First Cosmology Results Using Type Ia Supernovae From the Dark Energy Survey: Photometric Pipeline and Light Curve Data Release
D. Brout, M. Sako, D. Scolnic, R. Kessler, C. B. D'Andrea, T. M. Davis, S. R. Hinton, A. G. Kim, J. Lasker, E. Macaulay, A. Möller, R. C. Nichol, M. Smith, M. Sullivan, R. C. Wolf, S. Allam, B. A. Bassett, P. Brown, F. J. Castander, M. Childress, R. J. Foley, L. Galbany, K. Herner, E. Kasai, M. March, E. Morganson, P. Nugent, Y.-C. Pan, R. C. Thomas, B. E. Tucker, W. Wester, T. M. C. Abbott, J. Annis, S. Avila, E. Bertin, D. Brooks, D. L. Burke, A. Carnero Rosell, M. Carrasco Kind, J. Carretero, M. Crocce, C. E. Cunha, L. N. da Costa, C. Davis, J. De Vicente, S. Desai, H. T. Diehl, P. Doel, T. F. Eifler, B. Flaugher, P. Fosalba, J. Frieman, J. García-Bellido, E. Gaztanaga, D. W. Gerdes, D. A. Goldstein, D. Gruen, R. A. Gruendl, J. Gschwend, G. Gutierrez, W. G. Hartley, D. L. Hollowood, K. Honscheid, D. J. James, K. Kuehn, N. Kuropatkin, O. Lahav, T. S. Li, M. Lima, J. L. Marshall, P. Martini, R. Miquel, B. Nord, A. A. Plazas, A. Roodman, E. S. Rykoff, E. Sanchez, V. Scarpine, R. Schindler, M. Schubnell, S. Serrano, I. Sevilla-Noarbe, M. Soares-Santos, F. Sobreira, E. Suchyta, M. E. C. Swanson, G. Tarle, D. Thomas, D. L. Tucker, A. R. Walker, B. Yanny, Y. Zhang
FFERMILAB-PUB-18-540-AEDES-2017-0311D
RAFT VERSION J UNE
4, 2019
Preprint typeset using L A TEX style emulateapj v. 12/16/11
FIRST COSMOLOGY RESULTS USING TYPE IA SUPERNOVAE FROM THE DARK ENERGY SURVEY:PHOTOMETRIC PIPELINE AND LIGHT CURVE DATA RELEASE
D. B
ROUT , M. S AKO , D. S COLNIC , R. K ESSLER , C. B. D’A
NDREA , T. M. D AVIS , S. R. H INTON , A. G. K IM , J. L ASKER ,E. M
ACAULAY , A. M ÖLLER , R. C. N
ICHOL , M. S MITH , M. S ULLIVAN , R. C. W OLF , S. A LLAM , B. A. B ASSETT ,P. B
ROWN , F. J. C ASTANDER , M. C
HILDRESS , R. J. F OLEY , L. G ALBANY , K. H ERNER , E. K ASAI , M. M
ARCH ,E. M ORGANSON , P. N UGENT , Y.-C. P AN , R. C. T HOMAS , B. E. T UCKER , W. W
ESTER , T. M. C. A BBOTT , J. A NNIS ,S. A VILA , E. B ERTIN , D. B
ROOKS , D. L. B URKE , A. C
ARNERO R OSELL , M. C
ARRASCO K IND , J. C
ARRETERO ,M. C ROCCE , C. E. C
UNHA , L. N. DA C OSTA , C. D
AVIS , J. D E V ICENTE , S. D ESAI , H. T. D IEHL , P. D OEL ,T. F. E IFLER , B. F
LAUGHER , P. F OSALBA , J. F
RIEMAN , J. G
ARCÍA -B ELLIDO , E. G AZTANAGA , D. W. G
ERDES ,D. A. G
OLDSTEIN , D. G RUEN , R. A. G
RUENDL , J. G
SCHWEND , G. G
UTIERREZ , W. G. H ARTLEY ,D. L. H
OLLOWOOD , K. H ONSCHEID , D. J. J
AMES , K. K UEHN , N. K UROPATKIN , O. L AHAV , T. S. L I , M. L IMA ,J. L. M
ARSHALL , P. M ARTINI , R. M
IQUEL , B. N
ORD , A. A. P LAZAS , A. R OODMAN , E. S. R
YKOFF ,E. S
ANCHEZ , V. S CARPINE , R. S CHINDLER , M. S CHUBNELL , S. S ERRANO , I. S
EVILLA -N OARBE , M. S OARES -S ANTOS ,F. S OBREIRA , E. S
UCHYTA , M. E. C. S WANSON , G. T ARLE , D. T HOMAS , D. L. T UCKER , A. R. W ALKER , B. Y ANNY , AND
Y. Z
HANG (DES C OLLABORATION ) Department of Physics and Astronomy, University of Pennsylvania, Philadelphia, PA 19104, USA Kavli Institute for Cosmological Physics, University of Chicago, Chicago, IL 60637, USA Department of Astronomy and Astrophysics, University of Chicago, Chicago, IL 60637, USA School of Mathematics and Physics, University of Queensland, Brisbane, QLD 4072, Australia Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, CA 94720, USA Institute of Cosmology and Gravitation, University of Portsmouth, Portsmouth, PO1 3FX, UK ARC Centre of Excellence for All-sky Astrophysics (CAASTRO) The Research School of Astronomy and Astrophysics, Australian National University, ACT 2601, Australia School of Physics and Astronomy, University of Southampton, Southampton, SO17 1BJ, UK Graduate School of Education, Stanford University, 160, 450 Serra Mall, Stanford, CA 94305, USA Fermi National Accelerator Laboratory, P. O. Box 500, Batavia, IL 60510, USA African Institute for Mathematical Sciences, 6 Melrose Road, Muizenberg, 7945, South Africa South African Astronomical Observatory, P.O.Box 9, Observatory 7935, South Africa George P. and Cynthia Woods Mitchell Institute for Fundamental Physics and Astronomy, and Department of Physics and Astronomy, Texas A&MUniversity, College Station, TX 77843, USA Institut d’Estudis Espacials de Catalunya (IEEC), 08034 Barcelona, Spain Institute of Space Sciences (ICE, CSIC), Campus UAB, Carrer de Can Magrans, s/n, 08193 Barcelona, Spain Santa Cruz Institute for Particle Physics, Santa Cruz, CA 95064, USA PITT PACC, Department of Physics and Astronomy, University of Pittsburgh, Pittsburgh, PA 15260, USA Department of Physics, University of Namibia, 340 Mandume Ndemufayo Avenue, Pionierspark, Windhoek, Namibia National