L-PICOLA: A parallel code for fast dark matter simulation

TL;DR

L-PICOLA is a fast, scalable, parallel code for generating dark matter simulations that accurately reproduce key statistical measures, significantly faster than traditional N-Body simulations, aiding large-scale structure survey analyses.

Contribution

The paper introduces L-PICOLA, a novel parallel code capable of rapidly producing accurate dark matter simulations including non-Gaussianity and lightcone effects, outperforming existing methods.

Findings

Reproduces $z=0$ power spectrum within 2%

Reproduces reduced bispectrum within 5%

Runs three orders of magnitude faster than full N-Body simulations

Abstract

Robust measurements based on current large-scale structure surveys require precise knowledge of statistical and systematic errors. This can be obtained from large numbers of realistic mock galaxy catalogues that mimic the observed distribution of galaxies within the survey volume. To this end we present a fast, distributed-memory, planar-parallel code, L-PICOLA, which can be used to generate and evolve a set of initial conditions into a dark matter field much faster than a full non-linear N-Body simulation. Additionally, L-PICOLA has the ability to include primordial non-Gaussianity in the simulation and simulate the past lightcone at run-time, with optional replication of the simulation volume. Through comparisons to fully non-linear N-Body simulations we find that our code can reproduce the $z=0$ power spectrum and reduced bispectrum of dark matter to within 2% and 5% respectively on all scales of interest to measurements of Baryon Acoustic Oscillations and Redshift Space Distortions, but 3 orders of magnitude faster. The accuracy, speed and scalability of this code, alongside the additional features we have implemented, make it extremely useful for both current and next generation large-scale structure surveys. L-PICOLA is publicly available at https://cullanhowlett.github.io/l-picola