Artificial intelligence developed the formula for finding the mass of galactic clusters. article about it published In the Proceedings of the National Academy of Sciences.
The distribution of mass in the universe is the most important topic in astrophysics and cosmology. At the same time, it is very difficult to estimate the mass of distant galactic clusters, since the most convenient method – the study of orbital parameters – does not work with them. In this context, astrophysicists measure the parameters of galactic clusters using the Sunyaev-Zeldovich effect. Gravity compresses interstellar gas, and the scattering of radio emission photons by its hot electrons depends on the degree of this compression. In other words, the mass of a cluster can be estimated from how the cosmic microwave background radiation passing through it changes.
But the formula developed at the end of the 20th century was not ideal. Now Digvijay Wadekar of the Institute for Advanced Study at Princeton and his colleagues have used artificial intelligence and symbolic regression to find a more accurate formula. The tool basically goes through various combinations of mathematical operators such as addition and subtraction with different variables to see which equation best fits the observational data.
Vadekar and his collaborators fed the program with a cutting-edge universe simulation containing many galaxy clusters. Programs later written by mathematician Miles Cranmer sought and identified additional variables that could make mass estimates more accurate.
AI is useful for identifying new combinations of parameters that human analysts might miss. The program can better analyze large amounts of data by finding often unexpected influencing factors, while people’s attention is often limited to a few parameters. As a result, the program added a new term to the equation, and human researchers discovered it had a physical meaning. It turns out that the old formula incorrectly predicted the effect of supermassive black holes at the centers of galaxies on the microwave background.
During calculations using the new formula, it was revealed that the equation created by artificial intelligence reduces the spread in mass estimates of galaxy clusters by about 20-30 percent for large clusters compared to the old method.