Comparing Planck and WMAP: Maps, Spectra, and Parameters

Abstract

We examine the consistency of the 9 yr WMAP data and the first-release Planck data. We specifically compare sky maps, power spectra, and the inferred Λ cold dark matter (ΛCDM) cosmological parameters. Residual dipoles are seen in the WMAP and Planck sky map differences, but their amplitudes are consistent within the quoted uncertainties, and they are not large enough to explain the widely noted differences in angular power spectra at higher l. We remove the residual dipoles and use templates to remove residual Galactic foregrounds; after doing so, the residual difference maps exhibit a quadrupole and other large-scale systematic structure. We identify this structure as possibly originating from Planck’s beam sidelobe pick-up, but note that it appears to have insignificant cosmological impact. We develop an extension of the internal linear combination technique to find the minimum-variance difference between the WMAP and Planck sky maps; again we find features that plausibly originate in the Planck data. Lacking access to the Planck time-ordered data we cannot further assess these features. We examine ΛCDM model fits to the angular power spectra and conclude that the ∼2.5% difference in the spectra at multipoles greater than l∼ 100 is significant at the 3-5σ level, depending on how beam uncertainties are handled in the data. We revisit the analysis of WMAP’s beam data to address the power spectrum differences and conclude that previously derived uncertainties are robust and cannot explain the power spectrum differences. In fact, any remaining WMAP errors are most likely to exacerbate the difference. Finally, we examine the consistency of the ΛCDM parameters inferred from each data set taking into account the fact that both experiments observe the same sky, but cover different multipole ranges, apply different sky masks, and have different noise. We find that, while individual parameter values agree within the uncertainties, the six parameters taken together are discrepant at the ∼6σ level, with {χ }²}=56 for 6 degrees of freedom (probability to exceed, PTE = 3× {{10}^-10}). The nature of this discrepancy is explored: of the six parameters, {{χ }²} is best improved by marginalizing over {{{Ω}_c}{{h}²}, giving {χ }²}=5.2 for 5 degrees of freedom. As an exercise, we find that perturbing the WMAP window function by its dominant beam error profile has little effect on {{{Ω}_c}{{h}²}, while perturbing the Planck window function by its corresponding error profile has a much greater effect on {{Ω}_c}{{h}²}.