Interloper bias in future large-scale structure surveys

TL;DR

Future large-scale structure surveys face significant challenges from interloper emission lines, which can bias power spectrum measurements and cosmological inferences if not properly identified and mitigated.

Contribution

This paper quantifies the impact of interloper lines on survey measurements and develops a formalism to predict and correct for their bias in cosmological parameters.

Findings

Interloper fractions above 0.2% can bias power spectrum by over 10%.

A formalism predicts that 0.15%-0.3% interloper fraction can bias growth rate measurements.

Secondary line identification can effectively remove interlopers for PFS and WFIRST.

Abstract

Next-generation spectroscopic surveys will map the large-scale structure of the observable universe, using emission line galaxies as tracers. While each survey will map the sky with a specific emission line, interloping emission lines can masquerade as the survey's intended emission line at different redshifts. Interloping lines from galaxies that are not removed can contaminate the power spectrum measurement, mixing correlations from various redshifts and diluting the true signal. We assess the potential for power spectrum contamination, finding that an interloper fraction worse than 0.2% could bias power spectrum measurements for future surveys by more than 10% of statistical errors, while also biasing power spectrum inferences. We also construct a formalism for predicting cosmological parameter bias, demonstrating that a 0.15%-0.3% interloper fraction could bias the growth rate by more than 10% of the error, which can affect constraints on gravity upcoming surveys. We use the COSMOS Mock Catalog (CMC), with the emission lines re-scaled to better reproduce recent data, to predict potential interloper fractions for the Prime Focus Spectrograph (PFS) and the Wide-Field InfraRed Survey Telescope (WFIRST). We find that secondary line identification, or confirming galaxy redshifts by finding correlated emission lines, can remove interlopers for PFS. For WFIRST, we use the CMC to predict that the 0.2% target can be reached for the WFIRST H$\alpha$ survey, but sensitive optical and near-infrared photometry will be required. For the WFIRST [OIII] survey, the predicted interloper fractions reach several percent and their effects will have to be estimated and removed statistically (e.g. with deep training samples). (Abridged)