Our year 3 cosmology analysis of galaxy clustering and gravitational lensing is a massive effort from more than a hundred scientists. There will be 29 interconnected papers in all, with the main cosmology analysis at the bottom of the page.
Currently, 15 of the 29 papers have been released. Check them out!
More to come soon!
For questions, you can email lead authors, or firstname.lastname@example.org.
Dark Energy Survey Year 3 Results: Photometric Data Set for Cosmology
We describe the Dark Energy Survey (DES) photometric data set assembled from the first three years of science operations to support DES Year 3 cosmology analyses, and provide usage notes aimed at the broad astrophysics community. This Y3 “GOLD” catalog comprises nearly 5000 deg2 of grizY imaging in the south Galactic cap, including nearly 390 million objects, with depth reaching S/N∼10 for extended objects up to i ∼ 23.0.
Figure: Footprint of the DES Y3 Gold data (red) with DES Y1 (green) and SV (purple) overlaid.
Dark Energy Survey Year 3 Results: Deep Field Optical + Near-Infrared Images and Catalogue
Authors: W.G. Hartley (email@example.com), A. Choi (firstname.lastname@example.org), A. Amon (email@example.com), R. A. Gruendl, E. Sheldon, I. Harrison, G. M. Bernstein, I. Sevilla-Noarbe, B. Yanny, K. Eckert, H. T. Diehl et al.
We describe the Dark Energy Survey (DES) Deep Fields, a set of images and associated multi-wavelength catalogue (ugrizJHKs) built from Dark Energy Camera (DECam) and Visible and Infrared Survey Telescope for Astronomy (VISTA) data. We present a catalogue for the DES 3-year cosmology analysis of a subset of four fields with full 8-band coverage, with i-band depth of 25. The total unmasked area of 5.88 sq. deg. yields 1.6 million objects that are subsequently used in various parts of the DES 3-year weak lensing and galaxy clustering cosmology work such as the source galaxy redshift distributions (see Myles & Alarcon et al.), shear calibration simulations (see MacCrann et al.), and simulations of the survey transfer function (see Everett et al.). The galaxies will also be used as effectively noiseless templates (S/N>sqrt(10) x wide-survey galaxies) for shear measurements using the Bayesian Fourier Domain methodology (see Bernstein & Armstrong 2014). Beyond underpinning the Y3KP cosmology analysis, these data will also enable a wide variety of galaxy evolution studies.
Figure: Color image of approximately half of a DECam chip made from the grz filters at three different depths. The inset squares are a factor of 2 zoom-in of the interacting group located to the far right of the main image.
Dark Energy Survey Year 3 Results: Measuring the Survey Transfer Function with Balrog
Authors: S. Everett (firstname.lastname@example.org), B. Yanny, N. Kuropatkin, E. Huff, Y. Zhang et al.
We describe an updated calibration and diagnostic framework, Balrog, used to directly sample the selection and photometric biases of Dark Energy Survey’s (DES) Year 3 (Y3) dataset. We systematically inject onto the single-epoch images of a random 20% subset of the DES footprint an ensemble of nearly 30 million realistic galaxy models derived from DES Deep Field observations (see Hartley & Choi et al.). These augmented images are analyzed in parallel with the original data to automatically inherit measurement systematics that are often too difficult to capture with traditional generative models. The resulting object catalog is a Monte Carlo sampling of the DES transfer function and is used as a powerful diagnostic and calibration tool for a variety of DES Y3 science, particularly for the calibration of the photometric redshifts of distant “source” galaxies and magnification biases of nearer “lens” galaxies (see Myles & Alarcon et al.).
Figure: This figure shows the distribution of differences in recovered vs. injected i-band magnitude for Balrog galaxies as a function of the injected magnitude after basic science sample cuts. While ~99% of injections have magnitude responses that are contained within the third white contour (highlighted in the inset), there are long, non-Gaussian tails that capture many of the complex features of the DES measurement pipeline such as the effects from image artifacts, blending with other astronomical sources, and catastrophic modeling errors. This displays a narrow slice of the full DES transfer function which characterizes all of the selection and photometric biases of the survey’s measurements of the true cosmic signals in the sky.
