Bispectrum-window convolution via Hankel transform

Kevin Pardede, Federico Rizzo, Matteo Biagetti, Emanuele Castorina, E. Sefusatti, Pierluigi Monaco

Abstract

We present a method to perform the exact convolution of the model prediction for bispectrum multipoles in redshift space with the survey window function. We extend a widely applied method for the power spectrum convolution to the bispectrum, taking advantage of a 2D-FFTlog algorithm. As a preliminary test of its accuracy, we consider the toy model of a spherical window function in real space. This setup provides an analytical evaluation of the 3-point function of the window, and therefore it allows to isolate and quantify possible systematic errors of the method. We find that our implementation of the convolution in terms of a mixing matrix shows differences at the percent level in comparison to the measurements from a very large set of mock halo catalogs. It is also able to recover unbiased constraints on halo bias parameters in a likelihood analysis of a set of numerical simulations with a total volume of $100\, h^{-3} \, {\rm Gpc}^3$. For the level of accuracy required by these tests, the multiplication with the mixing matrix is performed in the time of one second or less.

Introduction

In modern cosmology, the Large-Scale Structure (LSS) of the universe serves as a primary laboratory for investigating dark energy, gravity, and the early universe. While the power spectrum (the 2-point correlation function) has traditionally been the primary tool for analysis, the bispectrum (the 3-point correlation function) provides critical additional information. It breaks degeneracies in galaxy bias parameters and offers insights into non-linear gravitational evolution and primordial non-Gaussianity.

A significant challenge in bispectrum analysis is the "window function" effect. Observations are constrained by the finite footprint and selection depth of galaxy surveys, which distorts the underlying physical signal. To compare theoretical models with observational data, one must convolve the model with the survey's window function. For the bispectrum, this involves a computationally expensive 6D integral. Standard numerical integration methods are too slow for Likelihood analyses (e.g., Markov Chain Monte Carlo), where the model must be recalculated thousands of times. This work introduces a high-performance method to perform this convolution exactly and efficiently.

Methodology: 2D-FFTLog and Hankel Transforms

The proposed method circumvents the 6D integral by transforming the problem into configuration space. In this space, the convolution becomes a simple multiplication with the window's 3-point correlation function.

Hankel Transforms and 2D-FFTLog

The core of the algorithm utilizes Hankel transforms. While FFTLog is a standard tool for 1D transforms in power spectrum analysis, the bispectrum requires handling two independent wavevectors ($k_1, k_2$) that define the triangle configuration. The authors implement a 2D-FFTLog algorithm to perform these transforms on logarithmically spaced grids efficiently.

Coordinate Systems and the Mixing Matrix

A key technical hurdle is the consistent definition of coordinate systems. The authors reconcile different conventions to ensure azimuthal angles are correctly handled during the transformation process.

Coordinate systems for bispectrum wavevector geometry.

The final output of this process is a mixing matrix. This matrix encodes the survey geometry and is pre-computed once for a given survey. Subsequent convolutions for any theoretical model are reduced to a simple matrix-vector multiplication, which executes in less than one second.

Validation and Results

The method was validated using several datasets:

  • Spherical Toy Model: A perfectly spherical window function allowed for an analytical comparison, isolating algorithmic systematic errors.
  • Pinocchio Mocks: 10,000 fast mock catalogs provided a massive volume to suppress statistical noise, confirming the method's accuracy at the percent level.
  • Minerva Simulations: A set of 298 N-body simulations ($100\, h^{-3} \, {\rm Gpc}^3$ total volume) was used in a full MCMC likelihood analysis. The method successfully recovered unbiased halo bias parameters.

The 2D-FFTLog approach significantly outperforms the common "1D approximation," which can exhibit 6-7% errors on large scales. In contrast, the new method maintains accuracy within ~1%, particularly in "squeezed" triangle configurations that are sensitive to primordial physics.

Discussion and Future Directions

The "mixing matrix" paradigm makes the bispectrum computationally "cheap" to evaluate, comparable to the power spectrum. This is a vital step toward including the bispectrum in the standard analysis pipelines of upcoming surveys like Euclid or DESI.

While the current work demonstrates success with spherical windows and real-space simulations, future work will focus on extending the method to handle realistic survey masks with holes and irregular boundaries. Additionally, further numerical testing of higher-order multipoles in redshift space and moving beyond the plane-parallel approximation for very wide-area surveys are necessary next steps.

In conclusion, the 2D-FFTLog method provides a robust and extremely fast solution for the bispectrum window convolution, enabling more precise cosmological constraints from the next generation of LSS data.

Summary updated

Abstract

We present a method to perform the exact convolution of the model prediction for bispectrum multipoles in redshift space with the survey window function. We extend a widely applied method for the power spectrum convolution to the bispectrum, taking advantage of a 2D-FFTlog algorithm. As a preliminary test of its accuracy, we consider the toy model of a spherical window function in real space. This setup provides an analytical evaluation of the 3-point function of the window, and therefore it allows to isolate and quantify possible systematic errors of the method. We find that our implementation of the convolution in terms of a mixing matrix shows differences at the percent level in comparison to the measurements from a very large set of mock halo catalogs. It is also able to recover unbiased constraints on halo bias parameters in a likelihood analysis of a set of numerical simulations with a total volume of $100\, h^{-3} \, {\rm Gpc}^3$. For the level of accuracy required by these tests, the multiplication with the mixing matrix is performed in the time of one second or less.
BibTeX
@article{pardede2022bispectrum,
  author = {Kevin Pardede and Federico Rizzo and Matteo Biagetti and Emanuele Castorina and E. Sefusatti and Pierluigi Monaco},
  title = {{Bispectrum-window convolution via Hankel transform}},
  journal = {Journal of Cosmology and Astroparticle Physics},
  year = {2022},
  doi = {10.1088/1475-7516/2022/10/066},
}
Figure