Power-spectrum Space Decomposition of Frequency Tomographic Data for Intensity-mapping Experiments

Abstract We present a Bayesian framework to establish a power-spectrum space decomposition of frequency tomographic (PSDFT) data for future intensity-mapping (IM) experiments. In contrast to most traditional component separation methods that work in the map domain, this new technique treats multifrequency power spectra as raw data and can reconstruct component power spectra by taking advantage of distinct component correlation patterns in the frequency domain. We validated this new technique for future IM experiments using synthesized mock data that contain bright foreground contaminants, IM signals, and instrumental effects at different frequencies. The PSDFT approach can effectively remove the bright foreground contamination and extract the targeted IM signals using a Bayesian approach in a power-spectrum subspace. This new approach can be directly applied to a broad range of IM analyses and will be well suited to future high-quality IM data sets, providing a powerful tool for future IM surveys.