Dark Energy Survey Year 3 Results: Weak Lensing Shape Catalogue
Authors: M. Gatti (email@example.com), E. Sheldon (firstname.lastname@example.org), A. Amon, M. Becker, M. Troxel, A. Choi, C. Doux, N. MacCrann, A. Navarro Alsina, I. Harrison, D. Gruen, G. Bernstein, M. Jarvis, L. F. Secco, A. Ferte, T. Shin, J. McCullough, R. P. Rollins, R. Chen, C. Chang, S. Pandey, I. Tutusaus, J. Prat, J. Elvin-Poole, C. Sanchez, et al.
We present and characterise the galaxy shape catalogue of DES Y3. Our self-calibrating shear measurement pipeline metacalibration builds and improves upon the pipeline used in the DES Year 1 analysis in several aspects. The final shape catalogue consists of 100,204,026 galaxies, measured in the riz bands, resulting in a weighted source number density of neff=5.59gal/arcmin2 and corresponding shape noise σe=0.261. We perform a battery of internal null tests on the catalogue, including tests on systematics related to the point-spread function (PSF) modelling, spurious catalogue B-mode signals, catalogue contamination, and galaxy properties to show that this catalog is ready for weak lensing cosmology.
Figure: Effective density of weak lensing sources in the DES Y3 metacalibration shape catalog, across the full footprint.
Dark Energy Survey Year 3 Results: Point-Spread Function Modeling
Authors: M. Jarvis (email@example.com), G. M. Bernstein, A. Amon, C. Davis, P. F. Léget, K. Bechtol, I. Harrison, M. Gatti, A. Roodman, et al.
We introduce a new software package for modeling the point-spread function (PSF) of astronomical images, called PIFF, which we apply to DES Y3. We describe the relevant details about the algorithms used by PIFF to model the PSF, including how the PSF model varies across the field of view. Diagnostic results show that the systematic errors from the PSF modeling are very small over the range of scales that are important for the DES Y3 weak lensing analysis. In particular, the systematic errors from the PSF modeling are significantly smaller than the corresponding results from the DES Y1 analysis. We also briefly describe some planned improvements to PIFF that we expect to further reduce the modeling errors in future analyses.
Figure: The rho statistics are the principal diagnostic we use to establish the quality of the PSF model, since they contribute directly as an additive systematic error in the cosmic shear correlation functions. There are 5 statistics presented in these two figures, which are all demonstrated to be small enough not to significantly bias the DES Y3 cosmology analysis. Other statistics are presented in the paper.
DES Y3 results: Blending shear and redshift biases in image simulations
Authors: N. MacCrann (firstname.lastname@example.org), M. R. Becker, J. McCullough, A. Amon, D. Gruen, M. Jarvis, A. Choi, M. A. Troxel, E. Sheldon, B. Yanny, K. Herner, S. Dodelson, J. Zuntz et al.
We characterize and provide calibration for the gravitational lensing lensing shear measurement pipeline used on the DES Year 3 data. This requires highly detailed and realistic simulations of the DES images, in which we inject artificial shear signals. By analysing our recovery of these injected signals, we can quantify the small but significant biases present in our estimation of the shear, due to subtle effects like blending of multiple galaxies, and derive corrections to be used in the weak lensing cosmology analyses. We derive a new formalism for including the effects of blending in the effective redshift distribution for weak lensing, which allows for a single detected object to contribute to multiple redshifts in the theoretical predictions of weak lensing statistics. Such an approach will become essential for upcoming, deeper datasets, where there will be increased blending and stricter accuracy requirements, such as the Rubin Observatory LSST.
