# 2018 CMB maps

## Overview

This section describes the CMB maps produced from the Planck data. These products are derived from some or all of the nine frequency channel maps using different techniques and, in some cases, using other constraints from external data sets. Here we give a brief description of each product and how it is obtained, followed by a description of the FITS file containing the data and associated information. All the details can be found in Planck-2020-A4[1] and, for earlier releases, in Planck-2013-XII[2] and Planck-2015-A09[3].

## 2018 CMB maps

CMB maps have been produced using four different methods: COMMANDER, NILC, SEVEM, and SMICA, as described in the CMB and foreground separation section and also in Appendices A-D of Planck-2020-A4[1] and references therein.

For each method we provide the following:

• Full-mission CMB intensity map, with corresponding confidence mask and effective beam transfer function.
• Full-mission CMB polarisation map, with corresponding confidence mask and effective beam transfer function.
• In-painted CMB intensity and polarisation maps, intended for PR purposes.

In addition, and for characterisation purposes, we include four other sets of maps from two data splits: odd/even ring and first/second half-mission. Half-difference maps can be used to provide an approximate noise estimates for the full mission, but they should be used with caution. Each split has caveats in this regard: there are noise correlations between the odd/even split maps, and unobserved pixels in both splits. Masks flagging unobserved pixels are provided for each split, and we strongly encourage use of these when analysing split maps.

In addition, for SMICA, we also provide a CMB map from which Sunyaev-Zeldovich (SZ) sources have been projected out, while SEVEM provides cleaned single-frequency maps at 70, 100, 143 and 217 GHz for both intensity and polarization.

All CMB products are provided at an angular resolution of 5 arcmin FWHM, and HEALPix resolution Nside=2048, with the exception of the SEVEM cleaned single-frequency maps which are provided at their native resolution, and in units of Kcmb.

For a complete description of the above data structures, see below; the content of the first extensions is illustrated and commented in the table below.

The gallery below shows the inpainted full-mission CMB maps (T, Q and U) from each pipeline. The temperature maps are shown at 5 arcmin FWHM resolution, while the polarization maps are shown at 80 arcmin FWHM resolution, in order to suppress instrumental noise.

### Product description

#### COMMANDER

Principle
COMMANDER is a Planck software code implementing Bayesian parametric component separation. Each astrophysical signal component is modelled in terms of a small number of free parameters per pixel, typically in terms of an amplitude at a given reference frequency and a small set of spectral parameters, and these are fitted to the data with an MCMC Gibbs sampling algorithm. A new feature in the Planck 2018 analysis is support for multi-resolution analysis, allowing reconstruction of both CMB and foreground maps at full angular resolution. Only CMB products are provided from Commander in the Planck 2018 release (see Planck-2020-A4[1] for details), while for polarization both CMB and foreground products are provided. For temperature, a dedicated low-resolution CMB map is also provided as part of the Planck likelihood package.
Resolution (effective beam)
The Commander sky maps have different angular resolutions depending on data products:
• CMB temperature and polarization and thermal dust polarization maps are provided at 5 arcmin FWHM resolution
• Synchrotron polarization maps are provided at 40 arcmin FWHM resolution
• The low-resolution CMB likelihood map is provided at an angular resolution of 40 arcmin FWHM.
The Commander temperature confidence mask is produced by thresholding the chi-square map characterizing the global fits, combined with direct CO amplitude thresholding to eliminate known leakage effects. In addition, we exclude all pixels brighter than 10mK in the 30GHz map, in order to remove particularly bright radio sources. Finally, we remove by hand the Virgo and Coma clusters, as well as the Crab nebula. A total of 88% of the sky is admitted for analysis.
The Commander polarization mask is produced in a similar manner, starting by thresholding the chi-squared map. In addition, we exclude all pixels for which the thermal dust polarization amplitude is brighter than 20µKRJ at 353GHz, as well as particularly bright objects in the PCCS2 source catalog. Finally, we remove a small region that is particularly contaminated by cosmic ray glitches. A total of 86% of the sky is admitted for analysis.

#### NILC

Principle
Needlet Internal Linear Combination (or NILC in short) is a blind component separation method for the measurement of Cosmic Microwave Background (CMB) from the multi-frequency observations of sky. It is an implementation of an Internal Linear Combination (ILC) of the frequency channels under consideration with minimum error variance on a frame of spherical wavelets called needlets, allowing localized filtering in both pixel space and harmonic space. The method includes multipoles up to 4000. Temperature and, E-mode and B-mode of polarization maps are produced independently. The Q and U maps of CMB polarization have been reconstructed from the corresponding E-mode and B-mode maps.
Resolution (effective beam)
The effective beam is equivalent to a Gaussian circular beam with FWHM=5 arcminutes.
For each needlet scale, we identify the frequency channel that contributes the most to the final reconstruction of CMB for that band. Then we scale the sky maps for 30GHz and 353GHz to that frequency channel to obtain the scaled-sky map and compute the root mean square (RMS) of full mission CMB map. The final mask is reconstructed from the union of all the masks obtained by setting a cut-off on the RMS value at different needlet scales.

#### SEVEM

Principle
SEVEM produces cleaned CMB maps at several frequencies by using a procedure based on template fitting in real space. The templates are typically constructed as the difference between two close Planck frequency maps and then subtracted from the CMB-dominated channels, with coefficients that are chosen to minimize the variance of the cleaned map outside a considered mask. In the cleaning process, no assumptions about the foregrounds or noise levels are needed, rendering the technique very robust. A subset of the cleaned single frequency maps are then combined to obtain the final CMB map.
Resolution
The cleaned CMB maps for intensity and polarization are constructed at Nside=2048 and at the standard resolution of 5 arcminutes (Gaussian beam). The maximum considered multipole is for intensity and for polarization.
The confidence masks cover the most contaminated regions of the sky, leaving approximately 84 per cent of useful sky for intensity, and 80 per cent for polarization.
The point source masks contain the holes corresponding to the point sources detected at each raw Planck frequency channel in intensity and polarization. The number of sources detected are given in the upper part of Table C.1 of Planck-2020-A4[1]. There is one mask for intensity and another one for polarization per frequency channel. When using the Planck channels in the construction of the templates, these have been inpainted in the positions of the point sources given in these masks, to reduce the emission from this contaminant in the templates and its propagation to the final cleaned CMB maps.
The inpainting masks include the positions of the point sources that have been inpainted in the cleaned single-frequency maps. They contain point sources detected at the original raw data at those frequencies plus the sources detected in the cleaned frequency maps (see Table C.1 of Planck-2020-A4[1]). There is a mask for intensity and another one for polarization for each of the cleaned frequency maps (70, 100, 143 and 217 GHz) as well as the corresponding masks for the combined map. The latter are constructed as the product of the individual frequency masks of those cleaned channels that are combined in the final CMB map (i.e., the product of 143 and 217 GHz masks for intensity and of 100, 143 and 217 GHz for polarization). Note that the inpainted positions are not excluded by default by the SEVEM confidence mask, but only if they are considered unreliable with the general procedure used to construct the SEVEM confidence mask.

##### Foregrounds-subtracted maps

In addition to the regular CMB maps, SEVEM provides maps cleaned of the foregrounds for selected frequency channels (categorized as fgsub-sevem in the archive). In particular, for both intensity and polarization there are cleaned CMB maps available at 70, 100, 143 and 217 GHz, provided at the original resolution and Nside of the uncleaned channel (1024 for 70 GHz and 2048 for the rest of the maps).

#### SMICA

Principle
SMICA produces CMB maps by linearly combining Planck input channels with multipole-dependent weights, including multipoles up to . Temperature and polarization maps are produced independently. In temperature, two distinct CMB renderings are produced and then merged (hybridized) together into a single CMB intensity map. In polarization, the E and B modes are processed independently and the results are combined to produce Q and U maps.
Resolution (effective beam)
The SMICA intensity map has an effective beam window function of 5 arc-minutes which is truncated at .
The SMICA Q and U maps are obtained similarly but are produced at =1024 with an effective beam of 10 arc-minutes.
A confidence mask is provided which excludes some parts of the Galactic plane, some very bright areas and masked point sources. This mask provides a qualitative (and subjective) indication of the cleanliness of a pixel.

Common masks have been defined for analysis of the CMB temperature and polarization maps. In previous releases, these were constructed simply as the union of the individual pipeline confidence masks. In the 2018 release, a more direct approach has been adopted, by thresholding the standard deviation map evaluated between each of the four cleaned CMB maps. This standard deviation mask is then augmented with the Commander and SEVEM confidence masks, as well as with the SEVEM and SMICA in-painting masks.

In addition, we provide masks for unobserved pixels for the half-mission and odd-even data splits, as well as an in-painting mask. The latter is not intended for scientific analysis, but for producing visually acceptable CMB representation for PR purposes.

In total, we provide the following masks:

• COM_Mask_CMB-HM-Misspix-Mask-Int_2048_R3.00.fits -- Temperature half-mission missing pixels mask with fsky = 96.0%. This should be applied in analyses of the half-mission split temperature maps.
• COM_Mask_CMB-HM-Misspix-Mask-Pol_2048_R3.00.fits -- Polarization half-mission missing pixels mask with fsky = 96.1%. This should be applied in analyses of the half-mission split polarization maps.
• COM_Mask_CMB-HM-Misspix-Mask-Int_2048_R3.00.fits -- Temperature half-mission missing pixels mask with fsky = 98.1%. This should be applied in analyses of the half-mission split temperature maps.
• COM_Mask_CMB-HM-Misspix-Mask-Pol_2048_R3.00.fits -- Polarization half-mission missing pixels mask with fsky = 98.1%. This should be applied in analyses of the half-mission split polarization maps.

#### CMB-subtracted frequency maps ("Foreground maps")

These are the full-sky, full-mission frequency maps in intensity from which the CMB has been subtracted. The maps contain foregrounds and noise. They are provided for each frequency channel and for each component separation method. They are grouped into 8 files, two for each method of which there is one for each instrument. The maps are are at Nside = 1024 for the three LFI channels and at Nside = 2048 for the six HFI channels. The filenames are:

• LFI_Foregrounds-{method}_1024_Rn.nn.fits (145 MB each)
• HFI_Foregrounds-{method}_2048_Rn.nn.fits (1.2 GB each)

To remove the CMB, the respective CMB map was first deconvolved with the 5 arcmin beam, then convolved with the beam of the frequency channel, and finally subtracted from the frequency map. This was done using the in harmonic space, assuming a symmetric beam.

The CMB-subtracted maps have complicated noise properties. The CMB maps contain a noise contribution from each of the frequency maps, depending on the weights with which they were combined. Therefore subtracting the CMB map from a frequency channel contributes additional noise from the other frequency channels. This caveat is particularly important for polarization, for which the noise in the cleaned CMB maps is large. After subtraction this noise term is perfectly correlated between frequency channels, with a perfect blackbody spectrum with T=2.7255K. Caution is therefore warranted when using these maps for scientific analysis.

