The deconvolved distribution estimator: enhancing reionisation-era CO line-intensity mapping analyses with a cross-correlation analogue for one-point statistics

TL;DR

The paper introduces the deconvolved distribution estimator (DDE), a novel statistical method that enhances CO line-intensity mapping analyses by suppressing interloper noise and improving signal detection through a Fourier-space deconvolution technique.

Contribution

The DDE extends the voxel intensity distribution (VID) by deconvolving joint and individual VIDs to reduce interloper bias and noise, offering improved sensitivity for future CO mapping observations.

Findings

DDE significantly improves constraining power over traditional methods.

Fisher forecasts show DDE matches combined one- and two-point statistics.

DDE effectively suppresses uncorrelated noise and interloper biases.

Abstract

We present the deconvolved distribution estimator (DDE), an extension of the voxel intensity distribution (VID), in the context of future observations proposed as part of the CO Mapping Array Project (COMAP). The DDE exploits the fact that the observed VID is a convolution of correlated signal intensity distributions and uncorrelated noise or interloper intensity distributions. By deconvolving the individual VID of two observables away from their joint VID in a Fourier-space operation, the DDE suppresses sensitivity to interloper emission while maintaining sensitivity to correlated components. The DDE thus improves upon the VID by reducing the relative influence of uncorrelated noise and interloper biases, which is useful in the context of COMAP observations that observe different rotational transitions of CO from the same comoving volume in different observing frequency bands. Fisher forecasts suggest that the theoretical sensitivity in the DDE allows significant improvements in constraining power compared to either the cross power spectrum or the individual VID data, and matches the constraining power of the combination of all other one- and two-point summary statistics. Future work should further investigate the covariance and model-dependent behaviour of this novel one-point cross-correlation statistic.