Generating approximate halo catalogues for blind challenges in precision cosmology

Lehman H. Garrison, Daniel J. EisensteinMonthly Notices of the Royal Astronomical Society 2019 · DOI

TL;DR

This paper introduces a novel "warping" technique to efficiently generate realistic cosmological simulations by slightly modifying existing ones. This method allows for high-precision halo catalogs with minimal computational cost, crucial for blind cosmological challenges where simulations must be reused without revealing underlying parameters.

What Problem Does This Paper Address?

Cosmological N-body simulations are essential for understanding the universe's large-scale structure, but they are computationally very expensive. This limits our ability to thoroughly explore the vast parameter space of cosmological models. Furthermore, in "blind challenges" – where researchers try to infer cosmological parameters from mock data without knowing the true values – simulations are typically single-use. Once a simulation's cosmology is revealed, it cannot be reused, necessitating new simulations for each challenge. This paper addresses the need for a method to efficiently generate multiple, high-precision mock catalogs from a limited number of expensive N-body simulations, specifically for small cosmological variations, and to enable the reuse of simulations in blind challenges.

Methods

The core of the paper is a "warping" procedure that modifies an existing simulation's halo catalog to match a target cosmology. This is achieved in four main steps:

  1. Initial Condition Residuals: The method starts by leveraging the difference in initial conditions (ICs) between the original and target simulations, scaled to the catalog redshift. This accounts for large-scale structure differences.
  2. Halo Property Rescaling: Halo radii and velocity dispersions are rescaled based on their abundance to match the target cosmology. This step addresses internal halo structure differences.
  3. Real-Space Transfer Function: A key innovation is fitting an isotropic "transfer function" in Fourier space. This function displaces halos to precisely match the target halo power spectrum in real space. This is done by calculating a "late-time displacement field" from the matter density and then applying a k-dependent transfer function to it.
  4. Redshift-Space Transfer Function: A similar transfer function is fitted for velocities to match the redshift-space monopole and quadrupole power spectra, accounting for peculiar velocities and line-of-sight effects.

The method uses a pseudo-HOD (halo occupation distribution) prescription to assign galaxies to halo particles, which helps suppress shot noise.

Datasets

The authors use halo catalogs from the AbacusCosmos suite of simulations. Specifically, they utilize two phase-matched simulations: AbacusCosmos_1100box_00-0 (original) and AbacusCosmos_1100box_01-0 (target). These simulations have 1440³ particles in 1100h⁻¹ Mpc boxes and differ by a 1% change in σ₈ (the amplitude of matter fluctuations). They use halo catalogs at z=0.5 and employ 10% subsamples of halo and field particles.

Key Insights and Ideas

  • Halo Bias Cancellation: A surprising initial finding is that a 1% change in σ₈ leads to a significant, but almost perfectly canceling, change in halo bias. This means that simply using IC residuals overshoots the target power spectrum, highlighting the need for a more sophisticated method like the transfer function.
  • Transfer Function as a Universal Tool: The paper demonstrates that a transfer function measured on one pair of phase-matched simulations can be effectively applied to another pair with different initial phases, significantly increasing the reusability of simulations.
  • Precision over Speed for Small Changes: The method prioritizes achieving high precision (0.1% accuracy) for small cosmological variations, rather than broad accuracy for large changes. This is crucial for modern precision cosmology.
  • Importance of Redshift-Space Warping: While real-space warping gets close, matching redshift-space statistics (monopole and quadrupole) requires dedicated velocity warping, including velocity dispersion matching and a velocity transfer function.

Key Results

  • The real-space power spectrum monopole can be matched to the target cosmology with 0.1% accuracy up to k = 1h Mpc⁻¹.
  • When the transfer function is applied to a new realization with different initial phases, the accuracy degrades slightly to 0.3%.
  • Similar accuracy is achieved for the redshift-space monopole and quadrupole.
  • In all cases, the achieved accuracy is better than the sample variance of the 1.1h⁻¹ Gpc simulation boxes.
  • The fitting process for the transfer function is computationally intensive but manageable, taking about 1.2-2.3 hours on a 24-core machine for a single measurement.

Limitations and Future Work

  • Halo Finder Dependence: The current work uses simple Friends-of-Friends (FoF) halos, which have known issues like over-merging. Future work will test the method with more sophisticated halo finders like Rockstar or spherical overdensity.
  • Parametrization of Transfer Function: The transfer function is currently parameterized in 10 discrete bins, which allows flexibility but can be computationally intensive. Future work will explore lower-dimensional parametrizations, which could accelerate convergence.
  • Complex Cosmology Changes: The current study focuses on a 1% change in σ₈. Future work will test the method's applicability to more complex cosmology changes that affect the power spectrum shape more significantly.
  • Interpolation of Transfer Functions: The authors suggest that interpolating transfer functions between cosmologies could further maximize the utility of simulations, allowing for even more efficient generation of mock catalogs.
  • Internal Halo Structure: The current method primarily modifies halo positions and velocities. Future extensions could incorporate changes to internal halo structure, such as concentration.

Summary updated

Abstract

ABSTRACT We present a method for generating suites of dark matter halo catalogues with only a few N-body simulations, focusing on making small changes to the underlying cosmology of a simulation with high precision. In the context of blind challenges, this allows us to re-use a simulation by giving it a new cosmology after the original cosmology is revealed. Starting with full N-body realizations of an original cosmology and a target cosmology, we fit a transfer function that displaces haloes in the original so that the galaxy/HOD power spectrum matches that of the target cosmology. This measured transfer function can then be applied to a new realization of the original cosmology to create a new realization of the target cosmology. For a 1 per cent change in σ8, we achieve 0.1 per cent accuracy to $k = 1\, h\, \mathrm{Mpc}^{-1}$ in the real-space power spectrum; this degrades to 0.3 per cent when the transfer function is applied to a new realization. We achieve similar accuracy in the redshift-space monopole and quadrupole. In all cases, the result is better than the sample variance of our $1.1\, h^{-1}\, \mathrm{Gpc}$ simulation boxes.
BibTeX
@inproceedings{garrison2019generating,
  author = {Lehman H. Garrison and Daniel J. Eisenstein},
  title = {{Generating approximate halo catalogues for blind challenges in precision cosmology}},
  booktitle = {Monthly Notices of the Royal Astronomical Society},
  year = {2019},
  doi = {10.1093/mnras/stz600},
}
Figure