The frequency maps from which the CMB have been subtracted are:

• LFI_SkyMap_0nn_1024_R3.00_full.fits
• HFI_SkyMap_nnn_2048_R3.00_full.fits

Note that the zodiacal light correction described here was applied to the HFI temperature maps before the CMB subtraction.

### Production process

#### COMMANDER

Pre-processing and data selection
The primary Commander 2018 analysis is carried out at full angular resolution, and no smoothing to a common resolution is applied to the maps, in constrast to the procedure employed in previous releases. The temperature analysis employs all nine Planck frequency maps between 30 and 857 GHz, while the polarization analysis employs the seven frequency maps between 30 and 353 GHz. No external data are used in the 2018 Commander analysis.
Priors
The following priors are enforced in the Commander analysis:
• The 30 GHz zero-level is fixed to zero, while the 44 and 70 GHz zero-levels are fitted freely with uniform priors. HFI zero-levels are fitted with a strong CIB prior.
• Dipoles are fitted only at 70 and 100 GHz; all other are fixed to zero.
• Gaussian priors are enforced on spectral parameters, with values informed by the values derived in the high signal-to-noise areas of the sky
• The Jeffreys ignorance prior is enforced on spectral parameters in addition to the informative Gaussian priors
Fitting procedure
Given data and priors, Commander either maximizes, or samples from, the Bayesian posterior, P(theta|data). Because this is a highly non-Gaussian and correlated distribution, involving millions of parameters, these operations are performed by means of the Gibbs sampling algorithm, in which joint samples from the full distributions are generated by iteratively sampling from the corresponding conditional posterior distributions, P(theta_i| data, theta_{j/=i}). All parameters are optimized jointly.

#### NILC

Pre-processing
All sky maps are convolved/deconvolved in harmonic space, to a common beam resolution corresponding to a Gaussian beam of 5 arc-minutes FWHM. A very small preprocessing mask has been used on the temperature sky maps. Prior to implement the pipeline on the sky maps, the masked regions are filled using PSM tools which uses an increasing number of neighboring pixels to fill regions deeper in the hole. At each iteration it uses pixels at up to twice the diameter of the pixel times number of iteration. No preprocessing has been done on polarization sky maps.
Linear combination
Needlet ILC weights are computed for each of T, E and B, for each scale and for each pixel of the needlet representation at that scale. For each of T, E and B, a full-sky CMB map, at 5 arc-minutes beam resolution, is synthesized from the NILC needlet coefficients.
Post-processing
E and B maps are re-combined into Q and U products using standard HEALPix tools.

#### SEVEM

The templates used in the SEVEM pipeline are typically constructed by subtracting two close Planck frequency channel maps, after first smoothing them to a common resolution to ensure that the CMB signal is properly removed. A linear combination of the templates is then subtracted from (hitherto unused) map d to produce a cleaned CMB map at that frequency. This is done in real space at each position on the sky: where is the number of templates. The coefficients are obtained by minimising the variance of the cleaned map outside a given mask. Note that the same expression applies for I, Q and U. Although we exclude very contaminated regions during the minimization, the subtraction is performed for all pixels and, therefore, the cleaned maps cover the full-sky (although foreground residuals are expected to be particularly large in those areas excluded by the minimisation).

There are several possible configurations of SEVEM with regard to the number of frequency maps which are cleaned or the number of templates that are used in the fitting. Note that the production of cleaned maps at different frequencies is of great interest by itself in order to test the robustness of the results, and these intermediate products (cleaned maps at individual frequencies for intensity and polarization) are also provided in the archive. Therefore, to define the best strategy, one needs to find a compromise between the number of maps that can be cleaned and the number of templates that can be constructed.

Intensity

For the CMB intensity case, we have cleaned the 70, 100, 143 and 217 GHz maps using a total of five templates. In particular, three templates constructed as the difference of two consecutive Planck channels smoothed to a common resolution [30GHz – 44GHz], [44GHz – 70GHz] and [543GHz – 535GHz] as well as a fourth template given by the 857 GHz channel are used to clean the 100, 143 and 217 GHz maps. Before constructing the templates, the six frequency channels involved in the templates are inpainted at the corresponding point source positions detected at each frequency using the Mexican Hat Wavelet algorithm (these positions are given in the provided point sources masks). The size of the holes to be inpainted is determined taking into account the beam size of the channel as well as the flux of each source. The inpainting algorithm is based on simple diffuse inpainting, which fills one pixel with the mean value of the neighbouring pixels in an iterative way. To avoid inconsistencies when subtracting two channels, each frequency map is inpainted on the sources detected in that map and on the second map (if any) used to construct the template. Then the maps are smoothed to a common resolution, by convolving the first map with the beam of the second one and viceversa. For the fourth template, we simply filter the inpainted 857 GHz map with the 545 GHz beam. The cleaned 70 GHz map is produced similarly by considering two templates, the [30GHz – 44GHz] map and a second template obtained as [353GHz – 143GHz] constructed at the original resolution of the 70 GHz map.

The coefficients to clean the frequency maps are obtained by minimising the variance outside the analysis mask, that covers the 1 per cent brightest emission of the sky as well as point sources detected at all frequency channels. Once the maps are cleaned, each of them is inpainted on the point sources positions detected at that (raw) channel. Then, the Mexican Hat Wavelet algorithm is run again, now on the cleaned maps. A number of new sources are found and are also inpainted at each channel. The resolution of the cleaned map is the same as that of the original data. Our final CMB map is then constructed by combining the 143 and 217 GHz maps by weighting the maps in harmonic space taking into account the noise level, the resolution and a rough estimation of the foreground residuals of each map (obtained from realistic simulations). This final map has a resolution corresponding to a Gaussian beam of fwhm=5 arcminutes and Nside=2048 and the maximum considered multipole is . The monopole and dipole over the full-sky have been subtracted from the final CMB map.

In addition, the cleaned CMB maps produced at 70, 100, 143 and 217 GHz frequencies are also provided. The resolution of these maps is the same as that of the uncleaned frequency channels and have been constructed at Nside=1024 for 70 GHz and Nside=2048 for the rest of the maps. They have been inpainted at the position of the point sources detected in the raw and cleaned maps (these positions are given in the corresponding inpainting masks). The monopole and dipole over the full-sky have also been removed from each of the cleaned maps.

The confidence mask is produced by studying the differences between several SEVEM CMB reconstructions, which correspond to maps cleaned at different frequencies or using different analysis masks. The obtained mask leaves a useful sky fraction of approximately 84 per cent.

Polarization

To clean the polarization maps, a procedure similar to the one used for intensity data is applied to the Q and U maps independently. Cleaned maps at 70, 100, 143 and 217 GHz are also produced but, given that a smaller number of frequency channels is available for polarization, the templates selected to clean the maps are different. In particular, we clean the 70 GHz map using two templates and the rest of the channels using different combinations of three templates.

Following the same procedure as for the intensity case, those channels involved in the construction of the templates are inpainted in the position of the sources detected in the raw frequency maps. The sources are selected from a non-blind search, based on the Filtered Fusion technique, using as candidates those sources detected in intensity. These inpainted maps are then used to construct a total of six templates, one of them at two different resolutions. To trace the synchrotron emission, we construct a template as the subtraction of the 30 GHz minus the 44 GHz map, after being convolved with the beam of each other. For the dust emission, the following templates are considered: [217GHz – 143GHz], [217GHz – 100GHz] and [143GHz – 100GHz] at 1 degree resolution, [353GHz – 217GHz] and [353GHz – 143GHz] at 10 arcminutes resolution. The last template is also constructed at the resolution of the 70 GHz channel, in order to clean that map.

Different combinations of these templates (see Table C.3 in Planck-2020-A4[1] for details) are then used to clean the raw 70, 100, 143 and 217 GHz channels (at its native resolution). The corresponding linear coefficients are estimated independently for Q and U by minimising the variance of the cleaned maps outside a mask, that covers the point sources detected in polarization and the 3 per cent brightest Galactic emission. Once the maps have been cleaned, inpainting of the point sources detected at the corresponding raw maps is carried out. Then the non-blind search for point sources is run again on the cleaned maps and the new identified sources are also inpainted. The 100, 143 and 217 GHz cleaned maps are then combined in harmonic space, using E and B decomposition, to produce the final CMB maps for the Q and U components at a resolution of 5' (Gaussian beam) for a HEALPix parameter Nside=2048. The maximum considered multipole is . Each map is weighted taking into account its noise and resolution. In addition, the lowest multipoles of the 217 GHz cleaned map are down-weighting, since they are expected to be more contaminated by the presence of residual systematics.

The cleaned CMB maps at individual frequency channels produced as intermediate steps of SEVEM are also provided for Q and U, at their native resolution. The four pairs of Q/U maps have been inpainted in the positions of the detected point sources (given by the corresponding inpainting masks).

The confidence mask is constructed as the product of two different masks. One of them is obtained from the 353 GHz data channel and excludes those regions more contaminated by thermal dust. The second mask is constructed by thresholding a map of the ratio between the locally estimated RMS of P in the cleaned CMB map, over the same quantity expected for a map containg CMB plus noise. The combination of these two masks leaves a useful sky fraction of approximately 80 per cent.

#### SMICA

Intensity.

SMICA operation starts with a pre-processing step to deal with regions of very strong emission (such as the Galactic center) and point sources. The nine pre-processed Planck frequency channels from 30 to 857 GHz are then masked and harmonically transformed up to to form spectral statistics (all auto- and cross- angular spectra). Two different masks are used to compute the spectral statistics. The first one preserves most of the sky while the second preserves CMB-dominated areas. These two sets of spectral statistics are used to determine two sets of harmonic weights which are thus adapted to two different levels of contamination. Two CMB intensity maps are produced and then merged into a single intensity product. The merging process is devised so that the information at high Galactic latitude and medium-to-high multipole is provided by the CMB map computed from high Galatic latitude statistics (note that this map does not include the LFI channels) while the remaining information is provided by the other CMB map (which does include all Planck channels). See Planck-2020-A4[1] for more details.

Polarisation.

The SMICA pipeline for polarization uses all the 7 polarized Planck channels. The E and B modes of the frequency maps are processed independently by SMICA to produce E and B modes of the CMB map from which Q and U maps are derived. The foreground model fitted by SMICA is 6-dimensional which is the maximal dimension supported by SMICA when operating in blind mode, that is, assuming nothing about the foregrounds except that they can be represented by a superposition of 6 components with unconstrained emission laws, unconstrained angular spectra and unconstrained angular correlation. See Planck-2020-A4[1] for more details.

Summary table with the various masks that have been either been used or produced by the component separation methods to pre- or post-process the CMB maps.

Pipeline specific mask filename Field Description

### Inputs

The input maps are the sky temperature maps described in the Sky temperature maps section. All pipelines use all maps between 30 and 857 GHz in temperature, and all maps between 30 and 353 GHz in polarization.

