Info for
Prospective Students
Cattolica Students
International Students
Academic Staff
Alumni
Institutions, Companies and Professions
strumenti-icon
ARE YOU AN ENROLLED STUDENT?
YOU ARE A LECTURER OR STAFF MEMBER
IT
Seminar | 14 April 2025

Discovering the center of the Earth with atomistic "deep-learning" simulations

Brescia

Sandro SCANDOLO, ICTP, Trieste

Abstract
Buried more than 5000 km deep, the Earth's inner core is still largely unknown. Thanks to seismic data we know its density and pressure, and we know that it is a solid composed mainly of iron. But we don't know its crystal structure, nor why transverse seismic waves (so-called shear) propagate so slowly. Despite recent advances, it is still not possible to reproduce in the laboratory the pressure and temperature conditions of the Earth's center in a controlled manner. Atomistic first-principles simulations have been able to provide useful insights, but they are limited to simulation cells containing only a few hundred atoms. Thanks to "deep learning" techniques, we were able to extend the size of the simulation cells to a few million atoms, while maintaining a "chemical" accuracy in the description of the interactions between atoms [1]. This allowed us to propose that the inner core is composed of a cubic alloy (bcc) of iron and other light elements, the main one being silicon. This phase reproduces the seismic data better than any other hypothesis made so far [2,3], and opens new horizons in the understanding of the deepest layers of the Earth.

[1] Z. Li, S. Scandolo, Physical Review B 109, 184108 (2024); Computer Physics Communications 304, 109307 (2024)
[2] Z. Li, S. Scandolo, Geophysical Research Letters 51, e2024GL110357 (2024)
[3] Z. Li, S. Scandolo, Nature Comm., under review; arXiv:2409.08008.

Filename
Locandina Alla scoperta del centro della Terra con simulazioni atomistiche deep-learning.pdf
Size
191 KB
Format
application/pdf
Poster
scroll-top-icon