AI Boosts Precision in Dark Energy Analysis, Reveals Universe's Expansion Secrets

AI Boosts Precision in Dark Energy Analysis, Reveals Universe's Expansion Secrets
by Sophie Jenkins
London, UK (SPX) Mar 13, 2024

In a study by UCL and the Dark Energy Survey collaboration, artificial intelligence (AI) has significantly refined our understanding of dark energy, doubling the precision of how we gauge the universe's key characteristics over the past seven billion years. By analyzing a comprehensive map of both dark and visible matter, this research provides a groundbreaking method for more accurately defining the universe's makeup, including the density of dark energy.

Dark energy, a dominant force behind the universe's expansion, accounts for approximately 70% of the universe's total content, with dark matter and normal matter constituting 25% and 5%, respectively. Dr. Niall Jeffrey of UCL Physics and Astronomy highlighted the breakthrough, stating, "By employing AI to study simulations of various universes, we've managed to double our precision in estimating crucial universal properties. Achieving this through traditional means would necessitate quadruple the data, or an additional mapping of 300 million galaxies."

Dr. Lorne Whiteway, also of UCL Physics and Astronomy, discussed the implications of their findings on existing cosmological models, specifically the cosmological constant, and opened the door to alternative explanations, such as potential inaccuracies in our current understanding of gravity.

This investigation builds on the 2021 Dark Energy Survey analysis but with improved precision. The survey itself utilized weak gravitational lensing to study the distribution of matter across a quarter of the Southern Hemisphere's sky by examining the light distortion from 100 million galaxies.

By simulating various universe models with UK government-funded supercomputers and employing machine learning to analyze these simulations, researchers have developed a novel approach that leverages a wealth of data previously untapped. These advancements were made possible through the UK's DiRAC High Performance Computing facility.

Looking forward, projects like the European Space Agency's Euclid mission promise to expand our dataset on the universe's structure, potentially resolving the current discrepancy between the observed smoothness of the universe and predictions made from the cosmic microwave background analysis.

The Dark Energy Survey collaboration, a global effort led by UCL and hosted by the Fermi National Accelerator Laboratory, has cataloged hundreds of millions of galaxies. The 570-megapixel Dark Energy Camera, integral to this research, was stationed at Chile's Cerro Tololo Inter-American Observatory.

Dark Energy Survey Year 3 results: likelihood-free, simulation-based wCDM inference with neural compression of weak-lensing map Research Report:statistics