### CMB file names

The CMB products are provided as a set of five files per pipeline, one file covering some part of the entire mission (full mission; first half-mission; second half-mission; odd rings; and even rings), with a filename structure on the form

• COM_CMB_IQU-{method}-2048-R3.00_{full,hm1,hm2,oe1,oe2}.fits

The first extension contains the full-sky CMB maps in the fields called I_STOKES, Q_STOKES, U_STOKES. The full-mission files additionally contains an ASCII table with the effective beam transfer function in the second extension. The structure of each file is given as follows:

CMB R3.00 map file data structure
Ext. 1. or 2. EXTNAME = COMP-MAP (BINTABLE)
Column Name Data Type Units Description
I_Stokes Real*4 uK_cmb I map
Q_Stokes Real*4 uK_cmb Q map
U_Stokes Real*4 uK_cmb U map
Keyword Data Type Value Description
AST-COMP String CMB Astrophysical compoment name
PIXTYPE String HEALPIX
COORDSYS String GALACTIC Coordinate system
POLCCONV String COSMO Polarization convention
ORDERING String NESTED Healpix ordering
NSIDE Int 2048 Healpix Nside
METHOD String name Cleaning method (COMMANDER/NILC/SEVEM/SMICA)
Optional Ext. 2. or 3. EXTNAME = BEAM_TF (BINTABLE). ONLY FULL-MISSION DATA FILES
Column Name Data Type Units Description
INT_BEAM Real*4 none Effective beam transfer function.
POL_BEAM Real*4 none Effective beam transfer function.
Keyword Data Type Value Description
LMIN Int value First multipole of beam WF
LMAX_I Int value Last multipole for Int beam TF
LMAX_P Int value Last multipole for Pol beam TF
METHOD String name Cleaning method

All maps are provided in thermodynamic units (Kcmb</cmb>), with Nside=2048 and a nominal angular resolution of 5' FWHM.

### CMB simulations

End-to-end simulations corresponding to each of the CMB data products are provided in terms of 999 CMB realization and 300 noise realizations individually propagated through each pipeline. These files are called

• dx12_v3_{method}_{cmb,noise,noise_hm1,noise_hm2,noise_oe1,noise_oe2}_mc_?????_raw.fits
• dx12_v3_sevem_{freq}_{cmb,noise,noise_hm1,noise_hm2,noise_oe1,noise_oe2}_mc_?????_raw.fits for SEVEM cleaned cmb maps at single frequencies.
• dx12_v3_smica_nosz_{cmb,noise,noise_hm1,noise_hm2,noise_oe1,noise_oe2}_mc_?????_raw.fits for SMICA SZ-free cmb maps.

Note that only 999 CMB realizations are available, as one realization was corrupted during processing.

## Previous Releases: (2015) and (2013) CMB Maps

2015 Release of CMB maps

CMB maps

CMB maps have been produced using four different methods: COMMANDER, NILC, SEVEM, and SMICA, as described in the CMB and foreground separation section and also in Appendices A-D of Planck-2015-A09[3] and references therein.

As discussed extensively in Planck-2015-A01[4], Planck-2015-A06[5], Planck-2015-A08[6], and Planck-2015-A09[3], the residual systematics in the Planck 2015 polarization maps have been dramatically reduced compared to 2013, by as much as two orders of magnitude on large angular scales. Nevertheless, on angular scales greater than 10 degrees, correponding to l < 20, systematics are still non-negligible compared to the expected cosmological signal.

It was not possible, for this data release, to fully characterize the large-scale residuals from the data or from simulations. Therefore all results published by the Planck Collaboration in 2015 which are based on CMB polarization have used maps which have been high-pass filtered to remove the large angular scales. We warn all users of the CMB polarization maps that they cannot yet be used for cosmological studies at large angular scales.

For convenience, we provide as default polarized CMB maps from which all angular scales at l < 30 have been filtered out.

For each method we provide the following:

• Full-mission CMB intensity map, confidence mask and beam transfer function.
• Full-mission CMB polarisation map,
• A beam transfer function.

In addition, and for characterisation purposes, we include six other sets of maps from three data splits: first/second half-ring, odd/even years and first/second half-mission. For the year-1,2 and half-mission-1,2 data splits we provide half-sum and half-difference maps which are produced by running the corresponding sums and differences inputs through the pipelines. The half-difference maps can be used to provide an approximate noise estimates for the full mission, but they should be used with caution. Each split has caveats in this regard: there are noise correlations between the half-ring maps, and missing pixels in the other splits. The Intensity maps are provided at Nside = 2048, at 5 arcmin resolution, while the Polarisation ones are provided at Nside = 1024, at 10 arcmin resolution. All maps are in units of Kcmb.

In addition, for each method we provide three sets of files, each categorized by the "R2.0X" label as follows:

R2.02
 This set of intensity and polarisation maps are provided at a resolution of Nside=1024. The Stokes Q and U maps are high-pass filtered to contain only modes above l > 30, as explained above and as used for analysis by the Planck Collaboration; THESE ARE THE POLARISATION MAPS WHICH SHOULD BE USED FOR COSMOLOGICAL ANALYSIS. Each type of map is packaged into a separate fits file (as for "R2.01"), resulting in file sizes which are easier to download (as opposed to the "R2.00" files), and more convenient to use with commonly used analysis software.

R2.01
This is the most complete set of 2015 CMB maps, containing Intensity products at a resolution of Nside=2048, and both Intensity and Polarisation at resolution of Nside=1024. For polarisation (Q and U), they contain all angular resolution modes. WE CAUTION USERS ONCE AGAIN THAT THE STOKES Q AND U MAPS ARE NOT CONSIDERED USEABLE FOR COSMOLOGICAL ANALYSIS AT l < 30. The structure of these files is the same as for "R2.02".

R2.00
 This set of files is equivalent to the "R2.01" set, but are packaged into only two large files. Warning: downloading these files could be very lengthy...


For a complete description of the above data structures, see below; the content of the first extensions is illustrated and commented in the table below.

The gallery below shows the Intensity, noise from half-mission, half-difference, and confidence mask for the four pipelines, in the order COMMANDER, NILC, SEVEM and SMICA, from top to bottom. The Intensity maps' scale is [–500.+500] μK, and the noise spans [–25,+25] μK. We do not show the Q and U maps since they have no significant visible structure to contemplate.

Product description

COMMANDER

Principle
COMMANDER is a Planck software code implementing pixel based Bayesian parametric component separation. Each astrophysical signal component is modelled in terms of a small number of free parameters per pixel, typically in terms of an amplitude at a given reference frequency and a small set of spectral parameters, and these are fitted to the data with an MCMC Gibbs sampling algorithm. Instrumental parameters, including calibration, bandpass corrections, monopole and dipoles, are fitted jointly with the astrophysical components. A new feature in the Planck 2015 analysis is that the astrophysical model is derived from a combination of Planck, WMAP and a 408 MHz (Haslam et al. 1982) survey, providing sufficient frequency support to resolve the low-frequency components into synchrotron, free-free and spinning dust. For full details, see Planck-2015-A10[7].
Resolution (effective beam)
The Commander sky maps have different angular resolutions depending on data products:
• The components of the full astrophysical sky model derived from the complete data combination (Planck, WMAP, 408 MHz) have a 1 degree FWHM resolution, and are pixelized at Nside=256. The corresponding CMB map defines the input map for the low-l Planck 2015 temperature likelihood.
• The Commander CMB temperature map derived from Planck-only observations has an angular resolution of ~5 arcmin and is pixelized at Nside=2048. This map is produced by harmonic space hybridiziation, in which independent solutions derived at 40 arcmin (using 30-857 GHz data), 7.5 arcmin (using 143-857 GHz data), and 5 arcmin (using 217-857 GHz data) are coadded into a single map.
• The Commander CMB polarization map has an angular resolution of 10 arcmin and is pixelized at Nside=1024. As for the temperature case, this map is produced by harmonic space hybridiziation, in which independent solutions derived at 40 arcmin (using 30-353 GHz data) and 10 arcmin (using 100-353 GHz data) are coadded into a single map.
The Commander confidence masks are produced by thresholding the chi-square map characterizing the global fits, combined with direct CO amplitude thresholding to eliminate known leakage effects. In addition, we exclude the 9-year WMAP point source mask in the temperature mask. For full details, see Sections 5 and 6 in Planck-2015-A10[7]. A total of 81% of the sky is admitted for high-resolution temperature analysis, and 83% for polarization analysis. For low-resolution temperature analysis, for which the additional WMAP and 408 MHz observations improve foreground constraints, a total of 93% of the sky is admitted.

NILC

Principle
The Needlet-ILC (hereafter NILC) CMB map is constructed both in total intensity as well as polarization: Q and U Stokes parameters. For total intensity, all Planck frequency channels are included. For polarization, all polarization sensitive frequency channels are included, from 30 to 353 GHz. The solution, for T, Q and U is obtained by applying the Internal Linear Combination (ILC) technique in needlet space, that is, with combination weights which are allowed to vary over the sky and over the whole multipole range.
Resolution (effective beam)
The spectral analysis, and estimation of the NILC coefficients, is performed up to a maximum . The effective beam is equivalent to a Gaussian circular beam with FWHM=5 arcminutes.
The same procedure is followed by SMICA and NILC for producing confidence masks, though with different parametrizations. A low resolution smoothed version of the NILC map, noise subtracted, is thresholded to 73.5 squared micro-K for T, and 6.75 squared micro-K for Q and U.

SEVEM

Principle
SEVEM produces clean CMB maps at several frequencies by using a procedure based on template fitting in real space. The templates are typically constructed from the lowest and highest Planck frequencies and then subtracted from the CMB-dominated channels, with coefficients that are chosen to minimize the variance of the clean map outside a considered mask. In the cleaning process, no assumptions about the foregrounds or noise levels are needed, rendering the technique very robust. Two single frequency clean maps are then combined to obtain the final CMB map.
Resolution
For intensity the clean CMB map is constructed up to a maximum at Nside=2048 and at the standard resolution of 5 arcminutes (Gaussian beam).
For polarization the clean CMB map is produced at Nside=1024 with a resolution of 10 arcminutes (Gaussian beam) and a maximum .
The confidence masks cover the most contaminated regions of the sky, leaving approximately 85 per cent of useful sky for intensity, and 80 per cent for polarization.

Foregrounds-subtracted maps

In addition to the regular CMB maps, SEVEM provides maps cleaned of the foregrounds for selected frequency channels (categorized as fgsub-sevem in the archive). In particular, for intensity there are clean CMB maps available at 100, 143 and 217 GHz, provided at the original resolution of the uncleaned channel and at Nside=2048. For polarization, there are Q/U clean CMB maps for the 70, 100 and 143 GHz (at Nside=1024). The 70 GHz clean map is provided at its original resolution, whereas the 100 and 143 GHz maps have a resolution given by a Gaussian beam with fwhm=10 arcminutes.