Figure: A color image of a 1000 x 1000 pixel region of the same coadd tile for the real DES Year 3 data (left panel) and our fiducial simulation (right panel). The realistic joint distributions of observed galaxy morphology, flux and colors allows for a joint characterization of the impact of blending on the shear calibration and the effective redshift distribution.
Phenotypic redshifts with self-organizing maps: A novel method to characterize redshift distributions of source galaxies for weak lensing
Authors: R. Buchs (email@example.com), C. P. Davis, D. Gruen, J. DeRose, A. Alarcon, G. M. Bernstein, C. Sanchéz, J. Myles, A. Roodman, et al.
We introduce a new method of linking information from spectroscopy or other high-quality redshift estimates to wide field weak lensing source samples via photometric multi-band deep fields. Using mock catalogs, we show that this significantly reduces sample variance in redshift calibration. This methodology is at the same time a key to being able to use spectroscopic information for calibration of DES photometric redshifts without dominant selection biases.
Figure: Deep field multi-band photometry acts as an intermediary between redshift information and wide-field galaxy samples whose redshift distributions we calibrate.
Dark Energy Survey Year 3 Results: Redshift Calibration of the Weak Lensing Source Galaxies
Authors: J. Myles (firstname.lastname@example.org), A. Alarcon (email@example.com), A. Amon, C. Sanchez, S. Everett, J. DeRose, J. McCullough, D. Gruen, G. Bernstein, M. Troxel, S. Dodelson, A. Campos, N. MacCrann, B. Yin, M. Raveri et al.
We present the derivation and validation of redshift distributions for the source galaxies used in the DES Y3 weak lensing measurements. As the first application of this method to data, we validate that the assumptions made apply to the DES Y3 weak lensing source galaxies and describe our treatment of systematic uncertainties. In a first for redshift analyses, we construct an ensemble of redshift distributions whose variation encodes all uncertainties and combines information from galaxy color, position correlation functions, and shear ratios.
Figure: Ensemble of redshift distributions in four tomographic bins, as inferred from galaxy photometry. Each symbol shows the 95% credible interval of the probability of a galaxy in the weak lensing source sample and assigned to a given tomographic bin to have redshift z. The uncertainty on p(z) is due to biases in the secure redshifts used in the analysis, sample variance and shot noise in the galaxies in the DES deep fields, photometric calibration uncertainty for the DES deep fields, and the inherent uncertainty of the methods applied.
Dark Energy Survey Year 3 Results: Clustering Redshifts — Calibration of the Weak Lensing Source Redshift Distributions with redMaGiC and BOSS/eBOSS
Authors: M. Gatti (firstname.lastname@example.org), G. Giannini (email@example.com), G. M. Bernstein, A. Alarcon, J. Myles, A. Amon, R. Cawthon, M. Troxel, J. DeRose, S. Everett, A. J. Ross, E. S. Rykoff ,J. Elvin-Poole, J. Cordero, I. Harrison, C. Sanchez, J.Prat, D.Gruen, H. Lin, M. Crocce, E. Rozo et al.
We present the calibration of the Dark Energy Survey Year 3 (DES Y3) weak lensing source galaxy redshift distributions n(z) from clustering measurements, cross-correlating the weak lensing (WL) source galaxies sample with redMaGiC galaxies and BOSS/eBOSS galaxies. Two distinct methods for using the clustering statistics are described. The first uses the clustering information independently to estimate the mean redshift of the source galaxies within a redshift window, as done in the DES Y1 analysis. The second method establishes a likelihood of the clustering data as a function of n(z), which can be incorporated into schemes for generating samples of n(z) subject to combined clustering and photometric constraints. We apply the two methods to the DES Y3 data, and compare the clustering based constraints with the fiducial DES photometric redshift estimates.
