View source for Cosmology@Home

{{Infobox software
| name                 = Cosmology@Home
| logo                 = Cosmology@Home.jpg
| logo caption         = Cosmology@Home logo
| screenshot           = Cosmology@Home.png
| caption              = 
| description          = Cosmology@Home is a volunteer computing project on the BOINC platform that searches for cosmological models best fitting observational data, developed at the University of Illinois at Urbana-Champaign and later hosted at the Institut d'Astrophysique de Paris.

| status               = Inactive
| category             = Cosmology / Astrophysics
| compute              = CPU
| dependencies         = VirtualBox 4.3 or later (for camb_boinc2docker app); VT-x or AMD-v

| developer            = Marius Millea, Benjamin D. Wandelt
| author               = Marius Millea, Benjamin D. Wandelt
| sponsor              = University of Illinois at Urbana-Champaign; Institut Lagrange de Paris; Institut d'Astrophysique de Paris; NASA; Planck project
| released             = {{Start date and age|2007|06|26}}
| discontinued         = {{Start date and age|2024|01|10}}

| programming language = C, C++ (CAMB); Python (server tooling)
| operating system     = Windows XP or later, macOS 10.8 or later, Linux (64-bit)

| stats as of          = {{Start date and age|2017|06|30}}
| average performance  = 17494.51 GigaFLOPS
| active users         = 5086
| total users          = 70212
| active hosts         = 11612
| total hosts          = 170377

| rac                  = 
| credit per day       = 
| gpu performance      = 
| cpu performance      = 

| website              = {{URL|https://www.cosmologyathome.org/}}
}}

'''Cosmology@Home''' is a [[volunteer computing]] project that runs on the [[Berkeley Open Infrastructure for Network Computing]] (BOINC) platform. Its central goal is to compare theoretical models of the [[universe]] with observational data, and to identify the range of cosmological parameters that best fits measurements from instruments such as the [[Wilkinson Microwave Anisotropy Probe]] (WMAP), the [[Hubble Space Telescope]], and the [[Planck (spacecraft)|Planck satellite]].<ref name="wandelt_letter">{{cite web |url=https://www.cosmologyathome.org/wandelt_letter.php |title=Letter to Cosmology@Home users |publisher=Cosmologyathome.org |access-date=2011-02-20 |archive-url=https://web.archive.org/web/20120226155135/http://www.cosmologyathome.org/wandelt_letter.php |archive-date=2012-02-26}}</ref>

The project was originally hosted at the Departments of Astronomy and Physics at the [[University of Illinois at Urbana-Champaign]] (UIUC). In December 2016 it was transferred to the [[Institut Lagrange de Paris]] and the [[Institut d'Astrophysique de Paris]], both located at what was then known as [[Pierre and Marie Curie University]] in France.<ref name="WandeltLetter2016">{{cite web |url=https://www.cosmologyathome.org/wandelt_letter.php |title=Welcome Letter |last=Wandelt |first=Ben |publisher=Cosmologyathome.org |archive-url=https://web.archive.org/web/20161220130501/https://www.cosmologyathome.org/wandelt_letter.php |archive-date=2016-12-20 |access-date=2016-12-15}}</ref><ref name="MoveCompleted">{{cite web |url=https://www.cosmologyathome.org/forum_thread.php?id=7442 |title=Move Completed |last=Millea |first=Marius |date=2016-12-15 |publisher=Cosmologyathome.org |archive-url=https://web.archive.org/web/20161220130817/https://www.cosmologyathome.org/forum_thread.php?id=7442 |archive-date=2016-12-20 |access-date=2016-12-15}}</ref>

[[File:Cosmic Microwave Background (CMB).jpeg|thumb|right|300px|Temperature map of the [[cosmic microwave background]] (CMB) as measured by the [[Planck (spacecraft)|Planck satellite]]. Cosmology@Home used volunteer computers to generate theoretical CMB power spectra used in training the PICO algorithm, which in turn was applied to precisely this kind of data.]]

== Background and motivation ==

The discipline of [[physical cosmology]] seeks to determine the values of the fundamental parameters that govern the large-scale structure and evolution of the [[universe]]. These include quantities such as the [[Hubble constant]] <math>H_0</math>, the total matter density <math>\Omega_m</math>, the baryon density <math>\Omega_b h^2</math>, the dark matter density <math>\Omega_{\rm CDM} h^2</math>, the [[spectral index]] of primordial density fluctuations <math>n_s</math>, and the amplitude of those fluctuations <math>A_s</math>. Together these parameters define what is known as the standard [[Lambda-CDM model|Lambda cold dark matter]] (LCDM) model of cosmology.

