Scientists have discovered a vast structure made of dense material occupying the boundary between Earth’s liquid outer core and the lower mantle, a zone some 3,000 kilometers (1,864 miles) beneath our feet.
The researchers used a machine-learning algorithm that was originally developed to analyze distant galaxies to probe the mysterious phenomenon occurring deep within our own planet, according to a paper published on Thursday in Science.
Videos by VICE
One of these enormous anomalies, located deep under the Marquesas Islands, had never been detected before, while another structure beneath Hawaii was found to be much larger than previously estimated.
Scientists led by Doyeon Kim, a seismologist and postdoctoral fellow at the University of Maryland, fed seismograms captured from hundreds of earthquakes that occurred between 1990 to 2018 into an algorithm called Sequencer. While seismological studies tend to focus on relatively small datasets of regional earthquake activity, Sequencer allowed Kim and his colleagues to analyze 7,000 measurements of earthquakes—each with a magnitude of at least 6.5—that shook the subterranean world under the Pacific Ocean within the past three decades.
“This study is very special because, for the first time, we get to systematically look at such a large dataset that actually covers more or less the entire Pacific basin,” Kim said in a call. Though scientists have previously mapped out structures deep inside Earth, this study presents a rare opportunity to “bring everything in together and try to explain it in a global context,” he noted.
Earthquakes create seismic waves that travel through Earth’s interior where they become scattered and distorted by structures deep inside our planet. These warped patterns are captured in seismograms, which are recordings of wave activity inside Earth, enabling seismologists to capture rare glimpses of Earth’s inaccessible underworld.
The team focused on seismograms produced by shear (S) waves that travel along the boundary between Earth’s core and the lower portion of the mantle that borders it. These waves are the slower secondary waves that follow the initial tremors made by earthquakes, called primary (P) waves, and they generally produce clearer signals.
“We normally like to use S waves because they are larger in amplitude and the data is more or less clean because there is less P wave traffic,” said Kim. In particular, the team looked for the shear waves diffracting along the core-mantle boundary. “Because it diffracts along that surface, it’s a great phase to look for these tiny structures on top of the core-mantle boundary,” Kim noted.
When the shear waves hit these structures, they produce a type of echo-like signature known as a “postcursor” (there are helpful figures of this process on Kim’s website). These echoes indicate the presence of anomalies deep inside Earth called ultra low velocity zones (ULVZs), which are dense patches on the core-mantle boundary.
Nobody knows exactly how ULVZs are formed or what they are made of, but it’s clear that they have diameters of about a hundred kilometers and that they are dense enough to slow down waves that pass through them.
By running thousands of seismograms through Sequencer, Kim and his colleagues found that the strongest postcursor signals in their dataset emanate from under Hawai’i and the Marquesas Islands. This is tantalizing evidence of the existence of two “mega-ULVZs,” zones that stretch for about 1,000 kilometers, or more.
While the Hawaiian structure has been partially mapped out in previous studies, Kim’s team found that its dimensions are much larger than expected. Meanwhile, the mega-ULVZ detected under the Marquesas Islands represents “a previously unidentified localized wave-speed anomaly,” according to the study.
Mega-ULVZs are intriguing structures not only due to their size, but because they may be composed of exotic materials that date back to a time before Earth had a Moon. These huge anomalous chunks could be partially melted material that predate the Moon formation event, which scientists think was a gigantic collision between early Earth and a Mars-sized object more than four billion years ago.
“This is very interesting because this might indicate that mega-ULVZs are special and may host primitive geochemical signatures that have been relatively unmixed since early Earth history,” Kim said.
The new study demonstrates the applications of algorithms like Sequencer, which use a special type of process called unsupervised learning, in processing complex datasets like those found in astronomy, seismology, and countless other scientific fields. As opposed to supervised learning algorithms, which are trained to sort information based on known labels, unsupervised algorithms are designed to find insights in unlabelled datasets.
“What if we don’t really know what to look for in the dataset?” explained Kim. “This is the typical question we’d like to think about because the lower mantle, the target of our study, still has so many unknowns. It’s not really surprising to find almost anything in the lower mantle because we cannot actually go inside and take a look at it with our bare eyes.”
“When you use a sequencer, what it actually does is find additional information hidden behind this dataset,” he continued. “So, what we did here is find an optimal arrangement in the dataset itself. We’re not actually altering the dataset; we’re not doing anything but just rearranging and finding this optimal arrangement. That’s what Sequencer does.”
The team plans to continue developing this novel way of peering into Earth by examining higher-frequency waves that might yield finer details about the enigmatic structures on the core-mantle boundary. The researchers also hope to expand their dataset to seismograms produced under the Atlantic Ocean.
“We’re hoping that Sequencer will be able to basically let us use all of these diverse datasets and bring them together to look for these lower mantle structures systematically,” Kim concluded. “That is our vision going forward, to answer more questions about the lower mantle in general.”