Center for Supercomputing Applications, 1205 West Clark St., Urbana, IL 61801, USA Division of Theoretical Astronomy, National Astronomical Observatory of Japan, 2-21-1 Osawa, Mitaka, Tokyo 181-8588, Japan Institute of Astronomy and Astrophysics, Academia Sinica, Taipei 10617, Taiwan Cerro Tololo Inter-American Observatory, National Optical Astronomy Observatory, Casilla 603, La Serena, Chile CNRS, UMR 7095, Institut d’Astrophysique de Paris, F-75014, Paris, France Sorbonne Universités, UPMC Univ Paris 06, UMR 7095, Institut d’Astrophysique de Paris, F-75014, Paris, France Department of Physics & Astronomy, University College London, Gower Street, London, WC1E 6BT, UK Kavli Institute for Particle Astrophysics & Cosmology, P. O. Box 2450, Stanford University, Stanford, CA 94305, USA SLAC National Accelerator Laboratory, Menlo Park, CA 94025, USA Centro de Investigaciones Energéticas, Medioambientales y Tecnológicas (CIEMAT), Madrid, Spain Laboratório Interinstitucional de e-Astronomia - LIneA, Rua Gal. José Cristino 77, Rio de Janeiro, RJ - 20921-400, Brazil Department of Astronomy, University of Illinois at Urbana-Champaign, 1002 W. Green Street, Urbana, IL 61801, USA Institut de Física d’Altes Energies (IFAE), The Barcelona Institute of Science and Technology, Campus UAB, 08193 Bellaterra (Barcelona) Spain Observatório Nacional, Rua Gal. José Cristino 77, Rio de Janeiro, RJ - 20921-400, Brazil Department of Physics, IIT Hyderabad, Kandi, Telangana 502285, India Department of Astronomy/Steward Observatory, 933 North Cherry Avenue, Tucson, AZ 85721-0065, USA Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Dr., Pasadena, CA 91109, USA Instituto de Fisica Teorica UAM/CSIC, Universidad Autonoma de Madrid, 28049 Madrid, Spain Department of Astronomy, University of Michigan, Ann Arbor, MI 48109, USA Department of Physics, University of Michigan, Ann Arbor, MI 48109, USA California Institute of Technology, 1200 East California Blvd, MC 249-17, Pasadena, CA 91125, USA Department of Physics, ETH Zurich, Wolfgang-Pauli-Strasse 16, CH-8093 Zurich, Switzerland Center for Cosmology and Astro-Particle Physics, The Ohio State University, Columbus, OH 43210, USA Department of Physics, The Ohio State University, Columbus, OH 43210, USA Harvard-Smithsonian Center for Astrophysics, Cambridge, MA 02138, USA Australian Astronomical Optics, Macquarie University, North Ryde, NSW 2113, Australia Departamento de Física Matemática, Instituto de Física, Universidade de São Paulo, CP 66318, São Paulo, SP, 05314-970, Brazil Department of Astronomy, The Ohio State University, Columbus, OH 43210, USA a r X i v : . [ a s t r o - ph . I M ] J un Brout et al. : DES-SN Photometry Pipeline and Y1-Y3 Spectroscopic SNe Ia Light Curve Data Release Institució Catalana de Recerca i Estudis Avançats, E-08010 Barcelona, Spain Brandeis University, Physics Department, 415 South Street, Waltham MA 02453 Instituto de Física Gleb Wataghin, Universidade Estadual de Campinas, 13083-859, Campinas, SP, Brazil Computer Science and Mathematics Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831 (Received November 9, 2018; Accepted February 13, 2019)
Draft version June 4, 2019
ABSTRACTWe present griz light curves of 251 Type Ia Supernovae (SNe Ia) from the first 3 years of the DarkEnergy Survey Supernova Program’s (DES-SN) spectroscopically classified sample. The photometric pipelinedescribed in this paper produces the calibrated fluxes and associated uncertainties used in the cosmologicalparameter analysis (Brout et al. 2019-SYS, DES Collaboration et al. 2018) by employing a scene modelingapproach that simultaneously models a variable transient flux and temporally constant host galaxy. Weinject artificial point sources onto DECam images to test the accuracy of our photometric method. Uponcomparison of input and measured artificial supernova fluxes, we find flux biases peak at 3 mmag. Werequire corrections to our photometric uncertainties as a function of host galaxy surface brightness at thetransient location, similar to that seen by the DES Difference Imaging Pipeline used to discover transients. Thepublic release of the light curves can be found at https://des.ncsa.illinois.edu/releases/sn . Subject headings:
DES, techniques: photometry, supernovae, cosmology INTRODUCTION
The discovery of the accelerated expansion of the universe(Riess et al. 1998; Perlmutter et al. 1999) using Type Ia Su-pernovae (SNe Ia) has motivated the collection of ever-largerSN Ia samples in order to improve measurements of cosmo-logical distances and test the nature of dark energy. Con-straints from SNe Ia are best measured with a combination oflow ( z < .
1) and higher ( z > .
