Scientists discover dense rock structures near Earth’s core using AI



An international team of geophysicists used AI algorithm used in astrophysics to detect activities deep inside the Earth.

Share this Article:

An Artificial Intelligence (AI) algorithm meant to help astrophysics explore the galaxy has helped geophysicists discover a secret hidden on the core-mantle boundary. The core-mantle boundary is what divides the Earth’s molten iron core, and it’s rocky surface.

The core-mantle boundaries have remained one of the most enigmatic regions on Each that have previously not been explored due to several constraints; therefore there is only limited understanding of these areas. Geophysicists have studied seismic wave signals over the years to decipher the workings of the plate tectonics, on which living things thrive on, and the evolution of our planet. For example, by studying the seismic signals, researchers in the 20th century deduced that the Earth’s outer core must be in liquid-state because shear waves from an earthquake from one part of the globe could not be detected on the other side.

The algorithm, called the Sequester, has helped the researchers study a large seismic database, collected through different sources over the years, to find patterns in the region. The algorithm was co-developed by the study’s authors from Johns Hopkins University and Tel Aviv University to originally find patterns in radiation from distant stars and galaxies. For the geophysicists, Sequester studied sheer waves echo.

Seismic waves are generated underneath the Earth’s surface when an earthquake occurs. These waves travel distances and encounter a variety of rock density, temperature or composition which get reflected in echoes. Echoes change properties by what they encounter on their paths. The nearby echoes come back faster, while echoes from larger structures come back louder. By measuring the time and app of these returning echoes, also known as seismograms, scientists map out the physical properties of the geology beneath the surface of the Earth.

In a paper published in Science Magazine, the team described how the machine-learning algorithm Sequester studied over 7,000 seismographs, specifically, the Wasserstein metric, from hundreds of earthquakes with 6.5 or above magnitude, originating from the Pacific Ocean basic between 1990 to 2018. It then gasified this data to ‘Join the Dots’ to find the shortest path between all data points. This helped scientists derive a trend.

“By looking at thousands of core-mantle boundary echoes at once, instead of focusing on a few at a time, as is usually done, we have gotten a totally new perspective,” said Doyeon Kim, a postdoctoral fellow in the UMD Department of Geology and the lead author of the paper. “This is showing us that the core-mantle boundary region has lots of structures that can produce these echoes, and that was something we didn’t realize before because we only had a narrow view.”

Previously, scientists assumed that structures on core-mantle boundaries are rare. But the research enlightened them on quite on the opposite. “We found echoes on about 40% of all seismic wave paths…that means is the anomalous structures at the core-mantle boundary are much more widespread than previously thought,” said Vedran Lekić, an associate professor of geology at UMD and a co-author of the study.

The first anomaly was found underneath Hawaii. While the region is known to be a seismic spot, Sequencer and additional analysis found that the seismic wave anomaly may be a result of a mantle plume, an outward jutting hot rocky area arising from the Earth’s mantle. The area is estimated to be way larger and much denser than previously estimated because the seismic waves produced uniquely loud echoes.

The second discovery was more surprising because the researchers discovered a previously unknown zone underneath the Marquesas Islands, a remote island in French Polynesia. “We were surprised to find such a big feature beneath the Marquesas Islands that we didn’t even know existed before,” Lekić said. “This is really exciting because it shows how the Sequencer algorithm can help us to contextualize seismogram data across the globe in a way we couldn’t before.”


Previous News

IIT Guwahati to offer free online course on AI

Next News

Baidu exits the US-led AI alliance

Suggested Articles