Stars on the outskirts of galaxies rotate faster than predicted by the laws of physics if only visible matter is considered. This paradox was uncovered in the 20th century by Vera Rubin and Fritz Zwicky. The answer was the hypothesis of invisible mass — dark matter, which envelops galaxies and amplifies gravity. To find its exact location, astronomers need ultra-precise gravitational maps.
A new method turns a neural network into a cartographer. It takes a simple physical model of gravity, like a contour map, and, based on observational data, draws in the relief details — hills and valleys of gravity.
The accuracy impresses: on test models, the error was only 0.14% — that's like determining the height of Everest with an error of a couple of meters. Real data from the JWST telescope will be an ideal basis for such maps.
Dark matter creates an invisible gravitational landscape: stars, like marbles, roll down its slopes.