Figure: SOMPZ redshift distributions, as estimated in data, with and without clustering information (full-shape method). The bands encompass the statistical and systematic uncertainties of the distributions. Sampling from the joining posterior greatly reduces the point-by-point uncertainties in n(z), just as in the simulations. The clustering full-shape method is thus very successful at reducing the impact sample variance on SOMPZ estimators.
Dark Energy Survey Year 3 Results: Calibration of Lens Sample Redshift Distributions using Clustering Redshifts with BOSS/eBOSS
Authors: R. Cawthon (firstname.lastname@example.org), J. Elvin-Poole, A. Porredon, M. Crocce, G. Giannini, M. Gatti, A. J. Ross, E.S. Rykoff, A. Carnero Russell, J. DeRose, S. Lee, M. Rodriguez-Monroy et al.
This work describes the calibration of ‘lens’ galaxy redshifts in DES. In a method called ‘clustering redshifts’, we measure angular cross-correlations between DES galaxies and spectroscopic galaxies from the BOSS & eBOSS surveys. The signal from these correlations informs us about the redshift distribution, N(z), of the DES galaxies. We test and validate a new procedure to mitigate the ‘galaxy bias systematic’ error in this method. We constrain the mean and width of redshift distributions in eleven different photometrically-selected bins. We find biases on the mean redshift in most cases to be below |0.01|. Our uncertainties on the mean redshift range from 0.003 to 0.008 and our uncertainties on the width range from 4 to 9%. We discuss how these results calibrate the redshift distributions used in companion DES Year-3 papers.
Figure: The clustering redshift measurements for the eleven lens galaxy bins, five from the Redmagic sample, and six from the magnitude-limited (maglim) sample. Also shown are the photometric redshift estimates, and 1- and 2-parameter fits to the clustering data. In the 1-parameter fit, the photo-z is shifted to match the mean redshift measured by clustering. In the 2-parameter fit, the photo-z is shifted and stretched to fit the clustering data.
Dark Energy Survey Year 3 Results: Covariance Modelling and its Impact on Parameter Estimation and Quality of Fit
Author: O. Friedrich (email@example.com), F. Andrade-Oliveira, H. Camacho, O. Alves, R. Rosenfeld, J. Sanchez, X. Fang, T. F. Eifler, E. Krause, C. Chang, Y. Omori, A. Amon, E. Baxter, J. Elvin-Poole, D. Huterer, A. Porredon, J. Prat, V. Terra, A. Troja et al.
We describe and test the fiducial covariance matrix model for the combined 2-point function analysis of the Dark Energy Survey Year 3 dataset. Using a variety of new ansatzes for covariance modelling and testing we validate the assumptions and approximations of this model. These include the assumption of Gaussian likelihood, the trispectrum contribution to the covariance, the impact of evaluating the model at a wrong set of parameters, the impact of masking and survey geometry, deviations from Poissonian shot-noise, galaxy weighting schemes and other, sub-dominant effects. We find that our covariance model is robust and that its approximations have little impact on goodness-of-fit and parameter estimation. The largest impact on best-fit figure-of-merit arises from the so-called f_sky approximation for dealing with finite survey area, which on average increases the χ2 between maximum posterior model and measurement by 3.7%.
Figure: Impact of different covariance modelling choices on χ2 between measured 3x2pt datavectors and maximum posterior models. The dashed vertical lines and error bars indicate the 1σ fluctuations expected in χ2.
Assessing tension metrics with Dark Energy Survey and Planck data
Author: P. Lemos (firstname.lastname@example.org), M. Raveri, A. Campos, Y. Park, C. Chang, N. Weaverdyck, D. Huterer, A. R. Liddle, et al.
