Carnegie Mellon University
May 27, 2021

Dark Energy Survey Releases Most Precise Look at the Universe’s Evolution

Jocelyn Duffy
  • Associate Dean for Communications, MCS
  • 412-268-9982

New results from the Dark Energy Survey (DES) used the largest ever sample of galaxies over an enormous piece of the sky to produce the most precise measurements of the universe’s composition and growth to date. Scientists measured that the way matter is distributed throughout the universe is consistent with predictions in the standard cosmological model, the best current model of the universe.

Researchers from the National Science Foundation’s (NSF’s) Artificial Intelligence (AI) Planning Institute for Data-Driven Discovery in Physics and Carnegie Mellon University’s McWilliams Center for Cosmology were part of the multi-institution team, and brought Carnegie Mellon’s expertise in weak gravitational lensing, artificial intelligence and data science to the DES analysis.

Over the course of six years, DES surveyed 5,000 square degrees — almost one-eighth of the entire sky — in 758 nights of observation, cataloging hundreds of millions of objects. The results announced today draw on data from the first three years — 226 million galaxies observed over 345 nights — to create the largest and most precise maps yet of the distribution of galaxies in the universe at relatively recent epochs.

Since DES studied nearby galaxies as well as those billions of light-years away, its maps provide both a snapshot of the current large-scale structure of the universe and a movie of how that structure has evolved over the course of the past 7 billion years.

Exploring 7 billion light years of space with the Dark Energy Survey. 

To test cosmologists’ current model of the universe, DES scientists compared their results with measurements from the European Space Agency’s orbiting Planck observatory. Planck used light signals known as the cosmic microwave background to peer back to the early universe, just 400,000 years after the Big Bang. The Planck data give a precise view of the universe 13 billion years ago, and the standard cosmological model predicts how the dark matter should evolve to the present. If DES’s observations don’t match this prediction, there is possibly an undiscovered aspect to the universe. While the recently released results are consistent with the prediction, there remain hints from DES and several previous galaxy surveys that the current universe is a few percent less clumpy than predicted — an intriguing find worthy of further investigation.

 “In the area of constraining what we know about the distribution and structure of matter on large scales as driven by dark energy, DES has obtained limits that rival and complement those from the cosmic microwave background,” said Brian Yanny, a Fermilab scientist who coordinated DES data processing and management. “It’s exciting to have the most precise measurements of what’s out there and a better understanding of how the universe has changed from its infancy  to today.”

Ordinary matter makes up only about 5% of the universe. Dark energy, which cosmologists hypothesize drives the accelerating expansion of the universe by counteracting the force of gravity, accounts for about 70%. The last 25% is dark matter, whose gravitational influence binds galaxies together. Both dark matter and dark energy remain invisible and mysterious, but DES seeks to illuminate their natures by studying how the competition between them shapes the large-scale structure of the universe over cosmic time.

 DES photographed the night sky using the 570-megapixel Dark Energy Camera on the 4-meter Blanco telescope at the Cerro Tololo Inter-American Observatory in Chile, a division of the NSF’s NOIRLab. One of the most powerful digital cameras in the world, the Dark Energy Camera was designed specifically for DES and built and tested at Fermilab. The DES data were processed at the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign.

“These analyses are truly state-of-the-art, requiring artificial intelligence and high-performance computing super-charged by the smartest young scientists around,” said Scott Dodelson, the Director of the NSF AI Institute who co-leads the DES Science Committee with Elisabeth Krause of the University of Arizona. “What an honor to be part of this team.”

 To quantify the distribution of dark matter and the effect of dark energy, DES relied on two main phenomena. First, on large scales, galaxies are not distributed randomly throughout space but rather form a weblike structure due to the gravity of dark matter. DES measured how this cosmic web has evolved over the history of the universe. The galaxy clustering that forms the cosmic web, in turn, revealed regions with a higher density of dark matter.

 Second, DES detected the signature of dark matter through weak gravitational lensing. As light from a distant galaxy travels through space, the gravity of both ordinary and dark matter can bend it, resulting in a distorted image of the galaxy as seen from Earth. By studying how the apparent shapes of distant galaxies are aligned with each other and with the positions of nearby galaxies along the line of sight, DES scientists inferred the spatial distribution (or clumpiness) of the dark matter in the universe.

Analyzing the massive amounts of data collected by DES was a formidable undertaking. The team began by analyzing just the first year of data, which was released in 2017. That process prepared the researchers to use more sophisticated techniques for analyzing the larger data set, which includes the largest sample of galaxies ever used to study weak gravitational lensing.

