C2

From Planck PLA 2015 Wiki
Revision as of 15:45, 19 October 2012 by Fbouchet (talk | contribs)
Jump to: navigation, search

The main product to express cosmological constraints posed by Planck data is a Likelihood code. One of its by-product is the derived Maximum-likelihood CMB spectrum.

NB: This is very much a scientific product, whose development status will be described in a dedicated talk at the review.

Principles used to build Planck Likelihood[edit]

Approach taken:

  • Use of an hybrid approach for the likelihood
  • Pixel-based at low-ell
  • Fine-grained pseudo-CL at high-ell using independent groupings of detectors at 100, 143 and 217GHz

This allows easy “Planck-Ext” (stitching of high(er)-ell likelihoods, in particular ACT & SPT The Planck code, CLIK, provides a unique interface (and it allows internally to keep cross-checking with various approaches), at V4.1 now (V1 was mid-2011).

We derived from CLIK our “best” CTT, together with a basic (diagonal) error estimate, which will allow making plots & quick model checks, comparison with various sources of uncertainties comparison with the CTT derived from the CMB map(s) issued from Component Separation activities…



CLIK - The Planck C(l) LIKelihood code[edit]

[-Last modified 2012-10-10 -]

@@clik@@ is a package with both a code and data files to distribute the current planck likelihood. It is available from ESA through the [[1]] in under operational files, then clicking the supplementary data icon on the toolbar: Attach:supdaticon.png.

[++%bgcolor=#fdf center %the latest package version is 4.0%%++]

Here follows a description of the different releases of @@clik@@.



!! Version 4.0

Main improvements for this release are the addition of

  • CAMspec_v5.2dust hi-ell likelihood,
  • plik_dx9_v2 hi-ell likelihood and
  • an early october version of the actspt likelihood.

Partial compatibility between those likelihood has been implemented : all three share the same cib modeling and either CAMspec or plik can be this way used with actspt. CAMspec further use the same sz and szXcib modeling. plik neglect the sz and szXcib foregrounds for now (as they approximatively cancel in planck range of ells).

No update has been made on the low-ell likelihood. To recover the files (still compatible with the last version of the code) one has to retrieve clik v3.5 and use the file contained therein.

Beware that the previous CAMspec files will not work with the current release.

!!! Code

@@clik@@ code is at version 4.0b2

Known issues

  • none for now.

Changes:

  • from 3.5
 ** inclusion of actspt
 ** removal of dependency on gel
 ** removal/cleanup of lot's of legacy code
 ** better discovery of the hdf5 version (in particular version 1.8.X compiled in pure 1.6.X compatibility mode)
 ** better detection of some weird python install
 ** plik likelihood code has been improved and is now 20times faster
 ** correction for the C++ boilerplate
 ** correction of the clik_config scripts with the addition of laking libs

!!! Data

  • hi-ell
 ** plik_dx9_v2.clik (see TOBEWRITTEN)
 ** CAMspec_5.2dust.clik (see TOBEWRITTEN)
  • hi-ell non planck
 ** actspt_2012_10_08.clik (see TOBEWRITTEN)

Changes:

  • CAMspec has changed a lot and now include beam parameters
  • actspt is new
  • new easily modifiable foreground modeling module for plik.

An important note for actspt

As for the low-ell likelihood, actspt code relies on data being installed in a specific place. The file is usable as is, but will dump some data in /tmp at each initialization. If this is not a desirable behaviour, data can be dumped once and for all by following the procedure described here.


[++%bgcolor=#fdf center %Older package versions%%++]

!! Version 3.5

Main improvements for this release is the addition of CAMspec_v4.1 hi-ell likelihood. The lowlike likelihood is version is bumped to v1.1. Changes are only in the size of the clik file where unused data have been removed.

Beware that the previous CAMspec file will not work with the current release.

!!! Code

@@clik@@ code is at version 3.0

Known issues

  • plik likelihood is buggy on MacosX 10.7 when compiled with @@ifort 12.1.3@@. Gfortran is advised on macosx for now. Recommended configuration options for macosX is

%block bgcolor=#f0f9ff border='1px solid gray' padding=5px%[@ $> ./waf configure --install_all_deps --lapack_apple --gfortran @]

  • automatic installation of hdf5 can fail on some computers with old versions of gnu make, where make cannot determine the optimal number of threads to optimize hdf5 compilation.

Changes:

  • from 2.0
 ** inclusion of CAMspec
 ** correction of a regression on lapack install
 ** correction of a regression for external likelihoods
 ** addition of C++ boilerplate to @@clik.h@@
 ** correction of regressions in CAMspec and BOPIX when compiling with gfortran
 ** removal of unused variables in lowlike wrapper

!!! Data

  • hi-ell
 ** plik_v2.0.clik (see [here])
 ** CAMspec_4.1.clik (see [here])
  • low-ell
 ** PLow_v1.0.clik (see [here])
 ** lowlike_v1.1.clik (see [here])

Changes:

  • adding of CAMspec likelihood Beware that the previous CAMspec file will not work with the current release.
  • removal of unused data in lowlike, resulting in a smaller data file.

An important note for the two low-ell likelihoods

The bug described in release v3.0 has been solved. Nevertheless, one is strongly advised to follow the procedure described here in order to accelerate the initialization of the low-ell likelihoods.

Exemple of output[edit]

We'll give here an actual exemple of use and result of an exploration

Current Best-fit CMB spectrum[edit]

This is still evolving

Cosmic Microwave background

European Space Agency