Simulation data

From Planck PLA 2015 Wiki
Revision as of 15:06, 19 June 2013 by Lvibert (talk | contribs) (Overall description)
Jump to: navigation, search


The 2013 Planck data release is supported by a comprehensive set of simulated maps, including both a fiducial realization of the sky as seen by Planck and 1000-realization Monte Carlo (MC) sets of CMB and noise simulations, collectively known as FFP6.

The simulation process consists of

  • modeling the sky using pre-Planck data and generating an input sky map for each sky component for each detector that incorporates our best estimate of that detector's band-pass
  • simulating each detector's observation of each input sky component, following the Planck scanning strategy and using our best estimates of the detector's beam and noise properties, and mapping the results.
  • generating Monte Carlo realizations of the CMB and noise maps, again following the Planck scanning strategy and using our best estimates of the detector beams and noise properties respectively.

The first step is done by the Planck Sky Model (PSM), and the second and third by a suite of Planck Simulation Tools (PST). A brief description of these is given below.

The Planck Sky Model[edit]


Overall description[edit]

The Planck Sky Model, a complete set data and code to simulate sky emission at millimeter-wave frequencies, is described in detail in the Planck pre-launch PSM paper Delabrouille2012.

The main simulations used to test and validate the Planck data analysis pipelines (and, in particular, component separation) makes use of simulations generated with version 1.7.7 of the PSM software. Sky emission comprises the following components: CMB, thermal dust, spinning dust, synchrotron, CO lines, free-free, thermal Sunyaev-Zel'dovich (SZ) effect (with first order relativistic corrections), kinetic SZ effect, radio and infrared sources, Cosmic Infrared Background (CIB).

The CMB is a based on adiabatic initial perturbations, with the following cosmological parameters:

  • T_CMB = 2.725
  • H = 0.684
  • OMEGA_M = 0.292
  • OMEGA_B = 0.04724
  • OMEGA_NU = 0
  • OMEGA_K = 0
  • SIGMA_8 = 0.789
  • N_S = 0.9732
  • N_S_RUNNING = 0
  • N_T = 0
  • R = 0.0844
  • TAU_REION = 0.085
  • HE_FRACTION = 0.245
  • N_MASSLESS_NU = 3.04
  • N_MASSIVE_NU = 0
  • W_DARK_ENERGY = -1
  • K_PIVOT = 0.002
  • SCALAR_AMPLITUDE = 2.441e-9

All other parameters are set to the default standard of the Jan 2012 version of CAMB. In addition, this simulated CMB contains non-Gaussian corrections of the local type, with an f_NL parameter of 20.4075.

The Galactic ISM emission comprises 5 major components: Thermal dust, spinning dust, synchrotron, CO lines, and free-free emission. We refer the reader to the PSM publication for details. For the simulations generated here, however, the thermal dust model has been modified in the following way: Instead of being based on the 100 micron map of Schlegel, Finkbeiner and Davis (SFD; 1998), the dust template uses an internal release of the 857 GHz Planck observed map itself, in which point sources have been subtracted, and which has been locally filtered to remove CIB fluctuations in the regions of lowest column density. A caveat is that while this reduces the level of CIB fluctuations in the dust map in some of the regions, in regions of moderate dust column density the CIB contamination is actually somewhat larger than in the SFD map (by reason of different emission laws for dust and CIB, and of the highr resolution of the Planck map).

The other emissions of the galactic ISM are simulated using the prescription described in the PSM paper. Synchrotron, Free-free and spinning dust emission are based on WMAP observations, as analyzed by Miville-Deschenes et al., 2008. Small scale fluctuations have been added to increase the variance on small scales and compensate the lower resolution of WMAP as compared to Planck (in particular HFI channels). The main limitation of these maps is the presence at high galactic latitude of fluctuations that may be imputable to WMAP noise. The presence of noise and of added Gaussian fluctuations on small scales may result in a few occasional pixels being negative (e.g. in the spinning dust maps). Low frequency foreground maps are also contaminated by some residuals of bright radio sources that have not been properly subtracted from the templates of diffuse emission.

The CO maps are simulated using the CO J=1-0 observations of Dame et al. (2001). The main limitations are limited sky coverage, lower resolution than that of Planck high frequency channels, line ratios (J=2-1)/(J=1-0) and (J=3-2)/(J=2-1) constant over the sky. The CO in the simulation is limited to the three lowest 12CO lines.