A central observational probe for constraining these parameters is the [[cosmic microwave background]] (CMB), the relic thermal radiation left over from roughly 380,000 years after the [[Big Bang]]. The tiny temperature anisotropies across the CMB sky, characterised by the angular power spectrum <math>C_\ell</math>, encode information about the physical conditions of the early universe and the values of these parameters.<ref name="cmb_background">{{cite web |url=https://www.cfa.harvard.edu/research/topic/cosmic-microwave-background |title=Cosmic Microwave Background |publisher=Center for Astrophysics, Harvard and Smithsonian |access-date=2026-06-06}}</ref> Computing the predicted power spectrum for a given set of parameters requires running sophisticated Boltzmann codes such as [[CAMB]] (Code for Anisotropies in the Microwave Background), which can take many hours per model evaluation. To explore a realistic parameter space involving hundreds of thousands of candidate models, this demands an enormous amount of computational power - the exact problem Cosmology@Home was designed to solve.

== History ==

Cosmology@Home was created by Marius Millea, then a graduate student astrophysicist at UIUC, along with professor Benjamin D. Wandelt.<ref name="anderson_history">{{cite web |url=https://continuum-hypothesis.com/boinc_history.php |title=A brief history of BOINC |last=Anderson |first=David P. |access-date=2026-06-06}}</ref> The project launched for closed alpha testing on 30 June 2007, opened public alpha registration on 23 August 2007, and entered beta testing on 5 November 2007.<ref name="betathread">{{cite web |url=http://www.cosmologyathome.org/forum_thread.php?id=245 |title=Beta testing! |publisher=Cosmologyathome.org |archive-url=https://web.archive.org/web/20110725204627/http://www.cosmologyathome.org/forum_thread.php?id=245 |archive-date=2011-07-25 |access-date=2011-02-20}}</ref>

Early work units were straightforward CAMB runs on Windows and Linux. Over time Millea introduced a Dockerised application named '''camb_boinc2docker''' - leveraging VirtualBox-based virtualisation via the BOINC vboxwrapper - which allowed the science application to be shipped as a portable Docker container, greatly simplifying deployment and eliminating the need to compile separate native binaries for each supported platform.<ref name="faq">{{cite web |url=https://www.cosmologyathome.org/faq.php |title=FAQ/Troubleshooting |publisher=Cosmologyathome.org |access-date=2022-10-05}}</ref> This innovation also gave birth to the open-source tool '''boinc-server-docker''', which was originally developed for Cosmology@Home and has since been adopted as a general framework for hosting BOINC projects using Docker containers.<ref name="boinc_server_docker">{{cite web |url=https://github.com/marius311/cosmohome |title=The Cosmology@Home server (GitHub: marius311/cosmohome) |last=Millea |first=Marius |access-date=2026-06-06}}</ref>

In August 2017 a domain registration lapse following the server migration from UIUC to Paris caused the cosmologyathome.org domain to expire briefly; Millea posted updates on the BOINC forums under the temporary URL <code>cosmos.iap.fr</code> while the domain was recovered.<ref name="domainlapse">{{cite web |url=https://boinc.berkeley.edu/forum_thread.php?id=11777 |title=cosmologyathome.org domain temporarily down |publisher=BOINC message boards |date=2017-08-16 |access-date=2026-06-06}}</ref>

By the early 2020s the project entered a prolonged period of low activity. SSL certificate renewals lapsed repeatedly between 2022 and 2024, preventing users from logging in or downloading new work units.<ref name="certexpired">{{cite web |url=https://boinc.berkeley.edu/forum_thread.php?id=14836 |title=Cosmology's certificate has expired again |publisher=BOINC message boards |date=2022-11-09 |access-date=2026-06-06}}</ref> The project was eventually removed from the official BOINC active project list after the project administrator could not be reached.<ref name="boincsynergy_blog">{{cite web |url=https://boincsynergy.ca/blog/cosmologyhome/ |title=Cosmology@Home |publisher=BOINC Synergy |access-date=2026-06-06}}</ref> BOINCstats retired the project from tracking in October 2024.<ref name="boincstats_retired">{{cite web |url=https://www.boincstats.com/forum/10/13063,1 |title=Time to retire Cosmology@home |publisher=BOINCstats/BAM! |date=2024-10-13 |access-date=2026-06-06}}</ref>

