Tracing galaxy populations through cosmic time: A critical test of methods for connecting the same galaxies between different redshifts at $z < 3$

TL;DR

This study evaluates methods for linking galaxies across cosmic time, finding that constant number density selection outperforms stellar mass selection but still faces limitations, with velocity dispersion showing promise as a better tracer.

Contribution

It critically tests and compares galaxy selection methods across redshifts using semi-analytical models, highlighting the advantages of constant number density selection and exploring alternative tracers.

Findings

Constant number density selection better recovers galaxy properties than stellar mass selection.

Recovery accuracy depends on the chosen number density, with about 50% accuracy at a specific density.

Velocity dispersion may serve as a more reliable galaxy property tracer.

Abstract

Connecting galaxies with their descendants (or progenitors) at different redshifts can yield strong constraints on galaxy evolution. Observational studies have historically selected samples of galaxies using a physical quantity, such as stellar mass, either above a constant limit or at a constant cumulative number density. Investigation into the efficacy of these selection methods has not been fully explored. Using a set of four semi-analytical models based on the output of the Millennium Simulation, we find that selecting galaxies at a constant number density (in the range $-4.3 < \log\ n\ [\mathrm{Mpc}^{-3}\ h^{3}] < -3.0$) is superior to a constant stellar mass selected sample, although it still has significant limitations. Recovery of the average stellar mass, stellar mass density and average star-formation rate is highly dependent on the choice of number density but can all be recovered to within $<50\%$ at the commonly employed choice of $\log\ n\ [\mathrm{Mpc}^{-3}\ h^{3}] = -4.0$, corresponding to $\log M_\odot / h \sim 11.2$ at $z=0$, but this increases at lower mass limits. We show that there is a large scatter between the location of a given galaxy in a rank ordering based on stellar mass between different redshifts. We find that the inferred velocity dispersion may be a better tracer of galaxy properties, although further investigation is warranted into simulating this property. Finally, we find that over large redshift ranges selection at a constant number density is more effective in tracing the progenitors of modern galaxies than vice-versa.