Description
Dark matter halos, central to much of cosmology and astrophysics, are often summarised by their radial density profiles. These are typically assumed to have a fixed functional form (such as NFW) inspired by simulations, with a number of free parameters that can be constrained by fits to observational data. However, the approach of relying on simulations is undesirable, as the dark matter density profile depends on the adopted dark matter model (e.g. cold vs. self-interacting dark matter) and the baryonic physics implementation, both of which are highly uncertain.
To address this, we have developed a new method to constrain the functional form of halo density profiles directly from weak lensing data, without input from simulations. This is done using a novel symbolic regression algorithm called Exhaustive Symbolic Regression (ESR), which learns the optimal analytic expression to fit a set of data whilst using minimum description length, a tool from information theory, to penalise unnecessary complexity. Hence our algorithm can identify the halo profile with maximum empirical accuracy.
We apply ESR to weak lensing measurements from the Hyper Suprime-Cam (HSC) survey for X-ray-selected galaxy groups and clusters from the XMM-XXL survey. Our method identifies dark matter density profiles that provide a better fit than the standard NFW profile. With the influx of high-quality data from upcoming surveys, our method will improve inferences that rely on assumed density profiles and provide strong empirical constraints on the nature of dark matter.
