Asymmetries in the CMB anisotropy field

TL;DR

The paper tests whether large-scale CMB power is isotropically distributed by applying power-spectrum and N-point statistic analyses to the WMAP first-year data and comparing results with Gaussian simulations. It finds a significant hemispherical asymmetry, with the northern hemisphere showing reduced power and the southern hemisphere enhanced, and an axis of maximum asymmetry near the ecliptic. This anisotropy is reinforced by N-point correlation results, which show stark differences between hemispheres and persist across masks and frequencies, with COBE-DMR data showing a similar pattern. The findings challenge the assumption of statistical isotropy and invite exploration of potential systematic, foreground, or new physics explanations, though foregrounds and known systematics appear unlikely culprits.

Abstract

We report on the results from two independent but complementary statistical analyses of the WMAP first-year data, based on the power spectrum and N-point correlation functions. We focus on large and intermediate scales (larger than about 3 degrees) and compare the observed data against Monte Carlo ensembles with WMAP-like properties. In both analyses, we measure the amplitudes of the large-scale fluctuations on opposing hemispheres and study the ratio of the two amplitudes. The power-spectrum analysis shows that this ratio for WMAP, as measured along the axis of maximum asymmetry, is high at the 95%-99% level (depending on the particular multipole range included). The axis of maximum asymmetry of the WMAP data is weakly dependent on the multipole range under consideration but tends to lie close to the ecliptic axis. In the N-point correlation function analysis we focus on the northern and southern hemispheres defined in ecliptic coordinates, and we find that the ratio of the large-scale fluctuation amplitudes is high at the 98%-99% level. Furthermore, the results are stable with respect to choice of Galactic cut and also with respect to frequency band. A similar asymmetry is found in the COBE-DMR map, and the axis of maximum asymmetry is close to the one found in the WMAP data.

Figures (3)

• Figure 1: Top panel: The low-resolution Q-, V- and W-band co-added WMAP map, to which the extended Kp0 mask has been applied. Middle panel: Results from the local power spectrum analysis. The color of the large discs indicate the ratio between the $\ell=2$--63 power spectrum bin of the northern and southern hemispheres, as determined in a reference frame where the north pole pierces the center of the disc; light red/yellow indicates a low ratio, dark red a high ratio. The medium sized dots indicate the absolute value of the power spectrum estimated on a $9\hbox{$.\!\!^\circ$}5$ disc in the $\ell=2$--63 bin (i.e., not on the full hemisphere), and a dark blue dot means that this value lies below the lower $80\%$ confidence limit, while a green dot means it lies above the upper $80\%$ limit. Finally, the ecliptic poles are marked by the small dark-blue spots. The figure should not be interpreted as implying that data close to the galactic plane have been used in the analysis: the WMAP Kp2 mask has been applied for all power spectrum computations. Bottom panel: The results from the intermediate scale three-point analysis. Blue corresponds to low $\chi^2$, which again corresponds to small fluctuations in the three-point function, while red corresponds to high $\chi^2$. The full Kp0 mask has been applied to the data.
• Figure 2: Power spectra computed from the co-added V- and W-band WMAP data. The solid line (histogram) indicates the theoretical best-fit WMAP running index power spectrum. The dashed line shows the estimated power spectrum obtained by the WMAP team for the Kp2 mask. The black crosses and gray dots represent our estimates of the power spectra on the northern and southern hemispheres, respectively. Here, north and south are defined with respect to the axis that maximizes the asymmetry in the WMAP data for the corresponding hemispheres, such that the north pole is located at $(\theta,\phi)=(80^\circ,57^\circ)$. The gray bands indicate the 1 and $2\sigma$ confidence regions, as computed from the ensemble of 2048 Monte Carlo simulations. Formally, these error bounds differ between the hemispheres, but in practice, the difference is small and only the values from the northern hemisphere are shown.
• Figure 3: Pseudo-collapsed three-point and 1+3 four-point correlation functions (dotted line) as computed from the low-resolution co-added Q-, V- and W-band WMAP data for the Kp0 Galactic mask after imposing an additional constraint that $|b|>30^{\circ}$. The gray bands indicate 1, 2, and $3\sigma$ confidence regions as computed from a set of 5000 Monte Carlo simulations. The solid line shows the median.