== Science ==

=== Goal ===

The core scientific goal of Cosmology@Home is to find the set of cosmological parameters that best describes the observable universe, by comparing theoretical predictions with actual measurements. The project specifically targets the comparison of theoretical models against:

* Fluctuations in the CMB as measured by WMAP and the Planck spacecraft
* The universe's expansion rate as measured by the Hubble Space Telescope
* Large-scale structure data from galaxy surveys

Results from the project can also help scientists design future observational experiments and prepare analysis pipelines for upcoming telescopes and satellites.<ref name="wandelt_letter" />

=== Method: sampling parameter space with CAMB ===

For any given class of cosmological models, Cosmology@Home generates tens of thousands of sample "universes", each defined by a unique combination of cosmological parameters. Each sample is sent to a volunteer's computer as a work unit. The volunteer's machine runs the [[CAMB]] code to compute the CMB [[power spectrum]] for that parameter combination - a calculation that can take several hours on a standard processor. The result, a tabulated list of angular power spectrum values <math>C_\ell</math> as a function of multipole order <math>\ell</math>, is returned to the project server.<ref name="faq" />

=== PICO: machine learning from volunteer results ===

Once a sufficiently large training set of spectra has been collected from volunteers, the project uses a [[machine learning]] algorithm called '''PICO''' (Parameters for the Impatient Cosmologist), which was developed specifically by Fendt and Wandelt at UIUC for this purpose.<ref name="pico_apj">{{cite journal |last1=Fendt |first1=William A. |last2=Wandelt |first2=Benjamin D. |year=2007 |title=PICO: Parameters for the impatient cosmologist |journal=[[The Astrophysical Journal]] |volume=654 |issue=1 |pages=2-11 |arxiv=astro-ph/0606709 |bibcode=2007ApJ...654....2F |doi=10.1086/508342 |s2cid=16572972}}</ref> PICO learns to emulate the full CAMB computation by fitting multivariate polynomial functions to the training set. Once trained on approximately 20,000 examples (a process that takes about 30 minutes), PICO can replicate CAMB's output for any similar universe in just a few milliseconds per evaluation, rather than the hours required by CAMB itself.<ref name="faq" />

This emulation approach allows cosmologists to run a full Markov Chain Monte Carlo (MCMC) parameter estimation - involving hundreds of thousands of model evaluations - in a few hours on a single CPU, rather than the months that would be required using CAMB directly for each step.

[[File:WMAP 2010.png|thumb|right|250px|All-sky map of the CMB from [[WMAP]]'s seven-year data release. Measurements from WMAP were a primary observational dataset used in training and validating the PICO algorithm developed through Cosmology@Home.]]

A second, higher-accuracy paper extended the PICO algorithm to more general cosmological models and described the training methodology in detail, specifically demonstrating that PICO can be trained using the geographically distributed pool of hosts contributing to Cosmology@Home.<ref name="pico_arxiv">{{cite journal |last1=Fendt |first1=William A. |last2=Wandelt |first2=Benjamin D. |year=2007 |title=Computing High Accuracy Power Spectra with Pico |arxiv=0712.0194 |bibcode=2007arXiv0712.0194F}}</ref>

PICO was subsequently adopted widely by the broader cosmology community. It was notably used in the analysis of data from the [[Planck (spacecraft)|Planck satellite]]'s CMB power spectra likelihood tests,<ref name="planck2015">{{cite journal |collaboration=Planck Collaboration |year=2016 |title=Planck 2015 results. XI. CMB power spectra, likelihoods, and robustness of parameters |journal=Astronomy and Astrophysics |volume=594 |pages=A11 |arxiv=1507.02704 |doi=10.1051/0004-6361/201526926}}</ref> and in the analysis of South Pole Telescope CMB damping tail measurements.<ref name="spt_pico">{{cite journal |last1=Story |first1=K. T. |display-authors=etal |year=2013 |title=A Measurement of the Cosmic Microwave Background Damping Tail from the 2500 sq. deg. SPT-SZ Survey |journal=The Astrophysical Journal |volume=779 |pages=86 |arxiv=1210.7231 |doi=10.1088/0004-637X/779/1/86}}</ref>

=== Applications: the camb_boinc2docker app ===

Cosmology@Home ran two applications for volunteers:

