Dark matter halos and universal relations

One of the ultimate goals of physics is to find universal laws that govern nature. Such laws reveal the underlying order and simplicity of natural systems and are important because of their predictive abilities. A prominent example is the law of gravity. It not only permeates daily life in governing the motion of the things we see everyday but also governs the motion of the largest structures in outer space, ranging from stars to galaxies. In fact, this relation applies to all objects that possess mass which makes it a powerful tool for explaining the motion of a variety of systems.

Similarly, in cosmology and astrophysics, universal laws play an important role in understanding complex and mysterious structures such as dark matter halos. Dark matter is an invisible substance that constitutes around 85% of all matter in the universe. Halos of dark matter surround every galaxy in our universe, exerting gravitational pulls on other nearby dark and visible matter. Therefore, understanding the laws underlying their formation and structural shape will not only help unravel the mysteries surrounding dark matter but also provide important information about the formation and evolution of the galaxies they contain.

Unfortunately, this is a very difficult task because dark matter halos are characterized by many different properties. These include their radius, the masses of their gas and star constituents, the rate at which stars form, how fast they spin, and many more. With so many variables at play, it is extremely hard to find underlying relationships among these variables. However, recent advances in artificial intelligence enable the search for such relationships by making it easier to uncover patterns among large amounts of data.

The data that we use are simulations of dark matter halos and of the galaxies they contained provided by the Cosmology with Astrophysics with MachinE Learning Simulations (CAMELS) project. CAMELS contain 4,233 simulations and each represents a different Universe simulated with different cosmological and astrophysical parameters. By feeding this immense data to artificial intelligence algorithms, we have been able to identify a potentially new and universal relation between the mass of the halo and its twelve other properties. We have also found an equation that accurately approximates this relation, which consists of a power law using three of the halo properties: it’s radius, maximum circular velocity, and velocity dispersion.

Remarkably, this relation is accurate for halos of all sizes and can even accurately predict the masses of halos that are much larger and much smaller than the ones that the algorithm originally encountered. These include halos that are generated from simulations other than those from CAMELS, which the algorithm has never seen. In addition, the found relationship is still able to accurately predict masses of halos from other simulations that are run using a very different code and physics model. This is very surprising because the physics model that drives a simulation greatly affects the formation of the galaxies within the halos. Details regarding these results and other tests that verify the robustness of the found relation can be found in the paper linked below.

Overall, the accuracy of the relation’s predictions of the halo mass in all of the above-mentioned tests indicates that the relation may be a universal one. This result bears importance because it points to a more fundamental mechanism underlying the structure of dark matter halos that is not affected by cosmology, astrophysics, subgrid physics model of the simulations, and many more factors. While we speculate that this relation may be based on the gravitationally-driven nature of the halos, this is uncertain because of the manner in which the properties are defined in the simulations. Furthermore, the presented method obtaining relations among a high number of variables using artificial intelligence can also be very useful when applied to other datasets to reveal complex, hidden patterns. Details of this procedure can be found in the paper linked below.

Further Reading

Paper: https://arxiv.org/abs/2109.04484

Video: https://www.youtube.com/watch?v=_lIXL4-wkZ0

Post Author

Helen Shao

Undergraduate, Princeton University

Department of Astrophysical Sciences

Peyton Hall, Princeton NJ 08544, USA