Dark Energy Survey Simulations Huan Lin 1 Nikolay

Dark Energy Survey Simulations Huan Lin 1, Nikolay Kuropatkin 1, Risa Wechsler 2, Michael Busha 2, Matthew Becker 3, Bruno Rossetto 4, Luiz da Costa 4, Martin Makler 5, August Evrard 6, Andrey Kravtsov 3, for the Dark Energy Survey Collaboration 1 Fermilab, 2 KIPAC/Stanford/SLAC, 3 University of Chicago, 4 Observatório Nacional, Brazil, 5 CBPF, Brazil, 6 University of Michigan The Dark Energy Survey (DES) is a next generation optical imaging survey that will cover 5000 deg 2 of the southern sky using a new 520 Megapixel CCD camera on the 4 m Blanco telescope at the Cerro Tololo Inter-American Observatory. The DES will be the first survey to study dark energy by joining four complementary techniques - galaxy clusters, weak gravitational lensing, baryon acoustic oscillations, and Type Ia supernovae - on the same data. To prepare data management and science analysis pipelines for the survey, we are carrying out annual data challenges based on detailed catalog- and image-level simulations of DES performance. Here we describe our latest round of DES simulations, "Data Challenge 5" (DC 5). The DC 5 catalog simulations include: a deep, sky survey output of dark matter structure from a Gpc N-body simulation; a statistical galaxy catalog produced via an empirical method, "ADDGALS"; galaxy shape parameters from COSMOS data; weak lensing convergence and shear derived from dark matter ray tracing, plus strongly-lensed arcs, and; stars, based on the "Trilegal" model of the Milky Way. The simulated galaxy and stellar catalogs are used to populate synthetic DES images that account for a wide range of instrumental and observational effects, including: telescope and corrector optics, CCD detector performance, and atmospheric and weather conditions. Using grid computing resources at Fermilab, 3. 5 TB of simulated DES imaging data have been generated for DC 5, covering the 5 DES filters (griz. Y) over some 200 deg 2 of sky. Image Simulations Catalog Simulations “Addgals” Galaxy Catalog Gpc N-Body Simulation • • “Addgals” method assigns galaxies to dark matter particles in a way that reproduces the observed distributions and correlations among galaxy luminosities, colors, density, and clustering • Detector Model • Optics Model • Blanco telescope + DES optical corrector • Point spread functions (PSFs) computed on 1885 -point grid over focal plane Trained using observed SDSS and DEEP 2 galaxy samples • PSF from optical design (0. 27” FWHM) + other instrumental contributions (0. 4” FHWM, mainly CCD diffusion) • Can include optical misalignments (defocus, tilt) • Also includes “pupil ghost” and corrector geometric distortions Bias, read noise (10 e-), gain, bad columns, QE Flatfields from real DES CCD test data Saturated bleed columns Cosmic ray library from real DES CCD test data Cross talk (0. 001) between 2 amplifiers of each CCD Pixel size variations in “glowing edges” Zeropoints and color terms on CCD-by-CCD basis • Object Rendering • Observation Model • Photometric zeropoint variations drawn from SDSS data for both nominally-photometric and non-photometric conditions • Base simulation uses one “Carmen” run from the “Las. Damas” suite (http: //co-op. vanderbilt. edu/lasdamas/ ; Mc. Bride et al. 2010, in prep. ) of Nbody models evolved with Gadget • Atmospheric seeing (0. 9” median FHWM) from Blanco Mosaic-II Super. MACHO data; Moffat profile atmospheric PSFs • Simulation Volume: 1 Gpc/h periodic cube • Cosmological Parameters: m= 0. 25, = 0. 75, 8 = 0. 8, ns= 1. 0, flat CDM • Sky brightness variations • • Lightcone: 220 deg 2, z < 1. 