From Particles to Pixels: How many particles do I really need to construct stellar kinematic mock observational measurements?

TL;DR

This paper investigates how particle resolution affects the accuracy of mock stellar kinematic observations in galaxy simulations, providing guidelines for minimum particle counts to ensure reliable measurements.

Contribution

It introduces a method to determine the minimum particles-per-pixel needed for unbiased kinematic measurements in mock galaxy observations and implements a Voronoi-binning module for this purpose.

Findings

Minimum of 200 particles-per-pixel recommended to avoid bias in velocity dispersion estimates.

Binning mock images with sufficient particles improves the recovery of kinematic properties.

Resolution impacts the comparison of simulated and real galaxy kinematics.

Abstract

This work considers the impact of resolution in the construction of mock observations of simulated galaxies. In particular, when building mock integral field spectroscopic observations from galaxy formation models in cosmological simulations, we investigate the possible systematics that may arise given the assumption that all galaxies above some stellar mass limit will provide unbiased and meaningful observable stellar kinematics. We build a catalogue of N-body simulations to sample the range of stellar particle resolutions within the EAGLE Ref0050N0752 simulation box and examine how their observable kinematics vary relative to a higher-resolution N-body control. We use these models to compile a table of the minimum number of particles-per-pixel to reach a given uncertainty in the fitted line-of-sight velocity distribution parameters. Further, we introduce a Voronoi-binning module to the mock observation code, SimSpin, in order to meet these minimum numbers. Using EAGLE, we show the impact of this shot noise on the observed spin-ellipticity plane and the recovery of this space when observations are binned with increasing numbers of particles. In conclusion, we advise binning mock images to meet at least 200 particles-per-pixel to avoid systematically under-estimating the velocity dispersion along a given line-of-sight. We demonstrate that this is important for comparing galaxies extracted from the same simulation, as well as between simulations of varying mass resolution and observations of real galaxies.