A simulation calibrated limit on the HI power spectrum from the GMRT Epoch of Reionization experiment

TL;DR

This paper presents a new foreground removal method using singular value decomposition for the GMRT EoR experiment, resulting in a more robust but larger upper limit on the HI power spectrum at high redshift.

Contribution

It introduces a singular value decomposition approach to foreground subtraction, quantifies signal loss, and provides revised, more robust upper limits on the HI power spectrum.

Findings

Revised upper limit of (248 mK)^2 at k=0.50 h/Mpc

Foreground removal with SVD reduces assumptions and improves robustness

Best current constraints on reionization at z=8.6

Abstract

The GMRT Epoch of Reionization (EoR) experiment is an ongoing effort to measure the power spectrum from neutral hydrogen at high redshift. We have previously reported an upper limit of (70 mK)^2 at wavenumbers of k=0.65 h/Mpc using a basic piecewise-linear foreground subtraction. In this paper we explore the use of a singular value decomposition to remove foregrounds with fewer assumptions about the foreground structure. Using this method we also quantify, for the first time, the signal loss due to the foreground filter and present new power spectra adjusted for this loss, providing a revised measurement of a 2-sigma upper limit at (248 mK)^2 for k=0.50 h/Mpc. While this revised limit is larger than previously reported, we believe it to be more robust and still represents the best current constraints on reionization at z=8.6.