Skip to content

DESI DR1 Full Shape cosmology results

Overview

This repository contains cosmology chains and posterior maximization results for the DESI DR1 Full Shape cosmology results referenced in the paper DESI 2024 VII: Cosmological Constraints from the Full-Shape Modeling of Clustering Measurements, JCAP 2025, 07.

Data Access

Data URL: https://data.desi.lbl.gov/public/dr1/vac/dr1/full-shape-cosmo-params

NERSC access: 

/global/cfs/cdirs/desi/public/dr1/vac/dr1/full-shape-cosmo-params

Documentation

The primary directory contains three folders: cobaya/ contains MCMC chains and posterior samples obtained using the cobaya sampling code. iminuit/ contains posterior maximization results determined using the iminuit optimization code. combineharvesterflow contains chain data obtained with the CombineHarvesterFlow code to fit normalising flows to the DES 3x2pt and 6x2pt chains, re-weighting the DESI (FS+BAO) chains to compute the joint posteriors.

Within each of these directories, results for different cosmological model scenarios and combinations of datasets are arranged in subfolders named according to the [model]/[dataset] convention, where [model] and [dataset] can take different values as explained below.

Models

Model naming follows the Planck convention: the base \(\Lambda\)CDM cosmological model is denoted by the root base in the folder name, with additional free parameters – when present – indicated by suffixes attached to this root. The naming convention relating suffixes to models is as follows:

  • _w: A dark energy equation of state parameter \(w\) that is freely varied but restricted to be constant in time
  • _w_wa: A variable dark energy equation of state with two free parameters, \(w_0\) and \(w_a\)
  • _mnu: Free sum of neutrino masses, \(\sum m_\nu\), with only a non-negativity constraint, \(\sum m_\nu>0\)
  • _mnu059: Free sum of neutrino masses, \(\sum m_\nu\), with the additional prior that \(\sum m_\nu>0.059\) eV
  • _mnu100: Free sum of neutrino masses, \(\sum m_\nu\), with the additional prior that \(\sum m_\nu>0.1\) eV (Note: in all the above cases the neutrinos are treated as 3 degenerate mass eigenstates)
  • _nnu: Free effective number of relativistic degrees of freedom, \(N_\mathrm{eff}\)
  • _mu_sigma: Modified gravity parameters \(\sigma_0\) and \(\mu_0\), fitted using the isitgr code.

Suffixes can also be chained together to represent models with more than one free parameter, e.g. the folder base_mnu_w_wa/ contains results for the model where base \(\Lambda\)CDM is extended by allowing free sum of neutrino masses and a time-varying dark energy equation of state.

Datasets

Different datasets or likelihoods are indicated by strings present in the folder names, according to the dictionary below. For combinations of several datasets, the folder names are formed by chaining together building blocks representing different individual datasets, separated by _.

BAO likelihoods

In all cases above, the Full Shape and BAO data is often combined (e.g. desi-reptvelocileptors-fs-bao). The [X] placeholder specifies the subset of DESI data used and can take the following values:

  • bgs: BAO measured in Bright Galaxy Sample, at effective redshift \(z_\mathrm{eff}=0.295\)
  • lrgz0: BAO from the first Luminous Red Galaxy (LRG) redshift bin, at \(z_\mathrm{eff}=0.510\)
  • lrgz1: BAO from the second LRG redshift bin, at \(z_\mathrm{eff}=0.706\)
  • lrg: the combination of the above two redshift bins (lrgz0 and lrgz1)
  • lrgpluselg: BAO from the combined LRG and Emission Line Galaxy (ELG) sample at \(z_\mathrm{eff}=0.930\)
  • elg: BAO from higher redshift ELGs, at \(z_\mathrm{eff}=1.317\)
  • qso: BAO from the quasar sample (QSO), at \(z_\mathrm{eff}=1.491\)
  • lya: BAO from the Lyman-\(\alpha\) forest (and cross-correlation with quasars) at \(z_\mathrm{eff}=2.330\)
  • all: the combination of DESI Full Shape/BAO data in all the above redshift bins
  • all-nolya: Same as above, but excluding data from the Lyman-alpha forest

