Protocluster Discovery in Tomographic Ly$\alpha$ Forest Flux Maps

TL;DR

This paper introduces a new method for identifying protoclusters using tomographic Ly$ ext{a}$ forest flux maps, demonstrating high accuracy in simulations and robustness to varying sightline separations.

Contribution

The authors develop a computationally efficient technique to detect protoclusters in flux maps, validated with simulations and mock surveys, including large sightline separations.

Findings

Protoclusters create large flux decrements of about 10 $h^{-1}$Mpc.

The method achieves 90% purity and 75% recovery in noiseless maps.

It remains effective even with sightline separations up to 15 $h^{-1}$Mpc.

Abstract

We present a new method of finding protoclusters using tomographic maps of Ly$\alpha$ Forest flux. We review our method of creating tomographic flux maps and discuss our new high performance implementation, which makes large reconstructions computationally feasible. Using a large N-body simulation, we illustrate how protoclusters create large-scale flux decrements, roughly 10 $h^{-1}$Mpc across, and how we can use this signal to find them in flux maps. We test the performance of our protocluster finding method by running it on the ideal, noiseless map and tomographic reconstructions from mock surveys, and comparing to the halo catalog. Using the noiseless map, we find protocluster candidates with about 90% purity, and recover about 75% of the protoclusters that form massive clusters ($> 3 \times 10^{14} \, h^{-1} M_{\odot}$). We construct mock surveys similar to the ongoing COSMOS Lyman-Alpha Mapping And Tomography Observations (CLAMATO) survey. While the existing data has an average sightline separation of 2.3 $h^{-1}$Mpc, we test separations of 2 - 6 $h^{-1}$Mpc to see what can be tolerated for our application. Using reconstructed maps from small separation mock surveys, the protocluster candidate purity and completeness are very close what was found in the noiseless case. As the sightline separation increases, the purity and completeness decrease, although they remain much higher than we initially expected. We extended our test cases to mock surveys with an average separation of 15 $h^{-1}$Mpc, meant to reproduce high source density areas of the BOSS survey. We find that even with such a large sightline separation, the method can still be used to find some of the largest protoclusters.