning models to analyze over 2 million seismic records from 4,859 earthquakes recorded between 1990 and 2024, the team successfully detected approximately 175,000 ‘PKP precursor waves’. These extremely weak signals arrive at seismographs seconds before the primary PKIKP waves and had previously been too faint for traditional analytical methods to identify.

The study, published in the Journal of Geophysical Research: Solid Earth, revealed that these irregularities form continuous bands extending from the North Atlantic to Antarctica. Scientists suggest these deep structures could influence significant geological processes, including mantle plumes, earthquakes, tsunamis, and volcanic eruptions, by affecting how heat and materials circulate within the planet.