An improved estimator for non-Gaussianity in cosmic microwave background observations

TL;DR

This paper introduces a generalized, full-sky estimator for non-Gaussianity in CMB data that reduces variance and improves detection sensitivity, especially when combined with foreground cleaning techniques.

Contribution

It extends previous flat-sky, thin-surface estimators to full-sky CMB analysis, incorporating the full radiation transfer function for better non-Gaussianity detection.

Findings

The improved estimator reduces variance compared to the standard estimator.

Foreground cleaning enhances the estimator's effectiveness.

Performance is degraded by late-time ISW effects but can be mitigated with foreground tracers.

Abstract

An improved estimator for the amplitude fnl of local-type non-Gaussianity from the cosmic microwave background (CMB) bispectrum is discussed. The standard estimator is constructed to be optimal in the zero-signal (i.e., Gaussian) limit. When applied to CMB maps which have a detectable level of non-Gaussianity the standard estimator is no longer optimal, possibly limiting the sensitivity of future observations to a non-Gaussian signal. Previous studies have proposed an improved estimator by using a realization-dependent normalization. Under the approximations of a flat sky and a vanishingly thin last-scattering surface, these studies showed that the variance of this improved estimator can be significantly smaller than the variance of the standard estimator when applied to non-Gaussian CMB maps. Here this technique is generalized to the full sky and to include the full radiation transfer function, yielding expressions for the improved estimator that can be directly applied to CMB maps. The ability of this estimator to reduce the variance as compared to the standard estimator in the face of a significant non-Gaussian signal is re-assessed using the full CMB transfer function. As a result of the late time integrated Sachs-Wolfe effect, the performance of the improved estimator is degraded. If CMB maps are first cleaned of the late-time ISW effect using a tracer of foreground structure, such as a galaxy survey or a measurement of CMB weak lensing, the new estimator does remove a majority of the excess variance, allowing a higher significance detection of fnl.