Quantifying tensions — inconsistencies amongst measurements of cosmological parameters by different experiments — has emerged as a crucial part of modern cosmological data analysis. Statistically-significant tensions between two experiments or cosmological probes may indicate new physics extending beyond the standard cosmological model and need to be promptly identified. We apply several tension estimators proposed in the literature to the DES large-scale structure measurement and Planck cosmic microwave background data. We find that the parameter differences, Eigentension, and Suspiciousness metrics all yield similar results on both simulated and real data, while the Bayes ratio is inconsistent with the rest due to its dependence on the prior volume. Using these metrics, we calculate that DES Year 1 and Planck have 2.3 sigma tension under the Lambda-CDM paradigm. This suite of metrics provides a toolset for robustly testing tensions in the DES Year 3 data and beyond.
Figure: Tension estimates given by different metrics versus the corresponding Bayes ratio. Shaded regions highlight Jeffreys’ scale used to interpret the Bayes ratio, with the vertical line separating “Tension” to the left and “Agreement” to the right.
Dark Energy Survey internal consistency tests of the joint cosmological probes analysis with posterior predictive distributions
Author: C. Doux (email@example.com), E. Baxter, P. Lemos, C. Chang, et al.
We present the methodology for testing the internal consistency of the Dark Energy Survey (DES) measurements of cosmic shear, galaxy-galaxy lensing and galaxy clustering. We focus on data space tests to best identify potential inconsistent subsets using posterior predictive distributions. When applied to DES Year 1 data, we find overall good consistency and a good fit to ΛCDM. We test consistency by splitting the data into different probes, redshift bins and scales and find a small tension between large- and small-scale measurements (at the <2σ level).
Figure: Excerpt from Fig 2 of paper, comparing the posterior predictive distribution in blue and data in red for cosmic shear measurements when testing for 3x2pt goodness-of-fit (for both, we subtracted the best-fit and divided by the error for visualization).
Blinding multi-probe cosmological experiments
Authors: J. Muir (firstname.lastname@example.org), G. Bernstein, D. Huterer, F. Elsner, E. Krause, A. Roodman, et al.
This paper introduces the transformation applied to DES Y3 galaxy clustering and weak lensing measurements in order to hide the cosmology results until decisions about how to conduct the analysis were finalized. The goal of this kind of concealment, known as blinding, is to protect the DES analysis from bias that might be unconsciously introduced if experimenters are looking at how analysis choices influence how the results compare to their expectations. DES is the first analysis to use this method, which works by altering the two-point correlation functions that are input to parameter estimation. This paper shows that, for simulated DES Y3 analyses, the transformation can successfully change the best-fit cosmological parameters while preserving the internal consistency of different parts of data.
Figure: Effect of the blinding transformation on a simulated DES Y3 analysis.
Dark Energy Survey Year 3 Results: Optimizing the Lens Sample in Combined Galaxy Clustering and Galaxy-Galaxy Lensing Analysis
Authors: A. Porredon (email@example.com), M. Crocce, P. Fosalba, J. Elvin-Poole, I. Ferrero, E. Krause, X. Fang, T. Eifler, R. Cawthon, N. Weaverdyck, N. MacCrann, A. Carnero, et al.
We investigate potential gains in cosmological constraints from the combination of galaxy clustering and galaxy-galaxy lensing by optimizing the lens galaxy sample selection, using information from DES Y3. We explore easily reproducible selections based on magnitude cuts in i-band as a function of (photometric) redshift, zphot, and look to balance number density vs. photometric accuracy. Our optimal selection, the MAGLIM sample, satisfies i < 4 zphot + 18 and has ∼ 3.5 times more galaxies than our reference sample, REDMAGIC. Assuming a wCDM model and equivalent scale cuts for nonlinear effects, it leads to gains of 16% in σ8, 10% in Ωm, and 12% in w. In ΛCDM we find an improvement of 19% and 27% on σ8 and Ωm, respectively.
Figure: ΛCDM constraints from the combination of galaxy clustering and galaxy-galaxy lensing using the REDMAGIC (red) and MAGLIM (blue) samples as lenses. The MAGLIM constraints are tighter by 27% on Ωm, and 11% on S8 compared to REDMAGIC.