A team of astrophysicists from the United States and Korea has created a new dark-matter distribution map using a neural network-based deep learning method and the data on positions and velocities of galaxies in the local Universe.
The 3D density map of the local dark matter: X-mark at the center denotes the Milky Way Galaxy; dots denote galaxies, and arrows denote estimated directions of motion derived from the gradient of the reconstructed gravitational potential. Image credit: Hong et al., doi: 10.3847/1538-4357/abf040.
“The 80% of the matter in the Universe is in the form of dark matter that comprises the skeleton of the large-scale structure called the cosmic web,” said Dr. Donghui Jeong, an astrophysicist in the Department of Astronomy and Astrophysics and the Institute for Gravitation and the Cosmos at the Pennsylvania State University.
“As the cosmic web dictates the motion of all matter in galaxies and inter-galactic media through gravity, knowing the distribution of dark matter is essential for studying the large-scale structure.”
“However, the cosmic web’s detailed structure is unknown because it is dominated by dark matter and warm-hot inter-galactic media, both of which are hard to trace.”
In the study, Dr. Jeong and colleagues took a completely different approach, using machine learning to build a model that uses information about the distribution and motion of galaxies to predict the distribution of dark matter.
They built and trained their model using a large set of galaxy simulations, called Illustris-TNG, which includes galaxies, gasses, other visible matter, as well as dark matter.
They specifically selected simulated galaxies comparable to those in the Milky Way and ultimately identified which properties of galaxies are needed to predict the dark matter distribution.
“When given certain information, the model can essentially fill in the gaps based on what it has looked at before,” Dr. Jeong said.
“The map from our models doesn’t perfectly fit the simulation data, but we can still reconstruct very detailed structures.”
“We found that including the motion of galaxies — their radial peculiar velocities — in addition to their distribution drastically enhanced the quality of the map and allowed us to see these details.”
The researchers then applied their model to real data from the local Universe from the Cosmicflow-3 galaxy catalog.
The map successively reproduced known prominent structures in the local Universe, including the Local Sheet (a region of space containing the Milky Way, nearby galaxies in the Local Group, and galaxies in the Virgo Cluster) and the Local Void (a relatively empty region of space next to the Local Group).
Additionally, it identified several new structures that require further investigation, including smaller filamentary structures that connect galaxies.
“Having a local map of the cosmic web opens up a new chapter of cosmological study,” Dr. Jeong said.
“We can study how the distribution of dark matter relates to other emission data, which will help us understand the nature of dark matter.”
“And we can study these filamentary structures directly, these hidden bridges between galaxies.”
Sungwook E. Hong et al. 2021. Revealing the Local Cosmic Web from Galaxies by Deep Learning. ApJ, in press; doi: 10.3847/1538-4357/abf040