CosmoMC

CosmoMC
NASA / WMAP Science Team · Public domain · source
Name	CosmoMC
Developer	Anthony Lewis and collaborators
Released	2002
Programming language	Fortran 90, MPI
Operating system	Unix-like, macOS
Genre	Scientific software, Bayesian inference, Monte Carlo
License	Open-source (various distributions)

CosmoMC CosmoMC is a software package for Bayesian parameter estimation and model comparison used in cosmology. It performs Markov Chain Monte Carlo (MCMC) sampling to constrain parameters of cosmological models using observational datasets and theoretical predictions. The code has been widely employed by researchers working on cosmic microwave background analyses, large-scale structure, and dark energy investigations, and interfaces with spectrum calculators and likelihood modules from major experiments.

Overview

CosmoMC was developed to perform high-dimensional Bayesian inference in cosmological parameter spaces and to analyze data from experiments such as Wilkinson Microwave Anisotropy Probe, Planck, Atacama Cosmology Telescope, and South Pole Telescope. It connects theoretical predictions from Boltzmann solvers to observational likelihoods produced by collaborations like WMAP team and Planck Collaboration and has been cited alongside works by authors affiliated with institutions such as University of Cambridge, University of London, and University of Oxford. The package integrates with external libraries and tools common in astrophysics research, and its outputs are used in publications appearing in journals like Physical Review D, Monthly Notices of the Royal Astronomical Society, and The Astrophysical Journal.

Features and Capabilities

CosmoMC implements advanced MCMC techniques, including parallel tempering and fast sampling strategies, to explore parameter spaces for models such as ΛCDM, wCDM, and extensions with massive neutrinos or tensor modes studied in works by groups at CERN, Fermilab, and SLAC National Accelerator Laboratory. It supports likelihoods for datasets from Baryon Oscillation Spectroscopic Survey, Sloan Digital Sky Survey, Dark Energy Survey, and supernova compilations like those by the Supernova Cosmology Project and High-Z Supernova Search Team. The code can marginalize over nuisance parameters for instrument calibration and foregrounds as addressed by teams at Max Planck Institute for Astrophysics and Jet Propulsion Laboratory. CosmoMC offers covariance estimation, importance sampling, and posterior visualization compatible with analysis tools developed by groups at Harvard University, Princeton University, and Stanford University.

Methodology and Algorithms

At its core CosmoMC employs Metropolis-Hastings MCMC with proposals tuned by adaptive schemes and uses Gelman–Rubin diagnostics for convergence popularized in statistical contexts linked to researchers at Columbia University and University of California, Berkeley. It integrates likelihood evaluation pipelines that call Boltzmann solvers such as CAMB and interfaces older or parallelized implementations influenced by work at Institute for Advanced Study and Kavli Institute for Cosmological Physics. Techniques for sampling high-dimensional posterior surfaces trace conceptual lineage to algorithms discussed in literature from Alan Turing Institute affiliates and mathematicians associated with Imperial College London. CosmoMC supports parameter transformation, principal component analysis approaches used in studies by Lawrence Berkeley National Laboratory, and uses importance reweighting approaches referenced in publications from Yale University.

Implementation and Performance

The package is implemented in Fortran 90 with Message Passing Interface (MPI) parallelization strategies used on clusters at facilities like National Energy Research Scientific Computing Center and NERSC. Performance optimizations exploit shared libraries and parallel I/O conventions practiced at European Organization for Nuclear Research computing centers and have been benchmarked on hardware provided by Cray installations and university HPC clusters. CosmoMC's runtime scales with the complexity of likelihood modules from collaborations such as Planck Collaboration and the dimensionality of parameter sets (e.g., models including parameters motivated by Neutrino Oscillation experiments). Users running large chains have reported efficient wall-clock times when using scheduling systems like SLURM or PBS (software) on supercomputing resources.

Usage and Interface

Users configure CosmoMC via plain-text parameter files and compile-time options; this workflow resembles practices in projects at Los Alamos National Laboratory and Argonne National Laboratory. The code accepts likelihood modules provided by collaborations including Planck Collaboration and BOSS and can output chains consumable by plotting and analysis utilities used in research groups at MIT, Caltech, and University of Chicago. Scripts for post-processing and plotting are often based on community tools developed by researchers at Institute of Theoretical Astronomy-style groups and are compatible with statistical packages produced by teams at Oxford University Press-adjacent research units. CosmoMC supports restartable runs, checkpointing, and chain thinning to accommodate batch systems used at Center for Computational Astrophysics.

Development and History

CosmoMC originated in the early 2000s under the stewardship of Anthony Lewis and collaborators associated with institutions like University of Sussex and Cambridge University. Over time the codebase incorporated likelihoods and features added by contributors from research groups at Max Planck Institute for Astrophysics, University of Geneva, and CEA Saclay. Major updates coincided with releases of datasets from WMAP and Planck, and development discussions have occurred within working groups tied to International Astronomical Union meetings and cosmology conferences at KICP and Perimeter Institute. The project has inspired derivative tools and influenced software practices in cosmological inference across many institutions.

Applications and Impact

CosmoMC has been used to constrain parameters such as the Hubble constant in analyses that reference results from Hubble Space Telescope observations and large-scale structure measurements from 2dF Galaxy Redshift Survey and SDSS. Its outputs have informed debates involving datasets from Riess et al. and analyses discussed at workshops at Kavli Institute for Particle Astrophysics and Cosmology, shaping discussions about dark energy, neutrino mass bounds relevant to Super-Kamiokande and IceCube, and inflationary constraints tied to experiments like BICEP2. The code's role in reproducible cosmological inference has made it a staple cited in research spanning collaborations at European Southern Observatory, National Astronomical Observatory of Japan, and numerous university consortia.

Category:Cosmology software