1) redshift SNe. The trend inSN surveys over the last three decades has been towards widerand/or deeper rolling surveys where the same images are usedto both discover SNe and measure their light curves. Therolling search is conducive to forward modeling photomet-ric methods. So called ‘Scene Modeling Photometry‘ (here-after
SMP ), which simultaneously models a variable transientflux and temporally constant host galaxy, was first developedby Astier et al. (2006) and has been implemented for recentSN Ia cosmology analyses including for the Sloan Digital SkySurvey (SDSS; Holtzman et al. 2008, hereafter H08) and Su-pernova Legacy Survey (SNLS; Astier et al. 2013, hereafterA13), and as a crosscheck in Pan-STARRS (PS1; Scolnicet al. 2017).The Dark Energy Survey was conducted in two parts; awide-field galaxy survey (5,000 deg ) and a dedicated tran-sient search in the southern celestial hemisphere covering anarea of 27 deg (Bernstein et al. 2012, K15: Kessler et al.2015). The Dark Energy Survey Supernova Program (here-after DES-SN) has discovered tens of thousands of transients,of which ∼ . < z < .
2. A subset of ≈
500 SNe Ia from 0 . < z < . SMP pipeline, whichforward models SNe and their host galaxies to obtain theDES-SN lightcurves used for cosmological analysis. Thispaper is part of a series of 9 papers describing the analy-ses that lead to cosmological constraints from the spectro-scopic SNe Ia observed in the first three years of DES-SNand combined with a low-redshift sample (hereafter DES-SN3YR). These are: the DES-SN search & discovery (K15),spectroscopic follow-up (D’Andrea et al. 2018), calibration(Lasker et al. 2019), photometry (this work), simulations ofour dataset (Kessler et al. 2019), analysis of Host-SN corre-lations (Smith et al. in prep.), an inverse distance ladder H measurement (Macaulay et al. 2019), the blinded cosmologi-cal analysis and systematics validation (B18-SYS: Brout et al.2019-SYS), a Bayesian Hierarchical Method of cosmologicalparameter fitting (Hinton et al. 2019), and ultimately the un-blinded cosmological parameter constraints (DES Collabora-tion et al. 2018).Prior to implementing SMP , supernova candidates were dis-covered and located by the Difference Imaging pipeline, here-after
DiffImg (K15), which uses template images, degradeseither the template image or the search image to match theimage with worse seeing, and performs an image subtrac-tion to produce catalogs of transient detections.
DiffImg then creates candidates from multiple spatially coincident de-tections, and produces light curves from PSF photometry onthe differenced images.
DiffImg photometry is used in thereal-time analysis of light curves for the spectroscopic follow-up program, and has already been used in several analyses(Doctor et al. 2017; Soares-Santos et al. 2016; Soares-Santoset al. 2017).
SMP is not used for transient discovery becauseit would require modeling of all galaxies within the DES-SN footprint, which is not tractable for real-time transientsearches. However, because our
SMP pipeline does not de-grade images in the extraction of SN fluxes, it is ideal foruse in precision cosmology. The light curves presented hereare used in the DES-SN3YR cosmological parameter analysis(B18-SYS) and for obtaining cosmological constraints (DESCollaboration et al. 2018).We describe our implementation of the scene modeling con-cept, which is derived from the techniques used by SDSS(H08) and SNLS (A13) and has been developed specificallyfor DES-SN cosmology. Scene modeling methods have beenused extensively in other types of analyses such as crowded-field photometry (Riess et al. 2016, Schlafly et al. 2018). Inour implementation of
SMP , the transient flux and host galaxyare modeled simultaneously. The transient flux is allowed tovary over time and the host galaxy flux is fixed across all ob-servations.In order to evaluate the results of scene modeling photome-try, A13 moved nearby stars on their images to locations nearhost galaxies and treated them as fake SNe but did not mea-sure light curves. We have developed a unique approach inwhich we generated 100,000 artificial SN light curves that rout et al. : DES-SN Photometry Pipeline and Y1-Y3 Spectroscopic SNe Ia Light Curve Data Release 3
SMP Model Visual Representation F IG . 1.— Visual representation of the SMP process. The model is comprised of a temporally constant galaxy model and a temporally varying SN flux (deltafunction). Both the SN and galaxy are convolved with the PSF of each image in Fourier space to produce a model which can be compared to data. are inserted as point sources onto DECam images (hereafter‘fakes’). Injection of artificial point sources is one compo-nent of a multi-faceted plan to use fake SNe to trace bi-ases throughout the DES-SN cosmological parameter analy-sis. Here, they are used to check for flux biases introduced bythe photometric pipeline and to determine corrections for
SMP flux uncertainties. B18-SYS use fakes to characterize the out-put of
DiffImg and
SMP , which is needed for catalog-levelsimulations that are used to predict distance biases. B18-SYSalso present a full cosmological analysis of 10,000 fake SNethat have been “discovered” by the search pipeline, processedby the
SMP pipeline, and processed through our cosmologicalanalysis pipeline in the same manner as the real dataset.One outstanding problem in SN photometry that was dealtwith in previous surveys (e.g., R14: Rest et al. 2014, J17:Jones et al. 2017) is the underestimation of SN flux uncertain-ties when SNe are located near high local host galaxy bright-ness. R14 and J17 characterize the size of this effect by per-forming photometry at the location of the SNe when the SNflux is known to be zero. Here, we describe how we use ourextensive pipeline of fakes to assess the size of this effect forour analysis and model it precisely in catalog level simula-tions of our dataset.The outline of this paper is as follows. We discuss ourdataset, the preparation, and internal calibration of DES im-ages in Section 2. Our scene-modeling method is explained inSection 3. In Section 4 we show the results of our validationon fakes. In Section 5 we apply our pipeline to the DES-SN 3 year spectroscopic sample and present the light curvesused for our cosmological parameter analysis; the publicly re-leased light curve data can be found online . In Section 6 wecrosscheck the PSF model because it is not tested in our fakesanalysis. In Section 7 we discuss improvements to SMP andwe compare to
DiffImg and in Section 8 we give our con-clusions. DATASET AND IMAGE PRE-PROCESSING
The 3 Year Spectroscopic Sample
The DES-SN performed a deep, time-domain survey in fouroptical bands ( griz ) with an average cadence of 7 days perfilter covering ∼
27 deg over 5 annual campaigns from 2013to 2018 using the Dark Energy Camera (DECam: Flaugheret al. 2015). DECam exposure processing (Morganson et al.2018), DiffImg , and automated artifact rejection (Goldsteinet al. 2015) were run on a nightly basis.DES-SN observed in 8 “shallow” and in 2 “deep” fields, DES-SN Spectroscopic Sample Y1-Y3
SMP
Photometry Release: https://des.ncsa.illinois.edu/releases/sn with the shallow and deep fields having typical nightly point-source depths of 23.5 and 24.5 mag, respectively. Multipleexposures are taken each night with 3, 3, 5, and 11 (1, 1, 1,and 2) exposures taken in griz for the deep (shallow) fields(See D’Andrea et al. 2018). Images used in this analysis weretaken during the first three years of DES-SN, from Sept. 2013to Feb. 2016, in which we discovered roughly ∼ ∼ Image Processing
FirstCut
The DECam images used by the
SMP pipeline are firstpre-processed as part of the nightly single-epoch process-ing. This pre-processing stage, denoted FirstCut (Morgansonet al. 2018), accounts for crosstalk correction, bias subtrac-tion, bad-pixel masking (masking known problematic pixelsin the camera), and flat fielding. It also makes corrections toimage fluxes for CCD nonlinearity (Bernstein et al. 2017b)and the brighter-fatter effect (A13, Antilogus et al. 2014, andGruen et al. 2015), and it masks cosmic rays and satellitetrails.A sky level has been fit and subtracted using the princi-ple component analysis pipeline developed by Bernstein et al.(2017a). This procedure decomposes the image under the as-sumption that it is the sum of the astrophysical sources of in-terest, a zero-mean noise component, and a background com-ponent that is a linear function of a small number of sky tem-plates.
