Self-Interacting Dark Matter and the Origin of Ultradiffuse Galaxies NGC1052-DF2 and -DF4

Observations of ultradiffuse galaxies NGC 1052-DF2 and -DF4 show they may contain little dark matter, challenging our understanding of galaxy formation. Using controlled $N$-body simulations, we explore the possibility that their properties can be reproduced through tidal stripping from the elliptical galaxy NGC 1052, in both cold dark matter (CDM) and self-interacting dark matter (SIDM) scenarios. To explain the dark matter deficiency, we find that a CDM halo must have a very low concentration so that it can lose sufficient inner mass in the tidal field. In contrast, SIDM favors a higher and more reasonable concentration as core formation enhances tidal mass loss. Final stellar distributions in our SIDM benchmarks are more diffuse than the CDM one, and hence the former are in better agreement with the data. We further show that a cored CDM halo model modified by strong baryonic feedback is unlikely to reproduce the observations. Our results indicate that SIDM is more favorable for the formation of dark-matter-deficient galaxies.

Figure 1

Enclosed mass profiles before (dashed) and after (solid) tidal evolution for CDM, SIDM3, and SIDM5 benchmarks. The black arrows denote the upper limits on the total dynamical mass within $r=3.1$ and 7.6 kpc for the NGC 1052-DF2 galaxy at 90% C.L. [2, 3]. As a reference, values in the parentheses indicate deviations of the initial halo concentrations from the median of the concentration-mass relation ($z=0$) predicted in cosmological simulations [60].

Figure 2

Left: final stellar density profiles for the CDM, SIDM3, and SIDM5 benchmarks (solid), as well as the initial condition (dashed). The green band denotes the measured stellar density of NGC 1052-DF4 [4], and the arrows denote the half-mass radii of the stellar components from the simulations. Right: evolution of the bound halo masses for the benchmarks (solid). For comparison, we also show results without including stellar particles in simulations (dashed).