How to Zoom: Bias, Contamination, and Lagrange Volumes in Multimass Cosmological Simulations

TL;DR

This paper investigates how to select Lagrangian regions for multimass cosmological zoom simulations to avoid contamination, providing a formula for optimal selection and analyzing correlations with halo properties.

Contribution

It introduces a simple formula for selecting Lagrangian regions based on zoom level and assesses the relationship between Lagrangian volume and halo characteristics.

Findings

Contamination risk increases with zoom level.

A monotonic increase in Lagrangian volume prevents contamination.

No strong correlation between Lagrangian volume and most halo properties.

Abstract

We perform a suite of multimass cosmological zoom simulations of individual dark matter halos and explore how to best select Lagrangian regions for resimulation without contaminating the halo of interest with low-resolution particles. Such contamination can lead to significant errors in the gas distribution of hydrodynamical simulations, as we show. For a fixed Lagrange volume, we find that the chance of contamination increases systematically with the level of zoom. In order to avoid contamination, the Lagrangian volume selected for resimulation must increase monotonically with the resolution difference between parent box and the zoom region. We provide a simple formula for selecting Lagrangian regions (in units of the halo virial volume) as a function of the level of zoom required. We also explore the degree to which a halo's Lagrangian volume correlates with other halo properties (concentration, spin, formation time, shape, etc.) and find no significant correlation. There is a mild correlation between Lagrange volume and environment, such that halos living in the most clustered regions have larger Lagrangian volumes. Nevertheless, selecting halos to be isolated is not the best way to ensure inexpensive zoom simulations. We explain how one can safely choose halos with the smallest Lagrangian volumes, which are the least expensive to resimulate, without biasing one's sample.