SMICA

Principle
SMICA produces CMB maps by linearly combining all Planck input channels with multipole-dependent weights. It includes multipoles up to . Temperature and polarization maps are produced independently.
Resolution (effective beam)
The SMICA intensity map has an effective beam window function of 5 arc-minutes which is truncated at and is not deconvolved from the pixel window function. Thus the delivered beam window function is the product of a Gaussian beam at 5 arcminutes and the pixel window function for =2048.
The SMICA Q and U maps are obtained similarly but are produced at =1024 with an effective beam of 10 arc-minutes (to be multiplied by the pixel window function, as for the intensity map).
A confidence mask is provided which excludes some parts of the Galactic plane, some very bright areas and masked point sources. This mask provides a qualitative (and subjective) indication of the cleanliness of a pixel. See section below detailing the production process.

A number of common masks have been defined for analysis of the CMB temperature and polarization maps. They are based on the confidence masks provided by the component separation methods. One mask for temperature and one mask for polarization have been chosen as the preferred masks based on subsequent analyses.

The common masks for the CMB temperature maps are:

• UT78: union of the Commander, SEVEM, and SMICA temperature confidence masks (the NILC mask was not included since it masks much less of the sky). It has fsky = 77.6%. This is the preferred mask for temperature.
• UTA76: in addition to the UT78 mask, it masks pixels where standard deviation between the four CMB maps is greater than 10 μK. It has fsky = 76.1%.

The common masks for the CMB polarization maps are:

• UP78: the union of the Commander, SEVEM and SMICA polarization confidence masks (the NILC mask was not included since it masks much less of the sky). It has fsky = 77.6%.
• UPA77: In addition to the UP78 mask, it masks pixels where the standard deviation between the four CMB maps, averaged in Q and U, is greater than 4 μK. It has fsky = 76.7%.
• UPB77: in addition to the UP78 mask, it masks polarized point sources detected in the frequency channel maps. It has fsky = 77.4%. This is the preferred mask for polarization.

Additional pre-processing masks used mainly for inpainting of the frequency and/or cmb maps is show below in Masks

CMB-subtracted frequency maps ("Foreground maps")

These are the full-sky, full-mission frequency maps in intensity from which the CMB has been subtracted. The maps contain foregrounds and noise. They are provided for each frequency channel and for each component separation method. They are grouped into 8 files, two for each method of which there is one for each instrument. The maps are are at Nside = 1024 for the three LFI channels and at Nside = 2048 for the six HFI channels. The filenames are:

• LFI_Foregrounds-{method}_1024_Rn.nn.fits (145 MB each)
• HFI_Foregrounds-{method}_2048_Rn.nn.fits (1.2 GB each)

To remove the CMB, the respective CMB map was first deconvolved with the 5 arcmin beam, then convolved with the beam of the frequency channel, and finally subtracted from the frequency map. This was done using the in harmonic space, assuming a symmetric beam.

The CMB-subtracted maps have complicated noise properties. The CMB maps contain a noise contribution from each of the frequency maps, depending on the weights with which they were combined. Therefore subtracting the CMB map from a frequency channel contributes additional noise from the other frequency channels.

The second-order (kinematic) quadrupole is a frequency-dependent effect. During the production of the frequency maps the frequency-independent part was subtracted, which leaves a frequency-dependent residual quadrupole. The residuals in the component-separated CMB temperature maps have been estimated by simulating the effect in the frequency maps and propagating it through the component separation pipelines. The residuals have an amplitude of around 2 μK peak-to-peak. The maps of the estimated residuals can be used to remove the effect by subtracting them from the CMB maps.

Production process

COMMANDER

Pre-processing
All sky maps are first convolved to a common resolution that is larger than the largest beam of any frequency channel. For the combined Planck, WMAP and 408 MHz temperature analysis, the common resolution is 1 degree FWHM; for the Planck-only, all-frequency analysis it is 40 arcmin FWHM; and for the intermediate-resolution analysis it is 7.5 arcmin; while for the full-resolution analysis, we assume all frequencies between 217 and 857 GHz have a common resolution, and no additional convolution is performed. For polarization, only two smoothing scales are employed, 40 and 10 arcmin, respectively. The instrumental noise rms maps are convolved correspondingly, properly accouting for their matrix-like nature.
Priors
The following priors are enforced in the Commander analysis:
• All foreground amplitudes are enforced to be positive definite in the low-resolution analysis, while no amplitude priors are enforced in the high-resolution analyses
• Monopoles and dipoles are fixed to nominal values for a small set of reference frequencies
• Gaussian priors are enforced on spectral parameters, with values informed by the values derived in the high signal-to-noise areas of the sky
• The Jeffreys ignorance prior is enforced on spectral parameters in addition to the informative Gaussian priors
Fitting procedure
Given data and priors, Commander either maximizes, or samples from, the Bayesian posterior, P(theta|data). Because this is a highly non-Gaussian and correlated distribution, involving millions of parameters, these operations are performed by means of the Gibbs sampling algorithm, in which joint samples from the full distributions are generated by iteratively sampling from the corresponding conditional posterior distributions, P(theta_i| data, theta_{j/=i}). For the low-resolution analysis, all parameters are optimized jointly, while in the high-resolution analyses, which employs fewer frequency channels, low signal-to-noise parameters are fixed to those derived at low resolution. Examples of such parameters include monopoles and dipoles, calibration and bandpass parameters, thermal dust temperature etc.

NILC

Pre-processing
All sky frequency maps are deconvolved using the DPC beam transfer function provided, and re-convolved with a 5 arcminutes FWHM circular Gaussian beam. In polarization, prior to the smoothing process, all sky E and B maps are derived from Q and U using standard HEALPix tools from each individual frequency channels
Linear combination
Pre-processed input frequency maps are decomposed in needlet coefficients, specified in the Appendix B of the Planck A11 paper, with shape given by Table B.1. Minimum variance coefficients are then obtained, using all channels for T, from 30 to 353 for E and B.
Post-processing
E and B maps are re-combined into Q and U products using standard HEALPix tools.

SEVEM

The templates used in the SEVEM pipeline are typically constructed by subtracting two close Planck frequency channel maps, after first smoothing them to a common resolution to ensure that the CMB signal is properly removed. A linear combination of the templates is then subtracted from (hitherto unused) map d to produce a clean CMB map at that frequency. This is done in real space at each position on the sky: where is the number of templates. The coefficients are obtained by minimising the variance of the clean map outside a given mask. Note that the same expression applies for I, Q and U. Although we exclude very contaminated regions during the minimization, the subtraction is performed for all pixels and, therefore, the cleaned maps cover the full-sky (although we expect that foreground residuals are present in the excluded areas).

There are several possible configurations of SEVEM with regard to the number of frequency maps which are cleaned or the number of templates that are used in the fitting. Note that the production of clean maps at different frequencies is of great interest in order to test the robustness of the results, and these intermediate products (clean maps at individual frequencies for intensity and polarization) are also provided in the archive. Therefore, to define the best strategy, one needs to find a compromise between the number of maps that can be cleaned independently and the number of templates that can be constructed.

Intensity

For the CMB intensity map, we have cleaned the 100 GHz, 143 GHz and 217 GHz maps using a total of four templates. Three of them are constructed as the difference of two consecutive Planck channels smoothed to a common resolution (30-44, 44-70 and 545-353) while the 857 GHz channel is chosen as the fourth template. First of all, the six frequency channels which are going to be part of the templates are inpainted at the point source positions detected using the Mexican Hat Wavelet algorithm. The size of the holes to be inpainted is determined taking into account the beam size of the channel as well as the flux of each source. The inpainting algorithm is based on simple diffuse inpainting, which fills one pixel with the mean value of the neighbouring pixels in an iterative way. To avoid inconsistencies when subtracting two channels, each frequency map is inpainted on the sources detected in that map and on the second map (if any) used to construct the template. Then the maps are smoothed to a common resolution (the first channel in the subtraction is smoothed with the beam of the second map and viceversa). For the 857 GHz template, we simply filter the inpainted map with the 545 GHz beam.

The coefficients are obtained by minimising the variance outside the analysis mask, that covers the 1 per cent brightest emission of the sky as well as point sources detected at all frequency channels. Once the maps are cleaned, each of them is inpainted on the point sources positions detected at that (raw) channel. Then, the MHW algorithm is run again, now on the clean maps. A relatively small number of new sources are found and are also inpainted at each channel. The resolution of the clean map is the same as that of the original data. Our final CMB map is then constructed by combining the 143 and 217 GHz maps by weighting the maps in harmonic space taking into account the noise level, the resolution and a rough estimation of the foreground residuals of each map (obtained from realistic simulations). This final map has a resolution corresponding to a Gaussian beam of fwhm=5 arcminutes.

In addition, the clean CMB maps produced at 100, 143 and 217 GHz frequencies are also provided. The resolution of these maps is the same as that of the uncleaned frequency channels and have been constructed at Nside=2048. They have been inpainted at the position of the detected point sources. Note that these three clean maps should be close to independent, although some level of correlation will be present since the same templates have been used to clean the maps.

The confidence mask is produced by studying the differences between several SEVEM CMB reconstructions, which correspond to maps cleaned at different frequencies or using different analysis masks. The obtained mask leaves a useful sky fraction of approximately 85 per cent.

Polarization

To clean the polarization maps, a procedure similar to the one used for intensity data is applied to the Q and U maps independently. However, given that a narrower frequency coverage is available for polarization, the selected templates and maps to be cleaned are different. In particular, we clean the 70, 100 and 143 GHz using three templates for each channel. The first step of the pipeline is to inpaint the positions of the point sources using the MHW, in those channels which are going to be used in the construction of templates, following the same procedure as for the intensity case. The inpainting is performed in the frequency maps at their native resolution. These inpainted maps are then used to construct a total of four templates. To trace the synchrotron emission, we construct a template as the subtraction of the 30 GHz minus the 44 GHz map, after being convolved with the beam of each other. For the dust emission, the following templates are considered: 353-217 GHz (smoothed at 10' resolution), 217-143 GHz (used to clean 70 and 100 GHz) and 217-100 GHz (to clean 143 GHz). These two last templates are constructed at 1 degree resolution since an additional smoothing becomes necessary in order to increase the signal-to-noise ratio of the template. Conversely to the intensity case and due to the lower availability of frequency channels, it becomes necessary to use the maps to be cleaned as part of one of the templates. In this way, the 100 GHz map is used to clean the 143 GHz frequency channel and viceversa, making the clean maps less independent between them than in the intensity case.

These templates are then used to clean the non-inpainted 70 (at its native resolution), 100 (at 10' resolution) and 143 GHz maps (also at 10'). The corresponding linear coefficients are estimated independently for Q and U by minimising the variance of the clean maps outside a mask, that covers point sources and the 3 per cent brightest Galactic emission. Once the maps have been cleaned, inpainting of the point sources detected at the corresponding raw maps is carried out. The size of the holes to be inpainted takes into account the additional smoothing of the 100 and 143 GHz maps. The 100 and 143 GHz clean maps are then combined in harmonic space, using E and B decomposition, to produce the final CMB maps for the Q and U components at a resolution of 10' (Gaussian beam) for a HEALPix parameter Nside=1024. Each map is weighted taking into account its corresponding noise level at each multipole. Finally, before applying the post-processing HPF to the clean polarization data, the region with the brightest Galactic residuals is inpainted (5 per cent of the sky) to avoid the introduction of ringing around the Galactic centre in the filtering process.

The clean CMB maps at individual frequency channels produced as intermediate steps of SEVEM are also provided for Q and U, constructed at Nside=1024. The clean 70 GHz map is provided at its native resolution, while the clean maps at 100 and 143 GHz frequencies have a resolution of 10 arcminutes (Gaussian beam). The three maps have been inpainted in the positions of the detected point sources. Note that, due to the availability of a smaller number of templates for polarization than for intensity, these maps are less independent than for the temperature case, since, for instance, the 100 GHz map is used to clean the 143 GHz one and viceversa.

The confidence mask includes all the pixels above a given threshold in a smoothed version of the clean CMB map, the regions more contaminated by the CO emission and those pixels more affected by the high-pass filtering, leaving a useful sky fraction of approximately 80 per cent.

SMICA

A) Production of the intensity map.

1) Pre-processing
Before computing spherical harmonic coefficients, all input maps undergo a pre-processing step to deal with regions of very strong emission (such as the Galactic center) and point sources. The point sources with SNR > 5 in the PCCS catalogue are fitted in each input map. If the fit is successful, the fitted point source is removed from the map; otherwise it is masked and the hole is filled in by a simple diffusive process to ensure a smooth transition and mitigate spectral leakage. The diffusive inpainting process is also applied to some regions of very strong emissions. This is done at all frequencies but 545 and 857 GHz, here all point sources with SNR > 7.5 are masked and filled-in similarly.
2) Linear combination
The nine pre-processed Planck frequency channels from 30 to 857 GHz are harmonically transformed up to and co-added with multipole-dependent weights as shown in the figure.
3) Post-processing
A confidence mask is determined (see the Planck paper) and all regions which have been masked in the pre-processing step are added to it.

B) Production of the Q and U polarisation maps.

The SMICA pipeline for polarization uses all the 7 polarized Planck channels. The production of the Q and U maps is similar to the production of the intensity map. However, there is no input point source pre-processing of the input maps. The regions of very strong emission are masked out using an apodized mask before computing the E and B modes of the input maps and combining them to produce the E and B modes of the CMB map. Those modes are then used to synthesize the U and Q CMB maps. The E and B parts of the input frequency maps being processed jointly, there are, at each multipole, 2*7=14 coefficients (weights) defined to produce the E modes of the CMB map and as many to produce the B part. The weights are displayed in the figure below. The Q and U maps were originally produced at Nside=2048 with a 5-arc-minute resolution, but were downgraded to Nside=1024 with a 10 arc-minute resolution for this release.

Summary table with the different masks that have been used by the component separation methods to pre-process and to process the frequency maps and the CMB maps.

Commander 2015 (PR2) Used for diffuse inpainting of input frequency maps Used for constrained Gaussian realization inpaiting of CMB map Description
T_MASK NO NO T_MASK (the equivalent to PR1 VALMASK) is the confidence mask in temperature that defines the region where the reconstructed CMB is trusted. It can be found inside COM_CMB_IQU-commander-field-Int_2048_R2.01_full.fits.
P_MASK NO NO P_MASK is the confidence mask in polarization that defines the region where the reconstructed CMB is trusted. It can be found inside COM_CMB_IQU-commander_1024_R2.02_full.fits.
INP_MASK_T NO YES Three masks have been used for inpaiting of CMB maps for specific ranges: three different angular resolution maps (40 arcmin, 7.5 arcmin and full resolution), are produced using different data combinations and foreground models. Each of these are inpainted with their own masks with a constrained Gaussian realization before coadding the three maps in harmonic space.
INP_MASK_P NO YES Mask used for inpainting of the CMB map in polarization.
SEVEM 2015 (PR2) Used for Diffuse Inpainting of foregorund subtracted CMB maps (fgsub-sevem) Used for constrained Gaussian realization inpaiting of CMB map Description
T_MASK NO NO T_MASK (the equivalent to PR1 VALMASK) is the confidence mask in temperature that defines the region where the reconstructed CMB is trusted. It can be found inside COM_CMB_IQU-sevem-field-Int_2048_R2.01_full.fits.
P_MASK NO NO P_MASK is the confidence mask in polarization that defines the region where the reconstructed CMB is trusted. It can be found inside COM_CMB_IQU-sevem_1024_R2.02_full.fits.
INP_MASK_T YES NO Point source mask for temperature. This mask is the combination of the 143 and 217 T point source masks used for the inpainting of the foreground subtracted CMB maps at those two frequencies. These two maps have been combined to produce the final CMB map.
INP_MASK_P YES NO Point source mask for polarization. This mask is the combination of the 100 and 143 point source masks used for the inpainting of the foreground subtracted CMB maps at those two frequencies. These two maps have been combined to produce the final CMB map.
INP_MASK_T for the cleaned 100, 143 and 217 GHz CMB YES NO Three temperature point source masks used for the inpainting of the foreground subtracted CMB maps at the considered frequencies:
INP_MASK_P for the cleaned 70, 100 and 143 GHz CMB YES NO Three polarization point source masks used for the inpainting of the foreground subtracted CMB maps at the considered frequencies:
NILC 2015 (PR2) Used for diffuse inpainting of input frequency maps Used for constrained Gaussian realization inpaiting of CMB map Description
T_MASK NO NO T_MASK (the equivalent to PR1 VALMASK) is the confidence mask in temperature that defines the region where the reconstructed CMB is trusted. It can be found inside COM_CMB_IQU-nilc-field-Int_2048_R2.01_full.fits.
P_MASK NO NO P_MASK is the confidence mask in polarization that defines the region where the reconstructed CMB is trusted. It can be found inside COM_CMB_IQU-nilc_1024_R2.02_full.fits.
INP_MASK YES NO The pre-processing involves inpainting of the holes in INP_MASK in the frequency maps prior to applying NILC on them. The first mask (nside 2048) has been used for the pre-processing of sky maps for HFI channels and second one for LFI channels (nside 1024). They can downloaded here:
SMICA 2015 (PR2) Used for diffuse inpainting of input frequency maps Used for constrained Gaussian realization inpaiting of CMB map Description
T_MASK NO YES T_MASK (the equivalent to PR1 VALMASK) is the confidence mask in temperature that defines the region where the reconstructed CMB is trusted. It can be found inside COM_CMB_IQU-smica-field-Int_2048_R2.01_full.fits.
P_MASK NO YES P_MASK is the confidence mask in polarization that defines the region where the reconstructed CMB is trusted. It can be found inside COM_CMB_IQU-smica_1024_R2.02_full.fits.

Inputs The input maps are the sky temperature maps described in the Sky temperature maps section. SMICA and SEVEM use all the maps between 30 and 857 GHz; NILC uses the ones between 44 and 857 GHz. Commander-Ruler uses frequency channel maps from 30 to 353 GHz.

File names and structure

Three sets of files FITS files containing the CMB products are available. In the first set all maps (i.e., covering different parts of the mission) and all characterisation products for a given method and a given Stokes parameter are grouped into a single extension, and there are two files per method (smica, nilc, sevem, and commander), one for the high resolution data (I only, Nside=2048) and one for low resolution data (Q and U only, Nside=1024). Each file also contains the associated confidence mask(s) and beam transfer function. These are the R2.00 files which have names like

• COM_CMB_IQU-{method}-field-{Int,Pol}_Nside_R2.00.fits

There are 7 coverage periods:full, halfyear-1,2, halfmission-1,2, or ringhalf-1,2, and 4 characterisation products: half-sum and half-difference for the year and the half-mission periods.

In the second second set the different coverages are split into different files which in most cases have a single extension with I only (Nside=1024) and I, Q, and U (Nside=1024). This second set was built in order to allow users to use standard codes like spice or anafast on them, directly. So this set contains the I maps at Nside=1024, which are not contained in the R2.00; on the other hand this set does not contain the half-sum and half-difference maps. These are the 2.01 files which have names like

• COM_CMB_IQU-{method}{-field-Int|Pol}_Nside_R2.01_{coverage}.fits for the regular CMB maps, and
• COM_CMB_IQU-{fff}-{fgsub-sevem}{-field-Int|Pol}_Nside_R2.01_{coverage}.fits for the sevem frequency-dependent, foregrounds-subtracted maps,

where field-Int|Pol is used to indicate that only Int or only Pol data are contained (at present only field-Int is used for the high-res data), and is not included in the low-res data which contains all three Stokes parameters, and coverage is one of full, halfyear-1,2, halfmission-1,2, or ringhalf-1,2. Also, the coverage=full files contain also the confidence mask(s) and beam transfer function(s) which are valid for all products of the same method (one for Int and one for Pol when both are available).

The third set has the same structure as the Nside=1024 products of R2.01, but the Q and U maps have been high-pass filtered to remove modes at l < 30 for the reasons indicated earlier. These are the default products for use in polarisation studies. They are the R2.02 files which have names like:

• COM_CMB_IQU-{method}_1024_R2.02_{coverage}.fits

Version 2.00 files

These have names like

• COM_CMB_IQU-{method}-field-{Int,Pol}_Nside_R2.00.fits,

as indicated above. They contain:

• a minimal primary extension with no data;
• one or two BINTABLE data extensions with a table of Npix lines by 14 columns in which the first 13 columns is a CMB maps produced from the full or a subset of the data, as described in the table below, and the last column in a confidence mask. There is a single extension for Int files, and two, for Q and U, for Pol files.
• a BINTABLE extension containing the beam transfer function (mistakenly called window function in the files).

If Nside=1024 the files contain I, Q and U maps, whereas if Nside=2048 only the I map is given.

