Optimising the sample selection for photometric galaxy surveys

TL;DR

This paper introduces an automated pipeline to optimize sample selection in photometric galaxy surveys, significantly improving the precision of cosmological parameter estimation in 3x2pt analyses.

Contribution

It presents a novel, flexible method combining self-organising maps and iterative redshift-bin edge exploration for survey configuration optimization.

Findings

Optimized sample selection can double the dark energy figure of merit.

The method converges quickly to optimal configurations across different bins and cosmologies.

Optimal selection enhances survey efficiency, equivalent to quadrupling survey area.

Abstract

Determining cosmological parameters with high precision, as well as resolving current tensions in their values derived from low and high redshift probes, is one of the main objectives of the new generation of cosmological surveys. The combination of complementary probes in terms of parameter degeneracies and systematics is key to achieving these ambitious scientific goals. In this context, determining the optimal survey configuration for an analysis that combines galaxy clustering, weak lensing, and galaxy-galaxy lensing, the so-called 3x2pt analysis, remains an open problem. In this paper, we present an efficient and flexible end-to-end pipeline to optimise the sample selection for 3x2pt analyses in an automated way. Our pipeline is articulated in two main steps: we first consider a self-organising map to determine the photometric redshifts of a simulated galaxy sample. As a proof of method for stage-IV surveys, we use samples from the DESC Data Challenge 2 catalogue. This allows us to classify galaxies into tomographic bins based on their colour phenotype clustering. We then explore different redshift-bin edge configurations for weak lensing only as well as 3x2pt analyses in a novel way. Our method explores multiple configurations of perturbed redshift-bin edges with respect to the fiducial case in an iterative manner. In particular, we sample tomographic configurations for the source and lens galaxies separately. We show that, using this method we quickly converge into an optimised configuration for different numbers of redshift bins and cosmologies. Our analysis demonstrates that for stage-IV surveys an optimal tomographic sample selection can increase the figure of merit of the dark energy (DE) equation of state by a factor of $\sim$2, comparable to an effective increase in survey area of $\sim$4 for non-optimal photometric survey analyses.