Constraining Primordial Non-Gaussianity with Density-Split Clustering

TL;DR

This paper investigates how Density-Split Clustering (DSC) can enhance constraints on primordial non-Gaussianity (PNG) from large-scale structure surveys, demonstrating significant improvements over standard methods by capturing higher-order information.

Contribution

The study introduces a Fourier space analysis of DSC and compares its effectiveness with traditional halo power spectra, showing improved PNG constraints and exploring pipeline modifications.

Findings

Joint halo/DSC power spectra outperform halo power spectra in PNG constraints.

DSC captures higher-order information on small scales, improving sensitivity.

Applying DSC to halo fields does not enable sample variance cancellation on large scales.

Abstract

Obtaining tight constraints on primordial non-Gaussianity (PNG) is a key step in discriminating between different models for cosmic inflation. The constraining power from large-scale structure (LSS) measurements is expected to overtake that from cosmic microwave background (CMB) anisotropies with the next generation of galaxy surveys including the Dark Energy Spectroscopic Instrument (DESI) and Euclid. We consider whether Density-Split Clustering (DSC) can help improve PNG constraints from these surveys for local, equilateral and orthogonal types. DSC separates a surveyed volume into regions based on local density and measures the clustering statistics within each environment. Using the Quijote simulations and the Fisher information formalism, we compare PNG constraints from the standard halo power spectrum, DSC power spectra and joint halo/DSC power spectra. We find that the joint halo/DSC power spectra outperform the halo power spectrum by factors of $\sim$ 1.4, 8.8, and 3.6 for local, equilateral and orthogonal PNG, respectively. This is driven by the higher-order information that DSC captures on small scales. We find that applying DSC to a halo field does not allow sample variance cancellation on large scales by providing multiple tracers of the same volume with different local PNG responses. Additionally, we introduce a Fourier space analysis for DSC and study the impact of several modifications to the pipeline, such as varying the smoothing radius and the number of density environments and replacing random query positions with lattice points.