CMB R2.00 map file data structure
Ext. 1. or 2. EXTNAME = COMP-MAP (BINTABLE)
Column Name Data Type Units Description
I or Q or U Real*4 uK_cmb I or Q or U map
HM1 Real*4 uK_cmb Half-miss 1
HM2 Real*4 uK_cmb Half-miss 2
YR1 Real*4 uK_cmb Year 1
YR2 Real*4 uK_cmb Year 2
HR1 Real*4 uK_cmb Half-ring 1
HR2 Real*4 uK_cmb Half-ring 2
HMHS Real*4 uK_cmb Half-miss, half sum
HMHD Real*4 uK_cmb Half-miss, half diff
YRHS Real*4 uK_cmb Year, half sum
YRHD Real*4 uK_cmb Year, half diff
HRHS Real*4 uK_cmb Half-ring half sum
HRHD Real*4 uK_cmb Half-ring half diff
Keyword Data Type Value Description
AST-COMP String CMB Astrophysical compoment name
PIXTYPE String HEALPIX
COORDSYS String GALACTIC Coordinate system
POLCCONV String COSMO Polarization convention
ORDERING String NESTED Healpix ordering
NSIDE Int 1024 or 2048 Healpix Nside
METHOD String name Cleaning method (smica/nilc/sevem/commander)
Ext. 2. or 3. EXTNAME = BEAM_WF (BINTABLE) . See Note 1
Column Name Data Type Units Description
BEAMWF Real*4 none The effective beam transfer function, including the pixel window function. See Note 2.
Keyword Data Type Value Description
LMIN Int value First multipole of beam TF
LMAX Int value Last multipole of beam TF
METHOD String name Cleaning method (SMICA/NILC/SEVEM/COMMANDER-Ruler)

Notes:

1. Actually this is a beam transfer function, so BEAM_TF would have been more appropriate.
2. The beam transfer function given here includes the pixel window function for the Nside=2048 pixelization. It means that, ideally, . The beam Window function is given by

Version 2.01 files

These files have names like:

• COM_CMB_IQU-{method}{-field-Int|Pol}_Nside_R2.01_{coverage}.fits for the regular CMB maps, and
• COM_CMB_IQU-{fff}-{fgsub-sevem}{-field-Int|Pol}_Nside_R2.01_{coverage}.fits for the sevem frequency-dependent, foregrounds-subtracted maps,

as indicated above. They contain:

• a minimal primary extension with no data, but with a NUMEXT keyword giving the number of extensions contained.
• one or two BINTABLE data extensions with a table of Npix lines by 1-5 columns depending on the file, as described above: the minimum begin I only, the maximum begin I, Q, U, and confidence masks for I and P.
• a BINTABLE extension containing the beam transfer function(s): one for I, and a second one that applies to both Q and U, if Nslde=1024.

If Nside=1024 the files contain I, Q and U maps, whereas if Nside=2048 only the I map is given. The basic structure, including information on the most important keywords, is given in the table below. For full details, see the FITS header.

CMB R2.01 map file data structure
Ext. 1. or 2. EXTNAME = COMP-MAP (BINTABLE)
Column Name Data Type Units Description
I_Stokes Real*4 uK_cmb I map (Nside=1024,2048)
Q_Stokes Real*4 uK_cmb Q map (Nside=1024)
U_Stokes Real*4 uK_cmb U map (Nside=2048)
Keyword Data Type Value Description
AST-COMP String CMB Astrophysical compoment name
PIXTYPE String HEALPIX
COORDSYS String GALACTIC Coordinate system
POLCCONV String COSMO Polarization convention
ORDERING String NESTED Healpix ordering
NSIDE Int 1024 or 2048 Healpix Nside
METHOD String name Cleaning method (SMICA/NILC/SEVEM)
Optional Ext. 2. or 3. EXTNAME = BEAM_TF (BINTABLE)
Column Name Data Type Units Description
INT_BEAM Real*4 none Effective beam transfer function. See Note 1.
POL_BEAM Real*4 none Effective beam transfer function. See Note 1.
Keyword Data Type Value Description
LMIN Int value First multipole of beam WF
LMAX_I Int value Last multipole for Int beam TF
LMAX_P Int value Last multipole for Pol beam TF
METHOD String name Cleaning method

Notes:

1. The beam transfer function given here includes the pixel window function for the Nside=2048 pixelization. It means that, ideally, . The beam Window function is given by

Version 2.02 files

For polarisation work, this is the default set of files to be used for cosmological analysis. Their content is identical to the "R2.01" files, except that angular scales above l < 30 have been filtered out of the Q and U maps.

These files have names like:

• COM_CMB_IQU-{method}_1024_R2.02_{coverage}.fits

as indicated above. They contain: The files contain

• a minimal primary extension with no data, but with a NUMEXT keyword giving the number of extensions contained.
• one or two BINTABLE data extensions with a table of Npix lines by 1-5 columns depending on the file, as described above: the minimum begin I only, the maximum begin I, Q, U, and confidence masks for I and P.
• a BINTABLE extension containing 2 beam transfer functions: one for I and one that applies to both Q and U.

If Nside=1024 the files contain I, Q and U maps, whereas if Nside=2048 only the I map is given. The basic structure, including information on the most important keywords, is given in the table below. For full details, see the FITS header.

CMB R2.02 map file data structure
Ext. 1. or 2. EXTNAME = COMP-MAP (BINTABLE)
Column Name Data Type Units Description
I_Stokes Real*4 uK_cmb I map (Nside=1024)
Q_Stokes Real*4 uK_cmb Q map (Nside=1024)
U_Stokes Real*4 uK_cmb U map (Nside=2048)
Keyword Data Type Value Description
AST-COMP String CMB Astrophysical compoment name
PIXTYPE String HEALPIX
COORDSYS String GALACTIC Coordinate system
POLCCONV String COSMO Polarization convention
ORDERING String NESTED Healpix ordering
NSIDE Int 1024 or 2048 Healpix Nside
METHOD String name Cleaning method (SMICA/NILC/SEVEM)
Optional Ext. 2. or 3. EXTNAME = BEAM_TF (BINTABLE)
Column Name Data Type Units Description
INT_BEAM Real*4 none Effective beam transfer function. See Note 1.
POL_BEAM Real*4 none Effective beam transfer function. See Note 1.
Keyword Data Type Value Description
LMIN Int value First multipole of beam WF
LMAX_I Int value Last multipole for Int beam TF
LMAX_P Int value Last multipole for Pol beam TF
METHOD String name Cleaning method

Notes:

1. The beam transfer function given here includes the pixel window function for the Nside=2048 pixelization. It means that, ideally, . The beam Window function is given by

The common masks are stored into two different files for Temperature and Polarisation respectively:

Both files contain also a map of the missing pixels for the half mission and year coverage periods. The 2 (for Temp) or 3 (for Pol) masks and the missing pixels maps are stored in 4 or 5 column a BINTABLE extension 1 of each file, named MASK-INT and MASK-POL, respectively. See the FITS file headers for details.

The quadrupole residual maps are stored in files called:

• COM_CMB_IQU-kq-resid-{method}-field-Int_2048_R2.02.fits

They contain:

• a minimal primary extension with no data, but with a NUMEXT keyword giving the number of extensions contained.
• a single BINTABLE extension with a single column of Npix lines containing the HEALPIX map indicated

The basic structure of the data extension is shown below. For full details see the extension header.

Kinetic quadrupole residual map file data structure
Ext. 1. EXTNAME = COMP-MAP (BINTABLE)
Column Name Data Type Units Description
INTENSITY Real*4 K_cmb the residual map
Keyword Data Type Value Description
AST-COMP String KQ-RESID Astrophysical compoment name
PIXTYPE String HEALPIX
COORDSYS String GALACTIC Coordinate system
POLCCONV String COSMO Polarization convention
ORDERING String NESTED Healpix ordering
NSIDE Int 2048 Healpix Nside
METHOD String name Cleaning method

2013 Release of CMB maps

CMB Maps

CMB maps have been produced by the SMICA, NILC, SEVEM and COMMANDER-Ruler pipelines. Of these, the SMICA product is considered the preferred one overall and is labelled Main product in the Planck Legacy Archive, while the other two are labeled as Additional product.

SMICA and NILC also produce inpainted maps, in which the Galactic Plane, some bright regions and masked point sources are replaced with a constrained CMB realization such that the whole map has the same statistical distribution as the observed CMB.

The results of SMICA, NILC and SEVEM pipeline are distributed as a FITS file containing 4 extensions:

1. CMB maps and ancillary products (3 or 6 maps)
2. CMB-cleaned foreground maps from LFI (3 maps)
3. CMB-cleaned foreground maps from HFI (6 maps)
4. Effective beam of the CMB maps (1 vector)

The results of COMMANDER-Ruler are distributed as two FITS files (the high and low resolution) containing the following extensions: High resolution N$_\rm{side}$=2048 (note that we don't provide the CMB-cleaned foregrounds maps for LFI and HFI because the Ruler resolution (~7.4') is lower than the HFI highest channel and and downgrading it will introduce noise correlation).

1. CMB maps and ancillary products (4 maps)
2. Effective beam of the CMB maps (1 vector)

Low resolution N$_\rm{side}$=256

1. CMB maps and ancillary products (3 maps)
2. 10 example CMB maps used in the montecarlo realization (10 maps)
3. Effective beam of the CMB maps (1 vector)

For a complete description of the data structure, see the below; the content of the first extensions is illustrated and commented in the table below.

The maps (CMB, noise, masks) contained in the first extension
Col name SMICA NILC SEVEM COMMANDER-Ruler H COMMANDER-Ruler L Description / notes
1: I Raw CMB anisotropy map. These are the maps used in the component separation paper Planck-2013-XII[2].
2: NOISE not applicable Noise map. Obtained by propagating the half-ring noise through the CMB cleaning pipelines.
3: VALMASK Confidence map. Pixels with an expected low level of foreground contamination. These maps are only indicative and obtained by different ad hoc methods. They cannot be used to rank the CMB maps.
4: I_MASK not applicable not applicable not applicable Some areas are masked for the production of the raw CMB maps (for NILC: point sources from 44 GHz to 857 GHz; for SMICA: point sources from 30 GHz to 857 GHz, Galatic region and additional bright regions).
5: INP_CMB not applicable not applicable not applicable Inpainted CMB map. The raw CMB maps with some regions (as indicated by INP_MASK) replaced by a constrained Gaussian realization. The inpainted SMICA map was used for PR.
6: INP_MASK not applicable not applicable not applicable Mask of the inpainted regions. For SMICA, this is identical to I_MASK. For NILC, it is not.

The component separation pipelines are described in the CMB and foreground separation section and also in Section 3 and Appendices A-D of Planck-2013-XII[2] and references therein.

The union (or common) mask is defined as the union of the confidence masks from the four component separation pipelines, the three listed above and Commander-Ruler. It leaves 73% of the sky available, and so it is denoted as U73.

Product description

SMICA

Principle
SMICA produces a CMB map by linearly combining all Planck input channels (from 30 to 857 GHz) with weights which vary with the multipole. It includes multipoles up to .
Resolution (effective beam)
The SMICA map has an effective beam window function of 5 arc-minutes truncated at and deconvolved from the pixel window. It means that, ideally, one would have , where is the angular spectrum of the map, where is the angular spectrum of the CMB and is a 5-arcminute Gaussian beam function. Note however that, by convention, the effective beam window function provided in the FITS file does include a pixel window function. Therefore, it is equal to where denotes the pixel window function for an Nside=2048 pixelization.
A confidence mask is provided which excludes some parts of the Galactic plane, some very bright areas and the masked point sources. This mask provides a qualitative (and subjective) indication of the cleanliness of a pixel.
The raw SMICA CMB map has valid pixels except at the location of masked areas: point sources, Galactic plane, some other bright regions. Those invalid pixels are indicated with the mask named 'I_MASK'. The raw SMICA map has been inpainted, producing the map named "INP_CMB". Inpainting consists in replacing some pixels (as indicated by the mask named INP_MASK) by the values of a constrained Gaussian realization which is computed to ensure good statistical properties of the whole map (technically, the inpainted pixels are a sample realisation drawn under the posterior distribution given the un-masked pixels.

