On foreground removal from the Wilkinson Microwave Anisotropy Probe data by an internal linear combination method:: Limitations and implications

We study the Internal Linear Combination (ILC) method presented by the Wilkinson Microwave Anisotropy Probe (WMAP) science team, with the goal of determining whether it may be used for cosmological purposes, as a template-free alternative to existing foreground-correction methods. We conclude that the method does have the potential to do just that, but great care must be taken both in implementation and in a detailed understanding of limitations caused by residual foregrounds, which can still affect cosmological results. As a first step we demonstrate how to compute the ILC weights both accurately and efficiently by means of Lagrange multipliers, and we apply this method to the observed data to produce a new version of the ILC map. This map has 12% lower variance than the ILC map of the WMAP team, primarily because of less noise. Next we describe how to generate Monte Carlo simulations of the ILC map and find that these agree well with the observed map on angular scales up to lapproximate to200, using a conservative sky cut. Finally we make two comments to the ongoing debates concerning the large-scale properties of the WMAP data. First, we note that the Galactic southeastern quadrant is associated with notably different ILC weights than the other three quadrants, possibly indicating a foreground-related anisotropy. Second, we study the properties of the quadrupole and octopole (amplitude, alignment, and planarity) and reproduce the previously reported results that the quadrupole and octopole are strongly aligned and that the octopole is moderately planar. Even more interestingly, we find that the l=5 mode is spherically symmetric at about 3 sigma, and that the l=6 mode is planar at the 2 sigma level. However, we also assess the impact of residual foregrounds on these statistics, and find that the ILC map is not clean enough to allow for cosmological conclusions. Alternative methods must be developed to study these issues further.