Galaxy clusters are generated on the basis of cluster number counts, following the Tinker et al. (2008) mass function, for the cosmological parameters listed above. Clusters are assumed perfectly spherical, isothermal, and are modeled using the universal pressure profile of Arnaud et al. (2010). Relativistic corrections following Itoh et al. (1998) are included to first order. The simulated kinetic SZ effect assumes no bulk flow, and a redshift-dependent average cluster velocity compatible with the linear growth of structures.

Point sources comprise radio sources (based on extrapolations across frequencies of radio observations between 800 MHz and 5 GHz) and infrared sources (based on extrapolations in frequencies of IRAS sources). One small caveat that should be mentioned is that because of the unevenness of the radio source surveys, the equatorial southern part of the sky has less faint radio sources than the northern part. Although all the missing sources are well below the Planck detection level, this induces a small variation of the total emission background over the sky. Check the individual faint point source emission maps if this is a potential problem for your applications. See also the PSM publication for details about the PSM point source simulations.

Finally, the far infrared background due to high redshift galaxies has been simulated using a procedure is based on the distribution of galaxies in shells of density contrast at various redshifts (Castex et al., PhD thesis; paper in preparation). This simulation has been modified by gradually substituting an uncorrelated extra term of CIB emission at low frequencies, artificially added in particular to decorrelate the CIB at frequencies below 217 GHz from the CIB above that frequency, to mimic the apparent decorrelation observed in the Planck Early Paper on CIB power spectrum.

While the PSM simulations described here provide a reasonably representative multi-component model of sky emission, the users are warned that it has been put together mostly on the basis of data sets and knowledge pre-existing the Planck observations themselves. While it is sophisticated enough to include variations of emission laws of major components of ISM emission, different emission laws for most sources, and a reasonably coherent global picture, it is not (and is not supposed to be) identical to the real sky emission. The users are warned to use these simulations with caution.

PSM Products[edit]

The following products are available:

  • PR1_beta/HFI_SimMap_cmb_2048_R1.10.fits
  • PR1_beta/HFI_SimMap_co_2048_R1.10.fits
  • PR1_beta/HFI_SimMap_firb_2048_R1.10.fits
  • PR1_beta/HFI_SimMap_strongps_2048_R1.10.fits
  • PR1_beta/HFI_SimMap_faintps_2048_R1.10.fits
  • PR1_beta/HFI_SimMap_freefree_2048_R1.10.fits
  • PR1_beta/HFI_SimMap_synchrotron_2048_R1.10.fits
  • PR1_beta/HFI_SimMap_thermaldust_2048_R1.10.fits
  • PR1_beta/HFI_SimMap_spindust_2048_R1.10.fits
  • PR1_beta/HFI_SimMap_kineticsz_2048_R1.10.fits
  • PR1_beta/HFI_SimMap_thermalsz_2048_R1.10.fits

All files contains a single BINTABLE extension with either a single map (e.g., for the CMB) or one map for each HFI frequency (for all other components). In the latter case the columns are named F100, F143, … , F857. Units are microKCMB for the CMB, and MJy/sr for the others. The structure is given below for multi-column files.

FITS file structure
1. EXTNAME = 'SIM-MAP' : Data columns
Column Name Data Type Units Description
F100 Real*4 MJy/sr 100GHz signal map
F143 Real*4 MJy/sr 143GHz signal map
F217 Real*4 MJy/sr 217GHz signal map
F353 Real*4 MJy/sr 353GHz signal map
F545 Real*4 MJy/sr 545GHz signal map
F857 Real*4 MJy/sr 857GHz signal map
Keyword Data Type Value Description
COMP string component Astrophysical omponent
COORDSYS string GALACTIC Coordinate system
ORDERING string NESTED Healpix ordering
NSIDE Int 2048 Healpix Nside for LFI and HFI, respectively
FIRSTPIX Int*4 0 First pixel number
LASTPIX Int*4 50331647 Last pixel number, for LFI and HFI, respectively
FREQ string GHz The frequency channel
BAD_DATA Real*4 -1.63750E+30 Healpix bad pixel value
BEAMTYPE string GAUSSIAN Type of beam
BEAMSIZE Real*4 size Beam size in arcmin
PSM-VERS string PSM Versions used

The Planck Mission Simulations[edit]

Since the full focal plane (FFP) simulations

  • involve both HFI and LFI data
  • include large, computationally challenging, MC realization sets

they cannot be generated using either DPC's single-instrument cluster-based pipeline. Instead the PST consists of three distinct tool chains, each designed to run on the largest available supercomputers, that are used to generate the fiducial sky realization, the CMB MC, and the noise MC respectively. FFP6 was primarily generated on the Hopper and Edison systems at NERSC, with some of the LFI noise MCs generated on the Louhi system at CSC.

