Astrophysicists from the United States have used AI model SimBIG to calculate and refine the five main cosmological parameters with an accuracy that exceeds traditional methods.
Scientists have long been trying to study the parameters that control the processes in the Universe as accurately as possible. The more accurately the main cosmological parameters are measured, the closer to the real world the models created by scientists will become. This will also make it possible to confirm or refute generally accepted theories of the origin and evolution of the Universe.
Among the most challenging tasks at the moment are the refinement of the parameters that determine the amount and nature of dark matter, its source and the impact it has, as well as the description of the behavior of mysterious neutrinos. Such studies are usually very expensive and time-consuming.
«There are so many experiments and astronomical studies just to measure these six to ten parameters, and they cost several billion dollars», — notes one of the authors of the study, Professor of Astrophysics at the Flatiron Institute in New York Shirley Ho.
The study of cosmological parameters typically involves examining how galaxies are distributed and what their individual characteristics are through astronomical surveys. Astrophysicists use simplified models of the universe, comparing them with data obtained from surveys. However, this approach does not take into account some small details due to the high costs and problems associated with creating cosmological models with higher resolution.
In the new study, American scientists used artificial intelligence model Simulation-Based Inference of Galaxies (SimBIG), which was previously trained on 2 thousand simulated universes, each with different cosmological parameters. After that, the researchers added noise to these models to bring the simulated universes closer to our real one. The noise represents the imperfections of modern telescopes, as well as optical distortions.
The new AI model has learned to use small differences in the distribution of galaxies, such as the distance between individual galaxy pairs, to more accurately determine cosmological parameters. Then the trained model was provided with more than 100 thousand galaxies from the spectroscopic survey of baryon oscillations (BOSS) Sloan Digital Sky Survey. This number of galaxies makes up only 10% of the total survey.
The SimBIG AI model was able to use both large-scale and small-scale information from real data to refine cosmological parameters with an error of only ≤5%, which is twice as good as traditional methods.
Future research will collect more and more information about our Universe. SimBIG AI-like models will be able to help solve the most pressing issues today, such as the «Hubble voltage», which causes disagreements in estimating the expansion rate of the Universe.
In preparation for future research Shirley Ho, her colleague Liam Parker and other astrophysicists have teamed up to create a large collection of astronomical data. This work combines hundreds of millions of publicly available datasets from major astronomical surveys in an attempt to enable the development of large multimodal models specifically targeted to scientific applications.
The team of astrophysicists plans to apply the SimBIG AI model to 37 million galaxies to be surveyed LSST telescope at the Vera Rubin Observatory in Chile in the first year of operation. Provided that the accuracy is maintained, astronomers hope to verify the «Hubble voltage» and refine the parameter of the expansion rate of the Universe H0 in the early 2030s.
The results of the study are published in the journal Nature Astronomy
Spelling error report
The following text will be sent to our editors: