Principal Components of CMB non-Gaussianity

TL;DR

This paper develops a principal component analysis (PCA) framework for joint estimation of multiple non-Gaussianity signals in CMB data, accounting for secondary effects and biases, enhancing the analysis of primordial non-Gaussianity.

Contribution

It introduces a PCA-based method for joint estimation of various non-Gaussianity models in CMB data, including secondary effects, improving accuracy and computational efficiency.

Findings

PCA effectively constrains multiple non-Gaussianity amplitudes.

Bias from secondary non-Gaussianity can be quantified and mitigated.

Method validated with constraints on the DBI bispectrum.

Abstract

The skew-spectrum statistic introduced by Munshi & Heavens (2010) has recently been used in studies of non-Gaussianity from diverse cosmological data sets including the detection of primary and secondary non-Gaussianity of Cosmic Microwave Background (CMB) radiation. Extending previous work, focussed on independent estimation, here we deal with the question of joint estimation of multiple skew-spectra from the same or correlated data sets. We consider the optimum skew-spectra for various models of primordial non-Gaussianity as well as secondary bispectra that originate from the cross-correlation of secondaries and lensing of CMB: coupling of lensing with the Integrated Sachs-Wolfe (ISW) effect, coupling of lensing with thermal Sunyaev-Zeldovich (tSZ), as well as from unresolved point-sources (PS). For joint estimation of various types of non-Gaussianity, we use the PCA to construct the linear combinations of amplitudes of various models of non-Gaussianity, e.g. $f^{\rm loc}_{\rm NL},f^{\rm eq}_{\rm NL},f^{\rm ortho}_{\rm NL}$ that can be estimated from CMB maps. Bias induced in the estimation of primordial non-Gaussianity due to secondary non-Gaussianity is evaluated. The PCA approach allows one to infer approximate (but generally accurate) constraints using CMB data sets on any reasonably smooth model by use of a lookup table and performing a simple computation. This principle is validated by computing constraints on the DBI bispectrum using a PCA analysis of the standard templates.