Additional Image Preparation
After FirstCut, we perform additional image preparation.While we do not use
DiffImg photometry, we use a numberof the same modules as summarized below and described indetail in K15. For each exposure and CCD we perform the fol-lowing steps: i) compute an astrometric solution from a jointfit to a template image, resulting in improved relative astrom-etry between the different epochs, ii) determine a position-dependent PSF following the K15 options instead of thosefrom FirstCut, and iii) overlay the same fakes that were over-laid during the search. Additionally, we use a DES-derivedstellar catalog (described in Section 3.2.1 of K15.) instead ofan external catalog such as USNO-B (Monet et al. 2003).
Star Catalog
Calibrated tertiary standard star magnitudes from Burkeet al. (2018) are used for the DES-SN internal calibration
Brout et al. : DES-SN Photometry Pipeline and Y1-Y3 Spectroscopic SNe Ia Light Curve Data Releaseof each DES-SN image. Approximately 50 tertiary stan-dard stars lie within each DECam CCD image. Burke et al.(2018) have determined grizY magnitudes in the AB systemof these standard stars using the “Forward Global CalibrationMethod” (FGCM). The FGCM “forward” computes the frac-tion of photons observed for each star over repeated expo-sures by utilizing measurement of the instrument transmissionfunction, precipitable water vapor, observing conditions, anda model of the stellar source. In addition, using the passbandtransmission (instrument + atmosphere) vs. wavelength andthe spectral energy distribution (SED) of the source, correc-tions are applied to the stellar catalog fluxes (as well as tothe final SN fluxes). These SED-dependent “chromatic cor-rections” account for differences between SED and the meanstellar SED, and between atmospheric transmission of eachexposure and the mean atmospheric transmission. This cor-rection extends the FGCM calibration precision to be validover a wide color range ( − (cid:46) g − i (cid:46) METHOD
The
SMP method utilizes a set of calibrated DECam imagestamps centered at the location of a SN to constrain a modelfor a temporally varying SN and a temporally constant hostgalaxy (Figure 1). Here we outline the steps required to buildand fit the
SMP model.
Stellar Photometry
We use PSF-fitted photometry of the tertiary standard starsto determine the zero-point of each image. As discussedabove, the sky background in the FirstCut images was sub-tracted using PCA over the entire exposure. However, at thespecific locations of transient objects we check for residualnonzero sky background. Residual sky often occurs when themoon is bright, causing large sky gradients that are not cap-tured with PCA. We apply a second method of local sky back-ground and sky uncertainty estimation using concentric aper-tures of 40 and 60 pixels following Jones et al. (2015) andthe resulting sky and uncertainty are calculated in the samemanner for each tertiary standard star as well as for the SN.Biases are induced in PSF-fitted flux measurements whenthe astrometric solution of a source is incorrect or is uncer-tain (Rest et al. 2014). These biases are smaller for stars thanfor SNe because the stars have higher S/N and their positionsare better constrained. When computing photometric magni-tudes, in the limit of high S/N and a correct PSF model, thereis no astrometrically-induced flux bias if the astrometric so-lution and uncertainty are the same for both the stars and theSN itself. The bias in the zeropoint and the bias in the SN fluxwill cancel. Here, we discuss the expected photometric biasesin the real SNe dataset; in the fake dataset this is more subtleand is discussed in Sections 4.3.There are fundamental differences between stars and theSNe that must be accounted for. The stars may have mea-surable proper motion while the SNe do not. Additionally,the centroids of SNe have larger uncertainty because there arefewer epochs to constrain the position and the S/N is lower.Therefore, in modeling the SNe, we fix the location of the SNin R.A. and Dec. across all images (Section 3). While the SNposition is fixed (“fixed-position photometry”), we determinethe position of the stars for each image in order to account for stellar proper motions (“variable-position”). Proper mo-tions of the standard stars, which are estimated by linear fitsto the positions over 3 years of observations, have an RMS of ∼
10 mas per year.In order to be consistent in the application of the stellar po-sition in the photometry, Rest et al. (2014) and Scolnic et al.(2017) run fixed-position photometry on both the stars andSNe. In our pipeline we apply fixed-position photometry tothe SN but we apply variable-position photometry on the stars,and this inconsistency causes a small 1 − σ scatter in the recovered stellar mag-nitudes (hereafter ‘repeatability’) is plotted in Figure 2. Forthe brightest stars ( <
17 mag), the photometric uncertaintiesafter including Poisson noise analytically are 1 mmag, but theobserved measurement scatter is > F IG . 2.— Solid lines designate 1 σ scatter in the recovered stellar magnitude(repeatability) as a function of stellar catalog magnitude for each DECamband. Dotted lines designate the mean photometric uncertainties. There is afloor in the photometric repeatability of ∼ In order to demonstrate the size of the chromatic correctionsapplied to the tertiary standards in the SN fields, we comparethe un-corrected individual exposure (nightly) stellar photom-etry with the FGCM chromatically corrected stellar catalogmagnitudes (Figure 3). Differences are up to 4 mmag over thecolor range of the tertiary standards (0 . < g − i < Image Model Fitting
As in H08 and A13,