35 Mass resolution: Mp = 5 x 1010 Msun/h • Can model galaxies with accurate clustering all the way down to Mr = -12 • Atmospheric refraction: position offsets, plus stretching of PSF from differential chromatic refraction (DCR) • Java package reads object catalogs and observation information, and renders objects using shapelets (order n=15 136 coefficients per galaxy) • Shear and convergence due to gravitational lensing applied in shapelet space • Galaxies convolved with atmospheric and instrumental PSFs using FFTs • Instrumental signatures (flat field, read noise, gain, …) and sky background then added • Final images output in multi-extension fits format; 1 GB uncompressed per 3 -deg 2 DECam pointing Close-up of portion of simulated raw DECam focal plane Gravitational Lensing • • Ray tracing of Carmen N-body simulation provides values of the weak lensing convergence and shear at the location of each galaxy in the catalog Colors • Using the SDSS DR 6 training set, we measure P(SED | Mr , 5), the probability linking galaxy colors with r-band absolute magnitude and local galaxy density • 5 is the distance to the 5 th nearest galaxy • Colors are k-corrected; no other color evolution is included Weak lensing convergence and shear fields over small subset of simulation area • Also add small sample of arcs (strongly lensed galaxies) using massive halos from Carmen N-body simulation as lenses and galaxies from a HST/UDF (Sersic profile) catalog as sources Original pure simulated arc Blue shows convergence surface mass density; lighter ==> higher density Note simulated effects, such as flatfield features (“grind marks”, “tape bumps”; these all flatfield out), bright star artifacts (saturated bleed columns, diffraction spikes), cosmic rays, cross talk between amplifiers, “glowing edges”, and bad columns Final simulated arc after PSF convolution, rebinning to DES plate scale (0. 27”/pix), and addition of noise Red “whiskers” show shear field due to gravitational lensing Simulated 3 -Deg 2 DECam Focal Plane Galaxy Shapelets • Simulated galaxy shapes described by 2 D shapelet basis functions • Distribution of shapelet coefficients derived from COSMOS HST/ACS shapelets catalog (from M. Peeples & R. Massey) • Shapelets assigned to simulated DES galaxy using nearest neighbor matching to real COSMOS galaxy in space of griz magnitudes 62 2 kx 4 k Science CCDs (520 Megapixels) Also 8 2 kx 2 k Focus/Alignment CCDs are simulated, but not shown here (34 Mpix) A final 4 2 kx 2 k guide CCDs are on focal plane but not simulated since different readout cadence Example COSMOS HST/ACS image upon which DES galaxy shapelet distribution is based 1 GB per single 3 deg 2 DECam pointing Color composite made from simulated focal plane in g, r, i filters “Trilegal” Stellar Catalog and SDSS Photometric Standards • Use “Trilegal” model (Girardi et al. 2005) of the Milky Way’s stellar population to simulate positions of stars and their magnitudes/colors in the DES griz. Y bands • Include USNO-B stars at their real positions, but with magnitudes of nearest Trilegal star, to provide astrometric reference stars • • Data Challenge (DC 5) simulations Use real SDSS photometric standard stars derived from repeated observations of SDSS equatorial Stripe 82 • 2600 science images, multiple tilings in griz. Y filters over ~200 deg 2 of sky • 900 calibration images (biases, flatfields, standard stars) • 10 simulated observing nights • 3. 5 TB of simulated images, produced over 2 -week period using up to 225 Open Science Grid/Fermi. Grid compute nodes Example color-color and color-magnitude diagrams for simulated Trilegal stars • DC 5 now ongoing: tests data management pipelines and science analysis codes • Additional data challenges upcoming in preparation for start of DES observing in late 2011 DES Collaboration: Fermilab, University of Illinois at Urbana-Champaign, University of Chicago, Lawrence Berkeley National Laboratory, National Optical Astronomy Observatory, Spain DES Collaboration, United Kingdom DES Collaboration, University of Michigan, DES-Brazil Consortium, University of Pennsylvania, Argonne National Laboratory, Ohio State University, and Santa Cruz-SLAC-Stanford DES Consortium. DES Funding: DOE, NSF, STFC (UK), Ministry of Education and Science (Spain), FINEP (Brazil), and the Collaborating Institutions.