A Systematics-Robust 21 cm Foreground Removal Algorithm

The 21 cm line of neutral hydrogen is a promising probe of large-scale structure of the universe. However, cosmological 21 cm signal is extremely challenging to observe due to bright astrophysical foregrounds. Using the smooth spectral character of foregrounds, traditional linear filters can optimally remove foregrounds so long as the telescope response is accurately known. However, such methods are sensitive to telescope systematics such as antenna gain errors, leaving foreground residuals that dominate the signal. I will present a systematics-robust foreground removal technique that combines a traditional linear filter with a nonlinear method. We first roughly separate the signal and foregrounds using a linear filter. Then, by cross-correlating the residual-contaminated signal channel with the foregrounds, we estimate the systematics-induced contamination, which is subsequently subtracted. In simulations, we find the method suppresses contamination up to first order in systematic errors, making the experiment two orders of magnitude more robust to these errors. Our results suggest that the method can ease the requirement on telescope characterization for 21 cm cosmology experiments such as the Canadian Hydrogen Intensity Mapping Experiment (CHIME).