Inter-Scale Correlations as Measures of CMB Gaussianity

Abstract

A description of CMB temperature fluctuations beyond the power spectrum is important for verifying models of structure formation, especially in view of forthcoming high-resolution observations. We argue that higher-order statistics of inter-scale correlations, because of their low cosmic variance, may be effective in detecting non-Gaussian features in the CMB. Inter-scale correlations are generically produced in defect-based models of structure formation. We analytically study properties of general higher-order cumulants of Fourier components of homogeneous random fields and design a new set of statistics suitable for small-scale data analysis. Using simulated non-Gaussian fields, we investigate the performance of the proposed statistics in presence of a Gaussian background and pixel noise, using the bispectrum as the underlying cumulant. Our numerical results suggest that detection of non-Gaussian features by our method is reliable if the power spectrum of the non-Gaussian components dominates that of the Gaussian background and noise within at least a certain range of accessible scales.