 For example, calculating the redshift of a galaxy — the change in light’s wavelength due to the expansion of the universe — is a key step toward measuring how both galaxy clustering and weak gravitational lensing change over cosmic history.  The redshift of a galaxy is related to its distance, which allows the clustering to be characterized in both space and time.

 “We used a variety of artificial intelligence tools to calibrate the redshift distributions of the galaxy samples,” said Simon Samuroff, a physics postdoc at Carnegie Mellon who co-led the weak gravitational lensing analysis. “This was a huge effort that people put a lot of work into. We now have a method — based on Self-Organizing Maps —that nobody has used before, and it’s very robust.”

 Ten regions of the sky were chosen as “deep fields” that the Dark Energy Camera imaged repeatedly throughout the survey. Stacking those images together allowed the scientists to glimpse more distant galaxies. The team then used the redshift information from the deep fields to calibrate measurements of redshift in the rest of the survey region. This and other advancements in measurements and modeling, coupled with a threefold increase in data compared to the first year, enabled the team to pin down the density and clumpiness of the universe with unprecedented precision. 

 Along with the analysis of the weak-lensing signals, DES also measures other probes that constrain the cosmological model in independent ways: galaxy clustering on larger scales (baryon acoustic oscillations), the frequency of massive clusters of galaxies, and high-precision measurements of the brightness and redshifts of Type Ia supernovae. These additional measurements will be combined with the current weak-lensing analysis to yield even more stringent constraints on the standard model.

 “We used over 30 parameters to fit this complex data set,” said Andresa Campos, a physics graduate student at Carnegie Mellon who ran some of the hundreds of Markov Chain Monte Carlo analyses needed to extract constraints on the parameters. Campos and Dodelson also worked within DES to develop tools to assess consistency between DES and Planck data in these highly multi-dimensional spaces. “We used to be able to simply look at two measurements of a single number and tell instantly whether they were consistent. Now, in these multi-dimensional parameter spaces, we need the full power of Bayesian statistics and modern data science.”

 “One of the greatest discoveries we have made in this round is that the statistical power of DES and future data sets will require a new set of tools borrowed from, and developed in conjunction with, the data science community,” said Dodelson. The AI Institute focuses on applying AI and data science techniques to advance astrophysics, subatomic physics and biophysics. “Carnegie Mellon is well positioned to take a leading role in surveys in this new era.”

 DES concluded observations of the night sky in 2019. With the experience of analyzing the first half of the data, the team is now prepared to handle the complete data set. The final DES analysis is expected to paint an even more precise picture of the dark matter and dark energy in the universe. And the methods developed by the team have paved the way for future sky surveys to probe the mysteries of the cosmos.

 “The strength of massive surveys like DES to analyze the evolution of the universe more precisely is enhanced by this innovative use of AI tools, which leads the way to exciting results from future even larger surveys,” says Nigel Sharp, a program director in NSF’s Division of Astronomical Sciences.

 “The real legacy of DES will be the leaps forward that were made to obtain this key result; these will be critical for the next generation of cosmological experiments starting soon,” said Rachel Mandelbaum, a professor of physics at Carnegie Mellon and the spokesperson for the 1,000-member Dark Energy Science Collaboration that is charged with analyzing data from the upcoming Vera C. Rubin Observatory Legacy Survey of Space and Time.

The recent DES results will be presented in a scientific seminar on May 27. Twenty-nine papers are available on the arXiv online repository.

The Dark Energy Survey is a collaboration of more than 400 scientists from 25 institutions in seven countries. For more information about the survey, please visit the experiment’s website.Funding for the DES Projects has been provided by the U.S. Department of Energy, the U.S. National Science Foundation, the Ministry of Science and Education of Spain, the Science and Technology Facilities Council of the United Kingdom, the Higher Education Funding Council for England, the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign, the Kavli Institute of Cosmological Physics at the University of Chicago, Funding Authority for Funding and Projects in Brazil, Carlos Chagas Filho Foundation for Research Support of the State of Rio de Janeiro, Brazilian National Council for Scientific and Technological Development and the Ministry of Science and Technology, the German Research Foundation and the collaborating institutions in the Dark Energy Survey.Fermilab is America’s premier national laboratory for particle physics and accelerator research. A U.S. Department of Energy Office of Science laboratory, Fermilab is located near Chicago, Illinois, and operated under contract by the Fermi Research Alliance LLC. Visit Fermilab’s website at and follow us on Twitter at @Fermilab.

The DOE Office of Science is the single largest supporter of basic research in the physical sciences in the United States and is working to address some of the most pressing challenges of our time. For more information, please visit