Deep learning reveals six continuous bands at Earth's core‑mantle boundary

Scientists have overcome the inability to drill to the core‑mantle boundary (CMB) by applying deep‑learning techniques to seismic data collected during earthquakes. In a study published in *Journal of Geophysical Research: Solid Earth*, researchers analyzed a large set of a specific type of seismic wave and identified six continuous bands of irregularities at the CMB that had previously been detected only as isolated patches. The new method provides a more detailed and coherent picture of the boundary’s structure.

The CMB, located roughly 2,900 km beneath the Earth’s surface, marks the interface between the liquid outer core and the solid mantle. Its irregularities influence the planet’s magnetic field, mantle convection, and seismic wave propagation. By training neural networks on thousands of seismic waveforms, the team was able to extract subtle patterns that were invisible to traditional analysis. The resulting bands reveal a more complex, organized structure than the scattered anomalies reported in earlier studies, offering fresh constraints on the composition and dynamics of the deep Earth.

This breakthrough demonstrates the power of machine learning to enhance geophysical imaging where direct sampling is impossible. The refined map of the CMB may improve models of Earth’s thermal and chemical evolution and guide future seismic surveys aimed at probing the planet’s interior.