CMB multipole measurements in the presence of foregrounds

A. de Oliveira‐Costa, Max TegmarkPhysical 2006 · DOI · arXiv

Abstract

Most analysis of cosmic microwave background spherical harmonic coefficients ${a}_{\ell m}$ has focused on estimating the power spectrum ${C}_{\ell}=⟨|{a}_{\ell m}{|}^{2}⟩$ rather than the coefficients themselves. We present a minimum-variance method for measuring ${a}_{\ell m}$ given anisotropic noise, incomplete sky coverage and foreground contamination, and apply it to the Wilkinson Microwave Anisotropy Probe (WMAP) data. Our method is shown to constitute lossless data compression in the sense that the widely used quadratic estimators of the power spectrum ${C}_{\ell}$ can be computed directly from our ${a}_{\ell m}$-estimators. As the Galactic cut is increased, the error bars $\Delta{a}_{\ell m}$ on low multipoles go from being dominated by foregrounds to being dominated by leakage from other multipoles, with the intervening minimum defining the optimal cut. Applying our method to the WMAP quadrupole and octopole as an illustration, we investigate the robustness of the previously reported “axis of evil” alignment to Galactic cut and foreground contamination.

Introduction

In modern cosmology, the Cosmic Microwave Background (CMB) is typically characterized by its power spectrum ($C_\ell$), which quantifies the variance of fluctuations at different angular scales. While the power spectrum is a powerful tool for constraining cosmological parameters, it discards spatial information contained within the individual spherical harmonic coefficients ($a_{\ell m}$). These coefficients are essential for investigating large-scale "anomalies," such as the unexpectedly low power in the quadrupole and the peculiar alignment between the quadrupole and octopole—a phenomenon often referred to as the "axis of evil."

The primary obstacle in measuring these coefficients is the presence of the Milky Way. Galactic foreground emissions dominate the signal in the equatorial plane, necessitating the use of masks. However, masking the sky destroys the orthogonality of spherical harmonics, leading to "leakage" where signals from different multipoles become coupled. This paper presents a framework to navigate the tradeoff between foreground contamination and mask-induced leakage.

WMAP sky maps illustrating the challenge of Galactic foreground contamination and the resulting cleaned CMB maps.

Methodology

The authors develop a minimum-variance linear estimator to extract $a_{\ell m}$ values from imperfect data. By modeling the observed sky as $x = Ya + n$ (where $Y$ is the harmonic matrix and $n$ represents noise and unmodeled signals), they derive a weighting matrix $W$ that yields unbiased estimates $\hat{a} = Wx$ with minimal variance.

Crucially, this method is shown to be "lossless." Standard optimal quadratic estimators for the power spectrum $C_\ell$ can be derived directly from these estimated $a_{\ell m}$ coefficients, ensuring that no cosmological information is discarded during the compression from raw maps to multipoles. The analysis utilizes the 1-year and 3-year data releases from the Wilkinson Microwave Anisotropy Probe (WMAP), specifically focusing on the foreground-cleaned "TOH" map.

Results and Analysis

A key contribution of this work is the optimization of the "Galactic cut." By performing simulations with mock CMB and foreground maps, the authors identify a "sweet spot" for masking. If the mask is too small, residual Galactic foregrounds contaminate the measurement; if it is too large, the coupling between multipoles causes error bars to explode. For low-order multipoles ($\ell \le 5$), the optimal mask typically removes between 2% and 7% of the sky.

Stability of quadrupole (left) and octopole (right) coefficients across different Galactic masks, showing the growth of error bars due to multipole leakage at larger masks.

The analysis confirms that the $a_{20}$ coefficient—the quadrupole component aligned with the Galactic poles—is the most difficult to constrain because its spatial pattern closely resembles the Galactic foreground itself. Despite these challenges, the reported alignment of the quadrupole and octopole axes remains robust across different masks and data releases. In the WMAP 3-year data, the axes are separated by only $11.5^\circ$. The probability of such an alignment occurring by chance in the standard $\Lambda$CDM model is approximately 1-in-50, suggesting that these anomalies are likely physical features of the data rather than artifacts of foreground removal.

Discussion and Conclusion

While the minimum-variance estimator provides a rigorous way to quantify $a_{\ell m}$ and their associated errors, it faces computational challenges. The requirement to invert large covariance matrices limits its application to relatively low-resolution maps. Furthermore, even with optimal masking, small residual foregrounds (1–3 $\mu K$) persist.

Future missions, such as the Planck satellite, are expected to provide the higher sensitivity and broader frequency coverage needed to further refine these measurements. By providing a robust method for measuring individual multipoles, this work lays the groundwork for more detailed investigations into the large-scale structure of our universe and the potential physics behind observed cosmic anomalies.

Summary updated

Abstract

Most analysis of cosmic microwave background spherical harmonic coefficients ${a}_{\ensuremath{\ell}m}$ has focused on estimating the power spectrum ${C}_{\ensuremath{\ell}}=⟨|{a}_{\ensuremath{\ell}m}{|}^{2}⟩$ rather than the coefficients themselves. We present a minimum-variance method for measuring ${a}_{\ensuremath{\ell}m}$ given anisotropic noise, incomplete sky coverage and foreground contamination, and apply it to the Wilkinson Microwave Anisotropy Probe (WMAP) data. Our method is shown to constitute lossless data compression in the sense that the widely used quadratic estimators of the power spectrum ${C}_{\ensuremath{\ell}}$ can be computed directly from our ${a}_{\ensuremath{\ell}m}$-estimators. As the Galactic cut is increased, the error bars $\ensuremath{\Delta}{a}_{\ensuremath{\ell}m}$ on low multipoles go from being dominated by foregrounds to being dominated by leakage from other multipoles, with the intervening minimum defining the optimal cut. Applying our method to the WMAP quadrupole and octopole as an illustration, we investigate the robustness of the previously reported ``axis of evil'' alignment to Galactic cut and foreground contamination.
BibTeX
@inproceedings{oliveiracosta2006cmb,
  author = {A. de Oliveira‐Costa and Max Tegmark},
  title = {{CMB multipole measurements in the presence of foregrounds}},
  booktitle = {Physical},
  year = {2006},
  doi = {10.1103/physrevd.74.023005},
}
Figure