NILC

Principle
The Needlet-ILC (hereafter NILC) CMB map is constructed from all Planck channels from 44 to 857 GHz and includes multipoles up to . It is obtained by applying the Internal Linear Combination (ILC) technique in needlet space, that is, with combination weights which are allowed to vary over the sky and over the whole multipole range.
Resolution (effective beam)
As in the SMICA product except that there is no abrupt truncation at but a smooth transition to over the range .
A confidence mask is provided which excludes some parts of the Galactic plane, some very bright areas and the masked point sources. This mask provides a qualitative indication of the cleanliness of a pixel. The threshold is somewhat arbitrary.
The raw NILC map has valid pixels except at the location of masked point sources. This is indicated with the mask named 'I_MASK'. The raw NILC map has been inpainted, producing the map named "INP_CMB". The inpainting consists in replacing some pixels (as indicated by the mask named INP_MASK) by the values of a constrained Gaussian realization which is computed to ensure good statistical properties of the whole map (technically, the inpainted pixels are a sample realisation drawn under the posterior distribution given the un-masked pixels.

SEVEM

The aim of SEVEM is to produce clean CMB maps at one or several frequencies by using a procedure based on template fitting. The templates are internal, i.e., they are constructed from Planck data, avoiding the need for external data sets, which usually complicates the analyses and may introduce inconsistencies. The method has been successfully applied to Planck simulations[8] and to WMAP polarisation data[9]. In the cleaning process, no assumptions about the foregrounds or noise levels are needed, rendering the technique very robust. Note that unlike the other products, SEVEM does not provide the mask of regions not used in the productions of the CMB ma (I_MASK) nor an inpainted version of the map and its associated mask. On the other hand, it provides channel maps and 100, 143, and 217 GHz which are used as the building blocks of the final map.

COMMANDER-Ruler

COMMANDER-Ruler is the Planck software implementing a pixel based parametric component separation. Amplitude of CMB and the main diffuse foregrounds along with the relevant spectral parameters for those (see below in the Astrophysical Foreground Section for the latter) are parametrized and fitted in single MCMC chains conducted at $N_\rm{side}$=256 using COMMANDER, implementing a Gibbs Sampling. The CMB amplitude which is obtained in these runs corresponds to the delivered low resolution CMB component from COMMANDER-Ruler which has a FWHM of 40 arcminutes. The sampling of the foreground parameters is applied to the data at full resolution for obtaining the high resolution CMB component from Ruler which is available on the PLA. In the Planck Component Separation paper[2] additional material is discussed, specifically concerning the sky region where the solutions are reliable, in terms of chi2 maps. The products mainly consist of:

• Maps of the Amplitudes of the CMB at low resolution, $N_\rm{side}$=256, along with the standard deviations of the outputs, beam profiles derived from the production process.
• Maps of the CMB amplitude, along with the standard deviations, at high resolution, $N_\rm{side}$=2048, beam profiles derived from the production process.
• Mask obtained on the basis of the precision in the fitting procedure; the thresholding is evaluated through the COMMANDER-Ruler likelihood analysis and excludes 13% of the sky, see Planck-2013-XII[2].

Production process

SMICA

1) Pre-processing
All input maps undergo a pre-processing step to deal with point sources. The point sources with SNR > 5 in the PCCS catalogue are fitted in each input map. If the fit is successful, the fitted point source is removed from the map; otherwise it is masked and the hole is filled in by a simple diffusive process to ensure a smooth transition and mitigate spectral leakage. This is done at all frequencies but 545 and 857 GHz, here all point sources with SNR > 7.5 are masked and filled-in similarly.
2) Linear combination
The nine pre-processed Planck frequency channels from 30 to 857 GHzare harmonically transformed up to and co-added with multipole-dependent weights as shown in the figure.
3) Post-processing
The areas masked in the pre-processing step are replaced by a constrained Gaussian realization.

Note: The visible power deficit in the raw CMB map around the galactic plane is due to the smooth fill-in of the masked areas in the input maps (the result of the pre-processing). It is not to be confused with the post-processing step of inpainting of the CMB map with a constrained Gaussian realization.

Weights given by SMICA to the input maps (after they are re-beamed to 5 arcmin and expressed in K), as a function of multipole.

NILC

1) Pre-processing
Same pre-processing as SMICA (except the 30 GHz channel is not used).
2) Linear combination
The pre-processed Planck frequency channels from 44 to 857 GHz are linearly combined with weights which depend on location on the sky and on the multipole range up to . This is achieved using a needlet (redundant spherical wavelet) decomposition. For more details, see Planck-2013-XII[2].
3) Post-processing
The areas masked in the pre-processing plus other bright regions step are replaced by a constrained Gaussian realization as in the SMICA post-processing step.

SEVEM

The templates are internal, i.e., they are constructed from Planck data, avoiding the need for external data sets, which usually complicates the analyses and may introduce inconsistencies. In the cleaning process, no assumptions about the foregrounds or noise levels are needed, rendering the technique very robust. The fitting can be done in real or wavelet space (using a fast wavelet adapted to the HEALPix pixelization[10]) to properly deal with incomplete sky coverage. By expediency, however, we fill in the small number of unobserved pixels at each channel with the mean value of its neighbouring pixels before applying SEVEM.

We construct our templates by subtracting two close Planck frequency channel maps, after first smoothing them to a common resolution to ensure that the CMB signal is properly removed. A linear combination of the templates is then subtracted from (hitherto unused) map d to produce a clean CMB map at that frequency. This is done either in real or in wavelet space (i.e., scale by scale) at each position on the sky: where is the number of templates. If the cleaning is performed in real space, the coefficients are obtained by minimising the variance of the clean map outside a given mask. When working in wavelet space, the cleaning is done in the same way at each wavelet scale independently (i.e., the linear coefficients depend on the scale). Although we exclude very contaminated regions during the minimization, the subtraction is performed for all pixels and, therefore, the cleaned maps cover the full-sky (although we expect that foreground residuals are present in the excluded areas).

An additional level of flexibility can also be considered: the linear coefficients can be the same for all the sky, or several regions with different sets of coefficients can be considered. The regions are then combined in a smooth way, by weighting the pixels at the boundaries, to avoid discontinuities in the clean maps. Since the method is linear, we may easily propagate the noise properties to the final CMB map. Moreover, it is very fast and permits the generation of thousands of simulations to character- ize the statistical properties of the outputs, a critical need for many cosmological applications. The final CMB map retains the angular resolution of the original frequency map.

There are several possible configurations of SEVEM with regard to the number of frequency maps which are cleaned or the number of templates that are used in the fitting. Note that the production of clean maps at different frequencies is of great interest in order to test the robustness of the results. Therefore, to define the best strategy, one needs to find a compromise between the number of maps that can be cleaned independently and the number of templates that can be constructed.

In particular, we have cleaned the 143 GHz and 217 GHz maps using four templates constructed as the difference of the following Planck channels (smoothed to a common resolution): (30-44), (44-70), (545-353) and (857-545). For simplicity, the three maps have been cleaned in real space, since there was not a significant improvement when using wavelets (especially at high latitude). In order to take into account the different spectral behaviour of the foregrounds at low and high galactic latitudes, we have considered two independent regions of the sky, for which we have used a different set of coefficients. The first region corresponds to the 3 per cent brightest Galactic emission, whereas the second region is defined by the remaining 97 per cent of the sky. For the first region, the coefficients are actually estimated over the whole sky (we find that this is more optimal than perform the minimisation only on the 3 per cent brightest region, where the CMB emission is very sub-dominant) while for the second region, we exclude the 3 per cent brightest region of the sky, point sources detected at any frequency and those pixels which have not been observed at all channels. Our final CMB map has then been constructed by combining the 143 and 217 GHz maps by weighting the maps in harmonic space taking into account the noise level, the resolution and a rough estimation of the foreground residuals of each map (obtained from realistic simulations). This final map has a resolution corresponding to a Gaussian beam of fwhm=5 arcminutes.

Moreover, additional CMB clean maps (at frequencies between 44 and 353 GHz) have also been produced using different combinations of templates for some of the analyses carried out in Planck-2013-XXIII[11] and Planck-2013-XIX[12]. In particular, clean maps from 44 to 353 GHz have been used for the stacking analysis presented in Planck-2013-XIX[12], while frequencies from 70 to 217 GHz were used for consistency tests in Planck-2013-XXIII[11].

COMMANDER-Ruler

The production process consist in low and high resolution runs according to the description above.

Low Resolution Runs
Same as the Astrophysics Foregrounds Section below; The CMB amplitude is fitted along with the other foreground parameters and constitutes the CMB Low Resolution Rendering which is in the PLA.
Ruler Runs
the sampling at high resolution is used to infer the probability distribution of spectral parameters which is exploited at full resolution in order to obtain the High Resolution CMB Rendering which is in the PLA.

Summary table with the different masks that have been used by the component separation methods to pre-process and to process the frequency maps and the CMB maps.

Commander 2013 (PR1) Used for diffuse inpainting of input frequency maps Used for constrained Gaussian realization inpaiting of CMB map Description
VALMASK NO NO VALMASK is the confidence mask that defines the region where the reconstructed CMB is trusted. It can be found inside

COM_CompMap_CMB-commrul_2048_R1.00.fits and COM_CompMap_CMB-commrul_0256_R1.00.fits for low resolution analyses.

SEVEM 2013 (PR1) Used diffuse inpainting of input frequency maps Used for Constrained Gaussian realization inpaiting of CMB map Description
VALMASK NO NO VALMASK is the confidence mask that defines the region where the reconstructed CMB is trusted. It can be found inside
NILC 2013 (PR1) Used for diffuse inpainting of input frequency maps Used for constrained Gaussian realization inpaiting of CMB map Description
VALMASK NO NO VALMASK is the confidence mask that defines the region where the reconstructed CMB is trusted. It can be found inside COM_CompMap_CMB-nilc_2048_R1.20.fits.
I_MASK NO NO I_MASK defines the regions over which CMB is not built. It is a combination of point source masks, Galactic plane mask and other bright regions like LMC, SMC, etc. It can be found inside COM_CompMap_CMB-nilc_2048_R1.20.fits.
INP_MASK NO YES It can be found inside COM_CompMap_CMB-nilc_2048_R1.20.fits.
SMICA 2013 (PR1) Used for diffuse inpainting of input frequency maps Used for constrained Gaussian realization inpaiting of CMB map Description
VALMASK NO NO VALMASK is the confidence mask that defines the region where the reconstructed CMB is trusted. It can be found inside
I_MASK YES YES I_MASK defines the regions over which CMB is not built. It is a combination of point source masks, Galactic plane mask and other bright regions like LMC, SMC, etc. It can be found inside COM_CompMap_CMB-smica_2048_R1.20.fits.

