Haunted haloes: tracking the ghosts of subhaloes lost by halo finders

TL;DR

This paper introduces a particle tracking algorithm that recovers lost subhaloes in N-body simulations, significantly improving subhalo statistics and large-scale structure measurements by addressing halo finder limitations.

Contribution

The authors develop a post-processing particle tracking method to recover subhaloes lost due to halo finder failures, enhancing the accuracy of structure formation simulations.

Findings

Subhalo mass function increases by ~50%.

Halo correlation function doubles at small scales.

Tracking reduces reliance on orphan models.

Abstract

Dark matter subhaloes are key for the predictions of simulations of structure formation, but their existence frequently ends prematurely due to two technical issues, namely numerical disruption in N-body simulations and halo finders failing to identify them. Here we focus on the second issue, using the phase-space friends-of-friends halo finder ROCKSTAR as a benchmark (though we expect our results to translate to comparable codes). We confirm that the most prominent cause for losing track of subhaloes is tidal distortion rather than a low number of particles. As a solution, we present a flexible post-processing algorithm that tracks all subhalo particles over time, computes subhalo positions and masses based on those particles, and progressively removes stripped matter. If a subhalo is lost by the halo finder, this algorithm keeps tracking its so-called ghost until it has almost no particles left or has truly merged with its host. We apply this technique to a large suite of N-body simulations and restore lost subhaloes to the halo catalogues, which has a dramatic effect on key summary statistics of large-scale structure. Specifically, the subhalo mass function increases by about 50% and the halo correlation function increases by a factor of two at small scales. While these quantitative results are somewhat specific to our algorithm, they demonstrate that particle tracking is a promising way to reliably follow haloes and reduce the need for orphan models. Our algorithm and augmented halo catalogues are publicly available.