site stats

Astronomy Space

Daily Dose of Astronomy

Revolution in the Stars: 562 Potential Strong Lenses Identified Using AI Technology

Dive into the captivating world of astronomy as we explore the groundbreaking discovery of 562 new candidate strong lenses by an international team of astronomers using machine learning. This article unveils the powerful blend of AI and celestial observation, revealing the secrets of the universe through the lens of the Blanco Telescope and the Dark Energy Camera. Join us in this cosmic journey to understand the mysteries of gravitational lensing and its implications for our understanding of the cosmos.

High above the earthly realm, nestled amidst the Chilean peaks at 7,200 feet, lies a portal to the cosmic wonders – the Cerro Tololo Inter-American Observatory. Housing the colossal 4-meter Blanco Telescope since 1976, this observatory has been a silent witness to the night sky’s enigmatic beauty. But as the modern era dawns, a deluge of data challenges the limits of human analysis.

In an intriguing twist, a global team of astronomers has embraced the future, wielding machine learning to mine through the telescope’s vast image repository. Their quest? To uncover the faint echoes of distant galaxies from a staggering pool of 11 million celestial snapshots. This hunt focuses on a rare and intriguing phenomenon: strong gravitational lensing.

Imagine a colossal cosmic entity, nestled in the fabric of space, bending and twisting the light of a galaxy far beyond. This gravitational dance not only magnifies but also distorts the distant galaxy into surreal, stretched forms. These cosmic mirages, known as strongly lensed galaxies, are gateways to understanding the universe’s deepest secrets.

Despite the vast expanse of cataloged galaxies, only about a thousand lensed spectacles have been confirmed. Erik Zaborowski, a graduate student from Ohio State University and the lead author of this groundbreaking study, published in the Astrophysical Journal, emphasizes the rarity of these cosmic phenomena. The southern sky, largely unexplored, now reveals its treasures through the lens of the Blanco Telescope, equipped with the Dark Energy Camera (DECam).

These galactic enigmas offer insights into some of astronomy’s most profound puzzles, from the mysteries of dark matter to the universe’s expansion rate. The Southern Hemisphere, long overshadowed by its northern counterpart, is now under the rigorous scrutiny of astronomers.

The DECam, which first saw light in 2012, is an exceptional tool in this endeavor. Designed to capture vast sections of the sky, each image encompasses a view equivalent to 20 full Moons. With such a vast canvas, the challenge of sifting through astronomical data escalates beyond human capabilities.

Greg Mosby of NASA Goddard, not directly involved in the study but a practitioner of machine learning in astronomy, points to the inevitability of automation in data analysis. Machine learning, a subset of artificial intelligence, is revolutionizing various fields with its ability to learn from training data without explicit programming. In astronomy, this technology is increasingly deployed for image identification and analysis.

Zaborowski and his team embarked on an unprecedented journey, applying machine learning to the public data from the DECam Local Volume Exploration Survey (DELVE), which has recorded over 520 million cosmic sources since 2019. From this immense dataset, 11 million extended sources were selected for the search.

Before setting the algorithm loose on this cosmic hunt, the team engaged in a meticulous training process. They fed the model over 80,000 real galaxy images from DELVE, artificially inducing lensing effects on half. Additionally, over 3,200 false positives were included to refine the model’s accuracy.

Yet, the human touch remained integral. The model identified 50,000 potential lenses, which the astronomers painstakingly reviewed. The result? A discovery of 581 probable strongly lensed galaxies, 562 previously unreported, and eight potential lensed quasars.

While this marks a significant leap in the catalog of lensed galaxies, it also prompts a reevaluation of the role of giant databases and machine-learning models in astronomy. The scarcity of real lens images necessitates reliance on simulated data for training, raising concerns about the models’ effectiveness with unanticipated data types. Ben Metcalf of the University of Bologna, not involved in the study, voices reservations about this approach.

Despite these challenges, machine learning stands as a beacon of speed and efficiency, transforming months of tedious classification into mere hours. As data continues to proliferate, this study heralds a new era where machine learning not only accelerates discovery but also aids in deciphering the enigmatic messages these lenses convey about our universe.

Erik Zaborowski envisions a future where the abundance of lens data can unlock answers to fundamental scientific questions, propelling our understanding of the cosmos to new horizons.

Reference: skyandtelescope