Simulated Bars May Be Shorter But Are Not Slower Than Observed: TNG50 vs. MaNGA

TL;DR

This study compares the properties of galactic bars in the TNG50 cosmological simulation with observations from MaNGA, finding that simulated bars are shorter but not slower than observed, highlighting the impact of resolution on bar dynamics.

Contribution

It demonstrates that high-resolution cosmological simulations can produce bar pattern speeds consistent with observations, and clarifies the role of resolution in bar length and speed discrepancies.

Findings

Simulated bars have pattern speeds similar to observed bars.

Bars in TNG50 are approximately 35% shorter than observed.

Resolution effects influence the perceived slowness of simulated bars.

Abstract

Galactic bars are prominent dynamical structures within disk galaxies whose size, formation time, strength, and pattern speed influence the dynamical evolution of their hosts galaxies. Yet, their formation and evolution in a cosmological context is not well understood, as cosmological simulation studies have been limited by the classic trade off between simulation volume and resolution. Here we analyze barred disk galaxies in the cosmological magneto-hydrodynamical simulation TNG50 and quantitatively compare the distributions of bar size and pattern speed to those from MaNGA observations at $z=0$. TNG50 galaxies are selected to match the stellar mass and size distributions of observed galaxies, to account for observational selection effects. We find that the high-resolution of TNG50 yields bars with a wide range of pattern speeds (including those with $\geq 40~\mathrm{km}\,\mathrm{s}^{-1}$\,$\mathrm{kpc}^{-1}$) and a mean value of $\sim36~\mathrm{km}\,\mathrm{s}^{-1}\,\mathrm{kpc}$ consistent with observations within $6\,\mathrm{km}\,\mathrm{s}^{-1}$\,$\mathrm{kpc}^{-1}$, in contrast with previous lower-resolution cosmological simulations that produced bars that were too slow. We find, however, that bars in TNG50 are on average $\sim 35\%$ shorter than observed, although this discrepancy may partly reflect remaining inconsistencies in the simulation-data comparison. This leads to higher values of $\mathcal{R} = R_\mathrm{corot}/R_\mathrm{bar}$ in TNG50, but points to simulated bars being `too short' rather than `too slow'. After repeating the analysis on the lower-resolution run of the same simulation (with the same physical model), we qualitatively reproduce the results obtained in previous studies: this implies that, along with physical model variations, numerical resolution effects may explain the previously found `slowness' of simulated bars.