SMP uses a time series of image stampsfrom the data located at the position of the SN. We assume thatthe DECam pixel fluxes can be modeled from a temporallyvarying SN flux and a temporally constant galaxy model thatis modeled as a grid of pixels. In order to facilitate model rout et al. : DES-SN Photometry Pipeline and Y1-Y3 Spectroscopic SNe Ia Light Curve Data Release 5comparisons to all images simultaneously, all data images arescaled to a common zeropoint of 31.00 mag. Following H08and A13, the model is re-sampled to compare with the datasetand the data are never re-sampled to avoid correlated noise.A visual representation of the model is shown in Fig. 1. The“Model” images shown on the right hand side of Fig. 1 arecompared to data, and to constrain our model we minimizethe following: χ = (cid:88) i j , n ( S i j , n − D i j , n ) σ sky n , (1)for each pixel labeled with indices i and j , and exposure n . S i j , n are the modeled pixel fluxes and D i j , n are the data pixelfluxes. Equation 1 is weighted by the pre-computed variancein the sky counts ( σ sky n ) as motivated by A13 to preserve sta-tistical optimality for faint sources and avoid potential biasesdue to inaccuracy of the PSF model. However, because thedenominator of Eq. 1 does not include all sources of noise,we modify the photometric uncertainties output by SMP usingboth the analytical expectations of source and galaxy noise(Section 3.4), and we correct our uncertainties using resultson fake SNe (Section 5).For our model S i j , n , we define a temporally varying SN fluxfor each exposure n ( F n ) and a temporally constant grid offluxes ( g i j ) of size N × N ( N = 30). The SN and host galaxyfluxes per pixel are defined as follows: FSN i j , n = F n F n F n (cid:88) k i k j ˜ PSF k i , k j , n e π k i ( ¯ SN i ¯ SN i ¯ SN i − i ) / N e π k j ( ¯ SN j ¯ SN j ¯ SN j − j ) / N , (2) FGAL i j , n = (cid:88) k i k j ˜ PSF k i , k j , n ˜ g k i , k j ˜ g k i , k j ˜ g k i , k j e − π k i / N e − π k j / N , (3)and the model image S i j , n is defined as S i j , n = FSN i j , n + FGAL i j , n , (4)where ˜ PSF i j , n is the Fourier transform of the PSF evaluated atthe location of the SN. We vary the SN sky position in Fourierspace, where the SN point source is represented by a planewave at ¯ SN i , ¯ SN j in pixel coordinates relative to the centerof the galaxy model ( i and j ) which is defined to be the DiffImg
SN position. This formalism allows us to modelthe SN position at sub-pixel locations in Fourier space and toevaluate the likelihood in real space. The floated parametersin our fits are designated in bold font in Equations 2 & 3; theseparameters are F n , g i j , ¯ SN i , and ¯ SN j .We adopt a galaxy model on a grid of pixels with the same0.27 (cid:48)(cid:48) pixel scale as the DECam images. The reference centerof each data stamp is the position of the SN as determined by DiffImg . This position is an average of all epochs for whichthere was a
DiffImg detection. The reference center is at asub-pixel location, so as to facilitate comparison of our modelwith the data, we shift the galaxy model and the SN model foreach exposure by the difference of the center image pixel andthe reference center. This ZP of 31.00 is for internal
SMP computations only; the ZP in thepublic data files is 27.5.
In order to avoid degeneracies between the galaxy modeland the SN flux, we fix the model SN flux at zero for epochsoutside the observer frame range ∆ MJD peak >
300 days or ∆ MJD peak < −
60 days where
MJD peak is the derived date ofpeak flux from an initial light curve fit of
DiffImg photome-try and ∆ MJD peak = MJD exposure − MJD peak . We find that anyresidual SN flux beyond 300 days contributes to negligiblebiases in photometry ( < . F IG . 3.— Nightly (per exposure) tertiary standard star magnitudes com-pared to the FGCM pipeline catalog magnitudes as a function of the FGCMcatalog g − i color. The color binned mean of the magnitude residuals is shownin red. Implementation
We utilize a Markov Chain Monte Carlo Metropolis Hast-ings algorithm (Metropolis et al. 1953, Hastings 1970) to sam-ple the likelihood and we assume flat priors on each of ourmodel parameters with the exception of the SN R.A. and Dec.for which we assume a top-hat prior with radius 2 pixels thatis centered at the location of the
DiffImg fit sky position.For our model image stamps, we adopt a radius of 13 pixels(3.5 arcsec) around i , j , inside of which we compute χ fromEq. 1 using only pixels that fall entirely within the pre-definedradius. For each filter, we have a total of ∼
500 galaxy modelparameters and anywhere from 25 to 500 SN flux parame-ters; one for SN flux in each exposure that falls within ourdefined MJD range over which we fit SN fluxes. For our sam-pling algorithm, we do not employ more complicated algo-rithms such as emcee because the computation requirementsof our likelihood and the number of parameters make run-ning the required 2N walkers intractable. Instead, during thefirst 100,000 steps we optimize our steps in each parameter toachieve between 25% and 75% acceptance rate. We employa Geweke Diagnostic (Geweke 1992) test to ensure that ourchains for the SN fluxes have sufficiently sampled the pos-terior space. Our chains can run up to 2,000,000 steps. Thegalaxy model, which is represented as a grid of delta functionsin Fourier space, has power on all scales which can lead topoor convergence. For this reason we do not explicitly checkfor convergence of g i j , but rather we ensure convergence ofthe FGAL i j , n pixels in a 1 (cid:48)(cid:48) aperture centered at the location ofthe SN.The SMP fits are performed separately in each band. Whilethere could be added benefit in measuring the SN position byfitting all bands simultaneously, atmospheric refraction causes
