Planck 2015 results. X. Diffuse component separation: Foreground maps

TL;DR

Planck 2015 presents a unified Bayesian framework for diffuse foreground component separation by fitting a parametric sky model to Planck, WMAP, and Haslam data using the Commander code. The authors perform Gibbs sampling to map a high-dimensional posterior over sky components and instrument parameters, supplemented by nonlinear optimization to robustly locate maxima. The baseline model outputs full-sky maps for CMB, synchrotron, free-free, spinning dust (two components), CO, 94/100 GHz line emission, thermal dust, and SZ in temperature, and CMB, synchrotron, and thermal dust in polarization, with quantified residuals and extensive validation against external data. The work highlights degeneracies at low frequencies, calibration/bandpass uncertainties at high frequencies, and instrumental systematics in polarization, while delivering publicly available foreground products and demonstrating the power of joint, global Bayesian analysis for microwave sky studies.

Abstract

Planck has mapped the microwave sky in nine frequency bands between 30 and 857 GHz in temperature and seven bands between 30 and 353 GHz in polarization. In this paper we consider the problem of diffuse astrophysical component separation, and process these maps within a Bayesian framework to derive a consistent set of full-sky astrophysical component maps. For the temperature analysis, we combine the Planck observations with the 9-year WMAP sky maps and the Haslam et al. 408 MHz map to derive a joint model of CMB, synchrotron, free-free, spinning dust, CO, line emission in the 94 and 100 GHz channels, and thermal dust emission. Full-sky maps are provided with angular resolutions varying between 7.5 arcmin and 1 deg. Global parameters (monopoles, dipoles, relative calibration, and bandpass errors) are fitted jointly with the sky model, and best-fit values are tabulated. For polarization, the model includes CMB, synchrotron, and thermal dust emission. These models provide excellent fits to the observed data, with rms temperature residuals smaller than 4 uK over 93% of the sky for all Planck frequencies up to 353 GHz, and fractional errors smaller than 1% in the remaining 7% of the sky. The main limitations of the temperature model at the lower frequencies are degeneracies among the spinning dust, free-free, and synchrotron components; additional observations from external low-frequency experiments will be essential to break these. The main limitations of the temperature model at the higher frequencies are uncertainties in the 545 and 857 GHz calibration and zero-points. For polarization, the main outstanding issues are instrumental systematics in the 100-353 GHz bands on large angular scales in the form of temperature-to-polarization leakage, uncertainties in the analog-to-digital conversion, and very long time constant corrections, all of which are expected to improve in the near future.

Figures (4)

• Figure 1: Zodiacal light extrapolation from HFI to LFI and WMAP frequency channels in terms of full-sky mean brightness temperature. The dotted line shows the power-law fit to the HFI observations between 100 and 353GHz, $s(\nu) = 0.70\mu\textrm{K}_{\textrm{RJ}} (\nu/100\textrm{GHz})^{1.31}$, and the vertical grey lines indicate the central frequencies of the LFI and WMAP frequency bands. The intersection between the dotted and grey lines defines the extrapolation to low frequencies.
• Figure 2: $\chi^2$ (top) and residual maps, $\vec{d}_{\nu}-\vec{s}_{\nu}$ (bottom), for a Commander analysis that includes all Planck channel maps. These residual maps correspond to channels that are rejected from the baseline analysis due to instrumental systematics. No regularization noise has been added to the high-frequency channels in this case. The sharp ring-like features at high Galactic latitudes correspond to far sidelobe residuals; the broad features extending between the north and south ecliptic poles correspond to destriping errors; and the Galactic plane features correspond to calibration, bandpass, and modelling residuals.
• Figure 3: Processing masks (PM) used in the joint temperature analysis, including 99.6% and 61% of the sky, respectively. The former is used for calibration and bandpass estimation, and the latter for monopole and dipole estimation.
• Figure 4: Spectral energy densities (SEDs) for the main astrophysical components included in the present analysis, in brightness temperature. From left to right and top to bottom, panels show: (1) synchrotron emission; (2) free-free emission; (3) spinning dust emission; (4) CO line emission; (5) thermal dust emission; and (6) the thermal Sunyaev-Zeldovich effect. For each case, several parameter combinations are shown to illustrate their effect on the final observable spectrum. Vertical grey bands indicate the centre frequencies of the observations listed in Table \ref{['tab:data']}, but for clarity true bandwidths are suppressed. In each panel, the black dashed line shows the CMB brightness temperature corresponding to a thermodynamic temperature of $70\mu\textrm{K}$, the CMB rms at $1^$^∘$$ FWHM angular scale.