Cosmic microwave background (CMB) likelihoods

  • planck2018-lowl-TT-clik: CMB low-\(\ell\) (\(\ell\le30\)) TT Commander likelihood from Planck PR3
  • planck2018-lowl-EE-clik: CMB low-\(\ell\) EE simall likelihood from Planck PR3
  • planck2018-highl-plik-TTTEEE: Planck PR3 plik likelihood for TT, TE and EE at \(\ell>30\)
  • planck-NPIPE-highl-CamSpec-TTTEEE: alternative CamSpec CMB likelihood for \(\ell>30\) TT, TE and EE from Planck PR4
  • planck2020-hillipop-TTTEEE: alternative hillipop CMB likelihood for \(\ell>30\) TT, TE and EE from Planck PR4
  • planck2020-lollipop-lowlE: alternative lollipop CMB likelihood for \(\ell\le30\) TE and EE from Planck PR4
  • planck-act-dr6-lensing: CMB lensing data from the combination of Planck PR4 NPIPE and ACT DR6 (as provided by the ACT team)

The default combination of CMB likelihoods used in the DESI 2024 VII paper, and denoted as “CMB” therein, is planck2018-lowl-TT-clik_planck2018-lowl-EE-clik_planck2018-highl-plik-TTTEEE_planck-act-dr6-lensing.

Type Ia supernova likelihoods

  • pantheonplus: SNIa data from the Pantheon+ dataset
  • union3: SNIa data from the Union3 dataset
  • desy5sn: SNIa data from the Dark Energy Survey 5-year SN dataset

Weak lensing likelihoods

  • desy3joint: DESY3 3x2pt likelihood, tailored to modified-gravity (MG) analysis with similar conservative scale cuts as in DESY3 MG study
  • desy3-6x2pt: DESY3 6x2pt likelihood, including CMB lensing (Planck and SPT)

Additional external priors

  • schoneberg2024-bbn: A BBN prior on the baryon density \(\Omega_\mathrm{b}h^2\)
  • planck2018-ns10: weak Gaussian prior on the spectral index \(n_s\), centered at the Planck mean value \(n_s = 0.9649\) and with a width, $\sigma(n_s) = 0.042, chosen to be 10 times wider than the posterior width obtained from the Planck 2020 TTTEEE+lensing results.
  • planck2018-ns: Same as above, but with a Gaussian width that has the same precision as the Planck measurement.

These individual likelihoods and priors can be combined together in too many ways to exhaustively list here. Combinations can be identified by chaining together the appropriate roots using _ separators.

Data Model

The individual cobaya/[model]/[dataset]/ directories contain cobaya output files as described in full here (see also here for additional information). The output files in the folders are:

  • [prefix].[number].txt: ASCII formatted tables providing posterior samples, where [number] indicates the chain number from 1 to 4; [prefix] is chain for standard runs or chain.post.importance when any likelihoods have been added by importance sampling in post-processing
  • [prefix].input.yaml or [prefix].updated.yaml: YAML file recording the input settings provided for the MCMC run; the filename contains updated when the any component uses default settings not explicitly provided in the input, which are then recorded in this file.
  • chain.covmat: ASCII-formatted covariance matrix for sampled cosmological parameters, which can be used as a proposal pdf for new runs using the same settings (not present for post-processed runs)
  • chain.checkpoint: text file with summary information about the convergence of the MCMC run (not present for post-processed runs)

Chain data can easily be read and manipulated using the GetDist package, but can also be accessed by other commonly used chain-processing software.

The individual iminuit/[model]/[dataset]/ directories contain output files described in full here. The relevant files are:

  • bestfit.minimum.txt: ASCII-formatted file providing the parameter values and goodness-of-fit criteria at the maximum of the posterior (MAP)
  • bestfit.minimum: identical information to the file above, but recorded in GetDist-compatible formatting
  • bestfit.bestfit.txt: as for bestfit.minimum.txt, but recording the maximum likelihood position instead of the MAP (not present in all cases)
  • bestfit.bestfit: identical to bestfit.bestfit.txt but in GetDist-compatible formatting (not present in all cases)

The individual combineharvesterflow directories contain samples for joint DES 6x2pt and other likelihoods (DESI and CMB), obtained with CombineHarvesterFlow. Samples are exported in the standard GetDist format:

  • chain.txt: ASCII formatted tables providing posterior samples
  • chain.paramnames: text file containing the corresponding parameter names

Contact

Contact Enrique Paillas and Arnaud de Mattia for questions about this catalog.