Brout et al. : DES-SN Photometry Pipeline and Y1-Y3 Spectroscopic SNe Ia Light Curve Data Releasethe position of the SN to be color dependent, which is not ac-counted for in this work. A total of 41,004 jobs were run in-dependently in order to produce griz light curves for the 251SNe in the spectroscopic sample and 10,000 fakes. Each jobutilized a single FNAL processor and could take anywherefrom 5 to 48 hours to fit, with the latter occurring for deep-field z -band fits with up to 750 exposures. The vast majorityof the computation time is in the convolution of the galaxymodel with the PSF for each exposure. To improve fittingspeed, the PSFs were stored in Fourier space and the galaxymodel ( g i j ) is transformed to Fourier space and subsequentlyconvolved with the PSF requiring only n + F n for each expo-sure n by taking the mean of the MCMC chain. The error on F n is the standard deviation of the MCMC chain. For obser-vation sequences with multiple back-to-back exposures, wereport the weighted average flux and uncertainty among theindividual exposures. Uncertainties
Here we describe the treatment of the statistical uncertain-ties within
SMP to which an additional empirically observeddependence on host galaxy surface brightness is included inSection 4.4. There has been debate about the proper way to in-clude Poisson noise of the host galaxy and source in the pho-tometry fits (H08 and A13). H08 weight their fits according toexpected photon statistics, which includes the Poisson noiseof the host galaxy. A13 exclude the noise contribution of thehost galaxy and source in the fitting process. We have cho-sen the latter method (shown in Eq. 1) and correct our outputuncertainties using expected photon statistics after the fittingprocess following: σ = σ SMP fit + σ + σ , (5)where σ SMP fit is the uncertainty derived from the SMP
MonteCarlo chains which were computed using only the sky uncer-tainty, σ source is the Poisson noise of the SN, and σ hostgal is thehost galaxy Poisson noise. The host galaxy photon varianceon exposure n is approximated by σ , n = (cid:80) i j f gal i j , n × PSF i j , n (cid:80) i j PSF i j , n × NEA , (6)where f gal i j , n is FGAL i j , n expressed in photoelectrons fol-lowing: f gal i j , n = FGAL i j , n × ( ZP n − / . × Gain n , (7)and the noise equivalent area is NEA ≡ / (cid:80) i j PSF i j , n . Equa-tion 5 corresponds to our analytic expectation of the photo-metric uncertainties. Finally, we report the weighted aver-age uncertainty among the individual back-to-back exposures.Below we test the accuracy of our photometric extraction andcorrect σ stat for underestimation of the measurement noise. CORRECTIONS AND TESTS ON FAKE SUPERNOVAE
Fake SN Ia light curves are inserted onto DECam imagesat locations of real galaxies. Here we analyze a set of 10,000fakes that were discovered by
DiffImg and processed by
SMP . We optimize our pipeline for minimal photometric out-liers, check for biases in our photometric method, and applycorrections to our photometric uncertainties.
Fake Supernovae
The insertion of fake SNe at the image level and the subse-quent analysis of their measured fluxes is an important testof the photometric pipeline. It allows us to quantify mea-surement biases, compare
SMP uncertainties to the measuredminus true flux differences and determine uncertainty correc-tions, and optimize
SMP cuts to reject flux outliers. We simu-late a sample of SN Ia light curves and insert light curve fluxesonto DES-SN images using the measured PSF. Because weinsert an entire sample of SN Ia light curves, we are able tocharacterize biases in photometry as well as the propagationof these photometry biases to biases in measured distances.A13 moved nearby stars in their images to locations near hostgalaxies and treated them as fake transients, which preservesthe true PSF for each star, but it is difficult to trace photometrybiases to distance biases given that they have limited statis-tics of fake stars and do not model a sample of fake SNelight curve magnitudes. Additionally, A13 did not accountfor a position-dependent PSF when moving stars, whereas themethod described here does.Fake SN light curve fluxes are generated using the Su-perNova ANAlysis software package (
SNANA : Kessler et al.2009) in a Λ CDM cosmology ( Ω M =0.3). Light curve fluxesare overlaid as PSF sources onto the DECam images and pro-cessed with the DiffImg pipeline. A detailed descriptionof the simulation used for the fakes can be found in Sec-tion 2 of Kessler et al. (2019), but here we provide a briefsummary. The fake SNe span a wide magnitude range (from19 th mag to well below the detection limit) and redshift range(0 . < z < . PSFEx , and the flux in each pixel is varied by random Poissonnoise. Since we use a scaling of the modeled PSF to insert thefake transient, rather than the real PSF (i.e. moving real starsin the image), we separately check for potential PSF model-ing errors that are not included as a part of the analysis of thefakes.K15 inserted 100,000 fake SN light curves into the first 3years of DES-SN images. These fakes were used to monitorimage quality and ∼ DiffImg . However because
SMP is computationally expen-sive, for this first DES-cosmology analysis, only on a subsetof 10,000 fake SN light curves were processed by
SMP . Outlier Rejection
In order to reduce the number of photometric outliers, expo-sure quality requirements (cuts) were optimized on the sam-ple of fake SNe. We denote the fraction of 5 σ flux out-liers ( η σ ) when comparing the SMP fit flux ( F n ) to the truefake flux ( F True ). We remove exposures with poor data-modelagreement ( χ /ndof > .
2) and with poor seeing conditions(PSF
FWHM > η σ from 6 × − to 2 × − . Photometry Biases
Comparing the input photometry to the recovered photome-try ( ∆ F = F n − F True ), we measure photometric biases < . ∆ F / F True of rout et al. : DES-SN Photometry Pipeline and Y1-Y3 Spectroscopic SNe Ia Light Curve Data Release 7 F IG . 4.— a) Fractional flux residuals as a function of fake SN magnitude. All host galaxy local surface brightnesses are included. Comparison with theuncertainty in calibration non-uniformity from Burke et al. (2018) ( σ uniformity = 0 .
