A team of international researchers, including some from Lawrence Berkeley National Laboratory, published a study May 7 fusing astronomy and machine learning, and offering a novel glimpse into the depths of the universe.
The study deals with gravitational lenses — a rare cosmic phenomenon where a pair of galaxies acts as a massive telescope lens. When these galaxies are aligned relative to Earth, the gravity of the foreground galaxy bends the light from the background galaxy, resulting in a distorted and magnified image, according to Xiaosheng Huang, the study’s lead author and a UC Berkeley alumnus.
Using a neural network, the researchers combed through millions of telescopic images to find 335 potential gravitational lenses, Huang said. He added that this work granted the team — composed mainly of University of San Francisco undergraduates — prized access to the Hubble Space Telescope for research purposes.
“It’s like you have a telescope with eyepieces the size of a galaxy,” explained astrophysicist and study co-author David Schlegel.
Schlegel also heads a separate project, which aims to use gravitational lenses to make a three-dimensional map of the universe.
The goal of Schlegel’s project, for which this study served as preparation, is to shed some light on dark matter and dark energy, the particles responsible for the lenses’ gravitational effects and, according to Huang, “two of the most mysterious entities in the universe today.”
A new paper, which has not been officially published, adds to previous results by identifying more than 1,000 potential lenses.
Several campus undergraduate students were involved in this iteration of the project, from generating complex pattern-based models to developing new machine-learning techniques in order to further narrow the artificial intelligence’s lensing candidates by sight, according to Huang.
The project was conceived in 2018, and Huang’s ultimate goal is to analyze strongly lensed supernovae, which holds applications including dark matter analysis and calculation of the universe’s expansion rate. Juxtaposing images from multiple lenses to examine the same supernova becomes an “exquisite” way to map space and time, according to Schlegel.
Rising campus sophomore Saurav Banka joined the project through UC Berkeley’s Undergraduate Research Apprentice Program in spring 2020, and expressed surprise at being trusted with building and experimenting with key components.
“I was initially under the impression that only experienced grad students were assigned such components,” Banka said in an email. “I’m still amazed by the scale the research being done by Dr. Huang’s group — till now, we’re one of the leading teams in the detection of strong gravitational lensing.”
Rising campus junior Andi Gu echoed the sentiment, stating that he feels fortunate to have his work potentially play a part in answering the dark matter problem, which has baffled astrophysicists for more than half a century.
However, Huang is cautious to term the research groundbreaking.
“We’re opening doors,” Huang said. “It’s the analysis of these systems that will enable us to address these big questions.”
The images from the Hubble Space Telescope have allowed Huang and his team to conduct this analysis, with the findings expected to be published this fall.