While FFP6 is guaranteed to be internally self-consistent there are a number of differences with the real data that should be borne in mind, although all tests performed to date indicate that these are statistically insignificant:

  • FFP6 includes our best measurements of the detector band-passes, main beams and noise power spectral densities - known issues include the absence of side-lobes and the use of a single, independent, noise spectrum per detector for the entire mission.
  • FFP6 excludes all pre-processing residuals, assuming perfect calibration, transfer function deconvolution and deglitching.
  • FFP6 uses the HFI pointing solution for the LFI frequencies, rather than the DPC's two focal plane model.
  • FFP6 uses a different map-maker to HFI, and as a consequence implements very slightly different data cuts - primarily at ring boundaries - resulting in marginally different hit-maps.

Fiducial Sky[edit]

For each detector, fiducial sky time-ordered data are generated separately for each of its 10 PSM component maps and its strong point source catalogue using the LevelS software #reinecke2006 as follows:

  • The detector's beam and PSM map are converted to spherical harmonics using beam2alm and anafast respectively.
  • The beam-convolved map value is calculated over a 3-dimensional grid of sky locations and beam orientations using conviqt.
  • The map-based time-ordered data are calculated sample-by-sample by interpolating over this grid using "multimod".
  • The catalogue-based time-ordered data are calculated sample-by-sample by beam-convolving any point source laying within a given angular distance of the pointing at each sample time using multimod.

For each frequency, fiducial sky maps are generated for

  • the total sky (including noise),
  • the foreground sky (excluding CMB but including noise),
  • the point source sky, and
  • the noise alone

All maps are made using the MADAM destriping map-maker #[keihanen2010 interfaced with the TOAST data abstraction layer . In order to construct the total time-ordered data required by each map, for each detector TOAST reads the various component time-streams separately and sums then, and, where necessary, simulates and adds a noise realization time-stream on the fly.

HFI frequencies are mapped at HEALPix resolution nside=2048 using ring-length destriping baselines, while LFI frequencies are mapped at nside=1024 using 1s baselines.

CMB Monte Carlo[edit]

The CMB MC set is generated using FEBeCoP #mitra2010, which generates an effective beam for each pixel in a map at each frequency by accumulating the weights of all pixels within a fixed distance of that pixel, summed over all observations by all detectors at that frequency. It then applies this effective beam pixel-by-pixel to each of 1000 input CMB sky realizations.

Noise Monte Carlo[edit]

The noise MC set is generated just as the fiducial noise maps, using MADAM/TOAST. In order to avoid spurious correlations within and between the 1000 realizations, each stationary interval for each detector for each realization is generated from a distinct sub-sequence of a single statistically robust, extremely long period, pseudo-random number sequence.

Products delivered[edit]

A single simulation is delivered, which is divided into two types of products:

Six sky maps of the nominal mission at each HFI frequency
which is the basic simulation product. These files have the same structure as the equivalent SkyMap products described in the Frequency Maps chapter -- namely one BINTABLE extension with three columns containing 1) Signal, 2) hit-count, and 3) variance (TBC)
Three maps of the nominal mission containing 1) the sum of all astrophysical foregrounds, 2) the point sources alone, and 3) the noise alone
which are subproducts of the above, and are in the form of the PSM maps described in the previous section.

Note that the cmb alone is not delivered as a separate product, but it can be recovered by simple subtraction of the component maps for the total signal map. It is envisaged that some number of separate noise realizations may be delivered in the future.


<biblio force=false>

  1. References


Cosmic Microwave background

Planck Sky Model


(Planck) High Frequency Instrument

Flexible Image Transfer Specification

(Planck) Low Frequency Instrument

Data Processing Center

[LFI meaning]: absolute calibration refers to the 0th order calibration for each channel, 1 single number, while the relative calibration refers to the component of the calibration that varies pointing period by pointing period.

(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).

To be confirmed