* '''camb_boinc2docker''' - the primary and preferred application, requiring a 64-bit processor, VirtualBox 4.3 or later, and hardware virtualisation (Intel VT-x or AMD-v) enabled in the BIOS. This app packages CAMB in a Docker container and supports Windows XP or later, macOS 10.8 or later, and many recent Linux distributions.
* '''camb_legacy''' - an older application available to volunteers whose machines do not support virtualisation. It runs only on Windows and Linux and supports 32-bit processors. Results from this app were primarily used for testing the PICO algorithm rather than for direct comparison with observational data.<ref name="faq" />

== Technical infrastructure ==

Cosmology@Home ran on the standard [[BOINC]] server software stack. Millea developed '''boinc2docker''', a package that allows any BOINC project to distribute and run Docker containers as work units, combining the capabilities of boot2docker and the BOINC vboxwrapper mechanism. This project was later generalised into the '''boinc-server-docker''' framework, which packages an entire BOINC project server as a set of Docker containers for easy deployment.<ref name="boinc2docker_github">{{cite web |url=https://github.com/marius311/boinc2docker |title=boinc2docker: Run Docker apps on a BOINC project |last=Millea |first=Marius |publisher=GitHub |access-date=2026-06-06}}</ref> The project's server source code has been made publicly available.<ref name="cosmohome_github">{{cite web |url=https://github.com/marius311/cosmohome |title=cosmohome: The Cosmology@Home server |last=Millea |first=Marius |publisher=GitHub |access-date=2026-06-06}}</ref>

The project received support from both NASA and the [[Planck (spacecraft)|Planck project]].<ref name="server_status">{{cite web |url=https://www.cosmologyathome.org/server_status.php |title=Project status |publisher=Cosmologyathome.org |access-date=2022-10-05}}</ref>

== Milestones ==

{| class="wikitable"
! Date !! Event
|-
| 2007-06-30 || Project launched for closed alpha testing (invitation only)
|-
| 2007-08-23 || Public alpha registration opened
|-
| 2007-11-05 || Project entered beta testing<ref name="betathread" />
|-
| 2016-12-15 || Server and project moved from UIUC to the Institut Lagrange de Paris and the Institut d'Astrophysique de Paris<ref name="WandeltLetter2016" /><ref name="MoveCompleted" />
|-
| 2017-08 || Temporary domain outage following university server migration; domain cosmologyathome.org briefly expired<ref name="domainlapse" />
|-
| 2022 || SSL certificate lapses began preventing user logins<ref name="certexpired" />
|-
| 2024 || Removed from BOINC active project list; project retired from BOINCstats tracking<ref name="boincstats_retired" />
|}

== Publications ==

The following papers describe the science and methods directly associated with Cosmology@Home:

* {{cite journal |last1=Fendt |first1=William A. |last2=Wandelt |first2=Benjamin D. |year=2007 |title=PICO: Parameters for the impatient cosmologist |journal=The Astrophysical Journal |volume=654 |issue=1 |pages=2-11 |arxiv=astro-ph/0606709 |bibcode=2007ApJ...654....2F |doi=10.1086/508342 |s2cid=16572972}}

* {{cite journal |last1=Fendt |first1=William A. |last2=Wandelt |first2=Benjamin D. |year=2007 |title=Computing High Accuracy Power Spectra with Pico |arxiv=0712.0194 |bibcode=2007arXiv0712.0194F}}

== See also ==

* [[BOINC]]
* [[Einstein@Home]]
* [[MilkyWay@home]]
* [[LHC@home]]
* [[Cosmic microwave background]]
* [[Lambda-CDM model]]
* [[Wilkinson Microwave Anisotropy Probe]]
* [[Planck (spacecraft)]]

== References ==

{{Reflist}}

== External links ==

* [https://web.archive.org/web/20230000000000*/cosmologyathome.org Cosmology@Home on the Wayback Machine]
* [https://github.com/marius311/cosmohome Cosmology@Home server source code (GitHub)]
* [https://github.com/marius311/boinc2docker boinc2docker (GitHub)]
* [https://arxiv.org/abs/astro-ph/0606709 PICO paper on arXiv]
* [https://arxiv.org/abs/0712.0194 High Accuracy Power Spectra with PICO (arXiv)]

[[Category:BOINC projects]]
[[Category:Cosmology]]
[[Category:Distributed computing projects]]
[[Category:Volunteer computing]]
[[Category:University of Illinois Urbana-Champaign]]