006 mag) is shown. The shaded regions designate the 1 σ errors on the mean.b) RMS of the pull-distribution as a function of fake SN magnitude. − .
3% at faint magnitudes, which is included in the system-atic error budget of B18-SYS.There are three key differences between the analysis of theDES-SN dataset and that of the fake SNe. First, the astro-metric solution used to insert fakes (K15) is the same solutionthat is used to model the fakes within
SMP . Astrometric un-certainty is not simulated in the fake point sources. Second,K15 use zeropoints that were fit using aperture photometry toinsert fake fluxes onto images, while
SMP uses PSF fitting. Inorder to assess the accuracy of
SMP , we correct for the zero-point difference between the K15 and
SMP . Thus, our resultspresented here are insensitive to incorrect modeling of the ze-ropoint. B18-SYS discuss an independent method for validat-ing the zeropoint and internal calibration uncertainties. Third,the analysis of the fakes uses the same PSF model that wasused to insert the fakes. Inaccuracies of the PSF model arenot simulated in the fakes, and thus in Section 6 we perform acrosscheck of our PSF model.If the
SMP flux uncertainties are accurate, thenRMS( ∆ F /σ stat ) = 1. However, we observe that the RMSof the fakes is slightly above unity as shown in panel b) ofFigure 4. To characterize the excess scatter, we examine thedependence of the RMS on the local host galaxy local surfacebrightness ( m SB ). Host Galaxy Surface Brightness Dependence
We find that there is an underestimation of photometric un-certainties for SNe located in galaxies with high local surfacebrightness, as was seen previously in
DiffImg (K15). Ascale correction ( S ) is computed from the fakes as shown inFigure 5 that is required to bring RMS of recovered fake fluxesas a function of m SB to unity. This dependence (hereafter the Host SB dependence) has been seen in the past (K15, Scolnicet al. 2017). The source of the Host SB dependence is un-clear since we include host galaxy Poisson noise in our SMP uncertainty calculation (see Sec. 3.4). In
SMP , we find nosignificant bias in ∆ F /σ stat as a function of m SB .The inset of Figure 5 shows the results of SMP run on twoexample host galaxies, one bright and one faint. For the brighthost galaxy, visibly poorer χ distributions are seen across theimage stamp and structure can be seen in the residual stamp.To account for the increased scatter as a function of hostgalaxy surface brightness, K15 scaled their output SN fluxuncertainties. In SMP we apply the same method of scalingour SN flux uncertainties with multiplicative corrections ( S ).The SMP light curve photometric uncertainties ( σ F ) are givenby σ F = σ stat × S (8)where σ stat was defined as the co-added measurement uncer-tainty and S is the function of m SB , bandpass, and field shownin Figure 5. DES-SN SPEC SAMPLE Y1-Y3
In this work, we analyze the spectroscopically confirmedSN Ia subset of the data. As described in D’Andrea et al.(2018), 533 transients were targeted for spectroscopic classifi-cation, 251 of which were spectroscopically classified as TypeIa. We have run
SMP photometry on this sample, and showrepresentative examples of our resulting light curves across arange of redshifts in Figure 6. Light curve fits to the SALT2model are included to guide the eye, however we refer to B18-SYS for a detailed discussion of light curve fitting and light
Brout et al. : DES-SN Photometry Pipeline and Y1-Y3 Spectroscopic SNe Ia Light Curve Data Release F IG . 5.— Scale Correction (S) = RMS( ∆ F / σ stat ) as a function of m SB , for 10,000 Fake SNe Ia processed by SMP . The stars on the x-axis denote the mean localsurface brightness in the DES subset for each band. Inset: Examples of high and low m SB galaxies and SMP best fit models, data − model, and χ . curve quality cuts.A table of photometric measurements and uncertainties forthe DES-SN sample is available online in machine readableformat (see footnote on page 3). While all corrections to theflux uncertainties are included, we provide a separate tablelisting the uncertainty scales (S). CROSSCHECK OF THE PSF MODEL
As discussed in Section 3.1, any differences between pho-tometry of the standard stars and the photometry of the SNecan result in photometric biases. We explicitly check for bi-ases in photometry due to potential inaccuracies of the mea-sured PSF model because this is not accounted for in the anal-ysis of the fakes. This check is performed by comparing theratio of the stellar model stamps that were used to computethe zeropoints with the data stamps (model/data). The samemodel/data comparison is made for the
SMP galaxy + SNmodel. Any potential differences between the stellar ratiosand the SN ratios could lead to biases that are not canceled outby the zeropoint. In order to obtain sufficient S/N, we stackthe residuals for many fits where the SNe and stars are bright.In the top panel of Figure 7 we stack model/data stamps for3000 stellar fits of stars (19 < M star <
21) over 25 nights onthree different CCDs. We find that inaccuracies of the PSFmodel are limited to < .
3% in any given pixel. Additionally, as shown in the middle panel of Figure 7, we stack model/datastamps for the DES-SN SNe Ia and their host galaxies forepochs with 19 < M SN <
21 and find similar results althoughit is difficult to assess given the limited statistics of the spec-troscopic dataset ( ∼
300 stacked exposures). Finally, in thebottom panel of Figure 7, we show model/data stamps for fitsto the fake SNe sample. As expected, we do not observe thesame discrepancies between data and model because inaccu-racies in the PSF model are not simulated in our analysis ofthe fakes.To analyze the impact of the observed difference betweenour PSF model and the SN data, we correct the PSF model bythe stacked stellar residual stamps and then re-compute stellarphotometry. We find that this correction results in zeropointdifferences of < . DISCUSSION
The
SMP pipeline developed for DES-SN models the SNhost galaxy and SN transient flux simultaneously in order toextract a SN flux in each exposure. We have used 10,000fake SN light curves overlaid onto our images to quantify po- rout et al. : DES-SN Photometry Pipeline and Y1-Y3 Spectroscopic SNe Ia Light Curve Data Release 9 F IG . 6.— Representative light curves of DES SNe from the DES-SN3YR sample with photometric data provided by SMP and fits to the light curve dataprovided by SALT2 simply intended to guide the reader’s eye. SNe with C3 or X3 in the name are found in deep fields, the remaining SNe are found in theshallow fields. The fields are described in detail Section 2.1 of B18-SYS. Brout et al. : DES-SN Photometry Pipeline and Y1-Y3 Spectroscopic SNe Ia Light Curve Data Release
Stacked Stellar FitsStacked DES-SN FitsStacked Fake SN Fits F IG . 7.— Top Panel:
Ratio of stellar model to DECam data imagefor 3000 stacked cutouts of tertiary standard stars fainter than 19 th Mag.
