Estimates for the impact of ultraviolet background fluctuations on galaxy clustering measurements

• Published in: Monthly notices of the Royal Astronomical Society 2019-06, Vol.485 (4), p.5059-5072
• Main Authors: Upton Sanderbeck, Phoebe, Iršič, Vid, McQuinn, Matthew, Meiksin, Avery
• Format: Article
• Language: English
• ISSN: 0035-8711; 1365-2966
• DOI: 10.1093/mnras/stz741

ABSTRACT Spatial fluctuations in ultraviolet backgrounds can subtly modulate the distribution of extragalactic sources, a potential signal and systematic for large-scale structure surveys. While this modulation has been shown to be significant for 3D Ly α forest surveys, its relevance for other large-scale structure probes has been hardly explored, despite being the only astrophysical process that likely can affect clustering measurements on the scales of ≳Mpc. We estimate that the background fluctuations, modulating the amount of H i, have a fractional effect of (0.03–0.3) × (k/[10−2 Mpc−1])−1 on the power spectrum of 21 cm intensity maps at z = 1–3. We find a smaller effect for H α and Ly α intensity mapping surveys of (0.001–0.1) × (k/[10−2 Mpc−1])−1 and even smaller effect for more traditional surveys that correlate the positions of individual H α or Ly α emitters. We also estimate the effect of backgrounds on low-redshift galaxy surveys in general based on a simple model in which background fluctuations modulate the rate halo gas cools, modulating star formation: We estimate a maximum fractional effect on the power of ∼0.01 (k/[10−2 Mpc−1])−1 at z = 1. We compare sizes of these imprints to cosmological parameter benchmarks for the next generation of redshift surveys: We find that ionizing backgrounds could result in a bias on the squeezed triangle non-Gaussianity parameter fNL that can be larger than unity for power spectrum measurements with a SPHEREx-like galaxy survey, and typical values of intensity bias. Marginalizing over a shape of the form k−1PL, where PL is the linear matter power spectrum, removes much of this bias at the cost of ${\approx } 40{{\ \rm per\ cent}}$ larger statistical errors.