close
close

Astronomers use machine learning techniques to find quasars from the early universe in a sea of ​​data

Astronomers use machine learning techniques to find quasars from the early universe in a sea of ​​data
Astronomers use machine learning techniques to find quasars from the early universe in a sea of ​​data

AI helps search for cosmic jewels in the sea of ​​data

A detailed image from the Dark Energy Survey showing the field covered by one of the individual detectors of the Dark Energy Camera. Image credit: DES Collaboration/NOIRLab/NSF/AURA/M. Zamani

Quasars are extremely luminous galactic cores in which gas and dust falling into a central supermassive black hole emit enormous amounts of light. Due to their exceptional brightness, these objects are visible at high redshifts, i.e. great distances.

A higher redshift not only indicates that a quasar is farther away, but also that it lies farther back in time. Astronomers are interested in studying these ancient objects because they contain clues about the evolution of our universe in its early youth.

Candidates for high-redshift quasars are first identified by their color—they are very red—and then must be confirmed as such by looking at individual observations of their spectra. However, some high-redshift candidates may be falsely excluded from further study because their appearance is distorted by gravitational lensing.

This phenomenon occurs when there is a massive object, such as a galaxy, between us and a distant object. The mass of the galaxy curves space and acts a bit like a magnifying glass. This bends the path of the distant object’s light, resulting in a distorted image of the object.

While this alignment can be beneficial—the gravitational lens magnifies the image of the quasar, making it brighter and easier to see—it can also deceptively alter the quasar’s appearance.

Disturbing light from stars in the intervening lensing galaxy may make the quasar appear bluer, while the curvature of spacetime may make it appear smeared or multiplied. Both effects make it likely to be ruled out as a quasar candidate.

A team of astronomers led by Xander Byrne, an astronomer at the University of Cambridge and lead author of the paper that published these results in Monthly Notices of the Royal Astronomical SocietyThe aim was to rediscover the lens quasars that had been overlooked in previous studies.

Byrne set out to search for these missing treasures in the vast archive of data from the Dark Energy Survey (DES). DES was conducted using the Department of Energy-manufactured Dark Energy Camera mounted on the Víctor M. Blanco 4-meter telescope at the U.S. National Science Foundation’s Cerro Tololo Inter-American Observatory, a program of NSF NOIRLab.

The challenge was to find a method to extract these cosmic gems from the vast ocean of data.

The full DES dataset contains more than 700 million objects. Byrne pared down this archive by comparing the data with images from other surveys to filter out unlikely candidates, including objects that were likely brown dwarfs, which, while completely different from quasars in almost every way, can look surprisingly similar to them in images. This process yielded a much more manageable dataset of 7,438 objects.

Byrne needed to maximize efficiency in the search for these 7,438 objects, but knew that conventional techniques would likely miss the high-redshift quasars he was looking for. “To avoid prematurely discarding high-redshift quasars, we used a contrastive learning algorithm, and it worked beautifully.”

Contrastive learning is a type of artificial intelligence (AI) algorithm that uses sequential decisions to classify each data point into a group based on what it is or isn’t. “It may seem like magic,” Byrne said, “but the algorithm doesn’t use any more information than what’s already in the data. Machine learning is about figuring out which bits of data are useful.”

Byrne’s decision not to rely on human visual interpretation led him to consider an unsupervised AI process, meaning that the algorithm itself controls the learning process rather than a human.

Supervised machine learning algorithms rely on a so-called “ground-level truth” defined by a human programmer. For example, the process might start with a description of a cat and go through decisions like “This is/is not a picture of a cat. This is/is not a picture of a black cat.”

In contrast, unsupervised algorithms do not rely on this initial human-determined definition as the basis for their decisions. Instead, the algorithm sorts each data point by similarities to the other data points in the set. In this case, the algorithm would find similarities between images of multiple animals and group them as cat, dog, giraffe, penguin, etc.

Starting with Byrne’s 7,438 objects, the unsupervised algorithm sorted the objects into groups. The team used a geographic analogy and called the data groupings an archipelago. (The term does not imply spatial proximity between the objects. It is their properties that group them “close” to each other, not their positions in the sky.)

Within this archipelago, a small “island” subset of objects were grouped together as possible quasar candidates. Among these candidates, four stood out like gems from a pile of pebbles.

Using archival data from the Gemini South telescope, one half of the Gemini International Observatory operated by the NSF NOIRLab, Byrne confirmed that three of the four candidates on the “quasar island” are indeed high-redshift quasars. And one of them is most likely the cosmic prize Byrne was hoping to find – a gravitationally lensed high-redshift quasar. The team now plans to take more images to confirm the quasar’s lensed nature.

“If the discovery of a single lensed quasar is confirmed in a sample of four targets, the success rate would be remarkably high! And if this search had been conducted using standard search methods, this gem would likely have remained hidden.”

Byrne’s work is a clever example of how AI could help astronomers search through ever-larger troves of data. Massive influxes of astronomical data are expected in the coming years, with the ongoing five-year survey of the Dark Energy Spectroscopic Instrument, as well as the upcoming Legacy Survey and Space and Time, which will be run by the Vera C. Rubin Observatory starting in 2025.

More information:
Xander Byrne et al, Quasar Island – three new z ∼ 6 quasars, including a lens-shaped candidate, identified using contrastive learning, Monthly Notices of the Royal Astronomical Society (2024). DOI: 10.1093/mnras/stae902

Provided by NSF’s NOIRLab

Quote: Astronomers use machine learning techniques to find quasars from the early universe in a sea of ​​data (July 11, 2024), accessed July 11, 2024, from https://phys.org/news/2024-07-astronomers-machine-techniques-early-universe.html

This document is subject to copyright. Except for the purposes of private study or research, no part of it may be reproduced without written permission. The contents are for information purposes only.

Leave a Reply

Your email address will not be published. Required fields are marked *