Middle Panel:
Ratio of
SMP
SN + galaxy model to DECam data image for300 stacked cutouts of SNe in the DES-SN dataset brighter than 21 st Mag.
Bottom Panel:
The same ratio but for the results on the fake SNe Ia. tential biases in our photometry. We find that biases in pho-tometry are limited to 3 mmag, which is small in comparisonto the internal calibration uncertainties described in B18-SYS(6 mmag). Additionally, we find that errors in the PSF model-ing are sub-dominant to the photometric uncertainty budget.Finally, we correct our uncertainties for the host SB depen-dence.
The Host SB Dependence
The host SB dependence was first quantified for
DiffImg photometry in K15 and the excess scatter is also seen in the
SMP results. Because the host SB dependence is not uniqueto difference imaging photometry, we conclude it does notresult from the use of
SWarp (Bertin et al. 2002) which isused to co-add exposures nor is it from hotPants (Becker2015). Because the size of the dependence is similar in allbands, chromatic refraction likely plays a sub-dominant rolein the host SB dependence. The source of this additional scat-ter is likely due to a number of confounding sources similar tothe photometric repeatability floor for the stars. Atmosphericdistortions contribute a chromatic increase in flux scatter andastrometric errors could introduce un-modeled uncertainty inthe host galaxy itself. With improvements to the astrometricsolution expected in the coming analysis of the full DES-SN5 year dataset, we will be able to examine the dependence onastrometric quality.
Comparison To Difference Imaging
DiffImg was designed for DES-SN as a rapid transientidentification and
SMP was designed as a precision photomet-ric tool to be used for cosmology. Because they have been op-timized for different purposes, it is difficult to make a directcomparison. We find that the fraction of catastrophic photo-metric outliers ( η σ ) occurs at 0.02% for SMP in comparisonwith 0.08% for the
DiffImg pipeline. In addition, we com-pare the overall size of our photometric errors and find that theuncertainties output by the
SMP pipeline are slightly smallerthan those of
DiffImg (Figure 8). F IG . 8.— SMP and
DiffImg flux uncertainties with the 1-to-1 line drawnfor comparison. rout et al. : DES-SN Photometry Pipeline and Y1-Y3 Spectroscopic SNe Ia Light Curve Data Release 11
Future Work
A number of improvements can be made to our photometricpipeline and analysis of the fake SNe. There are two main as-pects of our fakes analysis that inhibit our ability to character-ize the full extent of our photometric pipeline. First, we knowthe precise PSF of our fake SNe since we use the same PSF tooverlay the point source and do the
SMP fitting. In the futurewe will vary the PSF and calculate the impact on photomet-ric repeatability, biases and the host SB dependence. Second,the method by which the fakes are inserted onto the images isnot representative of the true astrometric uncertainty becausethe fakes are inserted and modeled in
SMP using the sameastrometric solution. In the future we will vary the locationof the fake point source on each exposure by the astrometricuncertainty. The ability to simulate both of these effects willfacilitate the tracing of photometric biases due to the PSF andastrometry all the way to cosmological parameters.For future stage IV surveys in which calibration uncertaintyin the filter zeropoints approaches the < SMP pipeline, wewill also investigate applying a series of additive flux uncer-tainty floors dependent not only on m SB , but also on observingconditions. Lastly, we will also investigate the effects of bet-ter galaxy modeling and resampling tools such as GALSIM (Rowe et al. 2015). CONCLUSION
We have presented the photometric pipeline for the DarkEnergy Survey Supernova Program and made available theY1-Y3 Spectroscopic SN sample light curves that are usedin the cosmological analysis companion papers. This analysisuses the
SMP
Pipeline to measure fluxes of SNe in their galac-tic environments.
SMP was run on the 251 spectroscopicallyconfirmed SNe Ia and was validated on a sample of 10,000fake SNe Ia light curves injected as point sources onto DE-Cam images. We find that we recover flux values to within0.3% accuracy. We show improvement over the
DiffImg pipeline used for real-time transient discovery, however wefind that we still must correct for the underestimated uncer-tainties in high local surface brightness galaxies. The
SMP pipeline will be tested further on 40,000 fake SNe and ul-timately run on the full five year photometrically classifieddataset of ∼ Brout et al. : DES-SN Photometry Pipeline and Y1-Y3 Spectroscopic SNe Ia Light Curve Data Releasethe U.S. Department of Energy, Office of Science, Office ofHigh Energy Physics. The United States Government retainsand the publisher, by accepting the article for publication, ac-knowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publishor reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes.The UCSC team is supported in part by NASA grantNNG17PX03C, NSF grants AST-1518052 and 1815935, theGordon & Betty Moore Foundation, the Heising-SimonsFoundation, and by fellowships from the Alfred P. SloanFoundation and the David and Lucile Packard Foundation toR.J.F.: DES-SN Photometry Pipeline and Y1-Y3 Spectroscopic SNe Ia Light Curve Data Releasethe U.S. Department of Energy, Office of Science, Office ofHigh Energy Physics. The United States Government retainsand the publisher, by accepting the article for publication, ac-knowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publishor reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes.The UCSC team is supported in part by NASA grantNNG17PX03C, NSF grants AST-1518052 and 1815935, theGordon & Betty Moore Foundation, the Heising-SimonsFoundation, and by fellowships from the Alfred P. SloanFoundation and the David and Lucile Packard Foundation toR.J.F.