CMB from Planck 2018

Family of Planck CMB likelihoods. Contains interfaces to the official clik code and some native ones. You can find a description of the different likelihoods in the Planck wiki.

The Planck 2018 likelihoods defined here are: (new in 2.0)

  • planck_2018_lowl.TT: low-\(\ell\) temperature
  • planck_2018_lowl.EE: low-\(\ell\) EE polarization
  • planck_2018_highl_plik.[TT|TTTEEE]: plikHM high-\(\ell\) temperature|temperature+polarization
  • planck_2018_highl_plik.[TT|TTTEEE]_unbinned: unbinned versions of the previous ones
  • planck_2018_highl_plik.[TT|TTTEEE]_lite: faster nuisance-marginalized versions of the previous (binned) ones
  • planck_2018_highl_plik.[TT|TTTEEE]_lite_native: Python native versions of the nuisance-marginalizes ones (more customizable)
  • planck_2018_highl_CamSpec.[TT|TTTEEE][_native]: clik and native Python versions of the alternative high-\(\ell\) CamSpec likelihoods.
  • planck_2018_lensing.clik: lensing temperature+polarisation-based; official clik code.
  • planck_2018_lensing.native: lensing temperature+polarisation-based; native Python version (more customizable)
  • planck_2018_lensing.CMBMarged: CMB-marginalized, temperature+polarisation-based lensing likelihood; native Python version (more customizable). Do not combine with any of the ones above!


If you use any of these likelihoods, please cite them as:
Planck Collaboration, Planck 2018 results. V. CMB power spectra and likelihoods (arXiv:1907.12875)
Planck Collaboration, Planck 2018 results. VIII. Gravitational lensing (arXiv:1807.06210)

The Planck 2015 likelihoods defined here are:

  • planck_2015_lowl
  • planck_2015_lowTEB
  • planck_2015_plikHM_TT
  • planck_2015_plikHM_TT_unbinned
  • planck_2015_plikHM_TTTEEE
  • planck_2015_plikHM_TTTEEE_unbinned
  • planck_2015_lensing
  • planck_2015_lensing_cmblikes (a native non-clik, more customizable version of the previous clik-wrapped one)


If you use any of these likelihoods, please cite them as:
2015: N. Aghanim et al, Planck 2015 results. XI. CMB power spectra, likelihoods, and robustness of parameters (arXiv:1507.02704)


The Planck 2015 likelihoods have been superseded by the 2018 release, and will eventually be deprecated.


Some likelihoods cannot be instantiated more than once, or some particular two at the same time. This should have no consequences when calling cobaya-run from the shell, but will impede running a sampler or defining a model more than once per Python interpreter session.


To use any of the Planck likelihoods, you simply need to mention them in the likelihood block, or add them using the input generator.

The corresponding nuisance parameters will be added automatically, so you don’t have to care about listing them in the params block.

The nuisance parameters and their default priors can be obtained as explained in Getting help and bibliography for a component.


This likelihoods can be installed automatically as explained in Installing cosmological codes and data. If you are following the instructions there (you should!), you don’t need to read the rest of this section.


By default, the gfortran compiler will be used, and the cfitsio library will be downloaded and compiled automatically.

If the installation fails, make sure that the packages liblapack3 and liblapack-dev are installed in the system (in Debian/Ubuntu, simply do sudo apt install liblapack3 liblapack-dev).

If that did not solve the issue, check out specific instructions for some systems in the file in the folder [packages_path]/code/planck/code/plc_3.0/plc-3.01.

If you want to re-compile the Planck likelihood to your liking (e.g. with MKL), simply go into the chosen external packages installation folder and re-run the python waf configure and python waf install with the desired options, substituting python by the Python of choice in your system.

However, if you wish to install it manually or have a previous installation already in your system, simply make sure that clik can be imported from anywhere, and give absolute paths for each clik file, e.g.:

    clik_file: /your/path/to/plc_3.0/low_l/commander/commander_dx12_v3_2_29.clik
    clik_file: /your/path/to/plc_3.0/hi_l/plik/plik_rd12_HM_v22b_TTTEEE.clik

Manual installation of the Planck 2018 likelihoods

Assuming you are installing all your likelihoods under /path/to/likelihoods:

$ cd /path/to/likelihoods
$ mkdir planck_2018
$ cd planck_2018
$ wget ""
$ tar xvjf "data-action?COSMOLOGY.COSMOLOGY_OID=151912"
$ rm "data-action?COSMOLOGY.COSMOLOGY_OID=151912"
$ cd code/plc_3.0/plc-3.01
$ python waf configure  # [options]

If the last step failed, try adding the option --install_all_deps. It it doesn’t solve it, follow the instructions in the file in the plc_3.0 folder.

If you have Intel’s compiler and Math Kernel Library (MKL), you may want to also add the option --lapack_mkl=${MKLROOT} in the last line to make use of it.

If python waf configure ended successfully run python waf install in the same folder. You do not need to run, as advised.

Now, download the required likelihood files from the Planck Legacy Archive (Europe) or the NASA/IPAC Archive (US, outdated!).

For instance, if you want to reproduce the baseline Planck 2018 results, download the file COM_Likelihood_Data-baseline_R3.00.tar.gz from any of the two links above, and decompress it under the planck_2018 folder that you created above.

Finally, download and decompress in the planck_2018 folder the last release at this repo.

Interface for official clik code

Synopsis:Definition of the clik-based likelihoods
Author:Jesus Torrado (initially based on MontePython’s version by Julien Lesgourgues and Benjamin Audren)

Native CamSpec version

Synopsis:Definition of python-native CamSpec 2018 likelihood (not official Planck product)
Author:Antony Lewis (from CamSpec f90 source by GPE, StG and AL)

This python version loads the covariance, and cuts it as requested (then inverting). It can take a few seconds the first time it is loaded, but the inverse can be cached. The Planck likelihood code (clik) is not required.

Use dataset_params : { ‘use_cl’: ‘100x100 143x143 217x217 143x217’} to use e.g. just TT , or other combination with TE and EE.

Set use_range to string representation of L range to use, e.g. 50-100, 200-500, 1470-2500, or pass a dictionary of ranges for each spectrum.

##TODO: calPlanck vs Aplanck

Native lite version

Synopsis:Definition of python-native nuisance-free CMB likelihoods: e.g. plik_lite
Author:Erminia Calabrese, Antony Lewis

Nuisance-marginalized likelihood, based on covarianced and binned CL, with settings read from .dataset file.