Inputs

The input maps are the sky temperature maps described in the Sky temperature maps section. SMICA and SEVEM use all the maps between 30 and 857 GHz; NILC uses the ones between 44 and 857 GHz. Commander-Ruler uses frequency channel maps from 30 to 353 GHz.

File names and structure

The FITS files corresponding to the three CMB products are the following:

The files contain a minimal primary extension with no data and four BINTABLE data extensions. Each column of the BINTABLE is a (Healpix) map; the column names and the most important keywords of each extension are described in the table below; for the remaining keywords, please see the FITS files directly.

CMB map file data structure
Ext. 1. EXTNAME = COMP-MAP (BINTABLE)
Column Name Data Type Units Description
I Real*4 uK_cmb CMB temperature map
NOISE Real*4 uK_cmb Estimated noise map (note 1)
I_STDEV Real*4 uK_cmb Standard deviation, ONLY on COMMANDER-Ruler products
I_MASK Byte none Mask of regions over which CMB map is not built (Optional - see note 3)
INP_CMB Real*4 uK_cmb Inpainted CMB temperature map (Optional - see note 3)
INP_MASK Byte none mask of inpainted pixels (Optional - see note 3)
Keyword Data Type Value Description
AST-COMP String CMB Astrophysical compoment name
PIXTYPE String HEALPIX
COORDSYS String GALACTIC Coordinate system
ORDERING String NESTED Healpix ordering
NSIDE Int 2048 Healpix Nside
METHOD String name Cleaning method (SMICA/NILC/SEVEM/COMMANDER-Ruler)
Ext. 2. EXTNAME = FGDS-LFI (BINTABLE) - Note 4
Column Name Data Type Units Description
LFI_030 Real*4 K_cmb 30 GHz foregrounds
LFI_044 Real*4 K_cmb 44 GHz foregrounds
LFI_070 Real*4 K_cmb 70 GHz foregrounds
Keyword Data Type Value Description
PIXTYPE String HEALPIX
COORDSYS String GALACTIC Coordinate system
ORDERING String NESTED Healpix ordering
NSIDE Int 1024 Healpix Nside
METHOD String name Cleaning method (SMICA/NILC/SEVEM)
Ext. 3. EXTNAME = FGDS-HFI (BINTABLE) - Note 4
Column Name Data Type Units Description
HFI_100 Real*4 K_cmb 100 GHz foregrounds
HFI_143 Real*4 K_cmb 143 GHz foregrounds
HFI_217 Real*4 K_cmb 217 GHz foregrounds
HFI_353 Real*4 K_cmb 353 GHz foregrounds
HFI_545 Real*4 MJy/sr 545 GHz foregrounds
HFI_857 Real*4 MJy/sr 857 GHz foregrounds
Keyword Data Type Value Description
PIXTYPE String HEALPIX
COORDSYS String GALACTIC Coordinate system
ORDERING String NESTED Healpix ordering
NSIDE Int 2048 Healpix Nside
METHOD String name Cleaning method (SMICA/NILC/SEVEM/COMMANDER-Ruler)
Ext. 4. EXTNAME = BEAM_WF (BINTABLE)
Column Name Data Type Units Description
BEAM_WF Real*4 none The effective beam window function, including the pixel window function. See Note 5.
Keyword Data Type Value Description
LMIN Int value First multipole of beam WF
LMAX Int value Lsst multipole of beam WF
METHOD String name Cleaning method (SMICA/NILC/SEVEM/COMMANDER-Ruler)

Notes:

1. The half-ring half-difference (HRHD) map is made by passing the half-ring frequency maps independently through the component separation pipeline, then computing half their difference. It approximates a noise realisation, and gives an indication of the uncertainties due to instrumental noise in the corresponding CMB map.
2. The confidence mask indicates where the CMB map is considered valid.
3. This column is not present in the SEVEM and COMMANDER-Ruler product file. For SEVEM these three columns give the CMB channel maps at 100, 143, and 217 GHz (columns C100, C143, and C217, in units of K_cmb.
4. The subtraction of the CMB from the sky maps in order to produce the foregrounds map is done after convolving the CMB map to the resolution of the given frequency. Those columns are not present in the COMMANDER-Ruler product file.
5. The beam window function given here includes the pixel window function for the Nside=2048 pixelization. It means that, ideally, .

The low resolution COMMANDER-Ruler CMB product is organized in the following way:

CMB low resolution COMMANDER-Ruler map file data structure
Ext. 1. EXTNAME = COMP-MAP (BINTABLE)
Column Name Data Type Units Description
I Real*4 uK_cmb CMB temperature map obtained as average over 1000 samples
I_stdev Real*4 uK_cmb Corresponding Standard deviation amongst the 1000 samples
Keyword Data Type Value Description
PIXTYPE String HEALPIX
COORDSYS String GALACTIC Coordinate system
ORDERING String NESTED Healpix ordering
NSIDE Int 2048 Healpix Nside
METHOD String name Cleaning method (SMICA/NILC/SEVEM/COMMANDER-Ruler)
Ext. 2. EXTNAME = CMB-Sample (BINTABLE)
Column Name Data Type Units Description
I_SIM01 Real*4 K_cmb CMB Sample, smoothed to 40 arcmin
I_SIM02 Real*4 K_cmb CMB Sample, smoothed to 40 arcmin
I_SIM03 Real*4 K_cmb CMB Sample, smoothed to 40 arcmin
I_SIM04 Real*4 K_cmb CMB Sample, smoothed to 40 arcmin
I_SIM05 Real*4 K_cmb CMB Sample, smoothed to 40 arcmin
I_SIM06 Real*4 K_cmb CMB Sample, smoothed to 40 arcmin
I_SIM07 Real*4 K_cmb CMB Sample, smoothed to 40 arcmin
I_SIM08 Real*4 K_cmb CMB Sample, smoothed to 40 arcmin
I_SIM09 Real*4 K_cmb CMB Sample, smoothed to 40 arcmin
I_SIM10 Real*4 K_cmb CMB Sample, smoothed to 40 arcmin
Keyword Data Type Value Description
PIXTYPE String HEALPIX
COORDSYS String GALACTIC Coordinate system
ORDERING String NESTED Healpix ordering
NSIDE Int 1024 Healpix Nside
METHOD String name Cleaning method (SMICA/NILC/SEVEM/COMMANDER-Ruler)
Ext. 4. EXTNAME = BEAM_WF (BINTABLE)
Column Name Data Type Units Description
BEAM_WF Real*4 none The effective beam window function, including the pixel window function.
Keyword Data Type Value Description
LMIN Int value First multipole of beam WF
LMAX Int value Lsst multipole of beam WF
METHOD String name Cleaning method (SMICA/NILC/SEVEM/COMMANDER-Ruler)

The FITS files containing the union (or common) maks is:

which contains a single BINTABLE extension with a single column (named U73) for the mask, which is boolean (FITS TFORM = B), in GALACTIC coordinates, NESTED ordering, and Nside=2048.

For the benefit of users who are only looking for a small file containing the SMICA cmb map with no additional information (noise or masks) we provide such a file here

This file contains a single extension with a single column containing the SMICA cmb temperature map.

Cautionary notes

1. The half-ring CMB maps are produced by the pipelines with parameters/weights fixed to the values obtained from the full maps. Therefore the CMB HRHD maps do not capture all of the uncertainties due to foreground modelling on large angular scales.
2. The HRHD maps for the HFI frequency channels underestimate the noise power spectrum at high l by typically a few percent. This is caused by correlations induced in the pre-processing to remove cosmic ray hits. The CMB is mostly constrained by the HFI channels at high l, and so the CMB HRHD maps will inherit this deficiency in power.
3. The beam transfer functions do not account for uncertainties in the beams of the frequency channel maps.

## References

1. Planck 2018 results. IV. Diffuse component separation, Planck Collaboration, 2020, A&A, 641, A4.
2. Planck 2013 results. XI. Component separation, Planck Collaboration, 2014, A&A, 571, A11.
3. Planck 2015 results. XI. Diffuse component separation: CMB maps, Planck Collaboration, 2016, A&A, 594, A9.
4. Planck 2015 results. I. Overview of products and results, Planck Collaboration, 2016, A&A, 594, A1.
5. Planck 2015 results. VI. LFI mapmaking, Planck Collaboration, 2016, A&A, 594, A6.
6. Planck 2015 results. VIII. High Frequency Instrument data processing: Calibration and maps, Planck Collaboration, 2016, A&A, 594, A8.
7. Planck 2015 results. X. Diffuse component separation: Foreground maps, Planck Collaboration, 2016, A&A, 594, A10.
8. Component separation methods for the PLANCK mission, S. M. Leach, J.-F. Cardoso, C. Baccigalupi, R. B. Barreiro, M. Betoule, J. Bobin, A. Bonaldi, J. Delabrouille, G. de Zotti, C. Dickinson, H. K. Eriksen, J. González-Nuevo, F. K. Hansen, D. Herranz, M. Le Jeune, M. López-Caniego, E. Martínez-González, M. Massardi, J.-B. Melin, M.-A. Miville-Deschênes, G. Patanchon, S. Prunet, S. Ricciardi, E. Salerno, J. L. Sanz, J.-L. Starck, F. Stivoli, V. Stolyarov, R. Stompor, P. Vielva, A&A, 491, 597-615, (2008).
9. Multiresolution internal template cleaning: an application to the Wilkinson Microwave Anisotropy Probe 7-yr polarization data, R. Fernández-Cobos, P. Vielva, R. B. Barreiro, E. Martínez-González, MNRAS, 420, 2162-2169, (2012).
10. Wilkinson Microwave Anisotropy Probe 7-yr constraints on fNL with a fast wavelet estimator, B. Casaponsa, R. B. Barreiro, A. Curto, E. Martínez-González, P. Vielva, MNRAS, 411, 2019-2025, (2011).
11. Planck 2013 results. XXIII. Isotropy and statistics of the CMB, Planck Collaboration, 2014, A&A, 571, A23.
12. Planck 2013 results. XIX. The integrated Sachs-Wolfe effect, Planck Collaboration, 2014, A&A, 571, A19.

Cosmic Microwave background

Flexible Image Transfer Specification

Sunyaev-Zel'dovich

Full-Width-at-Half-Maximum

(Hierarchical Equal Area isoLatitude Pixelation of a sphere, <ref name="Template:Gorski2005">HEALPix: A Framework for High-Resolution Discretization and Fast Analysis of Data Distributed on the Sphere, K. M. Górski, E. Hivon, A. J. Banday, B. D. Wandelt, F. K. Hansen, M. Reinecke, M. Bartelmann, ApJ, 622, 759-771, (2005).

(Planck) Low Frequency Instrument

(Planck) High Frequency Instrument

Planck Sky Model

Data Processing Center

Planck Legacy Archive